Documente Academic
Documente Profesional
Documente Cultură
LIBRARIES
BRARY
NON-LINEAR MECHANICS
N.
MINORSKY,
Ph.D.
J.
W. EDWARDS
1947
ANN ARBOR
Copyright 1947 by
Originally Published,
J.
W. Edwards
Department of Commerce
Lithoprinted in U.S.A.
EDWARDS BROTHERS,
Ann
Arbor, Michigan
INC
FOREWORD
This report aims to bring to the attention of technical personnel,
both of the Navy and of other agencies engaged in the war effort, certain
mechanics does not introduce any new postulates but relies upon the same
The only
difference between these new methods and the old ones lies in the mathematThe old methods, such as that of small motions,
differential equations.
the new methods to obtain more accurate solutions without recourse to arti-
ficial simplification.
A few definitions and examples will be given so that the subject
order is
W
where
z,
*- (i)
& ^&
+
l)
t
+ *-' (()
f+
"* <()a:
Q(()
'
the dependent variable, constitutes the unknown function to be deis the independent variable, for example,
.
The coefficients p x
(0
...
t.
Q(t), designated
t.
According to
whether Q( t) =
non-homogeneous
or Q{
t)
pn{
t)
are constant; the linear differential equation is then said to have constant
coefficients
(a
+6
cos
pt)x
Although
VI
the theory of the Mathieu equation is much more difficult than that of equa-
The basic feature of such equations is that they obey the so-called
principle of superposition
will take place as if each component oscillation with a certain frequency existed alone, not influenced by the presence of oscillations with other frequencies.
ever, one must be certain that the differential equations are linear.
In this
If,
how-
for some other range the same differential equation ceases to give a
in' a
If the problem is investigated within that range the idealized linear differ-
If,
how-
ever, the same phenomenon is studied in a somewhat larger range, in which the
assumed idealizations cease to be true, the departures between the theory and
the observed facts become more and more evident, as the range of the investi-
gation is extended.
problem has been over-idealized, and a more correct form of differential equation must be employed.
usually non-linear.
describe the phenomenon mathematically by means of linear differential equaA few examples which follow illustrate this general situation.
Vll
THE PENDULUM
The exact differential equation of the simple pendulum is
6
-f Li
sin 6
+ 4(6)6 =
*w*f(i-V + ''H
resulting from the expansion of sin
all terms beginning with
6
2
fl
in a power series.
If we assume that
*< =
which gives
(9
[2]
+-%-6 =
= 2nYL/g.
jHl
-g-j.
4-
l(i LH - t) 6 /J
[3]
The spring constant in this case does not remain constant but diminishes with
increasing
6.
of the motion.
of sin
we obtain still greater accuracy but the problem becomes more comIn practice even the first approximation [3] gives an error of
1
plicated.
less than
+ B
+ B2 6 2 + D
h{6)
[4]
B lt B 2 are
viii
0,
where h
centric height.
As
done for a particular ship, the problem is reduced to a simple linear equation
i'e
+ B
+ Dh
[5]
If,
however,
B2 6
is of the
same order as
0,
16
+ B
+ B
;2
2
+ Dh
6]
cannot be avoided.
m
,
Pit)
71
where m, A, p L, and y are constants and P{ t) is a known function of time. Introducing the variable y = x/L, this equation acquires the form
y L(i
- vY
l
.
= QU)
7a;
<f>{y)y
Q(t)
7b
where
<^>{y)
LO
k is a
constant and
< y <
y.
R +
2
+ aR3 ~ R 3 ^-^
b,
[8] y are
where
certain constants.
r +
14 ZK
r + <t>w R =
91
ix
<f>{R)
aR 2
b/R
Sy ~
Equation [9] is also a homogeneous non-linear differential equation, where the non-linearity is found in both the spring-constant
= JR/2R.
Jd + Cia) - Mia) =
[10]
where J is the moment of inertia of the ship about the vertical axis through
its center of gravity,
C d
Ma
leads certainly to incorrect results; in fact this would mean that for an
a,
where
is a certain constant,
increases indefi-
through a maximum for a relatively small value of a and becomes zero thereafter, when the line of action of the resultant of the leeway forces recedes
Mia) =
a -
Jd + C a - iM -
M
2
)
a )a
[11
Ma
x
fi(l
x )x
[12]
(A
+ Bx 2 )x +
[13]
so
program, it was discovered that under certain conditions the blade angle on
the pump controlling the transfer of ballast between the tanks begins to
A detailed
It is suf-
31
of this report.
J<t>
(b
a x)
a3
+ c0 =
[14]
where J is the moment of inertia of the ballast, including tanks and ducts,
a x and a 3 are the coefficients in the expression giving the hydrodynamic
and
how to use these methods, answers can be given only for specific cases.
Each
must face the situation arising from the essential non-linearity of the phe-
The
present report introduces the student who is familiar with standard methods
CONTENTS
Introduction
PART
I
-
Chapter
1. Phase Plane and Phase Trajectories; Linear Oscillator 2. Phase Velocity 3. Theorem of Cauchy; Singular Points; Trajectories and Motions 4. Phase Trajectory of an Unstable Motion; Saddle Point 5. Phase Trajectory of an Oscillatory Damped Motion; Focal Point 6. Phase Trajectory of an Aperiodic Damped Motion; Nodal Point 7. Method of Isoclines 8. Systems with Negative Damping; Proude s Pendulum 9. Remarks Concerning Linear Systems
,
7
9
9 14
16 18 20
'
21 22
24 24 25
Chapter II
10. 11.
12. 13.
lA. 15.
General Properties of Non-Linear Conservative Systems Topology of Phase Trajectories in the Neighborhood of Singular Points A. Graphical Method B. Analytical Method Topology of Trajectories in the Phase Plane. Separatrlx Behavior of a Conservative System as a Function of a Parameter. Critical Values of a Parameter Poincare''s Theorem (14) (15) Motion of an Elastlcally Constrained Current-Carrying Conductor Relative Motion of a Rotating Pendulum
-
25 26
29
32 33 34 37
40 40 4l
Chapter III
16. 17. 18.
Questions of Stability
Introductory Remarks Stability in the Sense of Liapounoff Canonical Forms of Linear Equations; Characteristic Equation; Classification of Singularities (Poincare); Branch Points of Equilibria 19. Stability of Equilibrium According to Liapounoff 20. Liapounoff 's Theorem 21. Equilibrium of a Circuit Containing a Non-Linear Conductor (Electric Arc)
Chapter IV
-
42 48 51 55
62 62 66 68 75 76 77 78 79 82
22. Limit Cycles; Definition; Analytical Examples 23. Physical Examples 24. Topology of Trajectories in the Presence of Singularities and Limit Cycles 25. Further Properties of Limit Cycles; Indices of Poincare'; Theorems of Bendixson
....
A. Indices of Poincare B. First Theorem of Bendixson (The Negative Criterion) C. Second Theorem of Bendixson (5) 26. Parallel Operation of Series Generators 27. Stability of Periodic Motion
Chapter V
28. 29. 30. 31.
87
87 88 95 99 99 101 101 102
105
Introductory Remarks Transition of Singularities. Branch Points of Limit Cycles Self -Excitation of Thermionic Generators Self -Excitation of Mechanical and Electromechanical Systems A. Self -Excited Oscillations of a Mechanical Control System Case 1. Weak Amplification Case 2. Strong Amplification B. Self-Excited Oscillations in an Electromechanical System
-
Chapter VI
32. 33. 34. 35-
Introductory Remarks Lienard's Method Existence of Closed Trajectories in the Lienard Plane First Asymptotic Case: << 1
(j.
Xll
The Parameter p. Is Large 36. Second Asymptotic Case: 37- Limit Cycles in the Van der Pol and Llenard Planes
Ill
113
116
General Remarks. Limit Cycles of the First and Second Kind Differential Equation of an Electromechanical System Cylindrical Phase Trajectories of a Conservative System Cylindrical Phase Trajectories of a Non-Conservative System Closed Trajectories of the Second Kind in Non-Conservative Systems Oscillations of a Synchronous Motor
PART II
-
44.
Introductory Remarks
-
Chapter VIII
Method of Polncare
45. Condition of Periodicity 46. Expansions of Poincare'; Generating Solutions; Secular Terms 47. Systems With Two Degrees of Freedom 48. Stability of a Periodic Solution 49. Limit Cycle and Frequency of a Thermionic Generator 50. Bifurcation Theory For Quasi-Linear Systems 51. "Soft" and "Hard" Self -Excitation of Thermionic Generators; Oscillation Hysteresis A. Condition For a Soft Self -Excitation B. Condition For a Hard Self -Excitation
....
Chapter IX
167
52. Rotating System of Axes; Equations of the First Approximation 53. Topology of the Plane of the Variables of Van der Pol 54. Example: "Soft" and "Hard" Self -Excitation of Thermionic Circuits Equation of Lord Rayleigh; Froude s Pendulum 55- Example: 56. More General Forms of Non-Linear Equations
'
Chapter X
57- Introductory Remarks 58. Effect of Secular Terms in Solutions by Expansions in Series 59- Equations of the First Approximation 60. Non-Linear Conservative Systems .61. Examples of Non-Linear Conservative Systems A. Pendulum B. Torsional Oscillations of a Shaft C. Electrical Oscillations of a Circuit Containing an Iron Core
62. Systems With Non- Linear Damping of a Dissipative Type A. Linear Damping: fix) = Xx 2 B. Quadratic Damping: f(i) = bx C. Coulomb Damping: f ( x ) = i s^u (x) 2 D. Mixed Cases: f (x ) = ax + |3x 63. Systems With Non-Linear Variable Damping A. Van der Pol's Equation
'
B.
64. Existence of Limit Cycles; Systems With Several Limit Cycles 65. Stability of Limit Cycles; Critical Values of a Parameter; Systems With Several
66.
203 206
209
Chapter XI
67. Introductory Remarks, 68. Improved First Approximation 69. Applications of the Theory of the Improved First Approximation A. Variable "Spring Constant" B. Variable Damping
'
Correction For Frequency Approximations of Higher Orders Motion of a Conservative Non-Linear System With a Cubic Term Higher Approximations For Non-Linear, Non-Conservative Systems General Form of Equations of Higher Approximations
C.
209 210 2l4 215 216 217 219 222 223 229
xili
Chapter XII
233 233 234 236 238 239 239 240 24l 242 242 243 246
74. Introductory Remarks 75. Method of Equivalent Linearization 76. Principle of Equivalent Balance of Energy 77. Principle of Harmonic Balance 78. Examples of Application of the Method of Equivalent Linearization A. Non-Linear Restoring Force B. Non-Linear Disalpative Damping
C. E. F.
Non-Linear Restoring Force and Non-Linear Dlssipative Damping Non-Linear Conductors Thermionic Generators
References
PART III
79-
Introductory Remarks
-
Chapter XIII
80. 81. 82. 83.
Method of Complex Amplitudes Elements of the Theory of Linear Circuits Analogies Between Electrical and Mechanical Systems Application of the Method of Equivalent Linearization to the Steady State of a Quasi-Linear System 84. Application of the Method of Equivalent Linearization to the Transient State of a Quasi-Linear System 85. Non-Resonant Self -Excitation of a Quasi-Linear System 86. Resonant Self -Excitation of a Quasi-Linear System ...
Chapter XIV
87. 88. 89. 90. 91. 92.
-
267 269
272 272 274 278 283 284 285
289 289 293 294
Combination Tones; Subharmonics Equivalent Linearization for Multiperiodic Systems Internal Subharmonic Resonance Synchronization Internal Resonance of the Order One Method of Equivalent Linearization in Quasi-Linear Systems With Several Frequencies
-
Chapter XV
93. Equations of the First Approximation for a Periodic Non-Resonant Excitation 94. First-Order Solution of Van der Pol's Equation With Forcing Term 95- Improved First Approximation For a Non-Resonant External Excitation of a Quasi-
Linear System
96. Heteroperlodic and Autoperiodic States of Non-Linear Systems; Asynchronous
295
301
97. Equations of the First Approximation For an Externally Excited Resonant System 98. Fractional-Order Resonance 99- Parametric Excitation 100. Stability of Non-Linear External Resonance; Jumps
....
Chapter XVII
101. 102. 103. 104. 105. 106. 107. 108. 109.
Method of Mandelstam and Papalexi Resonance of the Order n ; Differential Equations in Dimensionless Form Periodic Solutions of a Quasi-Linear Equation With a Forcing Term Stability of Periodic Solutions Subharmonic Resonance of the Order One-Half for a Soft Self -Excitation Nature of Subharmonic Resonance of the Order One-Half for an Underexcited System Subharmonic Resonance of the Order One-Half for a Hard Self -Excitation Subharmonic Resonance of the Order One-Third Experimental Results
-
Chapter XVIII
Entrainment of Frequency
110. Introductory Remarks 111. Differential Equations of Van der Pol 112. Representation of the Phenomenon in the Phase Plane
xiv
Nature and Distribution of Singularities; Transient State of Entralnment Steady State of Entralnment Acoustic Entralnment of Frequency Other Forms of Entralnment
-
Chapter XIX
117. 118. 119. 120. 121. 122.
Parametric Excitation
Heteroparametric and Autoparametric Excitation Differential Equations With Periodic Coefficients Stable and Unstable Regions of the Mathieu Equation Physical Nature of Solutions Topology of the Hill-Melssner Equation Dependence of Heteroparametric Excitation on Frequency and Phase of the Parameter Variation Heteroparametric Excitation of a Dissipative System Heteroparametric Machine of Mandelstam and Papalexi Subharmonic Resonance on the Basis of the Mathieu-Hill Equation * Autoparametric Excitation
References
PART IV
127.
-
REIAXATION 0SCILIATI0NS
381
388
388 391 392 395 396 399 399 400 402
Introductory Remarks
-
Chapter XX
128.
Solutions of a Differential Equation in the Neighborhood of a Point of Degeneration Idealizations in Physical Problems Critical Points of Differential Equations; Basic Assumption Conditions of Mandelstam Remarks Concerning Systems of Degenerate Differential Equations
-
Chapter XXI
133. Periodic Solutions of Degenerate Systems of the First Order 134. Relaxation Oscillations in a Circuit Containing a Gaseous Conductor 135. RC-Multivibrator 136. System With One Degree of Freedom Descrlbable By Two Differential Equations of the First Order
405
410
410 4l4 415 417
Chapter XXII
...
137- Multivibrator of Abraham-Bloch 138. Heegner's Circuit; Analytic Trajectories 139- Transition Between Continuous and Discontinuous Solutions of Degenerate Systems
Chapter XXIII
140. Introductory Remarks 141. Qualitative Aspects of a Mechanical Relaxation Oscillation 142. Mechanical Relaxation Oscillations Caused By Non- Linear Friction
Chapter XXIV
143. Introductory Remarks 144. Elementary Theory of the Clock 145. Phase Trajectories in the Presence of Coulomb Friction 146. Electron-Tube Oscillator With Quasi-Discontinuous Grid Control 147. Phase Trajectories of an Impulse-Excited Oscillator
Chapter XXV
148. 149.
436 436
Parasitic Parameters on Stability of an Electric Arc Parasitic Parameters on Stability of Relaxation Oscillations
438 440
443
References
131,246 278,446
449
Index
INTRODUCTION
Practically all differential equations of Mechanics and Physics are non-linear; in the applications linear approximations are fre-
quently used. The Method of Small Motions is a well-known example of the "linearization" of problems which are essentially non-linear. With the discovery of numerous phenomena of self -excitation of circuits containing non-linear conductors of electricity, such as electron tubes, arcs, gaseous discharges, etc., and in many cases of non-linear mechanical vibrations of special types the Method of Small Motions becomes inadequate for their analytical treatment. In fact the very existence of these oscillations in a steady state indicates that there is an element of non-linearity preventing the oscillations from building up indefinitely, which by energy considerations is obviously impossible. In addition to this there is another important difference between these phenomena and those governed by linear equations with constant coefficients, e.g., oscillations of a pendulum with small amplitudes, in that the amplitude of the ultimate stable oscillation seems to be entirely independent of the initial conditions, whereas in oscillations governed by linear differential equations* it depends upon the initial conditions. Van der Pol (l)** was first to invite attention to these new oscillations and to indicate that their existence is inherent in the non-linearity of the differential equations characterizing the process. This non-linearity appears, thus, as the very essence of these phenomena and, by linearizing the differential equation in the sense of the method of small motions, one simply eliminates the possibility of investigating such problems. Thus, it became necessary to attack the non-linear problems directly instead of evading them by dropping the non-linear terms Although the early discoveries of Van der Pol invited the attention of mathematical physicists to these new problems, very little was done theoretically in the following years in the way of further generalizations of Van der Pol's original theory. On the other hand, the accumulation of experimental data continued at a rapid rate and each additional problem had to be treated mainly on its own merits and by methods which were not unified into one central doctrine. The situation was similar in some respects to that which existed in the early
''linear
stage of development of the theory of differential equations when, prior to the advent of Cauchy's general theorems of existence, a few isolated methods of direct integration were the only ones availahle.
Later, Cauchy's theorems made it possible to obtain a general theory of linear differential equations.
About fifteen years ago certain Russian scientists directed their attention to a further development of methods of Non-Linear Mechanics with a view toward obtaining a more general basis for a mathematical treatment of numerous experimental facts to which modern electronic circuits and apparatus contributed a large part. A line of approach was soon found in the classical researches of Henri Poincare who may be considered as a real forerunner of modern NonLinear Mechanics In his two treatises " Sur les courbes ddf inies par une equation differentielle" (2) and "Les methodes nouvelles de la mecanique ce'leste" (3) the great analyst opened two major avenues of approach to the solution of problems of Non -Linear Mechanics, i.e.,
.
1. 2.
These two trends persist in modern developments of NonLinear Mechanics, frequently supplementing each other in some respects. The adaptation of these general theories, developed originally for the purposes of Celestial Mechanics, to the problems of applied science in general is the work of Mandelstam, Papalexi, Andronow, Kryloff , Bogoliuboff , and a number of other Russian scientists working jointly in this field during the past fifteen years or so. The two major trends referred to, the topological method and the quantitative method of successive approximations, have their respective merits as well as limitations The topological methods are based on the study of the representation of solutions of differential equations in phase space; the latter is mapped by means of point-singularities and certain singular lines so as to obtain certain topological domains in which the form of the integral curves - the phase trajectories - can be investigated by relatively simple geometrical methods. The main advantage of the topological methods lies in the fact that insofar as they deal with trajectories , and not with laws of motion, they make it possible to obtain, so to speak, a bird's eye view of the totality of all possible motions which may arise in a given system under all possible conditions. Just as a topographic map of a locality gives an idea as to its three-dimensional form - its peaks, valleys, divides, and other features - a topological
.
picture of a domain of integral curves permits ascertaining at once in which regions of the domain the motions are periodic and in which they are either aperiodic or asymptotic. Likewise the critical thresholds or "divides" which separate the regions of stability and instability can be easily ascertained by these methods.
The principal l im itation of topological methods, as of other qualitative methods, is that they do not lend themselves readily to numerical calculations, insofar as they deal with geometrical curves, the trajectories, and not with the laws of motion which are of interest for numerical calculations. The quantitative methods, on the other hand, possess the advantage of leading directly to numerical solutions which are of importance in astronomical as well as in engineering applications. These quantitative methods, however, inevitably narrow the field of vision to a relatively limited region of the domain. This frequently limits the grasp of the situation as a whole, particularly if the system possesses critical thresholds, the separa trices, the branch points of equilibrium, etc . , at which the qualitative features of the phenomena undergo radical changes; such critical conditions are of common occurrence in Non-Linear Mechanics.
Both methods, however, very frequently supplement each other. The topological method permits a rapid exploration of the whole field of integration, and the quantitative method leads to numerical results, once a particular range of the problem has been selected for study. Another point of great importance in Non-Linear Mechanics is the question of stability. The fundamental theorem in this connection is due to Liapounoff (U). In Non-Linear Mechanics this theorem plays a role similar to the Routh-Hurwitz theorem for linear systems. Its formulation is closely related to the question of singularities of differential equations. An interesting feature of this theorem is the fact that, in a great majority of cases, it permits establishing criteria of stability for a non-linear system from the equations of the first approximation, which are linear; this fact simplifies the problem appreciably. This report reviews the progress accomplished in Non-Linear Mechanics approximately up to 19^0; its preparation was greatly facilitated by the availability of the following two works in the Russian language
1.
2.
"Theory of Oscillations," by A. Andronow and S. Chaikin, Moscow, 1937* "Introduction to Non-Linear Mechanics," by N. Kryloff and N. Bogoliuboff, Kief, 1937.
On a number of questions, particularly those treated in Part III, the original publications had to he consulted. In view of the fact that the literature, in Russian alone, comprises more than 2000 pages, to say nothing of the earlier publications of Poincare', Liapounoff, Bendixson (5), Birkhoff (6), Van der Pol, and others, no attempt was made to present a complete account of what has been accomplished in this field. For that reason the report is limited only to a few selected topics which seem to offer more immediate applications on the one hand, and which do not require too abstract mathematical generalizations on the other. It is believed that in its present form the report is within the grasp of an average reader having a general knowledge of the theory of ordinary differential equations in the real domain. The report as a whole falls into four major subdivisions: Part I is concerned with the topological methods; its presentation substantially follows the "Theory of Oscillations" (l9) The material is slightly rearranged, the text condensed, and a number of figures in this report were taken from the book. Chapter V, concerning Lie'nard's analysis, was added since it constitutes an important generalization and establishes a connection between the topological and the analytical methods, which otherwise might appear as somewhat unrelated. Part II gives an outline of the three principal analytical methods, those of Poincare', Van der Pol, and Kryloff -Bogoliuboff .* Part III deals with the complicated phenomena of non-linear resonance with its numerous ramifications such as internal and external subharmonic resonance, entrainment of frequency, parametric excitation, etc. This subject is still in a state of development, and the classification of the numerous experimental phenomena is far from being definitely established. Much credit for the experimental discoveries and theoretical studies of these phenomena is due to Mandelstam and Papalexi, and the school of physicists under their leadership.
The first four chapters of Part III represent the application of the quasi -linear theory of Kryloff and Bogoliuboff to these problems and the last three concern the developments of Mandelstam, Papalexi,
Andronow, Witt, and others, following the classical theory of Poincare'. Finally, Part IV reviews the interesting developments of L. Mandelstam, S. Chaikin, and Lochakow in the theory of relaxation oscillations for large values of the parameter . This theory is based on the existence of quasi-discontinuous solutions of differential equations at the point of their "degeneration," that is, when
(j,
* During the preparation of this report there appeared a free translation of extracts of
the Kryloff and Bogoliuboff text by Professor S. Lefschetz, Princeton University Press,
1943.
equation "degenerates" into one of lower order. A considerable number of experimental facts are explained on the "basis of this theoretical idealization. In going over this report the reader will notice that the electrical examples are more numerous than the mechanical ones The reason for this situation is twofold. First, electrical non-linear oscillations constitute generally useful phenomena that are utilized in radio technique, electrical engineering, television, and allied fields, whereas most of the known mechanical non-linear phenomena are of a rather undesirable, parasitic nature. Second, the determination of the parameters and characteristics is generally much easier in electrical than in mechanical problems, particularly when a mechanical system is relatively complicated. This does not mean, of course, that this state of things will always persist; in fact, mechanical engineers are becoming more and more concerned with non-linear problems (7), (8), (9), and the lack of any appreciable progress at present is not due to a lack of interest on their part but rather to the absence of a theory sufficiently broad to cover the various cases encountered in practice. It is difficult to write a relatively short report on a subject of this scope in a form that will be satisfactory to all readers. A mathematician will undoubtedly find a series of drawbacks in the presentation of purely mathematical matter; the proofs of a number of theorems are omitted. Frequently more rigorous theorems and criteria are replaced by rather intuitive definitions and statements, and so on. It is probable that the mere idea of applying the relatively complicated and laborious methods of Poincare' and Lindstedt for the purpose of explaining the well-known performance of a thermionic generator or of a simple mechanical system with non-linear damping or a non-linear spring constant may appear to an engineer pedantic and hence superfluous. On the other hand, it cannot be denied that in the past theoretical generalizations have been always most fruitful in the long run, although during the initial stages of development, they appeared to contemporaries somewhat laborious and confusing. It is sufficient to mention as an example the modern theory of electronics with its numerous ramifications in the fields of radio technique, television, controls, etc. In this case, between the spectacular experimental discoveries of Hertz, Marconi, Fleming, Lee de Forest, and others, there intervened the less spectacular but not less important theoretical work of Richardson, Langmuir, and Shottky on thermionic emission, which, in turn, was based on the earlier statistical theory of Maxwell and Boltzmann. This permitted later a more rigorous quantitative treatment of these phenomena. Very likely the present trend toward a codification of theoretical knowledge in the field of non-linear mechanics will bring about a further progress resulting from a more general viewpoint on the whole
.
subject. It must "be "borne in mind, however, that the period of codification of non-linear mechanics has existed only for ahout the last fifteen years. Hence, we are witnessing at present only the important initial stages of these studies rather than their final formulation. It is hoped that this review will attract the active attention of applied mathematicians, physicists, and engineers, for the new methods seem to affect all branches of applied science either "by offering more accurate solutions of old problems or by making possible an attack upon new problems beyond the reach of the older mathematical methods The preparation of this report was facilitated by valuable discussions and suggestions on the part of a number of mathematicians and mathematical physicists who showed an interest in this work. The writer is indebted to A. Franck and M. Levenson who were most helpful in connection with the critical editorial work and to Professors E. H. Kennard, W. Hurevicz, S. Lefschetz, and J. A. Shohat. The cooperation of the Applied Mathematics Panel of the National Defense Research Committee, Dr. W. W. Weaver, Chairman, and of the Applied Mathematics group at Brown University, Dean R. G. D. Richardson, Chairman is especially acknowledged.
PART
CHAPTER
1.
mi +
where
ex
0,
(c
>
0)
[1 .1
m
A
and
are constants.
Its integral is
x
= A
cos (cj
<(>)
where
and
t
</>
tions at
= 0; and
rax + ex =
kinetic,
[1
.3]
1-2 -ymx
2
,
Putting x = y,
form
where a = V2 h/c
and
fi
= VZh/m.
Equation
[1 .4]
is
(cj
<j>)
and noting
= A
coso) t;
Acj
smoj
[1-5]
The elimination of
72 + 7#-Y=l A A u)
[1-6]
= Vc/m.
[1 .4]
and [1.6]
Equation [1.6] was obtained from the solution of the dynamical Equation [1.1], whereas [1.4] results from the law of conservation of
energy, which is the first integral of the dynamical equation.
In both
press,
simply,
[1 .4]
plane.
cos
cj
is thus represented in
described by the representative point R situated at the extremity of the radius vector
r.
The projection of
R on
of the oscillator, and the projection on the y-axis gives its velocity y = x
at the instant
t.
another ellipse E x is obtained, concentric with the first one and having the same ratio of semi-axes /S/a = Vc/m = w
,
Thus, in this
case,
.1
Figure
.1
one to visualize at once all possible motions which may arise in a system
with zones separating motions of one type from those of an entirely different
type.**
Such initial conditions which are represented by points situated on the same trajectory are to be excluded inasmuch as they represent the same motion with a different phase; see Section 3.
** The representation of motion by trajectories in the {x,x) phase plane will be most frequently used Unless specified to in what follows. In some cases, however, different variables may be preferable. the contrary, by the phase plane we shall always mean the plane of variables x and i.
2.
PHASE VELOCITY
If r is
to R,
v =
dr/dt = ix + jy
With
where
Aq smu
Au
-*
cos cj Q
[2.1
0,
and cos u
being excluded.
= 0; this
This con-
of the ellipse.
point
V with
consideration.
3.
theorem can be found in any textbook of Analysis; for this reason it is sufficient to give here only its formulation.
order
j^
t
= f
x, x,
[5-1]
xQ
dx
dy-,
dy_,
,,
.,
Analytic at a given point means admitting a Taylor series about the point.
10
of-
the system
-jf
In applications,
t
fid,
Vi, V 2
>
Vnh
1, 2,
[3-3]
dynamical system.
Of partic-
ular importance is the case of [3.1] and system [3-2] in which the function / / i) , x does not depend on time t, and is a function of x, x, or y x y2
-,
yn,
alone.
question is "autonomous," that is, one in which neither the forces nor the
constraints vary in the course of time.
Henceforth we shall be concerned mainly with systems with one degree of freedom which are represented in the phase plane by phase trajectories
as previously explained.
|f
-*$.,>!
fj-M
>']
= Q (x, x)
dx
P(x,y)
dV
Q(x,y)
r, s i L ^ OJ
which gives the slope dy/dx of the tangent to the phase trajectory passing
through the point (x,y).
We shall often speak of the point R(x, y)
as moving along the trajectory*
,
point is governed by the law expressed by [3-^] or, more particularly, by Since time does not enter explicitly at the right in [3-4], the gen[3.4a].
eral solution of [3.4] can be written as
x
= =
<f>{i
x
x
y
y
= =
x(t)
ip(t
yit)
(x ,y
at
may also
).
11
tories [3-5], whereas the analytical methods outlined in Part II will be con-
In some cases it is
behavior of a given dynamical system; in such cases [3-5] gives a qualitative idea of the nature of the trajectories in various regions of the phase plane,
whereas Equations
conditions.
[3.*+]
possesses certain critical thresholds or "divides," in which case the application of [3-5] permits a preliminary qualitative study of the nature of the
trajectories in various domains of the phase plane before the actual integration of the dynamical Equations [>iMJ is undertaken.
In the following, we shall make use of the concept of singularities
of differential equations, and therefore we shall find it expedient to make
the following definitions.
A.
A point (x
Q(x
From these definitions it follows that an ordinary point is characterized by a definite value of the slope of the tangent to a phase trajectory passing through that point.
of the
with components dx/dt, dy/dt which determine a non-vanishing vector for any ordinary point [x,y) of the field.
)
Q{x y P(x ,y specifically, if the vector field is a velocity field, it is clear that a
=
,
= 0,
we have For a singular point {x ,y More hence, both components dx/dt, dy/dt vanish.
)
We
are thus led to identify the singular points in such a case with the points
of equilibrium of a dynamical system.
12
is a sing-
that a proper trajectory, that is, one which consists of more than one point,
t
= +
o
t
or
= -
= -
= +
or
It is to be under),
we
vortex point
Point
traj'ectory.
V in Figure
Sec-
tion
x = 0.
nection with equations of a more general form, in which case the trajectories,
while being closed, are not necessarily ellipses.
In all cases, however, a
vortex point
conditions
a.
tories.
.2.
A3 A
= +
<,
as shown in Figure
S and
two
A 4 S, approach
Figure 3-1
jectories,
S for
= -
which is equivalent to
A 3 and A 4
S.
Between these
13
S.
The repre-
A2
A3
for the four regions. In Section 4 we shall see that this sit-
Figure 3-2
ex =
of unstable motion.
ion for the existence of a saddle point will be given for systems of the
form [3.4].
3.
K 2 shown
in
Figure 3.2.
winds around a focal point F an infinite number of times as the point R ap-
proaches F.
= +
<,
If
jectories, as shown by the arrows in Figure 3.2, is reversed so that the tra-
x =
of a
2
{
cj
existence of a focal point in the case of the system [3-4] will be given in Section 18.
4.
nodal point
is a singularity which
curves Kj and
K of Figure
2
3.3,
so that when r
->
is approached
for
= +
or
= - oo,
Figure 3-3
14
2bx +
o)
x =
when
in Section 18.
It may be useful to add a remark concerning the physical interpre-
passing through an ordinary point {x Q ,y of the phase plane, we frequently understand by this that we wish to consider the motion as starting from given
Initial conditions x
,
= x
passes one and only one trajectory, is equivalent to saying that if the inithe subsequent motion is uniquely determined.
In fact, if
t
,
Geometrical-
passing through the point (x ,y ), although there exists an infinity of possible motions (C) on this trajectory when the arbitrary
is varied.*
It can also be stated that the solutions of [3-4] form a two-
4.
- ex =
0,
(c
>
0)
[4.1
= Ae rt + Be~
x,
rt
[4.2]
t
where
r = +
Yc/m.
The solution
thus,
f-
on ac-
Equation [4.1]
x
c
In what follows we shall use consistently this notation, i.e., C will designate a geometrical curve, the trajectory and (C) a motion of the representative point R on the trajectory C.
15
whence
dy dx
For
x = 0,
c
x y
[4.3]
0,
0,
0,
it is vertical.
For x =
= 0,
Figure 4.1
obtained by putting C =
which gives
y
The representative point R arriving from infinity will approach the origin
x = y =
and then will depart again into infinity following the direction of
the arrows
be
the angle measured from the position of stable equilibrium and y x = x x the
V* + V{x
) x
[4.4]
where
~ y
and
Vix^
0.
- V{tt)
> 0.
continues rotating in the same direction, i.e., y x keeps the same sign but
16
=
2.
7T.
h
1
- V{tt)
< 0.
close to x
3.
tt
h - V[n) = 0.
In this discussion the angle x x was measured from the stable equi-
librium position.
x lt
measured from
the point of unstable equilibrium S, the results obtained from the equation
of energy acquire a simple graphical representation shown in Figure 4.1
.
bolic trajectories situated in the upper and lower quadrants of Figure 4.1 those corresponding to Case 2 are represented by hyperbolic trajectories in
the right and the left quadrants of that figure; and the asymptotic Case 3,
in which the total energy h is just equal to the potential energy of the sys-
AjS and A 2 Sof Figure 4.1 along which the motion of the representative point
is asymptotic.
The inverse process, i.e., the motion of a pendulum placed initially in the unstable position of equilibrium, is also asymptotic; it is repre-
It is
seen that four asymptotic trajectories approach the saddle point, namely,
and
A2 S
for
= +
=,
= -
<x>,
5-
+ 2bx + u*x =
[5.1
provided
< 0.
x e
costal + a)
[5-2]
17
where x
o)
1
= Voj
- b
[5-1] and [5-2] and consider the representation of the motion in the phase
plane.
dx Jt
'
dy = ~ _ 2 2by _J X Jt
,
_
'
.,,
[5 3]
whence
dy_
= _ 2by + u Q x
y
dx
ential equation of the first order and can be integrated by introducing a new
+ 2bxy + w
v =
= Ce
tan
-l y -2
bx
">
" lX
[5-5]
dr/dt = ix + jy is obtained by substituting their expressions obtained from [5-2]. The absolute value of the
phase velocity is
\v\
yoj x
+ Abxyu
(1 4-
46 )y
-
[5-6]
The phase velocity is zero only at the origin, x = y = 0, of the phase plane,
which point coincides with the singular point of the differential equation as
seen from [5 A]
Equation [55] can be transformed into polar coordinates. The 2 left-hand term can be written as {y + bx) 2 + {u 2 - b 2 )x 2 = y + bx) +
(
cj
2
.
oj^x,
= y + bx,
[5-5] becomes
= Ce
tan
v = p
ip
sin ^,
= C e^
x
C = YC
X
[5-7]
6;
here however,
18
y--x
The
coordinates (from
y to u,
v)
original coordinates
x,
>
0,
correspond-
Figure
5-"
\Jj.
+ u
equation y + Zbxy + co x One may surmise that the point F towards which the spiral converges is an equilibrium position. We recognize in this
= C.
tions it is apparent that any trajectory of this kind approaches the focal
point for
= +
<=o;
of phase trajectories since through every ordinary point of the plane there
6.
u)
>
1
0,
- r
and
Tl
r2
Ae
Be
where
.
and
Differentiating we have
Ar
e~
Br 2 e~
Eliminating time
(xr l
+
x
y)
= C(xr 2 +
y)
,
[6.1]
v =
xr
+ y, u =
xr 2 + y
we get
[6.2]
Ca U
where a = r z /r
1
>
~
1,
r 2 being" the
Equa-
tion [6.2] represents parabolic curves tangent to the u-axis at the origin;
a for dv/du = Cau
1
is zero when u =
since a
>
is a
19
For C =
0,
i.e., v = 0; for
-* o
)/w
>
0.
These
the {x, y) -plane, the phase trajectories have the configuration shown in
Figure 6.1b.
The w-axis, v =
y = 0,
is represented in the
{x
-
xr x
xr 2
The trajectories
;
hence,
they will
become parallel to the v-axis, hence the asymptotic branches of phase trajectories approach parallelism with the line y =
-
xr 2 in the
i.e.,
{x,
y)-plane.
=.
the line y
0,
at
right angles.
We also find that the locus of the points in which the trajecrx + r2
Figure 6.1
point.
20
7-
METHOD OF ISOCLINES
Let
|f
-.,,),
|f -(..,)
dx
that of the trajectories.
P(x,y)
"
*_
U '^
Suppose that
isoclines.
s *.> -
'""
it; 1 .!
This equation clearly defines a curve in the (x,?/)-plane along which the
slope, dy/dx, of the trajectories remains constant.
dinary points since at singular points the slope of trajectories is indeterminate, so that [7*1] has no meaning at such points.
curve F{x,y) = a is traced, one draws along its length small lineal elements
a.
a lt
These
Starting from a point [x ,y a continuous curve can be traced following always the direction of lineal elements of the field. The curve so
)
obtained is clearly the trajectory passing through the point (x ,y ). The method is very valuable in cases where the explicit form of the solution of
the differential equation is not known.
It has been applied,
for instance,
- nil -
x )x
[7-2]
y\
^(l
x )y
dy dx
m(1
x )y
[7.3]
21
The only singularity of [1.2] is clearly the origin, x = y = 0; hence the method is applicable
Equation [7-1]
(n
a)
[7.*]
fix
For a fixed n and for a number of different values of the slope a, a series of curves [7-^] are
Figure 1.1
(1
for
fi
existence of a closed trajectory C to which the non-stationary spiral trajectories C" and C" approach both from the outside and from the inside of C.
The reader can easily check, by this method,
the principal types of
= 10.0.
8.
From
its jusb.
In
>
corre-
For
< 0,
proportion to velocity.
In electrical problems negative damping plays an im-
portant role; it is generally associated with circuits containing non-ohmic conductors, which exhibit a decrease in voltage
Figure 1.2
22
Froude (12) (13) discovered a similar effect in the case of a pendulum mounted on a rotating shaft with a certain amount of friction. It was obW.
served that a pendulum of this kind begins to oscillate with gradually in-
creasing amplitudes.
In addition
by the equation
l<f>
b<t>
C4> =
0,
<f>)
of angular velocities
of the pendulum.
Expanding the
function M(u>
bj>
+ C0 =
M'(u)
- 0I'(u)
that is,
l'<j>
M'(cj)\4>
c<t>
= Miu)
[8.1
The constant term M{u) on the right merely displaces the position of equilib-
of
<f>.
If,
i.e.,
were negative.
9.
with the representation of motion by phase trajectories; nothing new has been
learned so far, but the known facts were merely presented in a different way.
The connection between the singular points and the state of equilibrium of a
system will play an important role in what follows and it is useful to mention
at this stage that a linear approximation will have considerable importance in establishing the criteria of stability as will be shown in Chapter III.
25
problems studied so far, although quantitatively the relations may be somewhat different.
Thus,
in non-linear conservative systems the trajectories
around
vortex point are still closed curves but not necessarily ellipses.
Likewise the unstable equilibrium in non-linear systems is still characterized by a saddle point but the asymptotes of the latter are not straight
lines as they are in linear cases.
We shall investigate these questions in
detail in the following chapter but it is worth-while mentioning now that insofar as their local properties are concerned, non-linear systems behave not
These new closed trajectories are entirely different from those of linear
systems.
motion of a non-dissipative harmonic oscillator, whereas a corresponding pattern in a non-linear, non-conservative system is that represented by this
special closed trajectory of a new type, the limit cycle, as defined in Chapter IV.
The theory of limit cycles and associated phenomena is, thus, a do-
CHAPTER
II
The law of
differential equation.
10.
fix)
[10.1
It
o,
+00).
y;
fix)
[10.2]
IM
y
dx
[10 3]
.
The velocity
of motion is given by the ordinate y of the trajectory, and the phase veloc-
ity is
ds
~di
s^Tj'
,|/i
e^f
{
net]
From [10.3] it follows that the trajectories cross the x-axis iy = 0) at right angles and have horizontal tangents at points x which are roots of f[x), provided the trajectory does not cross the axis of abscissas
[y = 0) at these points.
= 0.
0,
f(x) = 0,
and y = d x/dt
= 0.
tion of equilibrium.
25
11.
A.
GRAPHICAL METHOD
For conservative systems the problem is simplified owing to the exFor a system of unit mass we have
2
\y +
V(x)
[11.1]
,
and the potential enwhich expresses that the sum of the kinetic energy, ?y ergy. V{x), remains constant. By definition V{x) = - J f(x)dx, where f{x) is
the restoring force.
V{x) <
since the
that is V'[x) = 0.
equilibrium.
The phase trajectories are symmetrical with respect to the x-axis. The trajectories cross the x-axis at right angles, that is dy/dx =
0,
2.
o;
provided y
h.
by drawing an auxiliary
(0)
i
h -
V{x),
y/-/2
"^T^^.
+
Vh
V(x)
i
\
Vh-V(x)
i
ii
ii ii
/ M2
i
show
--"""
(b)
balance of energy h
V[x) for a
Figure 11.1
26
The difference h
V(x) is
Mv
Figure 11.1b gives the phase diagram (coordiThe phase trajectories are
v tories in the right (or left) quadrants of Figure 4.1 representing the motion
in the vicinity of a saddle point when the pendulum does not have enough en-
ergy to overcome the "potential barrier" and is thus unable to go over the
In this case,
assumed in connection with Figure 4.1 is waived and the discussion, therefore, can be made general.
If the value of h is increased {h = h 2
),
the
point
is shifted to
(point Af2 ).
Figure
11 .2
minimum for x = x
x =
xl
;V{x
l )
0.
V(x)
is a singular point,
1 )
dinates
Vh
V{x) in Figure
then
Figure 11.2
ANALYTICAL METHOD
The topological picture of the
vof
Assume first that V{x) has a minimum for x = x lt and expand f(x) in
x = x1
=
-
a x (x
f
Xj)
3-
(x
x x)
1-2-3
(x
Xj)
[11.2]
Since V{x) =
constant of integration h
we get
27
V(x)
h,
F2
(z
Xl)
T^B
=
ix
Xl)
[n.3]
where
ai
/*C*i)
= -
V'ixJ;
a2
/"(x a )
= - V"\x^
x =
[11 .4]
+ ,
y =
O+77, this gives, upon the substitution of the value of V(x 1 + ) as given 2 by [n.3] into the integral of energy y /2 + V(x) = h, the equation
L 1
1-2
lie
=
1)
[11.5]
ax #
Y
0.
= h
1 )
have
=
at x = x
Since V[x
< 0.
1 )
is minimum,
V'{x
and
V"{x
>
Therefore a
7 "(is J
0,
2
For
h = h
an isolated point
=
-
rj
= 0,
trajectories.
For h
a >
[11 .53,
is
l
which is of the form
r,
u2 ^i_
i
[11.6]
2
'
2- + A- = 2 2
t 2
i
-"
where
= 2a; n
= Za/laJ
time.
2.
= 0.
ak
z
=
1
and
a k 0.
since
= V(x) lies
For h = h
rj
= 0,
)
a >
0,
[11 .6]
to the
k +
order of
becomes
T
oscillations.
even for small oscillations.
+
(fc
+ D!
_a
[11
'
6]
which represents a closed curve differing from an ellipse even for very small
It would be erroneous,
28
the phase plane will have the appearance of closed curves surrounding each
other around the vortex point V, Figure 11.2b, as previously found by the
graphical method.
If the potential energy V(x) has a maximum for x = x x the procedure
is similar, with the exception that a
ax
2
?7
1
= a2 =
ak
= 0)
is now positive,
/2
a,
form
m
with the previous notations.
77
,2
-*!
n.7]
= + Ya{$ and
77
Va[t as asymptotes.
77
is,
clearly
saddle point.
the maximum of potential energy are unstable and are represented in the phase
='=
a k _\
= 0;
a k + 0), we find by
a k + 2>
'
'
'>
^-
n this
fc
+1
ttfc$
(fc
+1)!
= a
[n.8]
such as asymptotic
;M
r
h = h.
1
/I h
1
= h"
r
i
flection point at x = x
y/V2
=-V"[x
l )
0.
term a k
of even order [k = 2, 4,
i"
as follows from
iz
Ah in this case.
the trajecx
of
Figure 11.3
29
V with
a saddle
point S.
proaching the point VS, and reverses the direction of its motion also in an
asymptotic manner.
The motion is ultimately unstable; this instability is
cence of a stable vortex point with an unstable saddle point were contaminating the process with an ultimate instability.
rium upon the coalescence of singularities is too complicated, and will not
be attempted here.
Summing up the results of this analysis we are led to the following two theorems of which the first is due to Lagrange and the second to
Liapounoff
THEOREM
is stable.
1.
If
V(x)
is
minimum
at the point x
the equilibrium
THEOREM
2.
If
12.
SEPARATRIX
ferent types.
V{x) + h.
between point
and
and to the right of point 13h > V(x) for all values of x.
2.
This corres-
ponds to the motion of a pendulum having an initial energy h greater than the
V mm
z
(x),
= h
]50
y/V2
ii
Figure 12.1
thus possible.
point
The tra2
are
and d
Figure 12.2
vortex point 4.
4.
If in a certain region,
e.g.,
- oo,
e.g.,
between
unstable equilibrium.
5.
h =
V[x x ) at
a certain x = x 1
3
The line
= h 3 is tangent to the
of its maximum.
and
is a saddle
,
point.
the
31
2,
gradually increases in
z
- h s is reached the
motion loses its periodic character because the time of approach to the saddle point is infinite.
"islands," is limited by the singular trajectories issuing from the saddle point
3-
The same conclusion is also valid for the separatrices issuing from
the other two asymptotes of the saddle point 3 and enclosing the vortex point
4
the
3
saddle point disappears and the phase trajectory encircles the saddle point
If,
however, the
and 4.
typified by those in the upper and lower quadrants, Figure 4,1, on the one
hand, and those in the left and right quadrants, on the other hand.
Further-
more, we are now able to correlate radical changes occurring in the topology
of phase trajectories with a critical value of
a
certain quantity h 3
and
11
in Figure 12.1a,
the separatrix
an d 11.
For h =
Ah (Ah >
0)
lines,
For h - h 3
Ah three
"islands" of periodic trajectories will appear around each of the three vortex points 8, 10, and 12.
We thus meet a situation which will be of great importance in what
follows.
of the phase trajectories undergoes a qualitative change" for a critical value of a certain parameter whenever we encounter two entirely different as-
The value
of the constant of energy h in the above examples illustrates this situaIn these examples the critical or "bifurcation" points at which these
tion.
changes occur are associated with certain values of the constant of energy
tion we shall consider a still more important case when a certain parameter
number
of
52
Before closing this qualitative study of the topology of trajectories in the phase plane it may be useful to mention the analogy existing
In both cases
e.g.,
13-
Consider
depending on a parameter
the solutions of
the differential equation represented by the trajectories undergo variations. By definition we shall designate as ordinary values of A those values which
continuous variation of X corresponds to a continuous variation of the trajectories in the phase plane without any radical changes in their topological
features. The values X; of the parameter X, at which the topological struc-
values .**
ft \\ J(x,k)
= - dV(x,X)
ox
* r
13. 1J
values of the potential energy, that is, by equating the expression [13.1] to
zero.
x,
For a
,
given value X
x2
x3
are
ix
33
It was
shown in Section
11
V xx
{x,k) .*
(Vxx > 0; f x < 0), the equilibrium is stable; hence the point of equilibrium
is a vortex point.
Figure 13-1
Furthermore for fx > the equilibrium is unstable. points of equilibrium are given by the equation
f(x,k)
and the criteria of stability by
Thus
fx (x,k)
Differentiating the equation f{x,k) =
dx
~dk
|
with respect to k we get
[13.2]
f x (x,k) f x (x,k)
at which two points
[13.3]
and B coalesce,
point x
Point D on Curve M.
x3
,
For k > k
equilibrium
x 3 and a
Point E on Curve M.
,
When k = k lt
double root x
dx/dk
oo,
The value of k = k
and x 2
For
indeterminate.
Poincare gives a
for
In the following the symbol /_ designates the partial derivative of / with respect to x.
34
with k fixed.
In such a case
f(x,k) decreases and hence f x {x,k) < 0, that is, V xx (x,k) > 0, which indicates the existence of a minimum of potential energy and, hence, stability.
The opposite condition occurs when the region of f{x,k) >
curve,
i.e.,
is above the
instability.
FAD and
("echange des
The points
of equilibrium in
librium exists.
A few simple examples illustrating the topological method outlined
in this and in the preceding sections are given in the following sections.
14.
and
carrying a current
i,
nite length carrying a current /; let a be the distance between the wall
constraining
/,
x being the
Wand
i.
is lim-
only.
the
is
/<3,X>
= -
*(*
^)
x)
[1A.1]
where k = 2Iil/k is a parameter; the first term kx is the force due to the
elastic constraint, and the second kk/{a
-
35
m:
It is equivalent to the system
k
y>
= m
a-x
ax + x)
ax +
k
(
r-,
i,
^i
t1^.3]
From
[1
dy dx
X [14.4]
y(a
y = 0),
given by
t hese p oints
9
y%-
X.
coordinates of
X < a /4.
Both roots
Xj
Substituting
V xx
(x,X), we have
for x x =
b,
fx {x lt k) <
0,
V"
IX
(x 1 ,X) > 0,
therefore,
Similarly, for x 2 = |- +
b,
f x (x 2 ,k) >
0,
V xx {x 2 ,k)
<
0,
and there-
point S.
the phase trajectories approaches infinity; hence the trajectories have the
asymptote.
We obtain its
mx -
/(x,X)
[14.5]
my +
2
-zkx
+ fcXIogU -
x)
[14.6]
0,
=-~+6.
Substituting
36
[14J]
Hence, the equation of the separatrix is
1 my
2
-(f +)
+*Xlog^
'A
14.8]
6
It
(a)
(b)
Figure 14.1
the motion is
for
jectories exhibit
asymptotically.
2.
k > a /4.
ax + k
has no real
The electrodynamic force exceeds the elastic force for all points of
fy
k approaches zero;
both singular points, assuming that the approach takes place from the region
2 where X < a 2 /4, approach each other and coalesce for k = a /4.
The separatrix
37
the mass of
Figure 14.2
mQ
a sin 9,
pendulum is
mQ
sin
cos 6.
mga
sin 6.
The
M{6,X) =
where X = g/Q 2 a is a parameter.
mQ
a (cos6
X) sinO
[15-1]
when
tion Is
*.%**
<n-
IB where / =
dulum.
6
mfl a (cos0
,2
k) sin 6
[15. 2]
go nd
't
ma
a;
Io)
- mQ
(cos$
X)sinO
[15-3]
Figure 15-1
whence
dto
m:2 a
(cos 9
X)
sin
de
Iu
[15-4]
The singu-
0,
0;
u>
0,
tt
(J 3
0, 9 3
cos
[15-5]
38
ly large value of Q.
by equation
f(6,k)
= mQ 2 a 2 (cos
A)
sin
[15-6]
= 0,
=
1
tt,
and cos
= A.
Figure
5 2
repre-
is shaded;
Figure 15-2
f{0,k) <
= cos
-1
0.
A,
0,
= 0,
77- ,
= -
7T,
f(0,k) >
(0 = 0,
A =
),
(0 =
7T,
A =
),
7r,
A =
at which there occurs the "exchange of stabilities" of Poincare. the curve cos
= A,
Its expression is
2/ sin
(
V
mS2
a
"
-f
cos
)-
[15-7]
The equation of
the separatrix is obtained from the condition that the phase trajectory passes
= t n,
0=0
u>
= mQ
a j
[si sin
+ 2A(cos0 +
Figure
T
1)
[15.8]
(B,
3
2
0=0, 0=0,
2
for which h = -
mkQ
its equation is
u>
^^[sin
+2A(cos0 -
1)]
[15-9]
0=0
for
becomes a
39
i2
# 0;
at the same
i
Vi and
around
-v
S~ ""v
\
C \
+ 1
\\ sN\v
/
T
)l
vv\ /
the coordi-
//
\\^~
s
xV
= 0.
/ _^ \v
Figure 15-3
9=0=0,
A and
decreases for
9=0.
u> i
the separatrix
+ 7T/2
is crossed,
the
-jr/2
X_
s *-~
~"*
\
-^
\
e
^-j
^
< X <
1
Figure 15.4
are P ss i b l e
a.
b.
Two periodic
)
and one
around the combination of one saddle point S and two adjoining vor-
: ,
V2
see
Chapter VII.
For X =
0,
i.e.,
Q >
tt/2
both separatrices
the
positions.
For
>
n,
The values
X=-1,X=0,
CHAPTER
in
QUESTIONS OF STABILITY
16.
INTRODUCTORY REMARKS
In the preceding chapter the question of the equilibrium of conser-
vative systems was investigated, and certain definitions were formulated. The
general problem of stability was studied by Liapounoff (4).
In this chapter
problem of the stability of equilibrium, postponing the question of stationary motion to a later chapter. The study made so far is still incomplete
since it concerns only periodic trajectories of a very special type, namely,
properties:
a.
= +
or
or for both.
periodic motions.
c.
from infinity. We shall be interested primarily in the stationary states of dynamical systems, that is, in the above specified Cases a and b.
As regards to
Case
c,
applications.
The tra-
The mo-
tions of Type
be defined later.
resentative point R approaches the focal and nodal points if they are stable
41
= -
o)
case do not approach this singularity although the potential energy is mini-
These considerations
17-
point R on this trajectory clearly represents a periodic motion. rical formulation of stability of the motion in the
sense of Liapounoff can be visualized as follows.
abruptly into R', a point which lies on a neighboring closed trajectory, C', as shown in Figure
17.1Let us consider the "perturbed motion" of
R
Figure 17
1
on C.
= RR',
origin-
ally small, remains small throughout the subsequent motion, the motion is stable in the sense of Liapounoff; if, however,
after a certain finite time, this condition ceases to be fulfilled, the motion is unstable.
It is to be noted that this statement requires that the curves C
t,
be fulfilled.
Only in a purely theoretical case when the motion of R takes place along the stable separatrix or asymptote, A X S in Figure 3.1, of a saddle point, the latter may be considered as a stable singularity. In Part IV we shall see that under special conditions R may follow a trajectory situated in the neighborhood of a stable separatrix and may give the impression for a limited time that a saddle point is a point of stable equilibrium. Hence one could better describe the stability of a saddle point by introducing the term almost unstable singularity.
42
Assume that
quantity A
r.
If,
for
= 0,
the points
R and
It follows that,
if the motion is to
radius in order that the distance between these points remains small for all values of
t.
We conclude therefore,
sense of Liapounoff only when they are strictly isochronous, that is, when
the period of motion of
R on C
C.
In gen-
vative systems so that such motions are generally unstable in the sense of
t,
although they may be considered as temporarthe motions are almost isochronous in some
region.
We shall see later that there exist special periodic motions ob-
and y = y(t, x
T.
2/(0)
We
if,
to any given
e
number
,
>
0,
another number
)\
|x
)\
x
\
<
e,
all
\y
t
,
<
implies \x(t, x
,
x(t, x 0> y
<
fi;\v(t, x
y(t, x
< v for
,
where
are the
definition
18.
CANONICAL FORMS OF LINEAR EQUATIONS; CHARACTERISTIC EQUATION; CLASSIFICATION OF SINGULARITIES (POINCARE); BRANCH POINTS OF EQUILIBRIUM
We shall now investigate the problem of stability of equilibrium
more systematically by showing its relation to the nature of the singularities of dynamical equations.
Let us consider the system of differential
equations
x
= ax +
by;
ex
+ dy
[18.1
^
where
a,
b, c,
d are constants.
ex
dx
ax
+ dy + by
[18.2]
= ax +
by
P'(x,y);
ex
+ dy + Q
(x,y)
[18.3]
with
dy dx
ex
ax
+ dy + + by +
[18.4]
where P'{x,y), Q'{x,y) are polynomials which have neither constant terms nor
linear terms in x and y.
= y =
= ax +
y,
6
Py;
rj
yx
6y
8 5
.
a P
with
y
6
=
0* where a,
(3,
= SJ;
= S 2??
[18.6]
where S
= ax +
= yx +
and
|
6y
[ 1
8 7
.
1 ]
and
77
from Equations
8 6
.
S^iax + Py)
S 2 (yx + dy)
= a(ax +
by)
+ P(cx + + 5(cx +
dy)
[18.8]
= yiax +
by)
dy)
(a
- S x + pc =
)
x )
0;
0;
y (a
- S2 +
)
dc
=
[18.9]
)
ab + P(d - S
yb + 6(d - S 2 =
6
44
quadratic equation
a
S
b
- S
[18.10]
or written explicitly
S(a +
d)
{ad
cb)
[18.11]
The
ct/fi
# S 2 and
a *
y
Conversely, if
y
6
= S 2 = a - d,
= 0,
form.
a
<5
0,
6$ A
A
-
A
/?
+
-
a.
7]
A
1
[18.12]
y = cA,
6
where A =
-
a
y
yS|
If we choose a =
cA,
)A,
[a - S 2 )A,
written as
=
Equations
[18.6].
[1
(S 2
a)
(S
a)
r\
cr\
[18.13]
By a similar procedure
can be reduced to a canon-
8 3
.
= SJ -
c(S 1
- S2
1
[cP'U.Ty)
)
(S 1
- a)Q'(^n)
[18.14]
S2
r]
c(S 1
- S 2 [cP'U,r]) +
)
(S 2
aJQ'U.ij)]
If the roots
then it can
be shown,
that
and
rj
Letting
= u
iv,
S 1 = a 1 + ib lf S 2 = a x
ib lt and equating
45
axu
bxv
-zQ'(u,v)
<-t
[18.151
v
26jC
cP'(u,v) +
L
(aj
a)Q'(u,v)
l
2(aj
a)u + 2b
v;
cu
[18.16]
and
77
in terms
into [18.13]. Once the possibility of the canonical transformation has been es-
tablished we can proceed with the analysis of the various cases arising from
the nature of the roots of the characteristic Equation [l8.ll].
We assume
* 0.
6
hereafter that S 1 # S 2 We further assume S S 2 * 0, that is, y a = 0, 1 8 reduces, if a # 0, to the simple equation When 2
.
a p
<5
dy dx
acx + ady
a(ax
by)
a
of the
THEOREM
1.
When
the roots
0.
same
sign, the
system [l8.l]
_ S
~dj
s1
[18.17]
I
Cr
[18. 18]
where a = S 2 /S 1 >
curves.
and C is a constant.
1,
For a >
ordinates
C->.
corresponding to
For a <
77
lar curve
= 0,
corresponding to C =
0.
follows from the canonical Equations [l8.6] that the representative point in
the phase plane approaches asymptotically the point
77
for
point is unstable
46
THEOREM
[l8.l] has a
2.
When
the roots
system
saddle point at $ = y =
In this case
drj
rj
[18.19]
where a =
s2
Si
>
0.
Integrating, we have
[18.20]
The phase trajectories are hyperbolic curves referred to the axes as asymptotes.
close to the -axis, will move away from the origin as follows from the first
Equation [18.6].
proach the origin, as follows from the second Equation [18.6], and then move
away from the origin along the branch approaching the -axis asymptotically
as shown in Figure 4.1.
previously defined.
the neighborhood of a saddle point are deformed and the asymptotes are in-
THEOREM
3.
When
0.
the roots
[18. l]
has a
focal point at * = y =
For real
complex,
and
77
Putting
u + iv and
77
[18.6] become
-7 at
tt = at
dv It
S, L
(a,;
+ ibMu +
l
iv)
[18.21]
du
It That is
= S
r]
(ttj
ibjiu
iv)
= a,w
v
b,v;
77 = dt
a,i;
b,u
[18.22]
Thus,
In the
dv du
ax u
+ M
it>
b^v
47
cos 0, v = r sin
4>
we find
dr
dj>
18.24]
&i
whence
r =
Ce
'
Thus,
0,
is an asymptotic point,
the focal
v,
add-
+ v
= p,
we find
1
dp
dt
CtiP
[18.25]
For a
<
focal point.
For a 1 >
focal
5-"
Figure 18.1
<
0,
one has a
> 9
one nas an
by putting p =
{a + d),
a
c
becomes then
+ pS +
18.26
48
as rectangular coordinates,
1
For
< 0,
I.e.,
ad
cb < 0,
= 0,
Hq =
regions
1
.
2 p > 0, p < ^q; the roots in this case are conjugate complex with
0,
focal points.
2.
p > 0, p
> hq; the roots are real and negative; this zone (oblique
p < 0, p
< ^q; the roots are conjugate complex with R[-S lj2
,
> 0,
p <
0,
> hq; both roots are real and positive; hence this region
ABCD
to B,
one passes from the domain of saddle points to that of stable nodal points
and from there to stable focal points; proceeding along BC the stable focal
CD
is inverse to AB.
gion is unstable.
The threshold p
hq =
between nodal and focal points, hence between the aperiodic motions and os-
19.
corresponding to
= 6(e)
such that
<
t
<
for
< +
oo
provided |x(<
)
- x
\
<
x
\
x
\
49
from that position will also remain small in the course of time.
We shall
let x = x
+ ,
||
be-
-/<*>
where f(x) is assumed to be analytic at the point x =
x its
x
.
H9.1]
Substituting for
|f
therefore f{x
/<*
*)
^ / (x o) +
*/'(<*>)
+ |t/" (;c o) +
,
'
'
[19-2]
is a singular point,
and
= 0.
4f = dt
where
aj
o'j*
2
2
'
[19-3]
f'(x Q ),
a2
=
2!
Upon dropping
in
9 3
ne has
[19.*]
ff-M
problems, the information obtained from the linear equations of the first
approximation is sufficient to give a correct answer to the question of stability of the non-linear system.
The solution of [19-^] is
|
Ce
'
where a
= f'(x
).
Hence, acit is
cording to Liapounoff, if a
unstable; finally, if a x
<
= 0.
For the present we shall limit ourselves to the more general case when a 1 #
50
's
line of argU'
[19-2] by ,
3
one has
dt
= a^ 2 +
a2
+
.
F(|)
19.5]
and F"(0) = 2a 2
This gives
dp
dT
where p =
2
1-2
t F"(6$)
[19.6]
it
-p
and
< 6 <
that is,
if a x <
~2
0,
F"{9$) <
0.
Hence, if p =
neighborhood of
if ^"(0) >
x =
Likewise,
stable.
f{x) is plotted
x.
= 0.
to the left of x
to the
con-
right of x
( )
Liapounoff s method
Mx N
in
by the tangenl
11
at x a x ix
provided a 1 #
0.
The di-
H
\
i.e.,
stab!
equilibrium at
I
(c)
(from negativ<
Figure 19.I
51
x > x
),
stable.
with f(x
= 0,
the question of
In fact,
ignate this case, shown in Figure 19.1c, as a half stable equilibrium, i.e.,
stable for x < x
to a
x
= 0,
20.
LIAPOUNOFF' S THEOREM
In physical problems the Liapounoff theorem is generally encountered
W'r
tions [20.1
]
'<*>'
ff"=
,
f'*
20
which are represented in the phase plane by trajectories, and the formulation of the Liapounoff theory depends then on the analytical form of Equa-
= Q{x1 y)
P(x,y)
dx
[2Q2]
The positions of equilibrium as was shown are identified with the singular
points.
when
P and Q
P
a
ax + by
ex + dy
is,
,y
of equilibrium is clearly
0,
Q{x,y) =
0.
In order to be
departure
xo
(,r?)
x =
+ I*
y = Vo + V
pn
drj
+ 2p 12 +
2
?7
p 22
r]
+
+
.20.3]
,
<7n
Ql2^V + 922^
52
where the non-written terms of Taylor's expansions are at least of third degree in
f,
x]
and a =
Px {x
,y
);
Py (x
,y
Qx
yn
d =
Qy
yQ
If
acteristic equation corresponding to the equations of the first approximation are different
from
zero, the equations of the first approximation always give a correct answer to the question of
stability of a non- linear
system.
equilibrium is unstable.
The starting point for this proof is to reduce the non-linear system to a canonical form by making use of
[
and [18.14],
'
Thus we have
du T7 =
bxU
(p n U
+
,
P 12 UV
+
,
2v
)
p 22 V
+
.
'
=
where p' ,n
[20.4]
S2 V
(q
nu
q 12
uv +
q 22 v
);+
-,
q'
22
|
2 where p = u +
||-
S u
x
+ S2 v2 +
4>{u,v)
[20.5]
2
.
curve
(t>[u,v)
0.
Write
v)
S u
x
S2
ip(u,v)
whence
1
2
dp_
dt
SjU* + S 2 v
,
i/f(u,v)
[20.6]
v =
S,
2
0,
where
< -AS-^u
0,
+ S2 v
point u = v =
where 0(0,0) = 0.
}|^|i{|<V
Let
6
+ S y>|
If the point
(20.7J
is initially
placed inside
53
Figure 20.1
Figure 20.2
In fact,
for all points of 6, with the exception of the origin, dp/dt <
so
that lim p = p
p
^
t
If
> 0,
then for
2
^
^
if
<
.
\S l
\S 2
\,
that
IS^
+ S2 v
2
|
\S X
U
dp
+ v
2
)
|Sj|
A
t
for
From [20.7],
we have
1
>
Po
\Sx\
> 0,
dp
Pit) P
(*o>
dt
dt >
y-^-lsJ^ 4''
4
<n)
which is impossible.
When S
z
= <j>{u,v)
has a minimum
v
<t>(u,v)
= 0.
with the
exception of point
a
As the region
Figure 20.2.
<5
such
<5
exists.
Fur=
ther,
Since
<f>[u,v)
dp/dt >
+ v
uf
+ ve
and p e
54
after a finite
Hence,
and
with two distinct tangents at the origin. region S has in this case two zones, say
II,
and
in which
</>{u,v)
>
Figure 20.3
IV,
in which
<t>{u,v) <
Before pro.
ceeding with the proof, it is first necessary to obtain the sign of d 2 p/dt 2
Differentiating Equation [20.5] and substituting du/dt and dv/dt from [20.4],
one obtains
L CL
jT2
Sx u
+ S2
+
=
</>
(u,v)
[20.8]
The surface
<j>
for which
</>
(O,O) = 0.
> 0.
We now can
gion
take a circle situated inside of both regions, S' and S, and drawn
It will be shown that it is impossible to deter-
initially at any point of this region except the origin, it would never reach
the boundary of
e
<5
exists.
<j>[u,v)
Since
>
0,
<5
for which
R
2
d p/dt
>
by assumption.
R will
e
start moving with an accelerated velocity and will reach the boundary of
in a finite time, as dp/dt is monotonically increasing.
This is contrary to
Hence, it is im-
<5
equilibrium is unstable.
The same conclusions reached in Cases 1,
2,
and
hold when we
ib
55
-j^ = a l %
v1
-jf
a1vl
[20.9]
vx
.
a l {u l
Vj)
i//
(u 1 ,v 1 )
[20.10]
where p = u
gree.
2
x
+ v
One can show as above that the existence of either maximum or minimum
If a
x
< 0,
> 0,
enables one to
If these conditions
are fulfilled, the theorem gives a correct answer at once, if not, one is
ABCD
intersects
Equation [18.11] become purely imaginary, and the equations of the first approximation of Liapounoff cease to be applicable.*
Omitting this exceptional case which generally corresponds to a
branch point of equilibrium (Section 27) in a great majority of practical
problems, the Liapounoff theorem yields the conditions of stability In a
An example is given in the following section.
21
where A is an electric
By Kirch-
Va
V = L-lf + at
di
lP(i);
E = RI +
V; I
dV
C-TTat
[21.1]
* It must be noted that, in the general theory of equilibrium, Liapounoff also considers in detail a series of particular cases which lie outside the range of validity of equations of the first approximation. We shall encounter one such case later. It is impossible, however, to give a full account of the Liapounoff theory here.
56
R
i
y-VWVW\j nmwmw\-\
J
Va
<M0
E
i
A
\ 'n
_L
Figure
21 .1
Figure
21 .2
Eliminating
dV
dt
E - V
Ri
di
V -
iff(i)
RC
dt
[21.2]
V = E - Ri
From Figure
21 .2
V =
0.(t o )
[21.3]
Va
V= E
Let
V and
are now
[21.4]
i
and
and
V = V +
Developing the function
find
<P(i
v;
ip{i)
we
+j) =
's
ip(i
jip'(io)
[21-5]
Following Liapounoff
mation are obtained by substituting the values [21.4] and [21.5] into [21.2].
Upon canceling the steady-state terms, we find
dv
v
C
''
dj_
di
where p =
i
RC
dt
[21.6]
ip'{i
The quantity
ip'{i
= p varies
with
is falling, p < 0;
designated as
57
[21 .6]
is
S* +
and its roots are
[JC
-- 4+
--
S
)
Zcd^)
2
[21.7]
*h
L + RCp +
2RCL
~ 2RCL VL
+ (RCp)
- 2RLCp - 4LCR 2
[21.8]
p,
dimensional manifold
C,
p) of
The parameters R, L,
AB and BC
of
Diagram in the
21 .3.
(R, p)
-plane,
Figure
square root in
as
[21 .8]
can be written
Figure
21 .3
(L
.2 - RCpY - (2RVLC)
- RCp) 2 - (2RVLC)
<
[21.9]
L - RCp + 2RVLC =
0;
L - RCp - 2RVLC =
[R, p) -plane,
[21.10]
1
shown by Curves
Equations [21.10].
The
and 2 respectively,
tribution of the complex roots of the characteristic equation, that is, the
focal points.
To distinguish between the stable and unstable focal points
58
L + RCp =
Since R, L, and C are positive, Equation [21.11] implies that p <
[21 .11
0.
It is
and p as
2
3,
Figure 21.3-
and
focal points.
range,
condition.
sible.
+ p = 0, Curve 4.
For when
+ p < 0,
the
It is
is positive.
Thus,
VL/C,
between Curve
The region
ABCD
in the direction
I.e.,
saddle points
>
->
>
un-
>
59
2.
L
that is
[21 .12]
RC ~
In the {L,p )-plane Curve
1
[21.13]
[21.13].
point.
Furthermore, the
R 2 C).
Straight Line
2,
L + RCp
= 0,
Figure
21 .4
2 2
is thus tan
/?
~\/RC.
stable focal points from that of the unstable ones; the self -excitation of
One concludes, there.
p > R.
R,
shown as Line
in Figure 21 .2,
Straight Line
60
discriminates easily between the stable and the unstable nodal points. Points
0(p = L
3.
0) and
M(p
R, L =
i.e.,
of focal points.
b.
+ p = 0.
points on the one hand, and the focal points on the other, represent
the zones of distribution of nodal points,
stable or unstable.
,
four-dimensional manifold
[R,
equilibrium in
in general,
R appears
as the
Va
of the arc
(tan a = R)
We shall
Figure
Points
1
,
21 .5
and
3,
p >
the
Point
is thus a po-
of equilibrium, Figure
Since in the
[V,
61
this point is expressed by the fact that |p| < R; that is, the absolute value
of the tangent to the falling characteristic of the arc is less than the pos-
equilibrium \p\< R, the equilibrium is stable; if \p\ > R, it is unstable. This criterion was found useful for a majority of applications. Later we
shall see that certain ambiguous cases were discovered in which this criteri-
In fact,
from the diagram of Figure 21.3, for instance, it follows that within a certain range of p <
0,
points).
points); in this latter region the phenomenon is stable in the sense that it
does not "run away," e.g., blowing fuses, but rather approaches a stationary
periodic motion.*
The above
study indicates that absolute instability, i.e., the zone of saddle points,
\p\
of unstable nodal points wedging in between the zone of saddle points and
otner singularities.
21 .5,
other two parameters L and C which appear dynamically in the process, that
is,
when
stability by taking into account both static and dynamic factors of equilibrium.
In each particular problem one has to ascertain first which particu-
* See Chapter
IV.
CHAPTER
IV
22.
occur around the vortex points; the phase trajectories form a continuum of
closed curves enclosing either one single vortex point or more generally, an
odd number of vortex and saddle points, the number of the former exceeding
the latter by one.
12.1
which is composed of several groups of vortex and saddle points, the con-
tinua of closed trajectories form "islands" in the phase plane limited by the
conditions; by varying these continuously one obtains the continuum of trajectories as long as they remain within the limits of the domain of periodicity
determined by separatrices. We shall now consider motions of an entirely different type observed
in autonomous non-linear and non-conservative systems.
Let
represented by equations
t
),
y = y( t)
= +
or for
= -
oo
ing on
approaches C.
t
>
one
is at
distance
L e
>
Intui-
A limit cycle C
if it is approached by trajectories
t
= +
it is called unstable
C
,
<*>,
and half
stable, or semi-stable
for
= +
= -
or vice versa.
Limit cycles met with in practice have the property that they are
63
proaches
t
or for
- oo,
erty in mind we can also say that the ultimate periodic motion on a limit
Cycle does
In this
possible to establish
~ = -
Vx*
(x
+
,
2\
[22.1]
dy
dt
Vx 2
(x
)]
dx 77 = dt
(1
2s
))
-jdt
dy
-x
y (1-r)
.
roo o1 22. 2 J
x,
r
y,
[22.3]
the second by x,
subtracting and
xy - r 6 we have
9
[22.4]
64
_ ,. Furthermore,
|
dr
-
r$
^m
dr
-
^
log
dr
2
-pry =
2*
log
A
being the initial value of
where
-=
1
-fr
o
is an integration constant, r n
Thus
7V-7 Ae +1
(
[22.5]
For
>
00,
r =*
<
>
of the circle r =
9=1,
the trajectories
= 0.
From
is
as a stationary solution.
equations
dx
-77-
= -
+ x(x
(
+
.
, s
1)
[22.6]
-j
a
t
dy
+ y{x +
,\ -1)
= r(r 2
1);
= -
[22.7]
[22.8]
Yl
- Ae 2(
If r
where
re
is
1
constant of integration.
<
then
<
and
=
1
.
->
VI
It is seen that for
t
\A\e
= 2t
[22.9]
-> - 00,
1
r =
inwards.
For
->
o,
r = 0,
65
For r
>
1,
we have
>
so that
For
-> -
outwards.
=
at
x (a
+
2
2/
)Mx
2\"2
/
+2/
2
2
-l)
-1)
1
[22.11]
dy
-tt at
= yix
,
+n)Mi +y
.
\2
-x
-,
Setting r
2
,'-i>Vg-i
1
[. 1?]
it
2
)
.
But
dw
u(w
l)
du w
du
it
+
1
du
7
(w
7T5" 2 l)
2cu
_ ,.
w
u
^ - 1 =
-
log 6
C +
That is
u
/"
>
[22.13]
Putting u
= v we obtain
(f + i).->-c."
r =
1
[22.14]
=0.
1
If r =
e,
Thus as
0,
i.e.,
>
>
is a stable limit cycle for the spiral trajectories inside the circle.
1
If,
however, r =
e,
then
>
0,
For
i>
>
r ->
~]
from
66
2t
>
0,
that is
- oo,
circle
r =
u(x
4-y
,
-Da;
.,
sin
5
1
a;'
y*
g-
[22.15]
dy
"77"
a;
/ /y (
+2/
,2
Dy
-,\
sin ^
= nr(r - D
sin
r2
r =* 1
[22.16]
which
6]
1).
were obtained in this manner and the resulting equations were then transformed into rectangular coordinates; hence the apparent complications.
It
must be borne in mind that the unearthing of limit cycles known, or suspected,
to exist in a given system is more difficult.
23.
PHYSICAL EXAMPLES
67
The
is of
4>
which a similar condition exists initially, the amplitude of oscillation eventually reaches a limit cycle.
A linearization of this kind, while giving an indication as to what does not give any information as to the final state of the
In other words, it does not permit the determination of
happens initially
stable oscillation.
the ultimate limit cycle towards which the initial process approaches asymp-
totically.
"cause," etc.
However, in view of
The final
amplitude depends only on the parameters of the system, but not on the initial
conditions.
A commonly encountered mechanism in which limit cycles exist is an
The expression "linearization" used here simply means dropping the non-linear terms from the differential equation. A somewhat different meaning is attached to this word by Kryloff and Bogoliuboff; see Chapter XII, Part II.
68
e.g., weight or main spring, released by the escapement; the impulses so re-
leased replenish the energy lost per half cycle of the oscillatory system
and thus allow for the "closing" of the phase trajectories which then become
limit cycles; see Part IV.
In fact,
if a clock is wound from rest, it is im-
Perhaps it is not too great an exaggeration to say that the principal line of endeavor of non-linear mechanics at present is a search for limit
cycles.
Mark (20) gave a theory of the performance of the heart considered as a relaxLikewise, V. Volterra ation oscillation mechanism possessing a limit cycle.
(21
)
24.
tion that a closed curve C in the phase plane be a trajectory is the existence
of at least one point singularity of a definite type inside it.
Figure 24.1
Figure 24.2
69
is a
Figure 24.3
almost self-evident.
-*
-*-
- o.
-* - oo
There is no es-
sential difference in considering nodal points when the nature of the trajec-
If the time origin is selected when the representative point is at the point
A,
and for
point A.
certain instant and disregard its past history, the state of the system is
then determined by the positive half-trajectory
{t
0)
"originating" at the
point A.
Very frequently this viewpoint is useful when we deal with the im-
Disregarding what occurred during the period of discontinuthe behavior of the system from that moment.
at
A and study
With this remark in mind we can say, for example, that to a set of
initial conditions represented by the point
a
half-trajectory winding onto the stable limit cycle C from the outside and
B there corresponds a
70
situations when a number of other singularities are present but for our immediate pur-
pose it will be sufficient to consider the simple case of one singularity surrounded by
one, or several,
limit cycles.
Figure 24.4
is surrounded
Cj,
etc.,
lines represent the stable limit cycles and those in broken lines, the un-
stable ones.
The fact that we assume that the limit cycles are circles is
In order to
wardly and outwardly stable and an unstable limit cycle is both inwardly and
outwardly unstable.
We can now formulate the following theorem* given here without
proof.
unstable and a cycle which is outwardly unstable is followed by one that is inwardly stable. The point singularity at the center is to be considered as a
degenerate limit cycle possessing only the outward stability (or instability).
We shall consider first a few typical examples which stress the sig-
case arises when a point singularity being originally unstable becomes stable
or vice versa.
C/
unstable, C 1 stable,
etc.).
It is apparent that,
This theorem is a particular case of a more general theorem formulated by I. Bendixson (5)- One can also find the proof of the proposition that two adjoining closed integral curves cannot both be stable in a recent paper by N. Levinson and O.K. Smith (22).
<
71
tionary motion will be reached not only when the system starts from rest but
also when it starts from any arbitrary initial conditions which are repre-
sented by a point of the phase plane inside the first unstable limit cycle
'
such as A.
limit cycle C
spontaneously by the system starting from rest since the oscillation cannot
develop beyond C
on which it becomes stationary.
the trajectories may approach the limit cycle C 1 if they originate from the
between C
'
and
C/
It is thus seen
C/
"barrier" for the initial conditions from which various stable limit cycles
such as Cj, C 2
is called soft
if
conditions into the annular region between two consecutive unstable limit
cycles in which
is situated.
in the following
it
manner.
stable limit
is said to
an unstable singularity.
In all
other cases
is said to
excitation is illustrated in Figure 24.5, in which the system is in stable equilibrium when at rest..
it cycle C
,
'
between F g and
Figure 24.5
72
(b)
(O
(d)
Figure 24.6
represented by a point
24.5,
A inside C
'.
A will approach
nates either at B or at D, which represent points of the annular region between the divides C
'
and
Cj,',
be of different kinds.
transition from stability to instability or vice versa, limit cycles may vary
in size, a stable limit cycle may coalesce with an unstable one with the dis-
appearance of both, limit cycles may shrink and coalesce with point singularities modifying the nature of the latter, etc.
Assume that there is an unstable singularity F M surrounded by a staas shown in Figure 24.6a.
;
'
between C
',
73
as shown in Figure 24.6c, although the actual trajectory of the system con-
tinues to be C
One possibility of frequent occurrence in physical problems is the case in which there exist no limit cycles beyond this stage.
Everything then
takes place as if the stable and the unstable limit cycles upon coalescing
approaches the stable focal point as shown in Figure 24. 6d, so that the selfexcited process disappears gradually.
The coalescence of limit cycles with point singularities will be
therefore,
x of
minous path traced by the electronic beam on the screen will give directly
the phase trajectory of the process.
"
|
I'l'l
trajectories.
Consider the
coupling LL'.
The oscillating
Figure 24.7
74
R2
It is
apparent first that the potential difference across R 2 represents the oscillating current
i
ditional details not shown in Figure 24.7, it was possible to start the phe-
nomenon from a given point (i,V) of the phase plane and also to vary the
Oscillograms a,
b,
c,
Oscil-
Oscillogram
amount of regeneration through a relatively weak coupling LL' below the critical value at which the energy input into the oscillating circuit becomes
Oscillogram
represents oscillation
Figure 24.8
75
latter unstable.
ferring the initial conditions to the different points of the phase plane the
stable limit cycles C
1
tory approaching C 2 from the outside has been recorded on the oscillogram.
25.
There exist criteria which rule out limit cycles in certain cases.
The method most frequently used for the establishment of the existence of limit cycles within a certain domain is based on the theorem of Bendixson
formulated below.
that Bendixson'
theorem
cannot be applied to systems with more than one degree of freedom; furthermore,
Poincare
has given a series of necessary criteria for the existence of limit cycles,
based on the theory of indices associated with closed curves.
The whole situation when considered from the point of view of de-
termining limit cycles can be best described in the words taken from the
"Theory of Oscillations" by Andronow and Chaikin:
"The present status of the theory establishing the
Indices of Poincare,
B.
trajectories of Bendixson,
There exists
We shall' not
go into this subject here but shall mention it in a later chapter where the
76
A.
INDICES OF POINCARE
direction.
x,
of a
RM
with R as
RM
RM
x'
"index
j "
of a closed
->>
number
of
complete revolutions
of
RM
The
RM
->
larity is meant the index of a closed curve surrounding the singularity and
lying in a vector field determined by the phase trajectories.
closed trajecb,
Figure 25-2a,
1
,
Figure
(a)
For such a^field the line integral fFds over a closed curve is zero.
77
1.
The index of a closed curve not containing singularities is zero. The index of a closed curve containing several singularities is
2.
always +
4.
The index of a closed curve, with respect to which the field vec1
By the second
conclusion the number of enclosed singularities must always be odd, the number of singularities with index +
by one unit.
12,
1
B.
periodic motions.
Let the motion in the phase plane be given by the equations
x
P(x,y);
if
= Q(x,y)
dP/dx
+
[25.1]
the expression
dQ/dy does
not change
a domain
of the
in that
domain.
/(Prf^-Q <**')-//(
with respect to this circuit.
dP
6Q
+ff)*<
[25.2]
y x dt)
within the area limited by the closed circuit, the integrand (dP/dx + dQ/dy)
changes its sign.
This is contrary to the assumption; hence, the closed
. .
78
curve cannot be a phase trajectory which implies that no periodic motion can
exist in D. One can also state a series of additional criteria based on the
impossible.
C.
<
2.
D limited by
two closed curves C x and C 2 , as shown in Figure 25.3, within which the half-
trajectory is confined and has no singularities either in D or on its boundaries, then the theorem asserts that there exists at least one stable limit
cycle in D.
For r <
1,
1,
dr^/dO < 0.
79
D or on
its boundaries.
Hence, there
As an example of such
yu(l
x )x
=
More specifically, for
ii
this equation admits a closed trajectory differing very little from a circle
of radius r =
2
D limited
and r 2
= 3,
circle.
If, however,
conditions specified by Bendixson 's theorem hold, then one is certain that at
least one periodic solution exists.
shall encounter the system
-ja t
dx
= - ay +
x{l
,.,
\ )
[25.*]
4t = A + ax + y(l a
t
2
)
where
is a constant and r
2
,
= x
+ y
2
.
It is apparent that,
for a suffi-
ciently large r
Furthermore, by
V = Vo + V,
[25-5]
t-^
t
dn
"
It is thus seen that the origin is an unstable focal point and since no other
singularities exist, the Bendixson theorem shows that there exists a stable
periodic solution contained in a finite domain.
26.
80
Field
Winding
and
A/VWWV
Figure 26.1
4>(i),
laws,
cuits
ip(i x )
i x
and
i 2
(r
+ R)i - Ri 2 1
L^
=
[26.1]
rp(i 2 )
(r
+ R)i 2 - Ri
- L
^f =
di
ip(ii)
(r
(r
+ R)i 2 Ri + R)i Ri 2
x 1
[26.2]
The variables
and
i 2
The
dP
di
x
,
l
,.
i
ir+R);
dio
dQ =
ip
(i 2 )
(r
+ R)
[26.3]
i/>i(i)
The condition of
)-plane.
linear equations
di x
~dt
^(tj)
(r
+ R)i - Ri 2
1
.26.4]
di 2
dt
ip(i 2 )
(r
+ R)i 2 Ri\
In fact, this inequality means that the initial supply of energy must be greater than its dissipation, see Section 21
81
that is by developing
Writing
(^Lo
-<*<><-.-'.
(p
one ha.
R)i 1 - L^T" - Ri 2 dt
[26.5]
Ri +
Y
(p
R)i 2
di 2 dt
Putting
= d/dt,
R L6)i Ri 2
1
[26.6]
- Ri! +
(p
- R - L6)i 2 =
i
1
The system [26.6] admits solutions other than the trivial ones
if the determinant is zero.
p
i 2
= 0,
- R - L6 - R
- R = - R - L6
[26.7]
r -
L6 =
R;
-
',-1
In practice generally p
6
2
(r
+ 2R)
= {p #U)
r)
[26.8]
- r
> 0; thus
Field
> 0.
As regards
6 lt
it may be
Winding
In the
and
add up
AAA/WW
i
In
i,
Figure 26.2
ergy and the other absorbs it so that no energy flows into the external circuit.
If the field connections are crossed as shown in Figure 26.2,
instead
82
xjj{i
(r
+ R)i 2 - Ri
- L
^=
[26.9]
^(t 2
(r
+ R)i - Ri 2 1
L^-
[p
(r
2i?)J
<5
= -
|-(p
r)
[26.10]
If p < r +
2i2,
there is
from a stable nodal point to a saddle point, which is in accordance with the
sequence of the zones of singularities shown in Figure 18.1
27.
If-
P(x,y)
and
|* =
Q(x,y)
[27-1]
= 0(0;
i/>(t)
[27.2]
x=x
where
yx
r)
[27.3]
It will
|(0 and
2
rj{t)
and
77
may neglect
2
,
3
,
rj
'.
83
neighborhood of
= P^X^yJt; +
P(x
,i/
)77
[27.4]
at
[27-5]
4^ = a
r
e():
+ d(0/y
t
where a(0
"
'
'
>
with
common period
7\
* 1 */
u (*);
v,
=
=
* 2
12
(0
[27.6]
U=
where / n
,
** 1 '/ (0; ai
h2t
e
v2
f 22 {t)
7\
and
fej
and h 2 are certain constants, real or complex, which are determined only to
integral multiples of
hj
h2
= /(<* + d)dt T
[27.7]
that is, the sum of the characteristic exponents is the average of the sum of
the diagonal coefficients of [27-5] taken over the period T.
This can be
proved as follows.
= iiJK^kjn 2 - Mi
d=
iifo
Mi
Mi
[27>8]
>
V2
$iV2where
<5
Uli=
[*
* 2 ]'
[fuftt -/12/21J
[h
h 2 ]t 6
[27.9J
84
(a
d)
-|
(h\
h2)
hi
h2
y\ {h
h 2 )dt
= jf( a +
4) dt
jf
j- dt
Since the last term on the right side of this expression has period T and
hence vanishes, one obtains the expression [27.71It should be noted further that if the system [27-51 admits a
{t) +
c2 f2
(t); rj(t) =
r)
1 1
(t) +
r]
{t)
that is,
$(t + T) = $(t);
+ T) = rj(t).
Cj(e
- 1)^ +
-
2<
(e
1)? 2
=
[27.10]
h it
ho
{e
D/7J
-I-
(e
1)77
Since
and
)
c 2
(^.^J
and ( 2 ,?7 2
or
Thus,
in the second,
hx
= 0.
A]
By differ-
dx
dy
[27.11]
a*
Comparing with
[Z~l
d y
= n* Q
s een
v
dx
{x
>
y)
di
+ Q * {x
.dy
'
y)
di
x
A]
it is
dyjdt satisfy
a
dyjdt
is
non-trivial solution of [27. 4] since x {t) and y%{t) are not constant.
Knowing that one of the characteristic exponents, say h 1> is zero, we can
h2
= y/[Ps (*!,!) + Q y {x
,y x )\dt
[27.12]
85
Following
tions Xj(0,
stability of equilibrium it can be shown that the motion defined by the funcVi(t) is stable if h <
The proof is
We shall
equilibrium, Chapter III, the real parts of both roots of the characteristic
Here, however,
the stability depends on the sign of the real part of the non-vanishing char-
acteristic exponent.
- n{~\
)x + x =
x
is known to possess
= 2 cos
t
,
= x
sin
t,
//
is very small.
yx
Thus,
4-
=
1
,
2 cos ^
CJ^fift);
= Ismvt
o> 2 (/i,
u 2 (/i,t)
where
2n/T ~
T~
when
4 cos
T
2tt,
/u
and
ij
(fi,t),
i)
1
zero uniformly in
*{1
)
>
0.
'
This gives
o)
3
P x (x ,y
= 0;
Q y (x ,y
1
1 )
=M1
h
vt +
{/i,t)]
and by [27.12]
T
0,
4 cos
a>
Afi
t)
dt
+ uAn)
Since
cj 4 {fi)
>
h <
)
which proves that the periodic motion in the neigha well-known fact.
borhood of (x ,y
2.
is stable,
Section 22.
stable
for the system [22.1], unstable for [22.6], and half stable for [22.11].
We
P,
=
(X
2
y'
1
(x
2x
)
y )2
>>.
'
Q,
(x
2
(x
2\
2y
i/T+7
86
= cos t, y
sin
t,
we
have
px
and
x i> V\)
= ~
2 c os
Q y (x
yx)
2 sin
2"
= -
-r
27T
f
J
sin
cos
t)dt
= -
is stable.
2
i/
1;
2 Q v = 3y +
Whence,
2n
f(4
- 2)dt =
>
r =
is unstable.
is
important to note that one is able to draw conclusions as to stability only when
0.
not applicable.
c.
first,
Px {x\, yi + Q y (x 1>yi =
)
)
so that h = 0.
dx
*Jt
It is observed that
dy
'
It
~ ~
Px
Qy =
0.
neither stable nor unstable but rather indifferent in the sense that if a
CHAPTER V
BIFURCATION THEORY (POINCARE)
28.
INTRODUCTORY REMARKS
It has been shown in Section 13,
14,
In this chapter
We shall not go into the investigation of these complicated cases but will
A preliminary qualitative
investigation of the subject of this chapter has been made already at the end
of Section 24 where the intuitive concepts of "sources" and "sinks" in the
phase plane were applied; here we shall follow the analytical method of
Poincare.
In practice,
earized" from the start so that some limited conclusions regarding stability
may be reached.
frequently
of the so-called "hard" self-excitation which has now been completely ex-
In other words,
29.
TRANSITION OF SINGULARITIES.
b,
c,
linear terms
P2
(x,y) and
parameter k.
We as-
sume further that the origin (0,0) is a focal point for all values of k under
consideration.
It was seen in Section l8 that,
5 = djU) + ib
1
(k)
[29.1]
52
By means of the transformations
=
[1
a t (k)
- ib^k)
=
=
a 1 {k)x
bx
(k)y
+ P 2 (x,y,k)
[29-2]
bi(k)x
a x (k)y
+ Q 2 (%,y,^)
where
and Q 2 are power series in x and y beginning at least with the terms
or a x > 0,
the second by y, adding the two equations, and transforming the resultant
dt
a x {k)r
+ P2 r
y,
cos
+ Q2 r
sin
[29.3]
similarly
d0
dt
~2 r
b^tir
+ Q2 r
cos
- P2 r
sine
[29.4]
Dividing Equation [29.3] by [29.4] one obtains the equation of the trajectories
dr _ de
"
'
r
'
a^^r + P
L6 1 (X)r
cos0 + Q 2 2 + Q 2 cos9 - P 2
sin 6
29.5]
sin
|a;|
and
\y\
{k).
Equation [29.5]
89
dr
a^k)
Je
TM~)
P2
cos 9
+ Q 2 sm6
P2
sin0
b
x
- Q 2 cos#
(k)r
cos 6
2 \
b~JkT~
iP9 sine
- Q9
6i(X)r
[29.6]
-f-
= R
{0,k)r
+ R 2 (6,k)r 2 + R 3 (0,k)r 3 +
[29.7]
where
*i
b.ik)
a x (P 2
sir\0
=
'
f?(
a : /P2 sin0
Q 2 cosfl\ P cos 8 + Q 2 + 2
i
sin 6
r'
fl,
Q 2 cos0\ +
)
aj
&(
(P 2
/P2
sni
Q2 cos
+ - Q 2 cos0)
[29.8]
cos g
6
+ Q2
5
:
sin g
1
cos6>
+ Q
sin
r"
Oj r
8)(P 2 ,2 4
sin
(9
Since it is desirable to investigate the transition from stable focal points to the unstable ones, the branch
18.1 must be followed; in this case q
=
0.
focal points to unstable ones the real part of the complex roots changes its
sign but the imaginary part does not.
In other words
b
1
(k) in Equations
[29.2] does not change its sign for k l < k < k 2 corresponding to the branch
BC in Figure
18.1
k in the interval
depending on either k or
R k {6,k)
are
periodic functions of
9.
6,
< p.
Thus
,k)
= f(0,r
v-Qitjftf,
A)
[29.9]
Substituting this solution into [29.7] one obtains the following set of
90
de
de
-Ij-
= uzR^e.k) + =
u z R (e,k)
l
R 2 (e,k)
[29.10]
2 Ul u 2
Since by assumption r
1;
M (0,X)
fc
[29.11]
for
fc
= 2, 3t
'
du
u
and, on integrating,
a {k)
x
b x (k)
o,(X)
u {6,k)
x
blM
[29.12]
Let r
it is apparent that the trajectory originating at the is a limit cycle if, and only if
ip(r
point
r =
6 = 0)
,k)=0
[29.13]
where
iP(r ,k)
f(27T,r ,\)
- f(0,r
,k)
f(27T,r ,k)
In this expression r
tp{r,
k) to des-
We have
<x\(k)r
+ a 2 U)r
+ a 3 U)r
a,U)
[29.14]
where
2n
-
(k)
u (2tt,X)
1
=
k
b,ik)
e
[29.15]
a k (k)
u k {27T,k),
2, 3,
91
a 1 (k
* 0.
solution r
and |x
borhood
-
= 0,
|
and X when
of limit
around k Q
2.
a^Xj
= 0.
= 0.
Furthermore
*s
a/U,,) * 0;
a 3 (X
)
*
1
where
X = X
a/ (X
.
which the stability of the singular point changes, there appears a limit
cycle or cycles.
the
,X)-plane.
,k) =
= 0;
X = X
0(r o
,X)
ot.ik)
+ a 2 (X)r + a 3 (X)r
=0
[29. l6]
0(r o ,X) =
(X
X )a;(X
(X
X )a.;(X )r
+ a 3 (X
)r
[29.17]
92
where eti(X
X = X
.
and
a^o)
dot 1
)
Since ^(Xq) *
ai(X
# 0.
that the first and the third terms on the right side of [29.17] are of the
first order, the second term is of the order 3/2 and can be neglected. Equation [29.17] then becomes
(A
X )a 1'(X
+ a3r
whence
2
ro
=~C^-^ "3
[29.18]
where
M -$$"+
-
1.
> 0,
ct
> 0;
3.
< 0, a 3 < 0.
)
= ai(X
>
creasing values of the parameter X the real parts of the characteristic roots
change from negative to positive for X
51
For
<
.
is real,
i.e.,
< 0.
Figure 29.1a.
In the region X
Equation [29.18] shows that the limit cycles exist only for X
-
fl
This
for
X^X
3.
The limit cycles in this case are manifestly stable since the
singularity is unstable.
By a similar argument one finds that the limit cycles exist for
see Figure 29. lc.
X^
in that region.
4.
93
Limit Cycle
Limit Cycle
*( ro> A )
Figure 29.1
plication of tbat theorem was made in Section 13 in connection with the question of the stability of equilibrium it can be also applied in this case as will be shown.
It is apparent that the curves of Figure 29.1
represent Equation
[29.17].
In the neighborhood of (X = X
<t>{r,k)
r = r
we can write
(X
X )a 1'(X
(X
X )a 2'(X )r
<x
(X )r
- 0(r o> X) =
0(r,X)
or
0(r,X)
(X
X )a 2'(X )(r X
)
<*
(X )(r
rQ
[29.19]
Recalling that r
and (X
and, hence,
order 3/2.
= a
(X )(r
[29.20]
94
> 0;
a3 <
0.
(2r,r
,X)
in applications.
From [29.20] it
for r < r
.
,k) = 0.
r
Figure 29.2
Since
ip[r ,k) = r
is
rp{r, k)
>
for r < r
the curve
for r > r
i//{r,k)
k) Is the
9 = 27r,
for r 1 < r
shown.
0(r
,X) > 0,
x
1$
For r 2 > r
ascertained by the more detailed study made in this section was possible only
by considering the non-linear terms.
tions which were sufficient for analyzing the conditions of the equilibrium
of the system are incapable of giving any information concerning the exist-
ct
(k) is
95
30.
Although this subject has been considered to some extent at the end
of the preceding section, we propose to investigate it in more detail by in-
/"N
\R
+ Ri +
1
C
J
dt
dt
[30.1]
in the usual notation; the coefficient k of mutual inductance between the anode and grid cir-
Figure 30.1
Vg
s
l
h=h+
where S lt S 2
,
vg + s 2 vg + s 3 vg +
[30.2]
S3
f{Vg
S 2 Vg
what follows.
h =
coefficients S lt S 2
where
,
s t vg + s 2 vg + s 3 vg
[30.3:
since the
In order to
v =
Vg /Vs
is the grid voltage beyond which the anode current I a does not
Putting S x =
/3 lf
S 2 Vs = y
x ,
and S 3 V S
where
/3 lt
y lt
have
96
"
[30.3] becomes
= vs^i v +
yy
<5>
[30.4]
of v,
the approximation I a =
{p v + y v
l l
2
)
is sufficiently accurate.
During the process of self -excitation the oscillabe considerable so that the third term
<$/
is justified.
is generally
= - 5
characteristic
Ia
= V ,(/V +
s
yv
6.V*)
[30.5]
in which
v =
fi lt
y1
and
<5j
are positive.
CV,
krjidu J
'
= rV-; CV
S
'
~r CV
S
jj dt
[30.6]
It
=jflt
xr =V,(0
,
+2y v1
361
v)v
'^*
130.7
we get
LCv + RC -
k(/3 1
2 7l v
- Sd^ 2
[30.8]
"cyclic" time r = u
where u
YT/LC
This gives
dv
''
dv
dt
dv dr
dr dt
dr
Uq
and similarly
d v
,2
,2
v
.2
(-0,
dt'
d-
"'o
dimensionless form
d v
dr 2
where
0(k)
/3(A)
+ 2y(k)v - 36(k)v
2
]
j^
[30.9
= {k0 - RC)oj q
l
y(k)
k yi u
6 (k)
= XS^q
[30.10]
97
dv
.
dr
w
[30. 11]
dw -r = dr
0{k)
+ 2y{X)v -
3<5(A)v
IV
]
dv
,
dr
= w
[30.12]
dw = dr
v 4-
pw
The point
v =
is a singular point.
equation are
oscillation which will be studied in Part IV, in which the starting of oscillations occurs in the manner of trajectories departing from an unstable nodal
point but we shall not treat this case here as the theory of these oscillations has not yet been established for the steady state.
For the oscillations developing from an unstable focal point (Section 8) the roots
e ^1,2
_ -
A
2
+ Jil _ - V 4
< 2.
Substituting for
fi
RC
01
<k<
U
R^ +
^1
"oi
p =
0i
- RC)u) =
that is
*o
^
w 2x
PC
[30.13]
Making use
= 1a
+ 2b
-
y;
Yl
a1 x
\4y(a 1 x
y)
126(a t x + I26{a
y)
\x
[30.14]
y
b
x
axy
- 4y(ajX +
b x y)
b x y)
\x
It is noted that the derivative with respect to X of the real part of the
roots S 1>2
for X = X
is positive; thus a
[k
>
0.
a 3 (X
<
P(k
and hence
a^Xj
= 0;
b^k^
= = 4y(X
= 1.
R 2 (6,k
R s (8,k
has the following form
sine cos
3
)
[30.15]
3
16y(X
sin
cos
12 6(k
2
)
sin
cos
du x de
du 2
de
4y(k
sine cos 6
.
[30. l6]
-^
= 2u R
2
(e,k
+ R
(e,k
u x (e,k
u 2 (e,k
-|y(X
)(l
cos 0)
U3
S6(k
)7T
99
= -
37r#Cw
i
(-g
)
< 0.
Thus, by
cycle.
X
pends on the curvature of the non-linear characteristic of the system, whereas the condition of stability depends on its slope at the point of equilibrium.
For this reason it was necessary to retain a cubic term in the series expansion of the plate current characteristic, Equation [30.5], in investigating
limit cycles, whereas for problems of equilibrium the equations of the first
were sufficient.
We have considered in this section only the case of a soft self-
subject here but will reserve it to a later chapter after we get acquainted
31
nation of the parameters of a mechanical system is generally a more difficult problem than that of electric circuits.
It is possible only in a few
particularly simple cases in which the chain of "causes" and "effects" can
be followed completely.
A.
Consider the following arrangement used for the study of antirolling stabilization of ships by the method of activated tanks (25).
Two
tanks are mounted on a pendulum and located symmetrically with respect to
100
the axis of oscillation; the tanks are connected by a U-tube and filled with
water.
placing the water in the system so formed; the blade angle a is controlled in
response to the angular motion of the pendulum as will be specified.
The system has thus two degrees of freedom, the angle
9
<f>
that
negligible.
the motion of
b<f>
c<t>
where
J,
b,
and
Assume now that the pump is made to act on the liquid column in the U-tube
and let the moment of the force exerted on the water by the pump be M(a)
,
Ma
3
3
.
this case is such that the blade angle is continuously adjusted to be propor-
4>
a3
3
.
the free oscillation of the pendulum is damped and the forced oscillation is
reduced in accordance with the linear theory; these features, are, however,
of no interest here.
Under
certain conditions the pendulum, the water in the tanks, and the blade angle
begin to oscillate or "flutter" spontaneously; the oscillation of the pendulum is, however, very small and will be neglected.
The oscillation of the
The blade angle in this particular case is made proportional to the angular acceleration of the pendulum. The latter, however, is in phase with the rate of flow through the tube. A complete study of the problem requires a consideration of the system with two degrees of freedom 8, comprising the pendulum and the water ballast. Insofar as this study is limited only to the motion of the latter, this complete study is omitted here. Account is taken only of its final conclusion, namely, the blade angle
<j>
101
blades, and also of the water, designated usually as "hunting" persists in-
J
that is
J'<j>
<f>
b<f>
4-
c0
= a^
*
az
<t>
(b
aj)0
a3
c<f>
=
b
[31.1]
-
ax
0.
It is to be
Hence,
a 1 (k) <
cation
04(A) > 0.
Case
Weak Amplification
-
and
c/J =
oj
we have
<j>
n'j>
p'0 3
<j
[31 .2]
form
d-4>
2a
.Jj.2
dr
+ J**\
n-^)\d^ +
nS +
1
d(f)
[51
5]
iVT
=0,
In
L31.4]
the origin,
<j>
<f>
= d<j>/dr
the origin,
= d<f>/dr= 0,
Case 2.
Strong Amplification
b
In this case
[04
(fi)
04(A) < 0.
,
- b
0,
a 3 /J = p'
d'4>
dr 2
^+
[31.5]
102
p'u>.
S 1)2
Hence for n /4 >
2
1
f
,
ff
n,2
-1
[31.6]
2
1
In the first case the control equipIn the second case, since
the singularities are unstable the trajectories approach a stable limit cycle
More-
a 1 [k 1
= 0.
b
The condition n /4 =
is equivalent to
2
1 )
/^J 2 u 2 =
b
"I
It is
the same both for stable and unstable operations of the control system.
The
a 1 (k)
characterizes the
The critical
'.
In this particuv,
l ,
the am-
By increasing either
so
We have,
If the
= 0,
a.
it can
,
Ma
3
ct
where
M M M
lf
3
>
0,
In such a case the self-excitation would appear in an entirely different manner, as will be explained in Part II.
Another example of the same kind is the well-known experiment (26) in which a series generator is connected to a separately excited motor.
103
= ax
as
where
is the current,
az
3 i
Koj
+ L
+ Ri
di
[31-7]
(ai
-R-Za^)f
t
=0
[31.8]
The quantity dto/dt can be eliminated from the equation by expressing that the
electrical power Kcoi absorbed by the motor serves to accelerate the rotor
Substi-
=
This equation
.is
[31.10]
where
[a 1
R)/L, n =
J>a z /L,
p =
K /LJ
2
> 0.
of Van der
Pol's type.
at
- (m -
ni )y
pi
[31.11]
Tt
dy
y;
2
jt
-
= ~
pi
+ my
and its roots are
[31J2]
mS
/// "'
+ p =
j
ra
+
_l_
i
y^
1
"< "m
/4
p >
/4 - p < 0.
m=
(a!
i?)/L > 0,
that is
ax
< R.
*>*;
The first condition is
a-
Pir) 2L
I
R\
b< LJ
w-w
ceding example.
104
or a
field which modifies the state of saturation in the machine and, hence, the
coefficients a
One can go a step further and determine the amplitude of the self-
excited oscillation but for this purpose we shall need the analytical method
of Poincare outlined in Part II.
CHAPTER
VI
32.
INTRODUCTORY REMARKS
In the preceding chapter we have investigated the principal prop-
non-conservative systems.
existence of such motions, a rather difficult question for which the rela-
tively simple criteria of Poincare and Bendixson give only limited information.
In Part II we shall enter into this question in more detail in
.
In this
N.
These methods occupy an intermediate position between the topological and analytical methods and thus present particular interest as a con-
fi(l
x )x
[32.1
+ f(x)x +
[32.2]
+ f(x,x)x +
g(x)
[32.3]
The functions / and g entering into these equations are subject to certain
of periodic solutions for [32.1] by Van der Pol rested upon the graphical
7-
analysis of the more general Equations [32.2] and [32.3] it is useful to consider the original Van der Pol Equation [32.1] from the physical standpoint.
It is apparent that
autonomous oscillatory system with one degree of freedom possessing a variable damping
- fi(1
-
)x.
been said regarding "negative damping" one concludes that for small values of
106
x the
stages the motion develops with gradually increasing amplitudes; for large x,
Ultimately a steady state is reached when absorption and dissipation of energy balance one another throughout the cycle. This is, in fact, in entire
essary to establish conditions for closed trajectories and to infer from them
that a periodic process is possible.
33.
LIENARD'S METHOD
(jf(x)x
o)
[33.1]
+ f(x)x +
[33.2]
and
x as
co
f(x)x + w
with
oj
tion x + x = 0.
< 2) or aperiodic
\C\
Chapter
I.
Setting
x = v,
=^v,
ax
+ f(x)v +
[33.3]
^L +
ax
Introducing a new variable y =
v
f( x )
,
*
v
o
X
[33.4]
+ F[x)
dx
- Fix)
107
xdx +\y -
F(x)] dy
[33.6]
dy
x y
dx
dx
v
- Fix)
t
dt
= un
dt
[33.7]
ferent from the customary (x,y= x)-plane considered previously so that the
x,
y = v + F(x)
In fact, the equation of the normal at the point (x, y) on the tra-
X)dx +
0,
y -
Y)dy =
There-
0,
X=
= F(x).
jectories of Lienard'
[33.6],
Equation
for x = x lt
whose coordinates
are 0, .F'fxJ.
by taking on the
F(x
1 )
By taking
,
other points x = x 2 , x 3
...
and by
Figure 33.1
Having a field of line elements and starting from a certain initial point
(x ,y Q
)
Since F(0) =
[33.5]
0,
is the origin x = y = 0.
x for x
and
y for y
108
2/
y -
F(x)
+
-
dx
+
-
dy
-
Region
2
3
+
-
+ +
when y = F(x).
Hence,
the
intersect the curve y = F[x); the abscissa of point C thus determines the
{x, y) -plane.
Instead of the above construction one can apply the method of isoclines,
see Section 7.
- Fix)
is thus on the
1
x.
and so on.
34.
viously defined.
Instead of Lienard'
+ f(x)x +
g(x)
[34.1
(1 ),
=1 f(x)dx
and G(x) =
g(x)dx.
The
109
1.
f(x)
F(x)
is an even function of x,
even.
2.
It is negative for
For x >
it
F{x)
> oo
with
x.
Figure 34.1
monotonic for
< x < x
Following the same procedure as in Section 33, one obtains the generalized Lienard equation
dy_
g(x)
dx
- F(x)
[34.2]
We shall show that there is one, and only one, closed trajectory of [34.2],
This
must necessar-
and (0,- y
and (0,- y
).
periodic solution will be proved if one shows that among all trajectories
there is one, and only one, with equal positive and negative intercepts
OA
and OB.
^(x,y)
jy
G(x)
= OB,
[31.3]
it is only necessary to
Thus k(0,y) =
jy
..2
.
OA
i^A]
fdk =
kB
- kA =
Fix)
d\
[34.5]
where, here and throughout the rest of the section, the integrals are curvi-
110
-"--G"
i
In
---^s'
"XJg'
x
\l
F(x)
J\ Jk
hI/7i
"^\S
/
sT
B
B'
'
*0C-A*
'
B"'
Figure 3^.2
\y
F[x)\
increases indefinitely.
Further-
more,
away from
bounded while its negative part due to the elements situated to the right of
H,
Hence,
A and B
dk =
and
3^.6]
35-
ii
x.
In
[35-1]
In the asymptotic case, when F(x) is very small, we may replace F(x) by fiF(x)
where
//
is a small number.
If
fi
= 0,
+ y
= constant.
If
fi
0,
Ill
x =
sin
<f>\
y =
cos
<j>,
[35.1]
dR
d<j>
H *
or
dR j- -
^F(i?sin0)sin0
[35-2]
Therefore,
[34.6] becomes
JV(i?sin0)#sin0 d0 =
o
2 Thus, for Van der Pol's equation, f(x) = x
[35-3]
F(x) = -y
%,
[35.3] is
fFCtfsin^) sin0
d<$>
[35-4]
Hence when n is small, R = 2 is the radius of the circle in the vicinity of which there exists a closed trajectory. Furthermore, a simple approximation
based on the smallness of ^ shows that along the closed trajectory the angular velocity
w *
is constant.
been established by means of a geometric method and from that point the argument has been analytic.
It will be shown in Part II that the same result can
36.
THE PARAMETER
(l
IS LARGE
proximation given in Part II, become too laborious owing to the poor convergence of the Fourier expansion representing the oscillation.
Lienard did,
regarding the wave forms of the oscillations for large values of n. propose to show this method.
Setting y =
fiz
We now
- F(x)]dz + -2 xdx =
[36.1]
112
For
fi
- F(x)]dz *
[36.2]
z
F(x) and
= constant,
to
dt
dz
fi
[36.3]
at
u)
dt
=
n[z
Fix)]
"1/7*,
~ means
since x is finite.
~F(x)
dz dx
by [36.2]
dF
dx
.,
which is finite.
But || = hence, || ~ -. Since the velocity remains j^ ||; small in this finite interval, we conclude that it takes a relatively long
Thus, we can conclude that on the branch
2 2
= .F(x) the
order of smallness,
1/a*
fiu Q f(x)x
U)
=
1
.
[36.4]
[36.4] becomes
/noj
f(x)x + w
[36.5]
'\//li
F(x)
0,
113
2 One can also see this from the second Equation [36.3], since dx/dt ~ l/u. 1/V We thus infer that on the branches z F(x) of integral curves, the represen.
l//i
and an accel-
"l//*
It takes,
We shall now investigate the motion on the other typical arcs for
which dz
zero.
d x/dt
2
2
0,
that is
z*
constant;
that is,
2
,
cj
jUO) a
f(x) x
2
;
[36.6]
~\/\i
0.
This almost horizontal branch in the (x, z)-plane is traversed very rapidly,
since dx/dt
/i
37.
nil
x )x
[37.1]
in the (x,t)-plane are shown in Figure 371 f r three values of the parameter
fi
(1).
37
the (x,x)-plane with which we were concerned throughout the first four chap-
37
by the graphical method of isoclines; these results are shown in Figure 37 '2.
0.1
(a)
10
(c)
Figure 37 !
114
0.1
lt= 10
Figure 37-2
Liapounoff
equations of the first approximation are
x - y,
In fact,
-
's
y =
x +
[37-2]
For
<
fi
< 2 the roots are conjugate complex with a positive real part, and
thus, an unstable focal point.
For n = 0.1, the limit cycle C differs very little from a circle of radius 2 and the trajectories approach
the origin is,
it both from the inside and the outside as shown in Figure 37 -2a.
1
For
=
fj.
.0,
the origin is still an unstable focal point but the limit cycle C difJ>~J
2b corresponding to
[
37 2
*
are
The trajectories
in the (x,x)-plane leave the origin along definite directions and approach
'2c
It is observed
that the limit cycle is now strongly distorted and acquires an elongated
+ F[x)], where
= dx/dt,
For n = 0.1,
Figure 37'3a, the limit cycle C is again nearly a circle as in the (x,x)-plane,
For larger values of
fi,
115
u=0.\
(a)
Figure 37
different kind of deformation as compared to that in the (x,x)-plane and acquires an almost rectangular form for
fi
One
recognizes in this figure the second asymptotic case, n discussed in the preceding section.
1,
of Lienard
of applications.
CHAPTER
VII
38.
GENERAL REMARKS.
<t>,
to within a
= 0,
generating line.
The general properties of this phase space are about the same as in
the case of a plane.
ture due to the fact that we are representing the trajectories on a cylinder
velocity
=
O
We thus conceive
= +
or for
= -
by
we
Figure 38.1
117
same situation which we have already studied in the phase plane; the only dif-
ference lies in the fact that the plane is wrapped on the cylinder.
FurtherThus,
for example, we can assert that inside a closed trajectory of the first kind
there must exist a number of singularities with the algebraic sum of their
and so on.
We cannot, however, assert this with respect to the closed trajectories of the second kind since these do not bound off any region.
The im-
portant property of these closed trajectories lies in the fact that they have
a period 2tt relative to
a limit
<f>
varies by 2n.
dinates the cylindrical phase space may be replaced by a plane with the origin left out.
In fact, if we set p = e*
,
where
<f>,
of the cylinder goes into a point M'{<t>,p) of the plane and the correspondence
between
M and
M'
that is,
topological.
+
oo,
Since
o,
p varies between
and
Intuitively speaking, the transformation Af->M" is equivalent to a flattening out of the cylinder into a plane.
This plane representation makes it clear
that the usual features of the planar sys-
Figure 38.2
closed trajectory of the second kind acquires a familiar feature, that is,
the origin behaves now in all respects as a point singularity.
39.
Ad + BO +
f{9)
[39-1]
118
where A, B, and
In the
A9 + B$ + C
f{e) and C sin e.
=M
[39-2]
Equations [391] and [392] are non-linear on account of the presence of the
terras
acter of the motion can be, in this case, either oscillatory or rotary accord"dead beat" or with dying-out oscillations
equation in a cylindrical phase space in which the existence of periodic trajectories of the second kind will be established. Introducing the "dimensionItL * * <-<-, _ B = a ^ A = j^= 3 0, [39.2] > 0, less time" r = u t = Vk-t and putting -=? " VCA Au) n C >o becomes
M -^
'
jf>+ft +-#-<>*
We shall make use of this form in what follows.
[39.3]
40.
'37
Hence
>
g-.-"-~**-^
3 az
z
fKU]
dz de
sing
[40.2]
0.
\z = 38 +
2
cos^
\k
[40.3]
where
^is
a constant of integration.
z
Hence
cos 8)
= Y2(38 +
[40.4]
3=0,
hence,
M=
0.
For
A;
= - 2,
= d8/dt = 0, and
9=0;
for
>
2 Since no confusion can arise, the usual notations d e/dt will be resumed.
119
Separatrix
Saddle
Point
Saddle
Point
Vortex
Point
Figure 40.1
Figure 40.2
6 - 0,
2=0.
< k < + 2.
For
n.
Thus far the discussion has been analogous to that of Section 12.
Let us wrap the plane figure on a cylinder of unit radius, see Figure 40.2;
the points
n and
ries of the second kind are those which go outside the separatrices.
cylinder they appear as surrounding the cylinder but not the singularities.
The curve y in Figure 40. la has maxima and minima rising on the average due to the presence
2. 0.
Let
V2(0O
z
+ cos 6) + k =
V2y
+ k.
of the term
fi6.
If the curve
The peri-
odic trajectories do not go around the cylinder, hence, they are not of the
second kind.
If
yS
= 1,
tion points.
exist, and the use of a cylindrical phase space offers no particular advan-
41
sider [40.2].
We assume in
120
fi
0;
j8
sine
[41
.1
j3
<
9t
-
2/C77-;
6j
(2k
1)tt
[4l .2]
where
= sin
fi
and
< 9
<
tt/2.
with coordinates
It is sufficient
= 0) and
Ai
with coordinates (9 = 9 jt
= 0).
say,
A i0 {9
= 9Q
= 0);
A j0 (9
=
1.
7T
= 0).
{
.
Group
6
9i
= 2k-n + 9
e
,
= 0.
Consider a slight
departure
from
We have 9 = 9
dz d9 dz
de
p az
e)
r^,
-zi
Developing sin (6
e
+ e) = sin
cos
= sin 9
az
z
e cos
[41.4]
=
dz
-az -,COS0
.
rfe
[41.5]
,,_
_,
+ aS +
cos0 o
0;
a>0
2
[4l .6]
< 4 cos 6
> 4 cos
it
Group Aj.
1
(2k +
)7T -
= 0.
az +
z
cos
fl
[-[m
yi
S2
aS-
cos0 o
[41.8]
121
S i,2
a
2
a
4
cos6 n u
[41.9]
42.
non-conservative system,
1
.
a.
>
0.
/S
>
{e) and
{6)
z 1 (9
2ir)
2tt)
{6),
{6 + 2n)
[6).
>
From
is
sin
[42.1]
6,
i.e.,
= 0,
only if
/?
<
>
= 0,
dz/d6 =
..
More-
sary to express the fact that this solution goes above the sinusoid of Figure 42.1.
0,
and hence,
z z
1
(0 O +
Z-it)
<
(6
).
{6
=
(
>
-
(1
+ fi)/ct,
6 )/a,
because
that is,
fi
sin
2.(6)
(a)
(b)
Figure 42.1
122
satisfying the
condition
z2
[0
+ 2n) >
(6
),
Figure 42.2.
ing from
G =
= (
)/a.
A curve
z
2
[$) issu-
the curve
(6) is
rising
clearly,
{6).
Beyond Further-
(6) decreases.
lf
cannot be lower
tion
M
)
the curve
-z
(e)
/
e
{6
2-n)
{$
Figure 42.2
in the
interval
). z
is unique.
In
fact,
and
61
+ 2n one has
+
2tt
\z
{9 1
2tt)
|-z
(0i)
= ~ otjzdO +
),
2tt/3
[42.2]
+ 2n) = z{6 1
whence
2-n
dB
277"/?
[*2.3]
{$) and
z 02
(0).
gularities, these solutions cannot intersect each other, so that one of them
is always greater than the other, e.g.,
0,
z Q1
0,
(0) >
02 (6)
for all
6,
whence
+
z
2rr
+ 2n
z
02
01
(e)de >
(8)d0
tion is unique.
2.
<
Proceeding as
1
(5
),
2tt)
=
and
Zi(Bq).
z
2
(0 o
such that
{0
+ 2n) ^
[0
),
123
on the surface of the cylinder passing through two adjoining saddle points,
A jf separated
by 2n.
The curve
A jl
has the
same slope at the next saddle point Aj 2 as the asymptote of the negative
slope.
z
2
One can show that, for sufficiently small values of a, the condition
+ 2n) >
z
2
{6
{0
is fulfilled.
damping is not too great, the rotary motion of the pendulum may become periodic when the energy communicated by the moment
the average to the energy dissipated by damping.
M per
slightly below or above this critical value the rotary motion will become either damped, i.e., the trajectory approaches a focal point, or will continue with an increasing angular velocity.
In the latter case a state of
periodicity of the second kind will eventually be reached when the energy
communicated to the system by the constant moment
energy dissipated by damping.
jectory of the second kind closed around the cylinder and not enclosing any
point.
unstable motion will have a tendency to approach a periodic rotary motion; in the latter case the motion will approach a definite
-Separatrices
-Saddle
Point
from it following one of the other two asymptotes either to a periodic trajectory of
the second kind or to a stable focal point.
Figure 42.3
124
43.
As. is
by the elec-
This
with the corresponding field of the salient poles on the rotor excited by a
direct current.
The manner in which both fields on the stator and rotor are
Figure 43.1
shown; this fact, on the basis of the Faraday-Maxwell theory concerning the
It is seen,
thus,
that for
O
Tr
If the resisting
increases
T and T r is reached.
by springs.
that which would exist if one had two rotating mechanical systems connected
It is apparent,
125
4>
Seporatnx
= n,
this similar-
Stable
Focal Point
interval
{n, 2n)
however,
when Nj comes in alignment with S 2 since a stable configuration reappears exactly similar to that which existed originally.
turbance causing a change
A<f>
It follows that,
if the dis-
of the angle
4>
point
If,
the representative point to a point A' on the other side of the saddle point
Sj
will approach another focal point B situated at an angular distance from the former point of equilibrium A.
2n
trically by Vlasov but not commonly encountered in connection with the industrial synchronous motors, that the rotor runs at a certain speed below the
synchronous speed.
hind the stator poles and the performance in this case is called asynchronous.
It is apparent that this mode of operation is possible only when a certain
= 2n.
second kind.
may arise.
It is generally im-
tions are characterized by the existence of limit cycles of the first kind.
With this physical picture of the phenomena involved we can now proceed with
a brief
outline of Vlasov'
theory.
126
is
cos
sin (y
mpE V
Q
2ojx
E
p)
sin
sin 4
dy
V
dy
~dt
4u
cos p
dt
mV
cos p
2uxR.E n
y sin(y
4-
2g
p)
sin(y
4-
p)
sinp
[43.1]
m
E
steady state,
y =
ie
dE
= f{i
where
V
tt
is the voltage applied to one phase of the stator, is the angle between vectors
1
- y
and V,
p = tan"
where
r is
M is
R
e
and
u)
E - K
mp kV^-cosp 2pux
2 2
()
a =
4u
t
my
a
5-
b,
2u>xR
2Mtox
c,
mpV
ft
[43.2]
<f>
= y + p at
= t,
dr 2
sin
cosp sin0(sin0
2ifsinp)
Kb
sin4p
cos
dcf>
ft
dt
if
cosp
if sinp
[43.3]
d0
d:
:43.4]
dz
ft
dr
if sin p
sin
Kb
sin4p
if cos p
cosp
cos p sin
sin
2 if sinp)
127
Consider now a cylindrical phase space with the axis of the cylinder parallel to the z-axis.
is replaced by
<f>
Equations
^=
where
v
p- =
sin0
/(*)*
[43.5]
r ACOSp
if sin
p >
and
f(</>)
= Kb Sm
P COSp
-^rcospsinp(sin0
- 2^sinp)
[43.6]
Equating the right-hand terms of [43.5] to zero, one obtains the singular
points.
If v > 1,
If
<
= sin"
v,
and
<f>
<f>
lt
= 0.
it is assumed that
0,
f{<t> 2
Setting
<f> 1
e,
d<j>= de
sin0
sin(0 :
e)
sin0j
4-
ecos^
^
Si, 2
%= -Wi + M)i
2
[43.7]
- f(<t> )S 1
+ cos
<f> l
= 0,
= \
2
)
[/W
/
<f>
(0 1 )
- 4cob0J
[43.8]
According as the sign of f {<f> 1 - 4 cos one has either a nodal point 1 or a focal point. In practice one usually encounters focal points. In such
cases the equilibrium is stable for /(0j) < point (0 2
= 0) is always a saddle point.
f{<t> x
> 0.
Section
4l
/(0) = 0,
128
E /V
y =
dE /di
and
order.
small.
put in a form
= /if
(</>)
There is a certain
relative order of magnitude of the small quantities K, y, izing each particular case. Under these conditions, ^3
[
b,
character-
form
f$
*;
sin0
+ ufQ (4>)z
fi
[^3.9]
-j^=
d(/>
~
z
sin0
[43.10]
= V2 (cos
4>
v<j)
h);
constant
[43.11
ies of the second kind, although there may be closed periodic trajectories
"the island,
"
limited by
oscillations, in general, are not of the "relaxation type" but are ordinary
Oscillations of
this kind are not important, however, because they are rapidly damped out by
the squirrel-cage damping arrangement,
here.
/u
Physically
this means that the performance, instead of being synchronous, becomes asyn-
chronous, that is, the motor begins to "slip" and settles on a sub synchronous
speed.
dq>
129
\z\
is small
No tra-
{</>)64>
<0
J
(<*)d^>0
_L.jm[t_CycJe
Figure 43.3
is
- nfM)
z(2tt)
z(0)
fi\f
{<f>)d<t>
v = Kb
,
sin4p cosp
.2
2 c cos p sin p
2n
[43.13]
Two cases will be considered according to the sign of the right-hand term
of [43.13].
a.
Kb -? cos
-
+ p
2 c
As a consequence we M
z
have z(2n)
z(0) < 0.
is large,
> 0,
the traz
is
small and greater than zero, they depart from the 0-axis as do
the trajectories of the conservative system [43.10] for any
z
> 0.
there exists at least one limit cycle of the second kind and it
is stable.
<
is no limit cycle.
b.
Kb |^g Hp cos p
z
+ 2
cos p sin 2 p
> 0.
Then
z(2tt)
z(0) > 0.
is
In this case the trajectories depart from the 0-axis when \z\
large,
> 0.
\z\
however,
-axis when
z
is small.
Hence, for
<
0,
150
theory of critical points, Vlasov has shown that, in addition to the point
singularities and limit cycles of the second kind, the operation of a synchronous motor may also exhibit limit cycles of the first kind.
In ordinary
industrial motors these steady oscillations of a "relaxation" type are generally eliminated by a suitable design.
131
REFERENCES
"On Relaxation Oscillations," by B. Van der Pol,
2,
(1)
Philosophical
Magazine, Vol.
(2)
November 1926.
(On
in Celestial Mechanics),
by H. Poincare, Gauthier-Villars,
Paris, Vol. 1,
1892.
(4)
907
(5)
1901
1912.
957
>
Chapter 2.
"Mechanical Vibrations," by J.
1
P.
Karman, Bulletin of the American Mathematical Society, Vol. 46, August 1940.
(10)
Mechanics), by
(11)
Appell, Vol. 1,
Paris.
Physics, USSR,
(13)
p.
1933.
1,
1894,
212.
(14)
H.
Poincare, Acta
1?2
(15)
"Differentialgleichungen" (Differential Equations), by L. Bieberbach, J. Springer, Berlin, Third Edition, 1930, Part 1, Chapter 4.
(16) (17) (18)
2,
1900.
Moscow,
(Russian),
(20)
937
ation" (Heart Beat Considered as a Relaxation Oscillation), by B. Van der Pol and M. Van der Mark, Onde Electrique, 1928, p. 365.
(21
Levin-
son and O.K. Smith, Duke Mathematical Journal, Vol. 9, June 1942.
(23)
2,
935
Telephones, December
(27)
925
Cartan, Annales
(Self-Excited Systems), by
133
(30)
of Technical Physics,
(31)
Engineering), by
Dreyfuss, 1911.
PART n
ANALYTICAL METHODS OF NON- LINEAR MECHANICS*
44.
INTRODUCTORY REMARKS
It is apparent that for practical applications the general quali-
tative methods reviewed in Part I are not sufficient, and that quantitative
mine the amplitude and phase of an oscillatory process with a certain prescribed accuracy once the general qualitative aspects of the phenomenon have
been ascertained.
In general, there exist no methods capable of yielding exact solu-
equation
x
ux =
nf(x,x,t)
[44.1]
if
= 0.
Polncare (1)** has shown that very near some of these solutions, when n = 0, periodic solutions of Equation [44.1] may exist for very small values of the
parameter n. The search for these solutions is the object of the method of small parameters of Poincare. This method, with its various ramifications,
constitutes the principal subject of Part II.
The scope of the quantitative methods available at present is
It is restricted, in fact, to the class of non-linear dif-
rather limited.
ferential equations of the type [44.1], and for this class it is further restricted by the condition that the parameter n should be very small, that is,
H
great and its applications in various branches of applied science are extensive.
Only in special cases when the parameter n is large does the theory
The text of Part II follows the presentation contained in the two treatises on Non-Linear Mechanics: "Theory of Oscillations," by A. Andronow and S. Chaikin, Moscow, (Russian), 1937, Chapters VII and VIII. "Introduction to Non-Linear Mechanics," by N. Kryloff and N. Bogoliuboff, Kieff, (Russian), 1937, Chapters X, XI, and XIII.
1^6
of Poincare cease to be applicable, and for such cases analytical methods are
practically unexplored.*
We shall frequently refer to Equation [44.1], with
fi
as quasi-
linear, which means that the solutions of this equation do not differ appre-
ciably from solutions of the corresponding linear equation when n = 0. The important point to be noted is that, although the solutions of [44.1] do not differ much from the solutions of the corresponding linear equation if
fi
is
very small, these periodic solutions do not exist in the neighborhood of any periodic solution of the corresponding linear equation but only in the neigh-
The establishment of
conditions for the existence of periodic solutions of [44.1] when n constitutes the crux of the theory of Poincare.
thus
procedure.
n
t
n
t
sin ut and
cos
u)t,
pansions.
any bound as
-> <,
Then
sina cosfit
+
,
cosctt sinfit
, at
(atf - -g-p +
t
,
cos fit
<*)
2!
sin/3*
[44.2]
fit,
cos
fit,
at and cos at is
<>,
If,
however, n
-> ,
and the whole expression [44.2] then becomes a periodic function of time with
period T =
2tt/u.
certain finite number of terms; hence one is generally confronted with secular terms.
In the original work of Poincare this difficulty was obviated by
137
astronomical purposes.
ionic generator oscillating with a frequency of several megacycles per second, it is apparent that in a few seconds such a generator will pass through
which eliminates the secular terms in each step of the recurrence procedure.
This was accomplished by Kryloff and Bogoliuboff
.
Andronow and
S.
Chaikin,
(4).
analogous in some respects and follow a method similar to the method of variation of constants of Lagrange.
Kryloff and Bogoliuboff, which is an attempt to simplify the problem by reducing the given non-linear differential equation to an equivalent linear one.
CHAPTER VIE
METHOD OF POINCARE
45.
CONDITION OP PERIODICITY
Consider a system of differential equations
x
ax
by
/jf^Xyy);
ex
+ dy + uf2 (x,y)
[45-1]
where f x and f 2 are the non-linear elements of the system and /i is a parameter. It will be assumed that f 1 and f 2 are analytic functions of their
1 ]
when n =
0.
(a
+ d)S +
(ad
be)
[45-2]
periodicity
a
0;
ad -
be >
[45-3]
Under these conditions the linear system admits an Infinity of periodic solutions of the form
x
= Kcosiot +
0);
= gKsiniut +
<t>
X)
[45.4]
where w = Vad
4>
be
is arbitrary and can be made equal to zero by a suitable choice of the ori.
gin of time
is then
[45-5]
x (t,K)
Kcoscot;
y (t,K)
= gKsin(ut +
X)
so that x
Let us
and let
For
x (0,K)
^
and
and
y(0,fi,K)
y (0,K)
yS
[45.6]
X
J3 1 (/u)
2 (j")-
It is obvious that
/3
(0) =
(0)
= 0.
fi,
fi x
and
/3 2
139
|jci'|,
l/?il,
and
\/3 2
are sufficiently
0<
<
t x
time.
and
/? 2
in the form
x{t,n,p ,P 2 ,K)\
x
y{t,n,/3 lt /3 2 ,K)
it is logical to assume
where
r(//)
/u
approaches zero.
Our chief objective is to show that under certain conditions periodic solu-
fi
is small
is clearly
,K) ,K)
= =
,
and
/3 1 [fi)
/? 2
(j")>
and
t[ju)
so as to satisfy Equations
so that
t(0)
/^(O)
(O)
[45.8]
The phase is arbitrary, however, so that it is possible to assume that one of the
s,
say
fi 1
equals zero.
Putting 2 =
fi,
written as
x(T
+ t,h,O,0,K) -
x(O,n,O,0,K)
= =
<j>(T,ti,0,K)
[^5.9]
ip(r,
i*,0,K)
system [45.1], with the conditions expressed by [45.3], has an infinity of periodic solutions corresponding to the arbitrary values of the integration constants
= U^i(r
fi,fi t
K) =
[45. TO]
iP(t,v,0,K)
fii/f 1
{T,n,/3,K)
140
/i
in
[n,K) =
t(/i)
and
x{j
{ii,K)
= 0, where
4>
and
ip
<j>
and
\p
in
which
dfi
e/i
Pin)
d lfx
e lfx
[45-11]
(j>
(t, n, P,K)
c/3
and ip^r,
n, P,K)
we get
<f>oi
at*
+ br +
+ + ^i T
+
+
=
[45.12]
^i
^oi
a i^
i^
'
=
and pin) from [45-11]
t{/j.)
= =
^oi
^oi
^( a
+ bd + +
^i^
cd
+
/i
^( a i
c i^i)
[45.13]
powers of
fi
0oi=0oiW =
a
^oi
= <V*) =
bxd
[45-14] [45-15]
fi
bd
cd x
0;
a1
c1d
must be equal to
zero and [45-15] that the system of the two equations must yield the values
of d and d x which determine the quantities t(/i) and p(fi) to the first order.
It is apparent that this is possible whenever the determinant
6i
C]
0,
With Expressions [45-12] taken into account, this is equivalent to the condition
50!
d0!
d/3
J =
dr
djh
dip,
6(t,P)
[45-16]
6t
dp
Hence, whenever the Jacobian [45-1 6] is different from zero, periodic solu-
provided
fi
If Equations
For
/i
141
For
\i
may exist in certain restricted regions of the phase plane in which the condition [.45.16] is fulfilled; the value of the Integration constant
amplitude, is determined by solving Equations [45.14],
K ls
the
when
fi
and
46.
fif(x,x)
[46.1
fi
= 0.
fi,
3 lt and
/? 2
>
we know by the
theorem of Poincare that this expansion converges in any arbitrary but finite
time interval provided these quantities are sufficiently small in absolute
value
We obtain
x
<j>
(t)
+ AB + Bp 2 +
X
Cfi
+ DP x n + E$ 2 n + Fm +
[46.2]
where
A, B,
are functions of
t.
expansion [46.2] with a periodic solution of [46.1] provided \n\, \P\Kil\\> and |/ff 2 (/i)| are small. Differentiating [46.2] with respect to t, we obtain
the following equations
x
<f>
(t)
+ Ap + B0 2 +
x
U)
+ A8 + BB +
1
[46.3]
we get
2
/<*,)
<x
)(y +
o
vd^ + \{x
*)
(^)o +
\ax%
<t>
\dxdx>
Substituting x
- x
= x -
(t);
x - x
= x -
4> Q
[46.3] into [46.4] and replacing x, x, and f{x,x) by their values [46.2],
[46.3], and [46.4] in the differential equation [46.1], one obtains a series
2
>
/Sj
P2
P\l*>
&zU->
f*
3 lt
one ob-
second order, one obtains nine differential equations, of which three are
A + A =
0;
B+B
[46.5]
142
+ C =
;
/<*.,,);
b +
;
Z>
(gU
jt
(|f)
[W.5]
;
f +
h^j
and ns^)
x
{t)
</>
{t) have
in the form
xQ
XQ
fi
y=
2
A/Sj AySj
+ BP 2 +
C/i
C//
+
+
Dfi x n
D^j'/I
+
+
= ^ =
+ =
5>S 2
[46.6]
5(0)
A(0)
[46.7]
C(0)
D(0)
E(0)
F(0)
C(0)
I>(0)
E(0)
F(0)
With these initial conditions the first two equations [46.5] have the solutions
A =
with period 2n.
cos
B =
sin
v =
V{t),
V(u) sinU
u) d\
[46.8]
Replacing V[u) in Equation [46.8] by the right-hand terms of the last four
equations [46.5], one obtains the following expressions:
A =
B
t
cos
A =
;
sin
sin
B =
C =
cost
t
\f\<j> (u),<j>
(u)}sm{t
u)du;
f[<j> (u),
(w)] cos(t
u)
du
[^6.9]
- b-] 'df ox
n "0
t
cos it
df -r-^sinw
.
ox'0
sin(
u)
du;
D =
j
df
dx,
cosu
df ^ sinw
.
dx
cos(
u)du
-/ dx
6
df -
t
.
sintt
df + ^r-r-cosw
,
dx
sin(i
u)du;
E =
F =
df -
dx Q
smu
.
df + -^r-cosu
,
dx
cosU
u)
du
df -/ Ldx c
^lc dx,
sin it
u) du;
af c + Slc dx o dx
cosit
-i
u)
du
145
where x
K cos
and x
sin
Replacing
1;
A{2tt)
B(2tt)
0;
1
2*
C{2n)
C{2n)
f(x ,x )cosu du
D{2tt)
1 df sin2w
.
df
sin
-x
0x n
u du;
D(2tt)
df
j dx
2n
cos
dx.
10/- du
-rrr-sm2u
dx
E(2rr)
df
j
dx
Q
sin
10/u
2
[46.10]
'
-z~^sm2u du;
ox
E(2tt)
1 df
2
dx
'
sin
9 2%
df + tp- cos 2 u du
,
dx
F(2tt)
j
2n
df r( df C(u)smu ^-L{u)smu du dx dx
,
F(2tt)
5
LyX
C(u)cosu
-j-1-
ox
C(u) cosu du
D
)
and
1/7
We have
-d
,, (J cos u) =
v
dw
df - cos u
cm
. ;
J sin u
df df dx df dx y = 3 = r-^ + ^
aw
dx
du
dx
du
and x
x
are
= Kcosu;
A"sinw;
= K sin u
0# -~du
du
hence
Kcosu
M. = A v Qf
du dX
r
sin
0/ it ^r^ cos u
dxr
and thus
1
costt
df f cos 2 u -
dx
sinw
or
du
! 0/ o = - sin 2%
dx
t-t-cos
df dx
w,
sinw
rJ
Iho.1
144
and similarly
1 K
-r
d if \ (/smw) du
df =
dx
sin
13/./ + K
^-T-sin2w
dx
cosw
, n .f [4-6.12]
-sr
/,/
2.
Di2n)
- }dxE(2tt)
du
--K>
^C(2tt);
Di27T)
=
j
oX
du
~ir
[46.13]
= -
E(2tt)
-^C(2n)
If C(2tt) =
and C(2tt) *
0,
l
D(2tt)
"df j = l-^-du;
= f^/ du \ _,
C(2tt)
K
[46.14]
E(2n)
If C(2tt) = C(2;r) = 0, one has
= -
J^C(2tt);
E(2tt)
2k
D(2n)
= f-^-du;
=
Z>(27r)
= =
f-^-dtt
[46.15:
(2tt)
0;
E(2n)
From the latter form of the expressions for D(2n) and D{2n)
4?4 and
it is
apparent that they represent the constant terms in the Fourier expansion of
W-, multiplied by
),
2-n.
C(2tt)
and C{2n)
as given
t,
)
cos
in the expansion
multiplied by n.
Hence, if f(x
xQ
cos
and x (t) =
1
sin
/S 2
the amplitude
and
so that Equations
0(t, fi,P v
p2 =
)
and
ip(r,
filt fi 2 )
[46. l6]
145
fi lt
and
One of the /?'s, however, is arbitrary and can be taken equal to zero, as
previously mentioned.
r and
/3 1 [/i)
If,
/Sj
0,
and
>
x(2n + r)
a
x(0)
t) t)
=
=
x(2n)
x(2tt)
+
4-
tx{2tt)
TJc(27r)
+
+
[46.17]
x(2tt
The coefficients
A{2n)
B{2tt)
fi
Con-
sidering r and
t)
x
2
(2tt)
A(2n)/3
B(2ir)02
C{2tt)^
D{2tt)P 1 h
+
2
E{2-n)p 2 n
+
[46. 18]
F(2tt)m
+ tx
(2tt)
tA(2tt)/3 1
tB(2tt)/3 2
tC{2tt)h
^x^n)
x(2tt
t)
x
2
(2tt)
A(2tt)0 1
+ B{2n)p +
2
1
C(2n)n
D{2tt)B 1 h
+ E(2n)0 2 n +
2
+
But x
F(2tt)m
+ tx
(2tt)
tA(2tt)0
tB{2tt)$ 2
tC(2tt)iu
+ ^-x'
(27r)
[46.19]
(27t)
Furthermore, A[2n) =
1,
A{2n) =
0,
B[2tt) = 0,
and B[2tt) =
t)
x(0)
= +
r K
+ rp 2 + +
C(2tt)^
+ +
C(2tt)tii
2
D(27r)/3 lfi
E{2tt)J3 2 h
F{2tt)h
[46.20]
and
x(2n
r)
x(0)
= - Kr +
rfi 1
+
+
C(2n)fi
C(2tt)th
+
2
D(2n)/3 lfi
E(2tt)8 2 h
F(2tt)^
[46.21
146
fi
Since
/a,
fi lt
82
[46.20] is C(27r)fi, and in [46.21] there are two terms of the first order,
-
Kt
and
C(27r)/u.
two equations
C(2n)
IfiKcosu,
Ksmu)smudu =
4>{K)
[46.22]
2k
C(2tt)h
Ko
/(iTcosw,
K sinw) cosudu =
if
fift(K)
[46.23]
duces to
D(27r)yS 1
F(2tt)h
[46.24]
Since r in this case is zero to the first order, we may proceed to the sec-
o/u
2
.
From [46.21
],
/? 2
= 0, and
where C{2n) =
0, C(2tt)
Dividing by
fi
Kov.
Z)(27r)/V
F(2tt)m
/3 X
expression, we obtain
F(2n)D(27v)
F(27t)D(2tt)
cj [45 2
K-D{2tt)
The correction r =
a/u
K**
2
,
/?j
and r =
ofi
the amplitude
in Equa-
Kcost +
fi
sin(t
u)du
cos
U\Itt)
[46.26]
147
where
if
C T+ T
f[x Q ,x
sinU
u) du
represents the first term of the Fourier expansion of the function appearing
on the right side of Equation [46.1].
r,
The
terms does not destroy the periodicity but merely accounts for a modification
of the period, as was explained in connection with Equation [44.2].
The
effect of the appearance of secular terms can also be ascertained from the
following example.
x(t)
K=
*~
[46.27]
/i.
fi
gives
K=0
\aK (0)cosKu(0)t
00
+ bK (0)sinKu)(0)t +
4-
+ n[a' (0)cosKu(0)t K
K=
aK (0)oj'(0)Ktsir\Ku)(0)t
+ b K (0) u
'
(0)
Kt
cos Kcj(0)t\
2
/i
K=
[46.28]
where a' b' b K {fi), K K and o>' designate the derivatives of the functions a K {/i) = and oj[/i) with respect to n in which the value has been substituted after
,
/j.
differentiation.
It is observed that,
the appearance of secular terms does not destroy the periodicity in view of
to
and x
cos
t
which amounts to the choice of generating solutions an 2 CH' in the form if cos 1 t and - K sin 1 t, instead of
2tt
2n
and
sin
t.
should be replaced by
where
it>
148
47.
tionary solutions in this case are periodic ones, and the topological representation of such motions in the phase plane does not present any particular
difficulty, as was shown in Part
I.
Although theoretically the extension of Poincare's method to systems- with several degrees of freedom follows the same argument, the practical
difficulties rapidly increase and the benefit derived from topological considerations disappears.*
Moreover, the stationary motions in systems of more
It thus becomes
play an important role in applications, and for that reason it may be of interest to give a brief outline of the method of Poincare in connection with
such systems, omitting the details which have already been explained.
The
calculations of Andronow and Witt (7) (8) are given in this and the following
sections
co
x ~y
fif(x,x,y,y;n)
figix, x,y,y;/Li)
oj
where
/i
In fact, a dynamical system with two degrees of freedom is generally reducible to a system of four differential equations of the first order, and its representation in a phase plane becomes generally impossible. Only in very special cases of the so-called "degeneration" defined in Part IV is a planar representation possible, but such "degenerate" systems possess entirely new features which are not investigated here.
149
.RcostL^
<f> Q
(t);
with period
T =
2 7T
or
x
[^7-2]
0;
Rcosto 2
ip
(t)
with period
2n
u)
2
(*)
+
77,
|;
77
[47.3]
and
4-
to?*
= 4/(0
,0 o ,O,O;O)+
fj + fj + fy + f-v+f^
g y rj
+Gz(f,
I,
V,V,)\
[47.4]
77
a>
rj
Ai[fif(0 o
,0 o ,O,O;O)
+ gj + # +
g y y\
g^/i
fi.
=p A + $B +
x
/? 3
C + P 4 D + n{E +
ySjF
+p G +
2
K + nL +
+ O 2 (0
lt
,0 3 ,0 4 ,n)]
+ d2 {0 v /3
,0 3 ,04 )
[47-5]
77
=p A +
x
B + 3 C + pj) + h\e +
2
fijF
/3 2
G + @3 H +
K+
/xL
+ O 2 (0 v
where A,
,P s ,/3 ,m)] a
+ Ot (0 v fo,fit ,0J
and A,
1
(0);
p2 =
|(0);
/? 3
77(0);
J3 4
77(0)
[47.6]
One of the /S's, as will be seen, can be assumed to be equal to zero; for example,
y8 2
= 0.
fi
system of
A,
L and
150
A = cosw^;
C =
0;
D =
0;
(p v
3 ,/? 4 =
)
F = F =
=
1
1
/(0 O '^o' O
'
O ;)
sinWjU
u)
du
1
i
/
o
(cosc^u -/x
Wj
sintdjtt
/x sint^U
)
u)
du
if
Wl
1
f
I
(coso; 2 w -/^
w 2 sino> 2 w
-/^) sintjjCi
w) o?w
-K"
sinw w
(
/_.
coscj 9 u /
sinw,(
u)
du
L =
\[Ef
i
/J
sin Wl
U -
u) d
[t7.7J
j4
0;
C =
cosoj 2
t;
D =
6> 2
sino> 2 ;
2 {fi
v P^,fi A )
^ F =
1
2
sr
(0 O
,0, 0; 0)
sinw 2 U
u)
du
F =
=
1
1
(coscjjW
gx
Wjsinw^
gf
I )sinj 2 (^
u)
du
if
(cosw 2 m
-gf
u 2 smcj 2 u
g^)sina> 2
u)
du
2 o
K =
L =
smto u
J
(
gy
coscj 2 w -g
smu) 2 {t
u)
du
fJ
ZT
9x
+ Edi + Eg y + Eg y + ^J sinw 2^ ~ w
>
rfw
i2tt = x( +
t)
,,
x(0)
'
0;
,., [x]
./2tt = x(^ +
t)
x(0)
=
[*7'.S]
[y]
=
(if
+ T) ~
^ (0)
0;
^]
+ T "
)
[
2>
(0)
defined by [47-8],
become
151
Ricoscj.t]
+ +
1 [A]
+ J[E] + P^F) +
fc[H]
[K]
fi[L]
[O z (0 u
,0 4 ,n)])
Ri-cj.sincj.t]
[A)
fi{l$]
X [F]
[H]
[K)
fi[L]
[O 2 (0 lt
,0 4 ,n)])
[47-9]
3
iC]
+ +
lD]
+
+
/i([E]
+ +
fijiF)
+ +
[H)
+ +
{K)
+ +
nil]
+ +
[
[O 2 (0 v
,0 4>fi)}) =
[C]
lD]
n{[E]
pJLf]
lH]
[K]
fi[L]
[O 2 (0 v
,04 ,n)}) =
In power series of t,
= =
a( VWl i(^
+ +
t)
'
A(0)
= =
an
a,r
a2 r
+ +
[47.10]
[A]
rj
A(0)
+ a^ +
a2 r
where a
and d
One obtains
- *^- +
- Ru>It +
PMo +
0,(0,
a ir)
+ +
^(e
+ +
e xT
0J
+
+
/3 3
ho
ko
+ +
nl o )
+ O 3 (0 v /3 s ,0 + O 3 (0 v
+ O 3 (0 lt + O 3 (0 v
3 ,/3 4
,fi,T)
= =
ffil
r)
fi(e
er
T +
^/
ho
ko
fii
,v,t)
(CO
+ Cxt) +
(d o
+ +
d lT ) +/x(e
+ +
T +
M
X
[47-n]
+ Psh +
0A +
4
/"*<>)
,0 4 ,u,T)
/? 3 (C
+ ^t) +
(d
fi(7
IjT
/O + 3 t +
tQ +
nT,)
,/3 4
M ,t)
=0;
;
0;
0;
d1
= ^
sin
o)
1
[47.12]
=
.
cos y
sin
C =
rf
cos y
where y = 2tt
1
O 3 [0 ,0 3 ,0 4 ,m), gives
1
t,
viz.,
= 4o +
0i<*i
03<*s
+^
)"5
+ O 2 {0\;0Sf
4 ,fi)]
[47.13]
152
where
*o
- R fo ~
*o
_Afi_.
- _k_.
^1% +
Introducing the value of r given in Equation [47.13] into the remaining equations [47
.11
]
,
/? 3
(cosy
1)
(- siny)
fie
+ O 2 {0 v + O 2 (0 v
,0 A ,u)
=
=
[7>1
(- w 2
siny)
/S
(cosy
1) 4- fitQ
,0 4 ,/i)
1 ]
to the first generating solution [47-2], it is necwhich reduces for n = and /S 4 (/i) approach zero as n apessary that the functions /3 1 (fi) 3 (//),
,
/ff
proaches 0.
=
o
f(Rcoso) u,
1
cj
1
Rsmu
u, 0,0; 0) sinc^jU
du
[47-15]
Prom this equation one obtains the amplitude of the periodic solution in the
neighborhood of n = odic solution is
0.
Jo
o
cos y
*o
1
sin
sm
1
U)
y
1
= 2/
(l
cosy) +
[47-16]
o 2
cos
f =
1
I
{cosu 1 u
fx
o) 1
sm(J l u
fj)
sin^w du +
[47.17]
If the determinant
If all determi-
153
<f>{t)
Rcosoj^ + ftA
+ n[E + ftF +
fttf
K + nL +
fiL
(ft,ft,ft,/i)]
[47- 18]
y
#(t)
K+
+ O2 (ft,ft,ft,0)]
where
by [M-7.7]; and ft, ft, and ft by [47-14]. Replacing ft, ft, and ft by their values in [47.18] and arranging the terms of the series according to the powers of fi, one obtains
X
is determined by [47.15]; A, E, F,
=
y
<f>{t)
fl<t>
+ M
M
<j>
fl
[47-19]
<t>{t)
fifa
<fi(t)
u;
^{t)
[48.1]
+ +
cj
= =
fi{fx u
+ fx u + fy v +
+
gx u
f^v)
[48.2]
v
u> 2
u(g x u
gy v
+
v
g y v)
g i are the
x,
variables in which
respectively.
x,
replaced by
<t>{t)
<j>{t),
J>[t)
and
4>{t)
'
Afi x
B/3 2
+ Cft +
Z)ft
/3's
where A,
ft;
w(0)
ft;
v(0)
ft;
i(0)
[48.4]
154
g^
the functions A,
can be assumed to be
A = A + A = A +
/uA,
+
-f
fi
A2 +
[48.5]
/uA 1
2T
fi
A2
4-
Substituting the expressions [48.3] into [48.2], taking into account their
form [48.5], and comparing the terms with the same powers of
fi,
one obtains
A lt
can be determined.
sions
A =
=
cosco^;
i
A = F'
x
etc.
sinw^;
1
i
B =
x
i
i
( sinwjW /x +
i
coswjit
-fA smu
{t
u) du;
etc.
C =
0;
= H;
etc.
[48.6]
D =
Q
0;
Z>\"= if;
etc.
A =
0;
A =
x
F;
etc.
u(0)
= =
[u]
= # i;
= ^
3
;
u(T)
v(T)
u(Q)
= =
[u]
= =
ip
[48J]
v(0)
[v]
v(0)
[v]
ip
where T =
-^-
dp
+
1
1 -
aT
d*l>i
diP,
dp 2
dp,
dP s
aT
dip 2
dP,
dip 2
dtp2
dp,
dP 3
dip 3
dp,
=
d\Pz
dp,
dih dp,
dP 2
dip,
dP 3
aT
1
-
djh dP,
djh dP,
[48.8]
diP,
dP 2
dp,
+ l-e
155
aT
e
IB]
[B]
[C]
[D] [D]
[A]
aT
e
[C]
[48.9]
[A]
[B]
[C]
+ l-e
[C]
[D] [Dl
[A]
[B]
+ l-e'
with [A] =
a(^4
t)
A(0), etc.
Putting
a 12 a 22
~^~
'13
'14
A{p)
a 21
a a S3
'24
=
+
P a 34 a 44
^41
[48.10]
"42
43
ap'
6p
cp
rf
[48.11]
with
a
--
an
-+-
a 22
-r
a,,
-r
a 44
a 99 + ^ ss + a 44 t a 33 +
;
a..
a..
[48.12]
A u + A 22 + A 33 + A 44
where
rf
4(0)
An
A 44
and
A 22
A 44 are
33
Since one of the characteristic exponents is always zero (9) because the equations are autonomous, d = 0, and the quartic equation thus re-
+ ap 2 + bp +
[48.13]
If the motion is stable in the sense of Liapounoff, the remaining three char-
acteristic exponents must have negative real parts, which means that the moduli of the quantities
e
aT
number p =
aT
p 2 )-plane is mapped into a half plane {z lt z 2 ) so that the circles, see broken line in Figure 48.1, transform into straight lines, see broken
z2
1^6
Figure 48.1
Figure 48.2
By this transformation the problem of finding the roots of [48.13] with moduli
situated inside the circle of the (pj p 2 -plane is reduced to that of determining the roots of the transformed equation having negative real parts. This
,
If this
(-c) +
(2b
3c)
+ z(-
4a
46
3c)
(8
+ 4a + 26 +
c)
[48.14]
(11) are
>
c
0;
26c
a6
ac
<
0;
+ 4a + 26 + c
>
[48.15]
These are the necessary and sufficient conditions for the roots of Equation
[48.14] to have negative real parts, or, which is the same, for the roots of [48.13] to have moduli less than two, which assures the stability of the peri-
odic motion.
26
3c > 0;
<
0;
(6
c)
a(b
c)
<
0;
4a
26
>
[48. l6]
[B],
can be devel-
for example
2
fi
[A]
= =
[A
+ fiUJ + +
/i[5j]
[A 2
bn
v +
(m)
a 12
[B]
[B
2
fi
[B 2
b 12fi
2 (fi)
The value of the determinant 4(0), in which are written only terms
//
fi,
is
157
b n fi
+
+
b\ 2 fi
b 22 n
+ + +
b n fi
b 2S /j
bu
A(0)
b 2l fi
b
b
nf +
n fi +
b Z2 n b i2 v
(cosy
l)
+
+
633/U
siny
Mn +
un +
o 2 sin
6 43/u
(cosy
1) 4- b
[48.17]
The values of
to account
6 12
Taking in
= =
ju2(l
cosy)(6 n
b 22 )
2 {fi)
(b
c)
a(b
c)
fi2(l
cosy) cosy(6 33
6 44 ) 4-
siny(w 2 6 34
4-
2 {fi)
+ 4a
2b
4-
lb
[48.18]
4(1
cosy)
4 O^u)
3c
4(1
cosy)
4-
O^/u)
If we consider
fi
1 ) 7r,
b 22
<
[48.19]
cosy(6 33
6 44 )
+ siny(w 2 6 34
^-j <
The values of
b 22
bn
Wl
I
(cos Wjit
/,.
Wj
sincjjtt -/j)
sincjJ
u) di
^22
PvZT
1
s * nu) i u
'
fx
cosWjW /i j coswJ
u)
'
du
6 33
Wl
(cosw 2 w -g
u^sinu^u
g-j s\nu 2
u)
du
<j 9
siny
[48.20]
6 44
=
J
1
{
sinu 2 u
gy
4-
cosu 2 u- gy)
cost
1
u)
du
a io.,siny
6 34
J ( smio
Wl
u gy
cos<j 2 w
g y j sin<J 2 (
uj
du
4-
a cosy
6 43
=
J
\cos
u gy
co
sin u) 2 u-g
cos
(j. 2
wl
dw a w 2 cosy
158
Introducing these values Into the inequalities [48.19]. one obtains the following conditions of stability
j" fx
l
du<0;
^g.duKO
R
cos
[48.21
u u\
x
and
fj.
= 0.
is determined by [47-15].
We shall return to this matter in Part III in connection with the question of stability of coupled electronic oscillators.
This problem is ex-
49.
thermionic generator.
of the
=
>
fi(0
- Z6v 2 )v
6
[49.1
and
>
0.
fi
is
<j>
(t)
Kcost;
2
(O
= - Ksint
[49.2]
~$6v v
and hence
3
f(v
,v
= - /3Ksmt + S6K
6K 3
=
2
cos
sint
[49-3]
$K -
-ft
and hence
[49.4]
K = |f
pends on the ratio
*
Thus the amplitude of the generating solution to the first approximation de-
yj.
2 The reader will note that by retaining the quadratic term y^v in Equation [30.5] one would have Equato 2n of the term Zyvv in Equation [46.22] tion [30.9] instead of Equation [4.9-1 ] The integral from vanishes, which proves that the term y-yv has no effect on the calculation of the stationary motion. This remark applies also to all even terms in the polynomial [30. 4]. See also Section 542,
159
is smaller.
->
indefinitely since the factor that eventually limits it is precisely the nonlinearity of the characteristic expressed by the term
[30.5].
2
,
in Equation
f- = dv
6dvv
and
dv
3<5v
so that
:r
= 66K
cos
sin
36K"
sin
and
-^i-
= $ -
3<5#
cos
jr 2
Z>(2tt)
j(/3
^^-J
= -
2tt/?
[^9.5]
D (2 7r)
period r = an
is
F(2tt)
[^9.6]
tt^ /?
[49J]
C{t)
= - -j-
sm3t
-^~smt =
--y- sm
st
1/_
2
sin*
l/
r
cos
do
(1-
+
2
\fj
--Mt
2,
.(i-*f). +
+
* +
o2
ip
[^9-8]
It is clear that the periodic solution
2 l/^r-
period is of the second order and hence can be neglected for small values of
j/.
The secular terms do not appear here in view of the fact that the cor-
l6o
50.
In order to
]
the equa-
pf(x,x,X)
[50-1]
fore; that is, there appear certain generating solutions in the neighborhood
of which periodic solutions exist when p is small.
lutions when the parameter k determining the state of the system varies and
of the generating
solutions is given by Equation [46.22], which can be written here in the form
\) sinu du
[50 .2
Putting
?*
Yp = 6(p'X)
=
IT
f(Vp
I
cosu,
Vp
du
[50-3]
p,
we have
<j>Ap,X) 9
2np
fr Vp x
sinw cosu
+ f-Vp x
cos
u\du +
J
2ir
2ir
2ttVp
/ cosu
+ -
2nVp
Z*
7=
J
sihw du
2ir
L-
/ sin u du by parts
^(p,X)
1 = wzlfidu 27r
o
[50.5]
161
rameter
k,
f(x,k)
0;
f(x,k) <
J
[50.6]
so that & p {p,k)
fi du <
<0.
Associating the function 0(p,A) with f(x,k), we may infer that the limit cycles are stable if
<t>{p,k)
0;
</>
(p,k) <
[50.7]
A proof for
the conditions under which stationary motion is stable may be found in Chap-
2
,
the square
= f
(x,k)x
[50.8]
fx {x,k) =
a {k)
a x (k)x
a 2 (k)x
[50.9]
To this class belong equations of the generalized Van der Pol type.
Substij/p
cos u
and x =
<fi(p,k)
ap
q2 p
q4 p
+
[50.10]
<t>Ap,k)
o^p
2
3a 4 />
8
a,,
a4
are functions of k.
51
It is observed that by
increasing the coefficient of mutual induction k between the anode and the
grid circuits, self -excitation starts smoothly as soon as a critical value
k = k
of this parameter is reached; for k > k
162
51
-1
representing
this "soft" case of excitation; upon decreasing X the phenomenon takes place in the opposite direction, as shown by the arrows.
The theory of soft self-
Figure
51 .1
Figure
51 .2
smoothly for X ^ X x
phenomenon is different; namely, for X = X : the self -excitation does not disappear; it disappears at X = X
< kx
called "hard."
hysteresis" (13).
Section 24.
51 -3
3F
+ JM + /<
IJdt =
[51.1]
where I a = f{V) is the non-linear function expressing the anode current I a as Let us approximate this function by a power a function of grid voltage V.
series in
We have
/
= fly) = a Q V +
2 V +
V3 +
<5
V4 +
[51 .2]
163
In-i
Figure
51 .3
Vt
is
the "saturation voltage," that is, a sufficiently high grid voltage beyond
Since
we can
V.
dt
[51.3]
V,
d
'
MX
dt 2
V = TTX
s
.
M"'
;
* --"
m\ X(1 M
LCx + RCx +
2fi
4-
3y
V+
45
V+
5e
7
=
4"!
dx
dt
[51 .^3
where w
K
cients
"
T)
2/?0
^X +
3r^2a;2
4<WV
5e <^
41 ]
dx
dr
/ff
,
of [5T.2] are small. One can take one of these coefficients, as a factor and write
for example
a x
,2
dr
V uo M Q
s
Ha^C
2x
3^, + 4^
2
5^
e
T
r3
dx
x4
Jd7
[51-5]
164
aM
;
RC
V
a(JO;
2M =
b(M);
^ =
Po
c(ikf);
V Po
2
d(Af);
V Po
i
e(M)
44 (XT
CT
[51 .6]
transformations, as
<p{p,M)
ax p
a' y 2
Mp +
a3
Mp
[51
J]
where
- RC
3V^
5V[
p,
one has
(p,M)
tfj
2a2y
Mp +
3a 3 e
Mp 2
[51 .8]
The discussion of Equations [51-7] and [51-8] yields the qualitative features
of the phenomena.
A.
> 0, y
< 0, and
= 0, Equation
[51-7] becomes
a2
\y
<fi(p,M)
(a 1
a2
\y
\Mp)p
(a
- RC -
\Mp)p =
[51 ,q]
In
a straight line p =
and a hyper-
bola
a 2 \y
\Mp =
[51-10]
equation
Mo
x
RC
= 0.
The value
=
1
[51-11]
165
<
to instability
{M >
and a sta-
ble limit cycle appears; the square of the amplitude of the latter increases
M+
ly strong coupling M.
The curves representing this case are shown in Figure 29. Ta and b.
The former represents the condition for
M>M
M< M
B.
> 0, y
e
,
>
and
the abso-
lute value of
[5.1
a^
and a 3 i"
= n, Equation
.6]
becomes
<f>{p,M)
= Uj + mMp - nMp
)p
[51 .12]
and a
curve
(Mct
- RC) +
mMp - nMp =
2
[51-13]
is clearly a 1 = 0, which
[
5T -11].
dp
_ dM_
00
dp
dM
2 _ a + mp - np M(m - 2np)
^,
J
'
M =
u u
RC
2 mr
[51-15]
An
[51 .11],
When
is compared with
from Equation
it is seen that
<
M= M
This
in the shaded area; whence, applying the criteria of Poincare, Section 13,
166
The
axis p =
is stable for
If
M< M
M>M
value
M
x ,
M= M
i
For
M= M
discontinuous jump
M N,
X
If
M>M
Figure
51 .4
and
M increases,
M>M
and
M decreases,
when
M= M M
This corresponds to
c.
M decreases
from
to
M= M
is
reached, the stable and the unstable limit cycles coalesce, and no limit cycle exists for
M<M
[x) = a x x - a 3 x
6
,
3
,
F2 (x)
= a x x + a 3 x*
where a x
a;,
y5
-.
which it is biased, each of these cases can be obtained in practice by a suitable adjustment of the circuit
CHAPTER
IX
vf(x,x)
y;
+ nf{x,y)
For
fi
[52.1]
=
,
one has
a cos
b sint,
with
sin
4-
6 cos
[52.2]
where a and
Va 2 +
2
.
If,
about the common origin of both systems, in this rotating phase plane Equa-
at a distance
OA
= Va 2 +
from origin 0.
The inclination of
6 =
OA
-1
tan
formation is nothing but the usual method of representing sinusoidal quantities by vectors, used in the theory of alternating currents.
We shall call
,
the rotating plane of the variables (a, b) the Van der Pol plane.
da
-\
db
dt
dt
7- sin
=
[52.3]
-j sin
at
at
z-f
cost
juf{acost
b sin
a sin
b cos
whence
j_
-jt
= uf(a cos t +
b sin
sin
b cos t) sin
[52.4]
-jt at
ix
f (a
cos
b sin
sin
b cos t) cos
Since n is small, by assumption, da/dt and db/dt are also small because f(x,y) is bounded.
In other words, a and
b
168
However, if x
and y are replaced by their expressions [52.2], the right sides of [52.4]
are periodic and can be expanded in a Fourier series so that [52.4] becomes
da Jt
"0 (a o
b)
2
~& (a, b)
(a
6) cos
(a, 6) sin
+ # 2 (a
6) cos It
[52.
db
dt
y"
i//
(a, b) cos
^j(a
b)
smt +
ip 2
{a ,b) cos 2t
explicitly,
da
It
<t>Aa,b)
2
'
db
dt
^ (a,b)
"
2
[52.6]
They are obtained from [52-5] by dropping the terras containing the trigonometric functions.
'
= -
= +
1
/(a
7T J
cos |
6 sin
sin
6 cos
sin I
[52.7]
to(a,b)
z"
77
/ (a
cos I
b sin
sin
b cos
cos 4
b,
= a
2
,
one obtains
-7 =
2
if -7
at
at
ir~
^^n
a cos
b sin
sin |
6 cos
( a
sin
6 cos
di
Putting
a cos
4-
sin
Kcos($ 0)
and
a
where
= tan
-1
sin
6 cos I
if sin
- 9,
^-,
one has
4^ at
/*
4>{K)
4r at
= f
0UO
[52.8]
169
where
<b(K)
2n
1
r*
/
I
K cos
K sin u
sin
u du
[52.9]
2*
ip
(K)
= + - 2nK
f(K cos
Ksmu)
cosw du
One notes that the definite integrals in these equations coincide with the
53.
trajectory is
existence of
a
circle; in the Van der Pol plane (a, b) the condition for the
^j
= n &{K) = 0.
<j>{K)
sinw du
of Poincare' s theory.
A root Ki of
^'(ifj < 0.
<f>{K)
{K
fv (K, cos u ,
sin
u ) du <
[53
"I
<f>'
{K
{ )
> 0.
ip(K)
2ttKJ
2"
/(Kcosu
K sin u)
cosu du
[53 -2
In this case
9 = $
= constant.
the plane of variables (a, b) of Van der Pol, in this case, is shown in Figure
53.1
.
for
6 =
is stable if
170
b
/Ki'
Stable
Limit Cycle
^Ki
^y\
a
\
/
Unstable
Circle
f
'
fI
W/
1
Stable
Circle
Stable
Focal Point
Unstable
Limit Cycle
Figure 53-1
Figure 53-2
If
one must apply the equations of transformation [52.2], where the variables a
and
are respectively
a cos
t
cos $
t
and
sin
This gives
t
b sin
= K
cos 6 cos
+
t
K
+
sin 6 sin
=
t
cos
[53.3]
?/
o sin
b cos
= K
cos
sin
fi^
sin 6 cos
= K
sin(t
i
where
is arbitrary.
This arbitrariness of
due to the fact that a point of equilibrium in the plane of variables (a, 6)
of Van der Pol corresponds to a circle in the phase plane of the variables {x,y).
ure 53.2.
Case 2.
-
ip
(K
==
2"
\
2ttK
f (K cos u
,
K sin u
cos
udu
Let
be the roots of
ifs(K)
of
= 0.
171
(a
by equations
the sign of
*
II
1C,.cos[/i0(A\M
];
O]
[53.4]
Stable
Limit Cycle
b)
-plane Is
The trajec-
K lf K
The
Figure 53-3
cos
b sin
=K
cos
Ml y rp (K^)]
- fxip (K
[53.5]
y
= -
a sin
6 cos
= -K
sin([l
j\t
O )
uniform.
the cor2
//
rection for frequency (and hence for the period) is of the order of
however, ip{K {
and
If,
fi,
0,
To sum up the results of this section, it can be said that the use
of the variables (a, b) of Van der Pol enables us, if the system is isochro-
nous, to represent a limit cycle by a point in the plane (a, 6), that is, by
a constant vector.
K
172
tions the representative point moves toward, or away from, the limit cycle
point along the length of the radius vector, since the phase angle
constant.
remains
The phase angle in this case has no particular physical signifiIf, however,
0,
ables (a,
x , y ) -plane
54.
EXAMPLE:
We now propose to apply the Van der Pol method to the problem pre-
Consider again
6
Equation [30.9].
/iy lf
and
/id lf
where n
is a small parameter.
v
2 n{P + 2y1 v 3d v )v
x
[54.1]
where &i,y lt and : are positive. In this case f[v,v) = (fi + 2yv
<5
36v 2 )v.
52 9
.
] ,
we have
4>(K)= +
((0 +
2yKcosu v =
36
cos u)
[54.2]
cos u, v =
<f>(K)
2ir
2ir
fiK
sin
udu + 2yK
cos
sin
udu 36K
3f22 u du u
2*r
-i
cos
sin
[54.3]
Since
2*
2t
sin
J
udu
tt
sin
u cos udu
sin
and
cos wsin
J
udu =
sin
J
udu
sin
J
udu n n =
173
we have
,,
_!(,*_). !(,_!^)
Prom Equation [54.4] this takes place for
[*.,
= 0.
and
|/|f For
if j
point.
excitation, one has to prove that the singularity is unstable and the limit
cycle is stable, see Section 29-
linear terms, is
v = y
and y +
v
v = n/3y,
that is
v
y,
= -
+ vpy
whose roots are
[5^-6]
- /ufiS
+1=0
ufi
up yr2
Si,2
^r
Since n is small, it is seen that the roots are complex, with a positive real
part.
= \ tt,
,
provided it is stable.
In order to establish the stability of the limit cycle
it is necessary to ascertain the sign of
<j>
l/4
y r*r,
K (K2
3/S)
One finds
<f>
K {K2 ) = j(fi -
<
The limit cycle is thus stable. In this discussion it has been assumed that the coefficients
/?,
y,
and
-7-.
d6
174
ip(K)=
-= Up + 2yKcosu -S6K
cos
u)Ksinucosudu
[54. 7]
\p(K)
= \0K\
1
00
2ir
2k
2k
sinucosudu
+ 2yK
sin
cos
udu 36K
cos ttsintidtt
[54.8]
4r
= 0.
radius vector of the representative point rotates around the origin with a
constant angular velocity, and the frequency correction is zero to the first
order
In order to investigate the variation of
as a function of time,
<t>(K)
its
mK - nK
where
and n =
g.
3<5
This gives
dK
mK -nK
where P -
S A
1 = ~
dK
vti mV U m K(l-pK')
,-
Vdt
$
dK
K(l-pK')
dK +
pKdK
zpz 1-pK
m/xdt
That is, d
logK-
dhj; log(1
pK 2
)]
= m/udt, or
d (} oe
=
Yi-.pK*)
mfxdt
a Vii
-pK*
^2
log log
~ ,/. _ Vi- pK
= mut
'
or
K
Vl
Finally,
-mfj t e
K
Vl
- pK 2
-pK
[54.9]
K
-
V(i-
pK
\g-2mu
+ pK
[54.10]
175
For
= 0,
\ tt and for r 7a
-> - <,
K+
0,
creasing
t,
Furthermore, for
->
<*>,
which shows
and
in
Equation [54 .10] certain initial and final values K' and
'
slightly re-
moved from the unstable focal point and the stable limit cycle respectively.
In such a case Equation [5^ .10] can be used for numerical calculations with
a view to ascertaining how rapidly the self -excited oscillatory process
corre-
limit cycle in the first approximation, as was noted in connection with the
2yK 2 [ sin u
2
cos
udu
in Equation [5^.8].
The inverse passage from the phase plane (a, b) of Van der Pol
to the ordinary phase plane {x,y) yields the expressions x =
y = x = -
K cos
and
sin
t,
where
Topologically, it corresponds to
stable limit cycle an unstable limit cycle exists, one is then concerned with
the so-called "hard" self -excitation of oscillations.
51
effect on the determination of limit cycles, one can drop them from the equation.
v+v
In this case f{v,v) =
/(iTcosw,
/u(0
36v
2
brjv
i
[5^.11]
(/?
+ ~$6v
^rjv )v
and
4
if sin w)
{/3
+ 36K
cos
brjK* cos u)
K sin u
176
<t>{K)
J_
oo
In 2n C
2
-9
>
we get
2ir
2n 2k
I/? if sin
udu +
36K
2jt
cos
sin wdtt
5ijK cos
o
2*
it
sin
wdw
2a-
2n
j3\ sin
[2udu +
Z6K
cos
o
it
sin
udu
(*
2
it
5rjK
cos
sin
udu
2n f
2n
.2
.
n;
o
cos
sin
udu =
.o
f
I
sin
udu
sin
udu =
ir
3 7T n 4 4
-i.lt
Zir
cos
it
sin
udu
cos
wdw
cos
udu =
3-5
it
This gives
*.- f
tive.
(/
!-
f^JT
)fl,
[54.12]
y
It is assumed that
and
77
are posi-
K=
0.
and making use of Liapounoff's equations of the first approximation, one finds
that the singularity is an unstable focal point
The limit cycles proper are given by the quadratic equation
^ -fas-
where S =
[5M3]
2
.
is
essentially positive.
2p
4p 2
where p 2 =
5- 77
and
<?
= 4-0
(*
) - M
M
2p
z
[54. 14]
= p 2 {S
2
) ,
(Sj,^) with
new origin at (
= -)
fix
" si
2o
177
[/jP-
/^^ A
For
/S
so
^577
quadratic equation p 2 S 2
are
]/q
2
\S|
'
qS
- J3
=
0'
,2
-4\P\p''
2p
Figure 54.1
2
2
j
0.
They correspond
line
fi
=
I
2p<
In the region O'O, where two roots exist, there are two limit
cycles,
= Vs^ and
= VsT>.
spect to K.
*W-f + f* -f
and S 2 for
/?
K ~
+
2
f S --f S
ft
3?
5p'
'
[5^.151
/^p
2
)
,
to the left of
q jr-^
Con-
o~~2-
The cor-
^1,2
2p
^
(i^.|)
2
.
.~l0~(q/2p
/?
= -
pf, op
>
and
<j>'(K 2
< 0.
cycles on the lower branch S x of the parabola are unstable and on the upper
If
fi
the point O', there is no self -excitation throughout the range O'O since the
unstable limit cycles S lt interposed between the stable focal points situated
on the /?-axIs and the stable limit cycles S 2
,
178
As soon as Point
at
which the unstable limit cycle disappears is reached, the stable limit cycle
J/4-
is reached abruptly; if
/?
If,
55.
EXAMPLE:
mx {a
fix
+ Kx =
[55
"
velocity
x;
By introduc1
equation
juf(x,x)
[55-2]
Assume that the damping terms are small enough to justify the intro-
/u.
In Rayleigh's equation /( x, x
= f{x) =
/?
mx
nx, where
as the re-
fi{mx
nx)
=0
and
ip{K)
where
>
and n >
0.
4>{K)
are
2tt J
~r
Z7T J
nK
sin
udu =
K \m
2^4/
-
nK
[55.^]
and
ip(K)
nK
sin
ucos
udu
mK
sin
cos
udu
[55-5]
179
From Equation [55.4] we see that the conditions for limit cycles are
k=
The first value
3
*-y
-
55 .6]
K=
If the non-
ny
+ x
= 0.
nS
= 0, whose roots
Vn 2
2
-4
\2
point.
For w < 2, one has an unstable focal point, and for n > 2, an unstable nodal
In both cases the point singularity
K=
is unstable.
K=
|/^
3n'
one
*+]
with respect to
K
[55J]
4>'(K)
=f ~\nK
9
Substituting
if
-^
in
,,/i/4ra
4m
the same sign, that is, of the form ( rax + nx 3 ), there would be no limit
cycle in this case, as is easily ascertained, although the oscillation would still be governed by a non-linear differential equation of a dissipative
type.
tion.
It can be shown that the oscillation of Froude's pendulum,
see
is the fact that the equation was overlinearized by dropping the non-linear
terms
180
M(u -
<j>)
M(u>)
4>M\oj)
+ ^7 M"(u) - trl"'(w) + 2! 3!
[55-8]
Mis
/,
keeping only the first two non-linear terms in the expansion, one obtains an
+
where
c
c<j)
k(j>
n<f>'
[55.9]
+ M'[u)
Putting
= - m, where
m
<j>
(f>
ra0
kcj>
ruf)
= n{m
<j>
<j>
n 6
whence
/(0)
Y^4>
+ k^ -
'
n
x
<i>'
and
f{Ksmu)
Hence
Ksinu
sin
K sin
.
.3
2t
2>r
2t
<f>(K)
m K\
2tt
sin
udu +
sin
udu n^K
sin
udu
that is,
<t>{K)=
-\
_
K=
K = \j^-
ne can calculate
and n 1
181
ratic term disappears from Expression [55 -TO] for the limit cycle because
J
56.
."
It
must be noted, however, that the existence of self -excited oscillations, that
is, of limit cycles, depends only on the conditions of the variable damping,
as was shown by Van der Pol (15) and generalized later by E. and H. Cartan
(l6) and Lienard (17)-
Van der Pol equation or for small x in the Rayleigh equation, is first negative then positive for larger values of the corresponding variable (x or x).
Physically this means that there exists, initially, an energy input into the
system which, in later stages of motion, becomes an energy drain from the
system, which then becomes dissipative.
tion, that is, of a limit cycle, depends thus on the average equality between
(nx
ce)x
/?x
+ yx 2 =
[56.1
which has variable damping, the term nx 2 x, as well as a variable spring constant, the term
(/?
+ yx)x.
rax
yS
(x
(a
n x )x
y x
= n4>(x,i)
[56.2]
It is
sufficient to divide it by
ra,
putting fi/m=
a>
1 n (a
- n x 2 )x Q
\
yQ x
Hf(x,x)
[56.3]
In The last operation, although convenient for practical calculations, is not altogether necessary. case one does not use it, the generating solutions should be taken in the form AT cos ut, - Ku sin uit - K sin instead of K cos t
, t
.
182
K cosu
and
equation [52-9] giving the limit cycles. The same procedure applies to the generalized Rayleigh equation with the variable spring constant.
In all cases the existence of limit cycles requires that the coef-
If this condition is
not fulfilled, no self -excited oscillations can exist in a steady state, and
the system behaves as purely dlssipative, while still non-linear.
CHAPTER X
THEORY OF THE FIRST APPROXIMATION OF KRYLOFF AND BOGOLIUBOFF*
57
INTRODUCTORY REMARKS
The method of Kryloff and Bogoliuboff is very similar to that of
sin w< of x +
w2 x =
0,
<j>)
Thus
(polar
in Kryloff and Bogoliuboff 's method, the "varied" constants are a and
(rectangular
coordinates).
from the point of view of applications, since it deals directly with the
may be helpful
EFFECT OF SECULAR TERMS IN SOLUTIONS BY EXPANSIONS IN SERIES
The difficulty arising from the appearance of secular terms has In the example given in Section
58.
46 that difficulty was avoided by a rather delicate change from the "old"
the correction for the period, which is not always obtainable as has been dem-
onstrated.
mation consists in abbreviating an infinite series by a few terms, the situation is still more difficult.
In order to see this point, consider again a quasi-linear differen-
+ u
+
fi
uf(x,x)
[58.1
1.
//,
x
Q
/ix
1
M~x 2
[58.2]
The subject matter of this and of the following two chapters is taken from the treatise of Kryloff and Bogoliuboff, Reference (5). This subject is also treated in the free translation by S. Lefschetz of the Kryloff -Bogoliuboff text, Princeton University Press, 19-43-
184
For
fi
= 0, x = x
= a cos
tot
For
<
/j.
small perturbation and to assume that this perturbation will also be felt in
the modification of the initial generating solution represented as a power
series of
is,
/u,
tion [58.1], limiting the expansion up to the power //^of the small parameter,
one obtains the following series of differential equations by equating terms
/u
u\
x
l
=
--
x\
+ u
/;(av
[58.3]
x
2
+ <A 2 =
This procedure has already been outlined in connection with Equations [46.5]
of the theory of Poincare.
It is easy to show, however,
+ M
Utox
Ae
where
cos
(<<T-T t+
)
[58.4]
and
<f>
are the constants of integration determined by the initial conone proceeds by substituting the expansion [58.2] into
ditions.
If, however,
]
Equations [58.3
Acos{tot
<f>)
one has
[58.5]
c/x
to
Aoj
sin (tot
<j>)
cos(tot
<t>)
[58.6]
185
and x
A(l
-^)cos(c^ +
<f>)
with time
+ <A(1 +
fix
[58.8]
which may be considered as the equation of motion of a mechanical mass attracted to the position of equilibrium by a force proportional to the distance, with a perturbation term proportional to the cube of the distance.
x =
fix,
u>
=
[58.9]
+
,
0)
= D 2X 3
>
.
From the first equation [58.9] x = A sin(o> + 0) this value for x into the second equation, one has
x
Substituting
11
+
<Jx
= -u A
sm(tjt
<f>)
= - T cj A 4
2
sm(o)t
0)
+ w A 3 sin3M +
2
<t>)
=-wMcos(w< +
4>)
j^
sin3((Ji
0)
whence
Asmicjt
</>)+
-~cjtA
cosM +
<t>)
-~sm^(utt +
<f>)
[58. TO]
Thus, this
Unfortunately, in
this case the exact solution is not known, as Equation [58.8] is not linear.
One can, however, affirm the correctness of the above statement by invoking
the law of conservation of energy, which holds in this case since the system
is conservative.
4- (1
fix
)u x
JI\J X
J 0JX
22^ +
flu
*'~
186
1.2. x +
1
0)
2,
H
jUU)
=
>
constant
l5o.ll
. J
fi
59-
which will
For
fi
= 0, Equation
whose solution is
x
sin
{u t
0)
a to cos {cj t
<f>)
59
where a and
For
]>
dsiniojt
0)
+ awcos(w +
0)
a<t>cos(ojt
4>)
[59-2]
]>
one has
[59-3]
dsinM +
0)
a<t>cos(u)t
]>
0)
one gets
aucos{u)t
0)
x,
0)
aw0 sinicot +
</>)
[59-^]
one has
0) a(j<j>sin{u)t
0)
+ nf
sin
{ut
0),
au
cos {o)t
0)
=0
[59-5]
one gets
0)
<f>),aa>cos{ojt
0)
cosM +
[59-6]
187
au
asm(cjt
0),
au cos (cot +
0) sin(a>
-I-
0)
[59-7]
fact that these first-order equations are now written in terms of the amplitude and phase as dependent variables.
tions [52.8] of Van der Pol and also with Equations [46.22] and [46.23] of
Poincare
Prom the form of the right side of Equations [59.6] and [59J], it is seen that both d and are periodic functions of time. Prom the fact that
the right-hand terms of these equations contain a small parameter n, one can
conclude that both a and 0, whjle being periodic, are functions which vary slowly during one period T = Zn/u of the trigonometric functions involved.
It is reasonable,
as constant during
one period T.
/(asin0, aco
cos
0)cos0
= K
(a)
+ /, Kn (a)
n
cos
n<f>
+ L n (a) sin
n<f>
[59.8]
<f>)
sin
= P
(a)
+ /, P (a) cos n
n=\
L
ncj)
+ Q n (a)sinn<f>
where
Kq(cl)
1 = - ff {asm
<}>
(f>dcf>
A( a =
)
o 27r
/(asin0,a<jcos0) sin0d0
(a)
1 = f (a sin \f
77"
a to cos 0) cos
cos
[59.9:
2*
L n (a) = /(a
77
sin
n0d0
PJa)
sin
cos
w0c?0
Q(a)
=
7T x
sin
n0d0
188
~K (a)-
U)
Q
CO
71=1
oo
7!
oo
Kia)cosn(u)t
0)
0)
[59- TO]
d0
dt
au)
P (a)
+ P aw
r-
{a)cosn(ut
0)
+
t
Q(a)sin n(ut
0)
=1
and
+ T, and considering
0(0
+T)-
a(t)
-^K
U)
<f>lt
t
a(t)
+ T) - 0(0
~PoU(t)} [59.n]
since
t
+ T
+ T
I
cosniut
0) dt
sinn(cjt
4-
<f>)dt
A<p
of
j7.=
fi
>
rf0
if (a);
=-^-P <a) aw
n
[59-12]
[59
1
it is seen
that the equations of the first approximation are obtained from the exact
equations by averaging the latter equations over the period, thus eliminating the rest of the Fourier series under the summation sign.
The analogy be-
tween Equations [59-12] and the "abbreviated" equations [52.6] of Van der Pol
should be noted.
Letting
oj
xp
cot
+ 0,
dcf>/dt
(a)
and
{a]
in [59 -9
da
~dt
E
u
2*
_L /(a sin0,
2tt
1
ao>cos0) cos0
c?0
= 0(a)
[59-13]
2*
dip
to
dt
au 2n
rf0
= Q{a)
[59-1^3
ip,
ip
189
60.
u>~x
+ nf{x) =
x.
da
~dt
2ttd
d<f>
dt
u +
M
'.naoj
f(a
[60.2]
Noting that
In
If (a sin0) cos0
o
d<f>
?Masin0) =
a
'o
where ^(x) =
/(^)rf^, Equation
[60
gives a = constant.
first approximation, it follows that the amplitude does not change in the
ori considerations, because the function / does not depend on the velocity x,
x.
Q(a)t
+ ^
ip
[60.3]
is a constant of integration.
in this case;
a sin Q(a)t
if;.
[60.4]
that is, the oscillations are not isochronous, but the dec-
(a)
= u +
na
\
2*
2n
<f>
f(a
sin0) sin0
d<f)
na
(j
sin
d<f>
y.
/ (a
sin
<f>)
sin
<j>
d<f>
2*
na
to"a sin0
+ uf(a
sin0)
sin0e/0
F(a
sin0) sin0
d<f>
[60.5
190
where
F(asin0)
u>
asin0
4-
juf(asm<j>)
[60.6]
On the other hand, the general form of a non-linear equation, in which the
0.
x + fif(x),
Comparing this
expression with Equation [60.6], one finds exactly the same result provided
one substitutes the generating solution a sin
the following theorem:
<f>
for x.
When
a system is conservative but not linear, the amplitude a remains constant and
term entering
to 2
into a linear
component
x and a non-linear
one, uf(x).
6l
PENDULUM
The differential equation for a pendulum is
+ 4- sin
0=0.
In
0.
is obtained.
gj
{0 - -g-) = 0,
~ where u 2
- -z~;
the differ-
= g/L.
Using Equa-
(a)
na
2 2
2*
3.3,,
a sin0
-^-J -\
sin0 d<p
'
2*
3 2n
<f>
0)
na
that is,
sin
d(f>
sin
d(j>
0)
/l
[1
,2>
a (a) =
or,
j/i
-^- -
w(i -
f^j
[61.1]
T(a)
= r(l +
^)
[61.2]
191
.001
x 1.006;
Furthermore, the latter method easily leads to correct results in more complicated cases for which exact methods are not available.
B.
.1
If
and
x
B2
J,
J2
lis
62
)
,
a certain function of
6l
say C{B l
= M{6) *
system are
el
+ c(e
,)
and
J2
- C(0 l
62 )
Figure
6l
.1
92
we obtain
J J2 Q +
X
+ J 2 )C(0) =
[61.3]
B +
C(d) = 0, where
6
K
3
,
+ J 2 )/J1 J2
3
where
C^
From this,
e
F{e)
= K c
K^^e
= me ne
,a)
=
7TCL
(masin^
no,
sin
0)sin0
dd>
a> n
v
K 4
C.a'
* C(e x
C is a function of
{$ x
2 ),
192
or
Q(a)
* w
v
to-
The non-linear frequency (and therefore the period) will thus depend on the
amplitude.
not isochronous.
C.
negligible resistance.
Figure
6l .2,
dt
dt
where
Figure
6l .2
by the equation
A0 +
C
B<P
[61.4]
Reducing it to
cj 2
A/C
= constant, and
1,
Q- (a)
naC
_,
(Aa
sinoi
+ Ba
sin
0) sin0
dcf>
-^[l
4A
'
that is,
=
L ,,,,,,
, ; ,./i
,.
M^! w d +
AA
^'^ A
[61.5]
The actual frequency Q{a) is thus increased in comparison with the frequency
o)
L with
oscillation.
62.
mx + Kx +
fix)
195
Dividing it by
and putting
K/m
x
4-
w2
x
we get
(j
fix) m
=
In this case
The expressions
and
entering into Equations [59 -13] and [59-1 4] of the first approximation, in
this case are
2* 2t
and
o
/(aw
cos0) sin0 d0
/(awcos0) sin0
J
d0 =
aw
2t
/(awcos0) d(awcos0)
J
0(awcos0) aw
d^
dt
= w
[62.1
[62.2]
),
where
ip
is
of the second order and therefore does not appear in equations of the first
approximation.
A.
LINEAR DAMPING:
In this case
2ir
2>r
/(awcos0) cos0 d0
= awAJcos 0d0 =
2
aukir
19^
- 1 2nmcj
Xa auXn = 2m
,
whence
At
2
[62.3]
2m sin(w^
i//
Q)
where
Wfc^l YKm
v
one notes that the first approximation gives the same expression for the am-
plitude as the exact solution; the difference between the expressions for
frequency
cj
2
,
B.
QUADRATIC DAMPING:
2 f(x) = bx
Since /(x) is an even function of x, and from physical considerations it should be an odd function of x, we should write the above expressior|
as f(x) = b\x\x.
2ir
In this case
2r
/ (awcos0)cos0rf0 = ba
,
o C
o)
Mcos0|
19
cos
<f>
2
d</>
2tt
*3
\
ba u
"\
cos
<f>
d<f>
4-
cos
3w
2
<f>
d<f>
'2
cos
4> d<j>
l ba
to
+
.8
O ba
,
cj
3J
= -
dt
znmo)
/(awcos0)cos0 a0
i
ibcja'
Zn m
that is,
d
da
(!)
dt
4bu>
a 2 dt
Snm
195
=
,
[62.4]
,
ar
*!^h
37rm
C.
COULOMB DAMPING:
f[x) = Asgn(x)*
f (a cj cos
<f>)
cos
d<f>
= A
cos
<f>
d<f>
+
3rr
cos0
2
d<f>
cos<j>d<f>
= \A
for a * 0.
[62.2]
Furthermore,
'o
/(acj cos 0)
cos,
<t>
d<j>
for a = 0.
By Equation
da
dt
2A
nmu)
if
a #
[62.5]
if
dt
2^
a,
t
Trmoj
[62.6]
2A
de-
2A
nmoj
t 1
.
D.
MIXED CASES:
2 f{x) = ax + 0x
+ Who = 0, where /,
K1) K
.
form x + u
The symbol sgn(i) designates a discontinuous function defined as follows: for x = 0. sgn(x) = - 1 for x < 0; and sgn(i) =
sgn(i) =
for x > 0;
** The writer is indebted to Dr. W.F. Durand for bringing this equation to his attention.
196
2ir
2tt
2x
3
fiau
cos0) cos0
d<f>
= aau
cos
d0 +
/?a
cos
d0
4-
cos
V d0
3?r 2
cos
d0
From this
a
=
27TW
aacon +
pa w
-r-
o + Sa
where S =
^fiaj/jn.
da
d(aS +
we get
+ ba
aS +
f) a
a d log
aS +
- jdt
Of
we obtain
a
a
aS +
or
ss + f
a
2
ao
a
-2
e
S +
=
1
f
[62.7]
- S
a
S +
y
,
For
= 0, one finds a = a
as in the case
causes a somewhat more rapid decay of amplitudes owing to the presence of the
term S
ing.
c en o -r
-9-
i n the
197
63.
SYSTEMS WITH NON-LINEAR VARIABLE DAMPING By the expression "variable" damping occurring in this section, we
shall understand a damping which, for small values of the determining variable
(either x or x as the case may be), is negative and becomes positive above a
to the system, and positive damping means withdrawal of energy from it, the
or limit cycles, are possible when the average supply of energy per cycle be-
tions governed by the following equations, which are solved here by the meth-
A.
iu(l
x )x
cj
=
ato
[63.1
We have f[x,x) =
fx(x x
.
x); furthermore,
cos0) is
/(a sin0, a
cos<)
t*(a
sin
0cos0
acos<^)
and
/(a sin0, acos^) cos^
fi(a' sin
cos
(j>
a cos
<f>)
>
^ =
dt
2jt
r-
2n
2jr
t*_
/(a
d<f>
=
ii'TT
a
L
sin
'
cos
d<t>
a
>
cos"0
d(f>
ATT J
which reduces to
da
dt
r2n
jua
'
-if(i-t) 4 2
V
6 '- 2
'
/(a sin0,
a cos0)
di///dt
sin
<t>
d<f>,
w=
that is,
ip
where
is arbitrary.
The solution of Van der Pol's equation to the first order is then
of the form
x
a sinU
ip
[63-3]
198
Prom the
preceding study of Van der Pol's equation we know that a limit cycle exists
in this case.
[63.2].
^-t)
d(log
4
that is,
da
2-
2)
fidt
^j4^-^T
ing for a
2
,
h
)/l
+ "t
2 a
77
6 '-*)
-jt
77
~T2 e"
"o
and solv-
Ml
a *_
1
ut
.
2
tt
b^
+ i
2
(
" -
1)
i.>"
[63 5
.
1)
=-
sin (t
4>n)
a sinU
i//
[63 .6]
|/1
+ ja U
1)
In fact, if
= 0,
= 0,
0.
->
00,
inside
199
B.
+
-
fi(-
2 a + 3x )x =
[63.7]
In this case
u>
"\
and f{x,x) =
ax +
fix
3
.
f(a
cos0) cos0
d<t>
= aa
00
2n
~
<f>
cos
<f>
d<f>
/3a
cos
d</>
= aan + 7r/?a
o>
=
2
one has
3 d=^(-{/?a 4
[63.8]
)
tem becomes linear and the amplitude, at least theoretically, increases indefinitely, the damping then being negative.
To find the mode of approach to the limit cycle, one must integrate
Equation [63.8].
tains**
jia_
t
ait)
=.-== l/l + 3 8 K I
\
=
2,
a (g
uat
~7
1}
[63.9]
"
This gives a(
t
t
_koo
)x
0,
ux
+ (V
x )x ml
2
=
1
.
where u z K/m, after which a change of the independent variable brings it to the above form, with u =
** In case Equation [63-7] is not reduced to a unit frequency and has the form
x
2
u)
+ (- a + 0x
)x
=
<t>),
the generating solution should be taken in the form aw cos in the following equation for o t )
<f>
(instead of a cos
ft
ait)
=
f/l
l| w * V--l)
[63.9a]
200
64.
groundwork for the investigation of an important subject, namely, the existence and stability of limit cycles.
It is useful for this purpose to recapitulate briefly the principal
sin
ip,
where a and
ip
2ttu)
f(a
sinci, acj
cos0) cos0
dcp
= 0(a)
[64.1
2*
U)
+
2iT(jja
5
f(a
sin0,
au
cos0) sin0
d(j>
= Q(a)
[64 .2
[64.3]
theory and the first equation [52.9] of Van der Pol's theory.
Q
and the identity
{a)
to
'=
2>r
/(a
sin0,
awcos0) sin0d0
0)
v
2* r 2 10
a sin
dd>
Thus
{a)
7TCL
>
d<j>
[64.4]
where F(x,x) = u 2 x + /uf(x,x) is the non-linear force appearing in the equation Likewise, Equation [64.1] can be transformed by means of the x + F(x,x) = 0.
(.2*
identity
aco
sin0cos0d0 =
0,
which gives
201
2<r
0(a)
d<f>
[64.5]
= =
0(a)
[64.6]
Q{a)
[64J]
where 0(a) and Q(a) are given by Equations [64.5] and [64.4] in terms of the
total non-linear force F{x,x).
K cos
when n =
0,
exists.
tude a 1
The question of the existence of limit cycles which are not generated by a harmonic x =
We now propose to investigate a practical case in which a limit cycle exists, as shown by experiment, and to show how this existence can be
Putting
Lie
G(x)
+ hjxdt =
C
[64.8]
where the constant parameters L and C designate the inductance and capacity
of the circuit respectively.
t,
dividing it by L,
we obtain
[64.9]
+ u 2 x + yG'{x)x =
Identifying the term jG\x)x with nf{x,x) of the general theory, which inci-
dentally imposes
202
2n
=-
\G'(a
sirup)
a co cos &
cos<pd<p
= 0(a)
[64. TO]
rp
co
4-
2n
I
G'(asin0) cos0
sin0 d<p
= Q(a)
[64.11
Integrating by parts, we see that the second term on the right side of Equation [64.11] is zero, hence,
ip
Q(a)
co
[64.12]
and
ip
cot
ip
where
\p
is arbitrary.
If we let
-J G'{a sin
<f>)
cos
d<f>
= R(a)
written as
a
0(a)
[64.13]
if G'(x) > 0, that is,
0,
if the voltage across the non-linear conductor increases with the current.
Thus, d < 0, and the final state of the system is x = 0, as is obvious from
: )
for
a certain amplitude a
Figures 64.1a, b,
in this case,
dissipatlve.
Figure 64.1b corresponds to the case when G'[x), and hence R[a)
are
negative for small amplitudes and become positive for larger ones.
ax
The root
x
.
If the
203
tR(o)
(b)
Figure 64.
arrows in Figure 64. lb.
limit cycle.
The amplitude a = aj thus corresponds to a stable
and a = a 3
applications.
critical value a = a
conveyed into the system, and this state exists until an amplitude a = a 2 is
reached.
In such a case if the initial amplitude a
0;
< ax
amplitude is a =
a
> a
acts as a kind
65.
STABILITY OF LIMIT CYCLES; CRITICAL VALUES OF A PARAMETER; SYSTEMS WITH SEVERAL LIMIT CYCLES
Let us first consider the question of the stability of limit cycles.
da)
= ^(aj)
6a
204
^(aj) 6a
[65.1]
If the initial dea
(a 1
< 0,
If
a (a 1
> 0, by a similar
occurrence is
0(0) >
[65.2]
which is equivalent to the existence of an unstable singularity in the presence of a stable limit cycle, see Part I, Chapter IV.
tion for a critical value of some parameter can easily be established by this
method.
In fact, assume that 0(a), in addition to a, also depends on a param-
and varying X.
When a value X = X
(O,X) > 0,
the amplitudes begin to grow from zero and the subsequent increase of ampli-
tudes from that moment will be determined by Equation [64.13], which now has
the form
d
0(a,X)
[65.3]
The limit cycle is reached for a value a l of the amplitude for which
0(a lt X) =
[65.4]
For a given value K = k lf the limit cycle will be determined from the equation
0(a 1 ,X 1
0(a,X)
0(a)
#i(o,X)
[65.5]
Differentiating
one has
0>,X) =
*.()
0j(a,X)
<.)
d[0 (a,X)] da
1
[65.6]
205
0(O)
0,(0, X)
[65.7]
<M0) >
since
[65,8]
0^0, A) >
0.
x
Consider a curve y = 0(a) and a straight line y 2 = ya shown in Figure 65.1. a (O) is the slope of the tangent to the curve 4>(a) at the origin, and
self -excitation occurs starting from rest (a = 0) only if the initial slope of the tangent to the curve 0(a) is greater than
the slope of the line a/A.
In
Figure 65.1
Figure 65.1,
1
l/A =
'*
tana
If
a
For
^ <
(O)
(O,A) > 0.
The equilibrium
condition is
= a x /k, which corresponds to the intersection of curves It is interesting to note that the amplitude, Equa-
= 0(a) and y 2 =
tion [64.10], holds in this case in spite of the fact that the frequency,
THEOREM
a 3> a 5
" ' '
1.
If
a6 ,
THEOREM
2.
is
sequence.
It should be borne in mind that we are considering only limit cycles that are generated by circles.
206
THEOREM
3.
The
of the
it
beyond
66.
a polynomial
f(i)
= A + Bi
i
4-
Ci
+ Di
+ Ei
+ Fi
[66.1
= a sin0, we have
9
</>
(a sin 0) cos
= 5 cos
2 Casing cos
3
2
+ 3D a
4 4
sin
2
cos
+ 4Ea 3
Forming the expression for R(a) =
sin
cos
+ 5Fa
)
sin
2
cos
/'(a sin
cos
d0, we have
i?(a)
=
7T X
i?
cos
d0
4-
73"
2Ca sin0
X
cos
d<j>
-\
3Da 2 sin 2
7T X
cos
d<j>
7T X
4"a sin
o
cos
d0
4-
7T X
5Fa
sin
cos
d0
[66.2]
"6
The second and the fourth terms on the right side of this equation are zero.
The remaining terms are
B
.
In
r
cos
d<f>
=B
[66.3]
2*
r-
2*
77X
3.Da
sin
cos
<j>
d<b
=
77
sin
L X
d0
X
SDa
rf0
3\
sin
5Fa
5Fa
sin
T 4r& f s.n
J
-
cos
rf0
7T
^^
0rf0
?* s.n
J
(j
/3
3-5\
)
5 p = - Fa
From this
R(a)
= B + |z)a + |-Fa 4
2
[66.4]
207
Setting
limit cycle is
AF = m
>
and
-frD
= n,
R(a)
= ma + na + B =
2
[66.5]
\ a 1,2
-nVn
- \mB
2
2
[66.6]
2ra
or a 2 must be
B<
0,
which expresses the fact that the slope of the characteristic f(i) is negative, that is,
"negative" resistance.
B >
0,
> 0.
are negative; hence, the amplitudes a x and a 2 of the limit cycles are imaginary.
B >
0,
2
< 0.
kBm
2.
3.
B B 5 B
and
pative.
4.
If
a!
> 0,
D
and
< 0, and
<
&r j^
amplitude
5.
(positive root).
>
a n2 > jfx-^r> the system behaves again as a dissipative one, and no limit cycles exist.
If
< 0, but
208
ties, and the above procedure permits ascertaining from the form of the char-
acteristic the behavior of the system Into which the non-linear element with
this particular characteristic is Introduced.
CHAPTER
XI
INTRODUCTORY REMARKS
In this chapter we review the extension of the method of Kryloff
derived from the classical methods of Gylden and Lindstedt used in celestial
mechanics.
As explained in Section 44, the object of these methods is to
eliminate the secular terms resulting from the resonance effect of subsequent
harmonics in the recurrence procedure by which the higher-order terms are determined.
The method can be summarized as follows.
+
T.
oj
nf{x)
[67.1
z(r)
J>
n
*>)
[67.2]
=0
where
z n {r)
with n =
1,
2,
We
Q =
where a n is constant.
Z ^
n n -
[67.3]
3"4 + "* +
Mf(z)
[67.4]
If one substitutes into this equation the series expansions [67.2] and [67.3],
11.
210
X,-^-#dr<
+ u~z n =
a ^~ 2 +
,
2
z.,
oj
,,, = - f (z
2l
- a2
d -^
z
d z, - o.!-^
6 7-5]
zn +
dr 2
where F{
z
,
w
z
x
,
" +
w>*o*ft
d~z
d
*
z,
J
z,
*H/
"* +
dr 2
dr 2
,
dr 2
It is
zQ
,
zn
z
1
,
is a certain polynomial in
zN
x
,
zn
apparent that if
and a
a lf
[N +
1)
ZM
n
z n (r)
and
Q =
1
^"a
n =
[67.6]
fi
(N
+ 1)
approximation.
The method of Lindstedt, which Kryloff and Bogoliuboff follow, con-
Ztt.
In fact, if these terms were left on the right side of Equations [67.5]* they
would account for the "resonance terms," which are of secular form, as pre-
viously defined.
By this meth-
od the difficulties encountered in the theory of Poincare (see Chapter VIII) in connection with the appearance of secular terms in the expansions are elim-
second or third approximation (and usually, in fact, the first) gives entirely
satisfactory results.
The subject matter of this chapter is considerably abbreviated, com-
'
68.
Equations [59-13] and [59-1^] of the first approximation were obtained by dropping the higher harmonics in the Fourier series on the right
211
undergo oscillations of
relatively
high frequency.
In order to take this into account, it is convenienent to consider
the quantities a and
parison with the rapidly varying trigonometric terms cos n(tot + 0) and
sin n(tot +
<f>)
.
{a)
~ and
n
d<j>/dt,
(a)
-~f,
V^ << K ?
fi
(a) s'mn(tot
0)
L
nto
0)
to
n =
.68.1]
^ P y
CO
(a) sinnitot
0)
nto
Q(a) cosndot
0)
ato
',
where a designates the first approximation for the amplitude given by Equation [59.13]-
fj_
n
yK
=
1
s\nn(tot
0)
L n cosn(tot +
0)
sin tot
CO
Pn sinnicot +
0)
Q cosn(a>
0)
a<j
n
[68.2]
If we let
1
2
^
n
y^ Pn
1
sin
n( to t
<fi)
Q n cosn(tot +
n
0)
/uS)
=
~
sinitot
+ +
0)cos//5
0)
cos(tot
+
0)
0) sinjuS
sin(<W
4-
since
//
is small.
y
2
OO
K n sinn(tot +
0)
L n cosn(tot +
0)
sin(a>
0)
0)
n =
[68.3]
212
2
fi
,
3
/u
,
0) r
one has
t asin(<j a 0) + u\
4.
S7 - -^ /> Kn
Vn
smniojt +
- L n cosn{cot + s
0)
,,
sin(a>
'
\..
0)
a//
V^-P o/ sinn(a> + 0)
n
Q cosnicot "
0)
cos(a>
0)
t
I
00.4]
,..,
If we put ut +
= r,
> *i
n
-K\,sinwT
L
n
cos tit
u(t)
y
....
[68.5]
Pn
sin
nr -- Q n cos nr cosnr
?<
and
u(t) sinr
v(t)cost = w(r)
[68.6]
a sinr
w(r)
[68.7]
= f{a sinr,
acj cost)
cost
sinr
^(a)
[68.8]
v'(t)
(a)
v'(t)cost
?<(t)cost
v(r)sinr
w'ir)
Loo .9
finds
w'(t)
= P
K
(a)cosr
(a)cosr
(a)
sinr
+ u cost
4-
.4-
v sinr
[68.10]
v'sinr
w"(r)
= P
(a)sinr
wsinr + vcosr
m'cost
[68.11]
Thus
w"(t)
+ w(t) = P
(a)sinr
K
-P
(a)cosr
wsinr
(a)cosr
4-
vcosr
u'cost
4
4
m'cost
v'sinr
v'sinr
u sinr
ycosr =
(a)sinr
[68.12]
213
Noting that
m'(t)cost + v'{r)smr
f(a
sinr, acocosr)
(a)
cost
(a)sinr
+ w(t) = f(a
sinr,
awcosr)
2K
(a)
cost
2P
(a)sinr [68.13]
/(asinr, acocosr)
= /
(a)
=
1
CO
fn {a) cos nr +
g n {a)sinnr
[68.14]
(a)
71
2n
/(a
dr
/n (a) =
1 Jf/(asinr,
,
nr dr
[68.15]
g n (a)
J
o
f(a
nr dr
and
{a) =
jf
{a) =
^g v
From
/(a
sinr,
awcosr) 2if
()
(a)
cost
2P
(a)sinr
fo
(f
n =
2
cosnr
gn
smnr)
[68.16]
w "(r) + w(t) = f +
\f
n =
2
cosnT
4-
gn
smnr)
[68.17]
w (t) =
/.
n = 2
{a cos
nr +
smnr)
214
(T w.. )
/o
-^A^ ^
s
n = 2
2 n
+_';
sinnT
t
68 18
-
Substituting this expression into Equation [68.7], one finally obtains the following expression for the improved first approximation
asinlwf
+
,
0)
,v
n
5-
,,
/u
(a) H
v^ fn
cos nicot
^
</>)
4-
#, sin?i(a>
re
+0)
r
[
68
. 1 9
.
1
* = 2
where a and
<t>
^j/i(a) and
4>
^9iia)
0,
co
L =
f
1
n cos
n (cot
2
<f>)
g n sinn(tot
&)
+ f
O^fi
2
)
[68.20]
where
is of the order of
ju
fif{x,x)
/uf\asm(ojt
0),
aocosiut
<f>)\
(m
whence, by [68.14],
x
to
fif(x,x)
O^in
2
)
2 2
(m
Thus the expression [68 .19] satisfies the original differential equation with
accuracy of the order of n 2 Furthermore, it can be shown, upon developing the expressions for
.
O^n
2
)
and
{m
),
<
69.
f(asinr)
n =
00
g n {a)sinnr
'
215
and
2
f
tf(a)
/(asinr) sinnT
rfr
as\n(ojt
fi
4>)
5-
ST /
n (a) sin
5
n(ut + &)
ii)*-*L n =2
Lo9.1
0(a)
(j
tf.(a)
/.i/ir
-I
that is,
Q
to the first order.
(a)
a>
+
a
a, (a) :
A.
to
where w 2 = w 2 +
fix
2
,
which gives
x
G)x
fix
[69.2]
so that
ft
j \x)
3
;
rt j(asmT) =
I3. -a sin3r
4
whence
,
fifj
la)
f/3 a 3
I
sin
1 a
sin
3r
\
I
sin
r r
3 = a
and
(a)
co
()
+ y^~1 0iM = w o +
^o^*
216
gf
(a)
= J
7r
f a sinr
sin3rj sin3r
dr = -
2
j
sin
3r
rf
a 6?rJ
3tt
3
.
sin
x a;
37ra
12tt
37T J
"I
a sm(a)
0)
-:
g"
(-2
r) sin3(o>
0)
asin(w
0)
2"
sin
3(wo^
0)
[69-3]
B.
VARIABLE DAMPING
Consider a differential equation of the form
x
to
fif(x)x
/() d.
F(acos0)
= ^F(a)cosw0
n =
[69.4]
a/(acos0) sin0
Putting
re
y =
nF(a) sinn0
77-
/(a
sin
t)
aw cost =
a>
nF
n {a)
smn [tot +
37J-
where
tion is
asin(
JT
^(g)sinw(o<
O)
[69.5]
217
rr/2,
00
a cos (Qt
Y w f?2 n
fi(l -
FAa)smn{Qt +
[69.6]
x,
x )x
=
u>
1,
F(x) = 4-
and
2
whence
.
^ ev F(acos0)
It follows that
a cos
33,
acos0 =
\
/a a(-j-
,\ l)cos0 +
3
cos
30
being zero.
^(a)
= a\jr-
/a 2
~\J
and
a (a) = y~-,
tne other
Fn (a)
acosit
- ^-sin3( +
[69.7]
fia
~2\
cosU
4-
O)
sin3U
[69.8]
C.
fi
= 0,
and
Si
0.
For quasi-
Q.
= acos(Qt +
+ 2* = 2
n
\A n cosn(Qt
L
+ Bn
sinn(Qt
[69.9]
J
fixx +
wV
fif(x)xx]dt
[69.IO]
218
since
x
co
+ nf{x)x =
xx
and
2
a;
.v
-r (xx)
xx dt +
x;
Hence
xx dt
dt
Likewise
j f(x)xx
for the same reason. Hence,
dt
x dt
co
x dt
[69-11]
Q' a 2
n U:
2
+ n
2
)
co
n =
JW,
TO
+ Bn
O
2
)
[69-12]
A p Bq
cos p( +
sin ?(> +
disappear when
It follows that
2
a
CO'
2
+ +
TO
Z'Ul =
2
2
+ Br
2
Zn U
TO
[69.13]
n
+ B n2
Equation [69.6],
has
An
and
Bn
-^ F A
n
(a)
and
Bn
219
K u
If we let
-o
+ _l 2
2
a>
n =
y ^
2 2
-l
n"
Fja) n
[69.14]
+
=
V 4^
n =
fi
'l
1
j^t>.W
or
= if 2 and
L
i^il
a
J
a;
1
k'ZA^t*-)
n =
2
^x^g ^^V^-)'
2
[69.15]
n = 2
'"
'
J L
n =
Q =
(j
_ jl_ 2^ 2 n
^U
=
2
rv
n
2
2 \
FKa) K\
a
:
-li
[69-16]
(a) =
jp = t and
as
Fn [a)
for n
3,
one has
ill 16
70.
+ u
nfix)
=
Let x = x(t) be
= 2-n/T.
Let
2ir.
z[Qt +
<f>)
= z[t)
with t = Qt +
4>.
x +
nf{x) =
0,
one obtains
Q 2 z + u2 z +
From now on the differential notation
respect to
r.
z
nf(z)
[70.1]
period
2ir.
series in u:
(r)
= ^TVz(t)
n =
and
Q2 =
n =
^^X
[70.2
220
with a
to
2tt
for
[70.1] with n =
requires that
2
.
Furthermore,
f(z)
f(z
fxz l
/u
f(z
fiZ 1
f'(z
fi
f'(
*V +
zx
j
2
//,
2
,
cj
=
f(z Q )
to
'z\
+ w2 z =
x
- a
u2 z 2 +
co
z2
= -
f'(z )z 1
- a2 z -
et
'z\
1
[70.3]
** + l
+ W
n +
F(z
Zl
-,Z
n)
- n +
Ct
n 'z\
a^
where F(z
z lt
zn
is a polynomial in
z lf
zn
lation on
r.
</>,
in the substitution r =
Qt +
<f>
so that
z(0)
=
z n {r)
,
[70.4]
The condition [70 .4] then yields the following conditions for
which are
for
=
,
1,2,3
[70-5]
Let
z 1#
zM
and 0^,
ra
[N +
1)
= 2^M n z n (r)
n -
and
Q =
2
^M
n
an
and hence
=0
x +
oj
1
x + nf{x)
)
to the order of n
x.
(N+
approximation of
221
= a cos t,
harmonic shall appear on the right side of Equations [70.3], for otherwise
some
zn
<
(z\
2j)
/(a cost)
4-
oncost
The function f(a cos r) can be developed in a Fourier series which contains
/(acosr)
=0
x )
fn (a)
cos
nr = /
(a)
/j (a)
cost
n =
fn (a)
cos
nr
[70.6]
io
(z\
= -
TO
J^fn (a)cosnr + =
2
A (a)
COST
(o)
unless the coefficient of cos r on the right side of Equation [70.6] is zero
Hence, the condition for the elimination of the secular term is
_.
a,
AM
1
[70.7]
One has
2i
= -
f
U)
(a)
^
71
fn (a)
cos nr
[70.8]
Z,
= A COST
j.
2.h^ a >
^r
~2
2^
/ ~^
cos
WT
r
z
[70.9]
is a constant.
z2
Substituting
and
A
,
so that
/ cos n+ J_ ^ fn2 w 2 -H n 1
_1_
'
"
6,
/o
[70.10
222
ct
and a 2
zn
[n ^ 3) should contain
automatically satisfied.
Assume that
can
z2
zn
and a lt a 2
<x
satisfying the
equation is
:y %>*J
~- <***\
" 2 tt n +
where
z
Q
* +
i)
= Fiz^z^
a iK +
oe
n+1 acosr
[70.11
= a cos r, as before.
Since
F{zq,
z
x
,
z lt
zn
)
zn
and
z\
zn
Putting F{z
,2 n
ct
n z\
ot
n_l z
5]
m=
bm
Zn +
""
\
2w +
^2
^0
27^ =
2
cos
mr +
U)
(n+i a
&i)cOST
[70.12]
an +
h a
C0SmT
[70.13]
Zn +
'
Z n+
w2
m=2
2^ b
rn
[70.14]
zn +
-^r[
~^
ra
y>
=
9 2
cos
mr
"
m 2
[70.15]
1.
the secular term having been removed again from the solution
71
ferential equation
x
fix
[71
.1
The integration of this equation by the method of the first approximation was
given in Chapter X.
22}
We propose now to determine the approximations of higher orders,
[t) = a cost,
cj
and a
Co
c^a
3 )
cost
cos
a^ja
and the solution of this equation gives
[71-2]
[71-3]
z2
z2
=
/
32
3
a cos rcos
a, a cos
,27 t +
1
2o
cos
3r
(2
Yz%
a
l
C0Sr
21 a5cos3r 128~
i28~
oe 2
a5cos5r
= -r^o a
4
.
t7l
Thus the
solution of Equation
[71
*]
is
^ =of
2
fi
21 i02l
acos3T+
s
,
To24
cos5r
acos(cjt +4>)
+ M g2
1
I
~M
32")
cos (3 U)t
^+^
i024
COs(5w<
[71-5]
where a and
The frequency
which a
1 ,
ce 2
n =
jua +
2
^^V
[71.6]
72.
(j
+uf(x,x) =
[72.1]
224
(f>
by
ut +
<f>
ip.
aco S
\ F(a) + Ar Y FM"*"* G
n
(a)cosn^
4-
n {a)%mn\p
sin^
f(>
-1
W<-GM: %-**)
The
[72.3]
Fn
the development of
/(acosr,
awsinr) =
^>^
n
TO
(a)cosnr
+ G n (a)smnr\
[72.4a]
G
where ^
{a)
x
0;
Q(a)t
+^
[72-5]
is an arbitrary constant.
a cos \Q(a)t
i/;
FAa)
+
ip
Q
]
jj_
+ Gn (a)sin w
G
[Q{a)t
+ #n
,.
n =
{a)
=
i//
0.
Q
,
In such a as is to be
If, however,
{a) =
has
only simple roots without being equal to zero identically, the solution
[72.6] has only one integration constant
ip
equation G {a) = 0.
1
G^a) =
0.
previously, the stationary oscillation in this case does not depend on the
initial amplitude
We now propose to establish the existence of periodic solutions,
that is, of limit cycles in non-conservative systems.
225
If x = z(Qt +
<f>)
differential equation
co z
+uf(z,Qz) =
[72J]
We can follow the same procedure as before, assuming solutions of the form
z
fiz x
2 /u
[12.8]
Q = +
where
zn
juQ 1
+ M Q2 +
Here we take
zn
as represented by their
Fourier series.
Forming
and
z,
z,
z,
and
into
One has
QQ z +
2
(j z
=
1
Q%z\+
a>
= -f(z =
,Q
- 2 -f
Sq'z 2
gj
z2
-f (z
t
,Q
)z 1
(z
,Q
)Q
z1
- 2Q Q 2 z [72.9]
- 2
- Q*z -
/i(2
,fl 2
fl
io
fl
J*+i+
a X+y= ~
fz(*o>
Q o*o)*nn
fi( z
o^
)Q
zn
- 2Q Q n+1 z
where
Z2
i2- z
l
F {z Q n
zn _
z n_
zn_
;i2
i2
nl
u
2
Fn
As before,
1,
zx
,
Q2
Hence
zn
(0)
= 0, where n = 0,
2,
z
2
:
,
so that
do not
and
Q into
u(z\ +
x )
fiacos t
00
au
sinr)
2(jQ a
1
cos
2cti2
acosT
[72.10]
226
Gv
l
=^ 2wa
[72.11]
Equation
w 2 (Zj +
zx
)= - F (a)
- \f =
n
2
(a)cos
nr +
Gja)
sin
nr
[72
z,
1
a, COST '
U-z
0)
sin
n r]
u 2 2*2 n=
[72.13]
ination of secular terms on the right side of Equation [72-9] for *2-
= ax
cosr+ u where
Gja)
1
= _ Fo^ + J_ 2 2j!
oj
yF i
1
n( a ) cos
nr +
n2
sin
nr
^,
w 2 (z 2 +
= f
4-
(a cost,
r + r + 2w,0 a
1
(acos r,
a wsin r) aj wsin
cos
4-
cj
Q 2 a cost +
v(r)
where v(r) is a periodic function containing the remaining terms of the third
equation [72.9].
series
~
v (r)
= 2^,
n=
r
\
cos
nr +
qn
smnr
fg
a 1 cos
r + /^a^sinr = -
(a
+ jua^cosr ,
(a
+ naj
wsin
f
where
a x cost
+ f a^sinr =
-
n=
F^{a) cos
nr + Gja)
sin
nr
[72.15]
the derivatives of
]>
Fn [a)
and G n [a)
Substi-
one gets
227
(j
2
('z
+ 2
= -
a
,
^F
00
'(o) cos
nr +
nr
CO
2a(i2 1 a 1
i2 2
a)cosr
4-
JF* p n cos
nr
4- 9 n
sinnr
[72.1 6]
and
ax
+ P 2 a = - -^ + %'
[72.17]
It follows that
G^T
]
"2.18]
condition G[(a) * 0, since only in this case does Equation [72.18] give the
determination of a
G x {a) *
a *= ~
\'{&
+ *.. -
if)
^]
oj ('z
(p
- a^o') +
[12.20]
z2
a 2 cosr
7J2"
(^0
a i Fo')
^Z'[^n-i^') cosw ^ + (9 n -a n=
2
G;)sinnT]
r
4^
[72.21]
Q3
is
determined by the condition for the absence of secular terms on the right
z
3
,
and so on.
228
Q n+1 from
the equations
a.GW=X
in which
| TO
and
UB+1 a +
fi
aJ=
)]
ri
[72.22]
and
rj
{a)
index n.
The expressions for the non-linear oscillation x and its frequency
(Qt
<t>)
+ MZ (Qt+
1
</>)
+ H NzN {Qt +
0)
[72.23]
Q =
oj
+ nQ +
x
QN
Following this procedure, one finds, as an example, for the second approximation
x
(a
mcl,) cos
{Qt
<b)
%- FAa) or
+
0)
,_
+
where
V_
io
2
^~i n= 2
y Fja)cosn(Qt +
o
,
</>)+
GJa)smn(Qt +
1
n2
2 ^,
F (a)
i
that the only difference between these expressions is in the amplitude of the
/ia :
Gx
(a)
= 0.
%.= j
If the limit cycle is reached,
G (a)+// 2 S
1
(a)
+ u 3 S 2 (a)+
[72.25]
2w
Gj(a)
+u z 'S (d) +
1
1 ,
N of
229
determination of the first harmonic of the limit cycle is only of the order th order and not of the N (N - 1
) ,
Hence, the
[N +
[N+
2).
= 0, one has
= acos(Qt +
<)
a) Q = u + uf^
Zcoa
[72.26]
For
N=
1, x
Q = ^ + /^4tT +
"^
UZM\
In order to establish the explicit expression for this approximation, one has
to determine a t and
Q 2 entering
73
quasi-linear system
x
+ u2 x +
fif(x,x)
[73.1]
zbliia)
[73.2]
JT= Ma)
a /
\
da
[73.3]
d^
Tt
nl \ Q{a) In
For the time being, the functions A(a) and Q[a) remain unknown.
fact, their determination constitutes the object of this procedure.
x=^-Q+~A da
dip
[73-4]
2^0
,2.
|4A 5a
4f dip
f A + ^M,4
da da da
[73.5]
x,
and x in the quasi-linear equation [73 -1] by their and [73-5L one gets
[73-*+]>
5^
d^da
5a 2
d^ da
[73.6] LM
J
i2
satis-
/i
satisfy the original quasi-linear equation [73. 1] to the same degree of approximation, conditions [73-3] being satisfied.
In order to apply the method of successive approximations, let us
(ip,a)
juz^ip^)
z 2 (ip,a)
A(a)
= uA
(a)
+ u 2A 2 (a) + +
n
2
[73-7]
Q{a)
u)
fiQ x {a)
{a)
d\
dlp2
*!
_i
_\..2_
& z o\
..
c.a
dz \(
dz
2z
(d
Z2
,
rl
dz
,(
dz
dz
- ZuU, 2u>Q
^
dfp2
J
Qxjjda
^
dip2
- 2Q A LU,A
a '*
X
di/jda
- A2 lQa2
ffiL
5^ 0a
M A _^
da Ai 5a
!
Ml
2
2(ji?2 2
|i
d^ 2
_ 2w ,4 2
_^L
d^da
[73.8]
= - E ,-2 fl2
0-2a,A
231
i^rr- + dip 2
\
*n +
,W = -En- Z"& n + rf d^
2
1
, 1
w i n +1 +
-^
dipda
,
where
En
is a function of Zj,
zn
jl n
Q lt
J2 n
and their
= a cos
ip.
obtains
(
\
8\
dip 2
+ zA u 2 = /(a cos ip
,
a usmip)
+ 2u)Q
acos
ip
+ 2g>A
simp
{
oo
= - J n=0
[F (a) n
cos
ip
+Gn (a)sinnip + 2 uQ
z
x
a co&ip
+ 2uA
sinip
[73-91
2wi2ja
- F^a) =
2o>A 1
- G^a) =
[73.10]
and
are determined.
(a^"
Zl
)
u2
= ~ F
(a)
nip\
[73- 11]
- "
$ W+
\jW^if^^
n=
2
r73.1l]
[73-10] and
from Equation [73-12] into the third equation [73.8], one has
S7^ r
%2
I
~^[F^ ^ C0S n $ + =
]
Gn( a) s nip +
sinip
+
where
2ti)Q 2 acosip
4-
2uA
[73-13]
{,)
a.
A 2 and
Q 2 from
the equations
2cjQ 2 a - Fj'%)
2toA 2
G['\a)
[73-1 *]
1
2^2
a cos
ip
+ nz^ip^) +
2
fx
z 2 (ip,a)
[73 .16
where a and
ip
|| = Mi(a) +
=
jjf0)
^
+
(a)
/<%
[73.171
uQ,{a)
2 ju
Q 2 (a) +
Q N (a) Q
1
[73-18]
| =
eu
^
2w
/"
and
^J
whence
- M *A 2 (a) + 3A 3 (a) +
^Jd)
[73.19]
w I? - - +
=
tttt
^
,
/^^ + ^
3
(a)
/i
NQ
N (a)
Equation [68.19].
Furthermore, since by the method of elimination of secular terms
the quantities
A2
An
z n {ip,a)
it is
apparent that the first equation [73-19] relates to the amplitude of the fun-
a (a) = |f = at
to
+ n
|^
2u>a
+ <u%(a) + *
[73.20]
CHAPTER
XII
INTRODUCTORY REMARKS
The method of Kryloff and Bogoliuboff outlined in Chapter X was
x = a
sin
ip
for a quasi-linear
equation [58. l] and by determining the functions a{t), the amplitude, and
ip{t)
,
2 As was mentioned, from the standpoint of formal accuracy of the order of n procedure the method resembles that of the variation of constants of Lagrange.
non-linear
/n.
Bogoliuboff have done and which is designated by them as the method of equivalent linearization.
The essence of the method is the determination of the
On the basis of formal procedure it is not clear why this particular determination of parameters leads to the possibility of approximating the
erally characterized by a certain Fourier spectrum of the component frequencies resulting from the non-linearity of the system.
If, however, one limits
Hence it
linearized problem the same oscillation which appears as the fundamental harmonic of the quasi-linear system.
a priori
the formulas giving the equivalent parameters follow directly from this prin-
parameters by postulating that the work per cycle done by a non-linear force
Fg
Viewed from this standpoint, the Principle of Harmonic Balance enables us to determine the equivalent parameters without actually writing the
2j>h
tions so obtained do not differ much from those of the corresponding linear
ered as a proof.
we shall see particularly in Part III, plays an important role in the quasi-
Thus,
several variables makes it possible to absorb the effect of an external periodic excitation by the equivalent parameter and thus to explain a number of
75.
tion
mx + Kx +
can be written x = a cos
ip
uf(x,x)
[75-1]
ip
are
-7 at
-r
Zntom
/(acosA
J
awsin0)sin0aV = 0(a)
[75-2]
TJr-flM at
where
2
[75.3]
(a) ==
u2
-\
irma
|/(acos0,
J
aa>sin0)
cos^d^
[75. ^J
As a matter of fact, this solution in Chapter X was taken as x = a sin , which merely reverses the sin ^ and cos ^ under the integral sign in the amplitude and phase equations. The notation in the present chapter complies with that used in the text of Kryloff and Bogoliuboff.
tf>
235
The second term on the right side of Equation [75 .^l is the frequency cor-
^-, &-i
Q 2 (a)
\
frequencies.
by the equations
n
irau)
in ?
/(acos0,
o
aa;sin0)
sin0d0
[75 5
K= K
-\
[75-6]
racy of the order of n 2 In fact, with values [75-5] and [75-6], the amplitude equation
.
[75-2] becomes
a
-J^a 2m
[75-7]
'4>
Q(a)
= ]/ m
r
[75-8]
One recognizes these expressions as the usual ones for the decrement and fre-
given by Equations [75-7] and [75-8], actually satisfies the equivalent line,
mx +
We have
x
Xx + Kx
[75.9]
dcos^>
asin^
ip
2m
acos^
.
a?sin^
K cos^ m
,
X
H
m
+
_ aQsmt//
.
.
X -
5,
2m
., .
gt
dQ
da
-1
sin0-
<5X 1 X + - ^ - 2m 9a 2m
2
a cosw
r_i.
.,_,
175
'0
# m
X
x
"
2m
2 2
0U M da 5a
. ,
sin
4-
aX X A TT- 77- ax 2m da 2m
C>A
\ml 2
/a\ /X\ a
cos
Substituting x, x, and x from these equations into the equivalent linear equation [75-9]. one sees that the quasi-linear differential equation
[75-"l]
2^6
i?(//
2
) ,
linear equation is satisfied with accuracy of the order of n 2 The transformation of the original quasi-linear equation
nix
+ Kx +
uf(x,x)
x + kx,
where
6
K
-
K.
k/2m=
a/a, as determined by
Equation [75-7] is the decrement, and Q VK/nt is the frequency of the equivIf one substitutes for X and K their expressions alent linear equation.
[75-5] and [75-6], one finds equations of the first approximation.
equations of the first approximation and the equivalent linear equation with
76.
Fg
= /if(x,x), by a
FL
lent parameters
Kx + K and
x
kx.
a small quantity of the second order from the solution of the original quasi-
Fg
and
FL
be the
same, that
/ujf(x,x)xdt
= kjx
dt
[76.1]
On
ip,
rp
and x =
aw sin
where
ip
cot
ip
and a and
ip
2tt/lo.
1 3
upon chang-
- aco
sin
\p,
becomes
237
2*
fx
/(c^cos
ip,
awsin^) asin^
dip
[76.2]
Xj
o
to
sin
dt
= ka
u)\sin
o
ip
dtp
= ka 2 u)7T
[76.3]
it
T o
/(acos0,
a<jsin0)
sin
(f>d<f>
r-^ in
This is precisely the first equation [75-5]> by which the parameter X was
originally defined.
It is thus seen that the introduction of the equivalent parameter
alent factor,
K lt
For
In this
are defined as
T T
W =
a
eicos$ dt
and
=
o
eisin$ dt
where
e,
i,
and
=jrJF(t)x(t)dt;
o
\F(t)x[t
o
~)
dt
Vyjf[x(t),x(t)]x(t
~j)dt
=jj[K,x(t) + kx(t)]x(t-j-)dt
[76.5]
\i
is
2^8
Substituting the generating solutions x = a cos {ut + 9), x = -awsln {ut + 9), and T = 2n/u, the left side of Equation [76.5] becomes
aufj.
LIT
J
1
/(acos0,
a u sm 6) cos
<t>d<f>
It follows that
K =
x
/(acos0,
7TCL J
aasin0)cos0d0
[76.6]
77.
F=
nf{x,x) and
FL
x + ~kx.
FL
can be written
FL
FL0
cos
{iot
respectively of
FL
+ $ L ), where FL0 and 6 L are the amplitudes and the phase The non-linear force F is represented by a Fourier series
cos {ut + 0)
,
If one makes
FL
FL0
and
6 = 9L
X and K can
x
be ob-
tained.
In fact,
FL = K
a cos {ut
BL )
- uHasm(ut +
9L )
[77 -l]
F=
r?
[/(acosr,
asinr)cosrrfT
cos {ut
0)
2"
+
TT
/(acosr,
-o
asinr)sinrrfr
sm(ut+6)
[77-2]
9)
9)
in Equa-
and
as before.
It is seen that both principles, that of the Equivalent Balance of
Energy and that of Harmonic Balance, are equivalent, because the work of
higher harmonics per cycle of the fundamental frequency is zero.
239
expressing the
equality of work per cycle for the non-linear and the equivalent linear
systems.
2.
and
in view of Statement
78.
method.
A.
mx +
where F(x) is of a quasi-linear type.
and n are constant.
F(x)
=
fix
3
,
[78.I
where
1
.
Since the system is conservative, the amplitude a remains constant, but the
one has
mx +
Hence f[x,x) = fix) =
/ux
3
ex
jux
[78.2]
KAa)
7ra
J
/(a
^-, whence
1
Q = V
ZT
.
V 0J +
= w 1 + V
-^^
* w
^T
= w
'
eta
1
)
where a = rnr^s
"''
240
amplitude a.
The oscillation is thus non-isochronous, although the ampliOne can easily discover this fact formally by calcu-
tude a is constant.
lating the expression for k from Equation [755]> which in this case gives
= 0.
B.
+ Kx +
infix)
b \x\x
in this case is
2k
,
2n
=
?rau)
J
/(
cloj
sin0)
sm&d& =
7racj
J
/(aocos0)
cos
0^0
where
2?r
2ff
\f(atjcos<t>) cos<j>d<f>
tj
|
|cos0|cos
o
0d0 =
a 2w 5
cos
3 d(f>
37T
COS
0rf0
+
2
COS'Vrf0
Hence, A =
-=r
r^.
-75
^ r^j
it is seen that
2jt
.
1
K =
x
jubato' f
77
.
1
sin0 sin<pcos0a0
Hence the non-linear correction for the frequency in this case is zero to the first order.
From the fact that^for quadratic damping the decrement varies with
the amplitude, one concludes that for large amplitudes quadratic damping is
more efficient, and for small amplitudes less efficient, than is linear damping,
-rz,
the motion
We have
a_
da
a dt
241
whence
"
~a? Jt ~
= St,
dt
\J
Hence, on integrating,
CLq
so that finally
a
4
'[78.4]
which coincides with Equation [62.4], obtained by the theory of the first
approximation.
C.
mx +
f(x)
c(x)
It will again be assumed that both f[x) and c[x) are quasi-linear, that is,
x + vip(x)
where X
are non-
and
ip{x)
linear terms
The application of the method of equivalent linearization gives
q = i//+
Y
2l
and
2m
+vi//{x), as ex-
10
where
7,
+ Kj + K 2 6 2 + Wh
sine
and the metacentric height of the ship, and K, and clents of resistance to rolling.
b,$
1
+ bJ 2 +
co
t =
6
2 >{2
D.
Lx
and a fixed
capacity C.
=
is
<P
= f(i), where
is the current.
For a sinusoidal
cos
(cot
\f(i
1
2*
7T
cos0)cos0d0
cos (cot
+ 0)
According
equation
4>
to.
= Le
where L e is the
Le =
\f(in
[78.5]
VUZ+TJC
178.6]
E.
NON-LINEAR CONDUCTORS
Consider a conductor and let the voltage drop
across its terminal
be
f{i)
If the current is of the form
i
cos
(cot
TT
f{i cos0)cos0
d<t>
cos(cot
+ 0) =
e
i
By putting
R =
e
f/(iCOS0)COS
1
7TI
J
rf0
If
energy; if
cos
d<f>.
(cot
+ 6),
-J f(e
cos^)
cos<j>
Here
243
1
it
f
j
dd>
[78.7]
electron tubes.
where
ia
f[E
+ e)
is the alternating
grid voltage.
If
e
is sinusoidal,
ia
will be
2*
\
f(E +
cos0)
cos<f>d<fi
cos(cot
0)
If one defines
1
7T
Se
en
[78.8]
tion
ia
f[E
i a
Se
F.
THERMIONIC GENERATORS
=
.
(M
DL)
-tj,
DL
-jj
takes
current is
di = s (m- dl)^I
e \
dt
By Kirchhoff's law,
1
ia
l +
ip.
where Ri R = Li L in the
iR
LJ?-mesh; hence
i' . L
In the
CL-mesh,
H'c<"
( T L iL
dt
Figure 78.1
2hh
whence
Substituting these values into Kirchhoff's equation and dropping the subscript L, one obtains
i
c = LCi^'.
at*
e
dt
dt
finally
LC ?* +
dt~
.R
se (m - dl)
ff
'
[78.9]
2LC
s(m-z)l)]
[78.10]
'^=
dt
Se
(M-DL)-^ *
2LC
^
e
,
-.
178.11
Se
(M-DL) =
j
e
[78.12]
the differential
equation is
L
where
e
+ * +-fJ"' -*&-#*&
[78 - 131
D*
0,
so that
j?[idt
[78.14]
2^5
Instead of
d~e
LC
JT-
+ (RC \"
de MS'/,//
e
[78.15]
when
S.
= RC
[78.16]
= RC MSe 2LC
[78.17]
Figure 78.2
If the
the decrement
ia
static curve
biasing voltage
aQ
ax e
,<
7i
[78.18]
where a lt
fi lt
ct,
1
Y\ e o
'
,/4
RC\
[78.19]
246
REFERENCES
"Les methodes nouvelles de la mecanique celeste"
(1)
(New Methods
in Celestial Mechanics), by H. Poincare, Gauthier-Villars, Paris, Vol. 1, 1892, Chapters III and IV.
(2)
Equations of the Theory of Perturbation), by And. Lindstedt, Memoires de l'Academie Imperiale des Sciences de St. Petersbourg, Vol. XXXI, No. 4, 1883.
(4)
Paris, 1930.
deux degres de liberte" (On the Mathematical Theory of Self -Excited Systems
with Two Degrees of Freedom), by A. Andronow and A. Witt, Journal of Technical Physics, USSR,
(9)
(Russian), 1934.
H.
(10)
mit negativen reellen Theilen besitzt" (On the Conditions under which an
247
(14)
05)
B. Van der Pol,
06)
925
07)
PART
NON-LINEAR RESONANCE
79.
INTRODUCTORY REMARKS
The object of Part III is to outline the present status of the theThe phenomena of non-linear resonance are far
more complicated and diversified than those of ordinary linear resonance, and
it does not seem possible as yet to give a unified picture of the whole sub-
ject.
procedure inevitably reflects the somewhat unsettled state of the whole subject and leads to a certain overlapping of topics. The reader will undoubtedly
they employ.
designating as internal resonance the case when the divisors in the generalized response function become small.
more or less equally divided between these two principal cases, whereas the
work of the school of Mandelstam and Papalexi centers mainly about the phenomena of external resonance, which are illustrated by numerous experimental
researches.
The reader will find it convenient to read Chapters X and XII of
Part II before reading Chapters XIII, XIV, XV, and XVI of Part III.
It is
I,
and particularly
250
II,
Part I.
The ess-
ence of Chapter XIX lies in the theory of differential equations with periodic
between the exceedingly complicated phenomena of non-linear resonance and various existing theories are probably incomplete at present, and It is hoped that this survey will serve as a stimulus for further generalizations.
CHAPTER
XIII
80.
exponential function
A few well-known
JU>t
advances its
3U
je.
phase by n/2
ut +
e
2;
whence
e
=
-n/2
.
r
.
3Ut
by
2.
Multiplication of
by j
JlJt
n
,
= j
tt/2.
vector / = ae
n n
on the vector ae
J0Jt
,
which mul-
Likewise,
[ju))
u n and
= d n/dt
= {ju)
= j n u> n
fx
f2
the operation
5-
is additive,
(ju))
that is,
n
,
{ju)f=
{jo))f x
+ [jcj)f 2
If,
<j>{j)
instead of
=
or
(joj)
example,
+ Bj
Af +
jBf.
The subject matter of this and of the following three chapters is taken from the treatise of Kryloff and Bogoliuboff, Reference (I).**
252
a.
without
/ is real.
6.
of the form
* (j)f
(t)
+ *&)/&) +
+ * n (;)/B U) =
[80.1
(j)I6
+ 4
(j)i x
,
+
Ix
,
<t>
{j)in
.
[80.2]
In
8l
y jlJ a
k
l aL dt k
y 2L
k
o Pk
^ik
UL dt
k
where I and
fi k
are
constant parameters.
is
n k
Z*
k
(jco)
E =
A^o
k
z{ju))I
[8l
-,
and
ZeuW = -^
k -
E = Y(ju)E
k
[81.2
2^ a k^^
253
where
Z(ju>)=^j
k'=0
[81 .33
Y(joj)
L..U
k =
Problems involving systems with one degree of freedom are thus re-
may be
For electric circuits this procedure is too well known to need emphasis here; consequently a few words about it will suffice.
For an induct-
ance L and a capacity C the values of the complex impedances are respectively
is a real quantity R.
the impedance equation is used; for a parallel one, the admittance equation.
L lt R lt and C
are in series,
one has
y = yl T z
1
+y
1
+ 7^7
Zl
77
Z<1
and so on. For systems with several degrees of freedom, there arises the question of coupling between these degrees of freedom.
The coupling factor can be
The
R lf
25^
function is a sinusoidal electromotive force E = E,e jut The second circuit contains L 2
.
R
2
and C 2
Mis
the coefficient
I*
Figure
8l .1
+ R1 i V
l
'
'
1
.
di -M^f- = E "* dt
jUl
e
[81.5]
dl2
r>
If. f
!
,. i,
,. *.
di\ dii
[81.6]
Rjl,,
- jMa>I 2 = E
jMuI, + (jL 2 u + R 2 +
whence
-^)h
-
r +
2
j(l 2 u,
'i
"
^y
^
+
J
c
(#i
+ JL^) R 2 + j[L 2 u) -
MV
Y
1
[81.7]
The cofactor of
the first circuit.
[8l
(jco)
of
M=
0,
Equation
J] gives
I1
=
R,
+ jL u
Mja>i 1 + (L 2 ju)
-
=
i2
[81.8]
)i 2
The condition which expresses the consistency of the system [81.8] for values
i
l
and
i2
= 0,
is
We follow the notation of Kryloff and Bogoliuboff in choosing the positive directions as shown in Figure 81.1.
255
A(ju)
that is,
= (LJu + jA-AlLiJto +
-^
2
V=
[81 .9]
(L^w 2
1)(L 2 C 2 oj
1)
MCC
1
u4 =
[81
JO]
Let
u
This gives
T~r"' LjCj
1 2 w 2 = TTr; L C
2 2
= L L
l
w 4 (l -
(u
+ u 2 )u 2 +
oj
[81 .11
^2 _
1
U>j
cj 2
\ )
fl + y\u
2
x
2 \ 2
u> 2
+
,
4ff
tj^g
.g
2(1
- /)
2
UJj
2x CJ 2 )
A
2(1
Y(CJ 1
- D 22x2 +
)
4ff
q^faig
.g^
io 2
of each circuit.
Very often the establishment of a formal analogy between the differential equations expressing two different types of problems permits a formal
transfer of known solutions of problems of one type to those of the other. The
It is preferable, however,
L~
jut = E + Rt + ^jidt = E Q e
o
[82.1]
Its equation is
d X -rrr
dt
dX
h -rr dt
_ jut ^ + kx = Fa e = F
,
256
v =
4^
at
dv
r;i
---,-
kjvdt F
[82.2]
~dt
One observes that Equations [82.1] and [82.2] are of the same form and that the following corresponding quantities indicate the analogy between electrical
(l;
(L,m)-
(R,h);
[g,k)\
(E,F)
[82.3]
h =
where x = Leo
- -^-
En
-i<t>e
VFTV'
<t>
[82.4]
Cu
= tan
-1
Lw_ oJ
R
The quantity
J<t>e
= Y
ft
is the complex admittance
(ju>)
of the circuit.
i<t>r.
[82.5]
|/V +
where
m = tan
i
(mu>
- )'
v4v
(moj
k)'
mo)
hco
<f>
Yh 2 u 2 +
(mu>
k)
In both cases
<f>
at
and
and
R and
compare the differentiated equation [82.1] with that of the mechanical system.
The condition of equivalence for these equations is
(i,x);
(L,m):
(R,h);
(,fc);
(ff,^)
[82.6]
257
rrig
-UOOffiTCRT
k2
k3
m3
-
'75'0'O'OTW^-
'OOWOW^-
F
(a)
i
L3
^0000'0"07T
^'000"0"OOOT^
(b)
Figure 82.1
Desig-
m lt
ra 2
and
3
by x lt x 2
and x 3 and
and x 3 =
jux 2
'
JOJX,
[82.7]
m m
jcov 1
+
^2
-.
Vj
3"
i>
J"
/
\v 2
vj = ,)
jojv 2
(v.
J>^
[82.8]
J"
jo)V 3
ju
(,
= F
v/jcj.
On the other hand, if, in the electric circuit of Figure 82.1b, the
current
i
L jui
1
1
1
1 (i
2
i
x
=
i
L 2 jui 2 +
,.
2
C 2 j(j
C 3 j(j
(i,
(i 3
[82.9]
L 3 jui 3 +
C 3 ju
In)
258
In this par-
83.
APPLICATION OF THE METHOD OF EQUIVALENT LINEARIZATION TO THE STEADY STATE OF A QUASI -LINEAR SYSTEM
small so that the theory of the first approximation describes the phenomena
= -R/2L.
Let us assume
= f[i).
In accord-
di
Re
is
R
where
i
2"
/(z'
cos
</>)
cos0
d<f>
[83.1
investigated.
Liu + R +
-pfr-
+ R =
e
[83.2]
Equating to zero the real and imaginary components of this equation, we obtain
two equations:
259
Lo)
7-J-
and
Ceo
R +
R=
[83-3]
The first equation [83-3] determines the frequency and the second the amplitude of oscillations in a stationary state.
Re <
= 0,
= 0,
that, is,
be noted that the method of equivalent linearization can only be applied when
the resistance
= -R/2L small,
in
which case the oscillations are feebly damped and the system is quasi-linear
and can be described by equations of the first approximation.
Let us now consider a linear circuit with admittance A{jo))
,
that is,
-tt-.
v,
and
>
Impedance will be
^>
" JUZ) +
0,
sTS)
83 4
-
'
that is,
SXa)
[83.5]
Again two equations are obtained by splitting the complex quantities into real
and imaginary components.
One of these equations determines the frequency and
the other the amplitude of the stationary oscillations.
= /<*)
where
ia
This
S(a)[e g
+ De a] =
S(a)e
[83.6]
where
eg
and
ea
the anode reaction (here n is the amplification factor of the tube), and S{a)
is the average transconductance,* a function of the amplitude a.
The quantity
S{a)
= -^jf(E +
acos<) cos^cty
[83.7
260
where
inductive coupling.
Designating by
A and B
and by
Figure 83.1
we have
= - Za (ju>)
M(ju)i a
[83.8]
From these equations and Equation [83*6] the admittance A(jto) of the circuit
is given by the expression
A(j cj) =
M(jcj) - DZjjo))
[83.9]
In a self -excited state the total impedance vanishes, and by Equations [83.4]
S(a)
[83.10]
In other words, the admittance of the external circuit must be equal to the
_9
M(jcj)
ea
Za (ju>)
[83.11]
K = D +
S(a)Za (ju)
[83.12]
This equation is established by Barkhausen (2) and can be applied to the circuit shown in Figure 83. 1.
By
(i
_ L Cu 2 +
)
jCRcj
[83.13
A and B
i1
AB = Za {ju))i a and the current i y in 2 = eAB /{R + jLu) = i a /[(l - CLu + jCRu].
is
e
)
eg
is
261
e
'
Mjb,ii
=
(1
lZ^+
)
jCRu
<
[83M]
whence
JlL
e
=
'
Mju>[{\
[(1
2
)
LCoj
+jCRg>]
_
'
Mjoj
AB
- LCu
+ jCRu](R + jLu)
R +
jLco
one finds, after separating the real and the imaginary components, the follow-
S(M - DL) = CR
1
[83.16] [83.17]
Since both
and
is of the second
S(a t )
= jf(E + acos0)cos0rf0 =
If
CR _
ia
[83.18]
f{E
3
+ acos
<j>)
= f(E
=
4-
acos0)
k^cosfi
k 2 (a
cos</>)
/c
(acos0)
+''
where
iQ
f{E
and k lt k 2
are constants.
..
r-
2*
\
2*
\i n
cos d>d<j>
k,a \cos
<j>d<t>
2*
2*
3
k2 a
2
I
cos
rf0
k3 a
3
1
cos
d</>
[83.19]
The first and the third terms of this equation are zero.
kx
+ k 3 a
mW
*.)
'
83 2 i
-
262
It is seen that the quadratic term of the polynomial approximating the exper-
ia
The ampli-
fc
and cubic terms and the parameters of the circuit, as was found previously by
84.
APPLICATION OF THE METHOD OF EQUIVALENT LINEARIZATION TO THE TRANSIENT STATE OF A QUASI-LINEAR SYSTEM
The method of complex amplitudes can also be generalized for the
transient state of
pt
,
quasi-linear system.
tial functions
where p =
+ ju.
The quantity
is the decrement if it
namely,
df
Thus,
for example,
^j- = jtf =
+ jio)f =
con-
It follows,
-
<f>
= tan
= tan
is an Increment. -1
If, however,
<
.
0,
that is, if
is a decrement,
ji
tt/2
> >
0,
</>
-> 7r/2;
<f>
if, however,
the
and
>
0,
which means
that all higher derivative vectors approach the in-phase condition with the
t)
=
[8l .9
]
[84.1]
Consider, for example, the self -excitation of a simple {L,C,R)circuit, as shown in Figure 84.1, closed on a non-linear conductor AT which has
the equivalent resistance
(a) .*
+ R =
e
[84.2]
263
+ ju
when
but is small.*
The linearized impedance of the
steady state is
(ju)
+ R = Lju +AR.+ R
e
t)
Cju
=
L
Figure 84.1
This gives
LCp
Dividing by LC and putting
1
(R
2
+ RJCp +
,
[84.3]
j-~ =
2
io
we have
p
_L +
R + R
+ w
'
whence
+ R = - R
2L
*-
+
\
w o2 ~
R +
\
R.\ 2
J
2L
jo)
[84.4]
where
R+ R
2L
w
0,
is the frequency
Self -excitation
.
{0)
<
that is, if
R < -R e {0)
<5
We shall call
-^j-
the linear
decrement.
<5.
RAa) Ro J
|i? e
[84.5]
(0)|
>
|-R
|,
that is,
<
0,
in which case it follows from [84.4] that the amplitudes increase iniThe function
|i? e (a)|
tially.
(aj) =
and
oscillation reaches
stationary state.
equivalent resistance
{a)
When a
->
a1
for which
{a)
* R
0.
84.2 in which
N is
sistance
{a)
In this expression, 6 is the decrement if it is positive and the increment if it is negative; this corresponds to the notation of Kryloff and Bogoliuboff
264
Z(jQ)
= LJQ + R +
x
-^
jQ
+
(1
C 2 jQ
2
)
- L 2 C2 Q
+ jR 2 C2 Q
[84.6]
N is
present,
{a)
to
in the preceding
MMMSiSLr
L2
^AAA/W
It is apparent
-AAAAA/V
R.
form
Zip)
A(p)
B(p)
Figure 84.2
[84.7]
A(p)
Let
= [^C^ +
(R 1
+ R e )C
l][L 2 C2 p
+ R 2 C2 p +
l]
- M^O,? =
4
[84.8]
L&
= u
2
i
>
M
'
^2^2
U) 2
YLJT 2
p2
2
= R C u =
2 2
R,
L 2 io 2
Equation
where p x and p 2 are small dimensionless factors of the first order. [84.8] becomes
Since p 1 and p 2 are small, we can introduce a small parameter p. by setting We t nus obtain the following characteristic equation p 1 = /i^j and /> 2 = ^ 2
-
A(p,p)
(4
^J-
+ i)(4 +
l)
^-^2" =
[84.10]
pp + p 2 p 2 +
x
one obtains a system of equations from which the subsequent terms p lt p 2 can be calculated. These equations are
265
Mp
A v {p ,0)p A P (P ,0)P 2
x
,o)
Vp
,0)
[84.11]
A pp (Po> 0) P?
+ 2V(Po.)Pi + ^>o>0) =
and p x
,
mp,
Introducing this value of p into Equation [84.10] and making use of Equations
[81.12] and [81.13], one obtains the following equations
2
.,
(p
+ QfHp +
o.
fl 2
-o
^m V>>0
A
.
- 2p(2 P + = ^
Q, -i
+ fl 2 --2^
)
[84<13]
^-^(5
It is apparent that,
l}+^+l)
+ /up,,
Q2
namely,
Q ~
i
Q 2,-0)2 -5Pi"i7T5 - Ql 2 Qf
1
2
1 J.
g y
2. u
Q2 - u 2 m, ^Ql - m 2 Ql
oj,
2
y 1-V
x
[84.14]
P
Q 2
p u
i
G)
i
o2 ~ Q%
Q\
-Ql
g*
Q p2 " 2 Q\ - Ql
1 1
From these expressions it follows that the frequency and the decrement during the transient state are given by either pair of the following expressions:
Q =
Q\
ojo
Qi
2
\
o),
rn
Q,;
Q 2 depends on
Isochronous
266
(i
- >),() -
i^
+ EM)
2L
*
If
(f^fe
rf
[.it]
when
/Oj
(0)
<
0,
self-excitation
with frequency
occurs.
can be written as
|fl(0)|
> \R
i\
[84.18]
where
R 01
This criti-
cal value is
R.
01
- -
(*i
+ R
L Q{ *L 2 Q?
Y
[84.19]
GJ?)
(1
) <J
(a)
=
(
Q? -
u> 2
Q 2
Si
gi 2Li
Q* - u* R 2
2 - 2
x
2L2/V
[84.20]
The decrement
^1
(a 1
when
(a 1
),
as previously shown.
(1
2
<7
)<Ua)
= at - q>2 Ri + Q* - Q 2L
2 2 X
2\
u)i
R%
Q* 2L 2
[84.21]
dj(a)
d 01 1
RM
#01
[84.22]
and similarly
dJa)
6 02 1
t
RM
R 02
J
[84.23]
where
j and
<J
01
and and
d 02
J? 01
Q2
and
R 02
(a)
determining p x from the second equation [84.11] depends on the condition On the other A p (p ,0) 0, as follows from the theorem on implicit functions.
hand, the expression
267
gl
V^O.O) =
2p Q (2pQ
+ 2? + Q2 2 2 ufuf
becomes
2jQ (Q
1
2
1
- Q2
or
2jQ 2 (Q? - Q2
OJ?U> 2
2
jQ 2 respectively. Hence, the condition A (p ,0) * fulfilled only when Q = Q 2 that is, when both circuits are not tuned in
for p
=
x
jQ
or p
is
85.
Z(p,n)
+ ur
(a)
=
fir e {a)
[85.I]
=
{a)
namely,
B{p,M)
where
A{p,/Lt)
and B(p,/u) are certain polynomials prime to each other; see EquaFor the system shown in Figure 84.2, the quantity A{p,n) is
tion [84.7].
B(p,/l)
= Cpt^ +
'
(lln
-22
l)
I
[85.3]
U>
268
A(p,m)
[85.5]
then zero.
=
jQ Q
[85.6]
Let p =
(p
+ o2 )0(p)
where
<t>(jQ
0.
A(p,0)
+ fiA^O)
[85.8]
contains the capacity and inductance terms; the secSince this term is proportional
It follows,
In view of
^|
q|
and
P>
^[
q\
are odd,
<t>{p)
is even.
A{p,fi)
+ uB(p,u)r =
e
[85.9]
= jQ + np +
x
[85.10]
= 2jQ
<f>{jQ
[84-.
11], we find
= jo ~
A,(jQ
p
,0)re
B/<f>
tities
S
"'" and
,
fe
'
=
'
[85.12]
are real.
From Equation [85. 11] we obtain the frequency and the decrement,
namely,
269
Q = Q
and
<5
+ kR
[85-13]
[85. 14]
- jQ Q and
e
[85.15]
6{Q) <
R
where
(0)
<
= -t-<*
(a)
at
and
Si
is
other frequencies of the system, the frequency and the decrement of a quasi-
Q =
[85.17]
mits neglecting the dissipative parameters, which are assumed to be small quantities of the first order, as far as the frequency determination is concerned, since the error is of the second order, that is, of the order of
In other words, the quasi-linear systems in a non-resonant state are also
2
fi
.
quasi-isochronous
86.
frequencies
and
Q2
This condition
shall be called the resonant self -excitation of a system with several degrees
of freedom.
Q2
and
quencies very small, say of the first order of smallness, it is necessary that
the difference of the component, non-coupled frequencies
cj
1
cj
of each de-
].
M of
mutual inductance
270
must also be small, that Is, the coupling between the two degrees of freedom
must be rather "loose."
fiP;
= nQ
[86.1]
+ u^x +
x
.
l)[x
+ m(1 +
/xP)^ 2 x
(1
fiP)
2
]
n Q x
[86.2]
where x = p/co
x = j +
/ix
1
and equating the coefficients of equal powers of u, one obtains the following equation for x 1
+
(2jx 1
j^)(2jx 1
+ JU +
2P)
- Q2 =
2x 1 {( l
2jP)
+ Q2 +
2/P>-!
f2
=
Q
2
[86.3]
+
fi
= -Wand
= N,
:
= - ^(1lA) +
4Y2
yM - 4N +
2
Vl6P 2 M 2 + (M 2 - 4A0 2 +
j [l
jy^
|/4JV
+ Vl6P
+ (M - 4A0 2
2
[86.4]
Putting
(R,
+ R,- R
2 2
)
C2
co
i[g
(l
*)*]
if,
and
[86.5]
A{R e +
R x
tf 2 )
C>
2 2
(l
^J=d
2
+ R 2 + R + L^Yrj +
e
Yd
AL X
co x
2(co 2
tij
uj4
t]
Vd 2
2
r,
[86.6]
d as
q =
Qi ( a )
"i
+"2 + ^i
2
e
/TTTl^Tp
[86J]
<5
^(a)
= R + R + R x
L^Vq
4L,
V<*
2
rj
271
or
Q = Q 2 ia) =
6
^-^ - T /+ R2 + R +
e
rj
K^T7
!86.8]
R,
dn(a)
Ligjj/j 4L,
Krf
The essential feature of these expressions is that both the frequency Q[a) and
the decrement d[a) are non-linear functions of
{a)
When
~ u2 ~
u, the
Developing the expressions [86.7] and [86.8] in terms of the parameters of the system, we obtain the following two sets of expressions for Q{a)
and
<5(a):
1
If (R e +
R )C u >
2
1
2g, then
2
Q
and
{a)
x
u;
d^a)
R + R + R - L
e
1
ioYiR e
+ R - R 2 C 2 co 2 2
1
4,,
4L,
[86.9.
QJa) =
R + R + R + L u\{R + R - R
to;
<5
(a)
2)
C 2 oj 2 -
4;
AL
2.
If [R e +
Q,(a)
= u
+ jVig 2 - (R +
e
- R
2 2)
C 2 u2
<M a
= R +
e
)
R%
and
[86.10]
QAa)
oj
|VV
(/?,
+i2 1
fl
2)
C>
2
(5
(a)
i? e
/?!
+
X
i?2
4L
system with two degrees of freedom introduces a radical change in the behavior
of the system. In a later chapter it will be shown that resonance in a multi-
CHAPTER XIV
SUBHARMONICS AND FREQUENCY DEMULTIPLICATION
COMBINATION TONES; SUBHARMONICS
The non-linearity of an oscillatory system accounts for the appear-
87.
ceives sounds which are not contained in the emitted acoustic radiation, and
he has shown how the slightly funnel-shaped form of the tympanic membrane of
to*x
= -
/3x
X(t)
[87.I
the function
is of the form
= a x cos
u>
+ a 2 cos
u 2 t.
oj
and u 2
the
frequencies
cj
co
and
tones.
*.-/(*)
where
ia
[87-2]
We shall be inter
It has been
shown in preceding chapters that this experimental relation can always be ex-
axv
a2 v
a3 v
[87-3]
The number of terms in this series depends on the degree of the approximation
desired, and it is shown (4) that in practice the coefficients decrease with
In order
simplify the argument, let us assume that the impressed grid voltage is of
form
v
th
= Hcosc^ +
cos
cj
t)
[87-4]
that is, it is composed of two alternating voltages in series having the same
amplitude
oj 2
Substituting [87. 4]
intj
Equation [87
] ,
one obtains
273
ia
a^icoscj^t
costJ 2
a 2 k (cosco l t
cosu 2 t)
a 3 k (cos^t
+ cosu 2 t)
[87.5]
co
and
to 2
2u 2 u 3w 2 2u
,
+
x
u) 2
and u x
u2
3w 1?
u2
2wj
- u) 2
2oj 2 +
x ,
and
2cj 2 -
u)
The current
will con-
u lt w 2
2u
2u> 2
Slj 1
3u 2
(j 1
o>
d) x
<j
2u x +
=
u>
2o) x
oj
2o 2
+ u v 2u 2
cj
1
oj 2
2n
x T20, where
= 100 cycles
ia
f2
will
100, 120,|
It is seen that the combination frequencies spread out on both sides of the
impressed frequencies,
= 100 and
f2
= 120.
enon, attention is usually centered on the frequencies lower than the im-
/=
ma) 1
+ nu 2
[87.6]
In Equation [87.6],
(0,3);
(1,1);
formations.
As follows from the foregoing analysis, the combination frequencies
and
u2
which are
the combination frequencies changes, but the relation [87.6] holds for any
27^
value of
co
or
a 2
If,
provided, the high-frequency part of the spectrum of the combination frequencies can also be recorded.
The relation [87.6] is applicable throughout the
v =
cosw^
a 2 cos
u2
is fairly rectilinear,
This shows that the combination frequencies are due to the non-linearity of
the circuit.
If a combination harmonic of amplitude
the current
cos
i nj7n
(nQ +
1
mQ2 )t +
n<f> Y
4-
m<^ 2
[87-7]
j(n!{
+ mQ2
88.
decrement were given, not by one set of relations, but by two such sets. Thus
for example, we always have two pairs of relations
Q =
and
Qjia);
= =
6 x {a)
Q =
ber of pairs of solutions:
(a);
6 2 {a)
Q =
Q^a);
6 x (a)
Q =
i2(a)
;
d n (a)
On the basis of the method of the first approximation, Chapter X, this means
that the most general oscillatory system is characterized by oscillations of
the form
i
ix
dx cos
ip 1
i2
0-2
cos ^2
in
a n cos
ip n
From now on we shall designate the component frequencies by capital letters 3^ and Q2 , reserving the small letters oj 1 and u 2 for the frequencies of uncoupled linear systems, as was done in the preceding chapter.
275
It is recalled that
\p,
the
the frequency
i k
.
$k =
Qk t
<f>
The most general forms of the equations of the first approximation for both a k and
ip
are
>
7T
is,
i
= "
** (afc)a * ;
^T
^ (a
* );
2
>
'
88
ix
i2
4-
in
a cos^j
x
a 2 cos^2
an cos&n
[88.2]
the system shown in Figure 84.2 and note that its impedance across the ter-
N
+
2
is
2j6 01
(#i
Si
CjQu>
Q)(Q2 2
- Q2 +
2jp2 io 2 Q)
2j6 02 Q)
L
- Q2 +
ftfi OJ
where
and
Q2
and
<J
02
tained from Equation [84.6] by making use of Equations [84.10] and [84.14].
Since for quasi-linear systems the decrements d 01
put
6 Q1
nx
6 Q2
n 2
and we obtain
2
^m "
7(.n\
,
,
_ (i
&
+
2
2jj
{uj 2
2
uQ)W
CjSico
- Q 2 + 2J! 2M Q) - Q 2 + 2juhu 2 Q)
a
rfi
L
o j,,
J
where x f 2 and ^ are small. This impedance is impressed on conductor having the characteristic
e
non-linear
= -
Fit)
= -
Mf(i)
[88.5]
If u = 0,
so that
+
<f>
i2
a 1 cos(Q 1 t
fa)
a 2 cos{Q 2
4> 2 )
[88.6]
where a lt a 2
4>
and
are constants.
276
If p>
= -
uf(i)
= -
jufiii
i2)
= - ufa^osi^t +
4> x )
a 2 cos(Q 2 t
\Jj x
+
ip
2 )1
[88.7]
and
namely,
a 2 cos
\p 2 )
f{z)
so that
/(a^cos^ + a 2
cos
if/ 2
XX A
*> cos
^^1 + m ^)
[88.8]
where the index % runs from zero through the positive integers and where
ha,
0.
and n were
= -
XX A
x
n,,
{nQ
mQ
)t
n<f> x
m<f> 2
[88.9]
i n,
vm m
uA n,m
l
n,m
cos
{nQ x
+ mQ 2 )t +
m<i> 2
[88.10]
Z j{nQ +
rnQ2 )\
and the total current will be the sum of all these harmonics.
It is apparent that the amplitude of each combination harmonic will
be
j{nQ x +
mQ
2)
+ mQ 2
+
co
2
2
(ni2x
+ mQ2
+
''
2.2/
4/i
^ {nQ^ +
1
mi2 )
n,m\
\[qI
(nQx
+ m
2 2
2) ]
4//^ 1\nQ 1
+ mQ2 f
(/
[fi2
nei-
mQ
2 2
)
or (wi2 1 +
ra42 2 )
Q
,
becomes
fi.
nQ
+ (m
t '\)Q 2 or
(n +
i2 x
mQ
2,
becomes
small and of the first order, or zero, both the numerator and the denominator
277
|e
BjW
may be finite.
or for n = 0,
0i);
(Q 2 t
<f> 2 )
[88.12]
Hence, aside from the two corresponding harmonics of current, all others are
Thus we have
/uA lo cos{Q 1 t
fa)
fa)
[88.13]
Linearization, we obtain
fiA xli
'
= - o L.TT
F{a
X X
2>r2,r
cos fa
ip 1
dfadip 2
[88.14]
M01 =
tt^ 2n d
^(aicos^!
ft
a 2 cos ^ 2 )cos
ip 2 dfadi/> 2
= =
tt~9
1
2t2>t
27r
\F{ai cos fa
6 6
,
.
4-
a 2 cos
ip 2 )
cos
ip 1
dip 1 dtp 2
fl i
[88.15
one has
fiA10
^'aj
and
/uA 0tl
= ife
and, by [88.13],
e
= -
{Reix
R' 'i2 ) e
[88.16]
This equation shows that in the example of non-resonance considered here the
and
i2
resonant system simplifies the problem because the variables can be separated
We obtain the following systems of equations of the first approxi-
-df-d}h
dt
6ai
x>
'
H " 6a
da2
__
..
[88.17]
"
o'
''
4h.
dt
.-
"
\j>
o"
=
+ fa and
2 t
<t>
>
278
[88.18]
at
at
In view of
[88.16],
d')i 1
+ Z{jQ" +
6")i 2
R'eiy
+ R e"i2 =
[88.19]
whence
Z{jQ'
6')
R' e
0;
Z{jQ" - 6") + Re =
[88.20]
Because
<f> 1
and
a = a&
Q" = Q 2
6'
6,(1
- 1^);
6"
<J
2 (l
- |^)
[88.21
where
i? 01
and
R 02
Re
'
and
"
e
Q and Q 2 respectively.
absence of resonance,
i
a^osiQit
0i)
a 2 cos(Q 2 t
<f>
2)
[88.22]
--4-irK
#=-4-B"
88 2
-
"
This result can be generalized for systems with n frequencies; however, this
89.
upon by an exciting force with two or more component frequencies has a discrete spectrum of combination frequencies which are more numerous than the
exciting frequencies.
On the other hand, a linear system with several degrees of freedom
279
one particular circuit of the system that is coupled to other circuits, the
excitation of this circuit occurs through couplings and, by virtue of the nonlinearity, is resolved into combination frequencies which, in turn, react on
the other circuits and cause combination frequencies in these circuits if they
are non-linear.
pear in the process of self -excitation but only those for which the initial
decrement
<Jy(0)
<
0.
fication by providing a set of differential equations of the first approximation in which the variables can be separated.
A new complication arises, however, when two frequencies of the
spectrum are nearly the same and coalesce at the limit; we shall call this
internal resonance of the system.
It occurs whenever two or more component
in this definition merely emphasizes the fact that the actuation is effected
dition.
^m
71 o\
"
(St
&
o
f L
Qll
J
N with
characteristic
[89.2]
= -
F(i)
= -
uf(i)
l
Q and Q 2
of
the linear system is not far from being a rational number; this fact is ex-
Q =
2
Q
o
net.
with
s
[89.3]
/u
frequencies
?!
and
r Q2 = Q
o
i1
i2
where
ix
and and
i2
= a 2 cos +
<i>
i(j
i//
and ^ 2
<j)
Q2 t
respectively; and a x
a2
<f>
lt
and
<f>
are constants
'
280
but very small we can still use the principle of superFor n position to the first order of approximation, as was just shown, and write
e
= - Hf
dicos {Q x t
<f>i)
a 2 cos
( Q-J, +
4> 2 \
[89.4]
Thus
(V
Qit
\
<f>
)\
\ = ?
\A m cos
Q
Til
+ Bm
sin
7H Q
[89-5]
m>
where
Am
and
Bm
a2
<t>
and
<f>
~ M Am
/
cos
Q
s
x t
+ Bm
sin
Q
s
x t
4
with frequencies ^-Q and
1
[89.6]
fi
in
lects in the first approximation all harmonics I m for which Expression [89.5] reduces to
e
r and
s,
HM
/u(A s cos
xt
+ BgSmQrf)
[89.7]
x t
= Cr cos( Q = C
and
s
xt
+
<f>i)
2)
+ A-sinf Qx
DsSiniQrf
2)
[89.8]
cos
x t
+
,
BsSinQit
cos(i2^
<f>\)
Dr C
,
2t
Cr
(s0
{s<f> 2
r< t>\)\
+ +
a2
rrdr
D r = jflai
C =
s
cos 1st
rfah
a 2 cos rrj
sin
rrdr
[89.9]
J/foi cos st +
"
a 2 cos \rr
+ {s<f>
2
r^j)
cos
srdr
D =
s
if(a 1 cos st
a 2 cos rr
+ y(s0 r^)
2
jsinsrrfr
281
If we put
s<t> 2
r<f> x
Rl'
=
7ra 2 J
1
a, cos (sr 1 \ I
+ rl
)
a 2 cos
cos
rrdr
2n-
-,
Ye " =
2"
jFaiCosfs-?
a 2 cos rr
sin
rrrfr
[89.10]
R =
'
\F
ira,i
a, cos
L
st
a 2 cos \rr
\
cos
/
srdr
Y = e
'
F
TTdi J
ai cos
L
st
a 2 cos [rr
*
+ )
S
-
sin
srdr
Q-^t
f? r sin
QM =
=
2j
Y "a
e
sin
( fi^
2)
(#?
[89.11]
^(^sCosP^
Bssin^i)
=
=
RgdicosiQit
4>\)
YJai sinCQ^
fa)
(R;
+ jYe'HiCOS^t +
fa)
]
[89-12]
it becomes
= ~
[(R' e
jYe'Ki
{Re"
+ jY ")i2
e
[89-13]
first approximation.
',
6 "
be the frequencies
- =
dt
da 2
<5'a i;
dt
d\p 2
~dt
6"a 9
[89.14]
Q'\
Q"
t
where, by [88.6], ^ 1 =
t
x
<t> x
and
i//
282
d<l1
A'n
da<l
X"n
[89.15]
7r
" r
i;
_
d<
fl
For dynamical equilibrium the impedance drop and the impressed voltage must
balance each other for both oscillations of the linearized systems; this gives
Z(jQ'
*')ti
+ Z{jQ" -
5")iz
{Re
jYJH-l
(R?
+ jY ")i 2 =
e
[89.16]
Hence
Z{jQ'
<$')
(R:
jYe)
0;
Z{jQ"
6")
{R'J
+ jY ") =
e
[89. 17]
It is noted that these equations are the same as Equation [88.20] except that
is now replaced by
jYe
[89.17] is identical to that for Equation [88.20], provided one does not sepa-
Pi
= O +
t
6t
_ RM)
Rqi
fl'.-.flj.+
^jjjk
" = 2 +5 2 _pl
[89.I8]
=
<f>
aiCosffli*
0i)
a 2 cos( Qx t
<j>
[89. 20]
)
where a 1# a 2
<j>
and
da
dt
1 _
.L
/
R' \ Re\ e
da 2
.(. x /,
R? K
e
L
02'
[89.21]
dfa
rf*
_*i_'V'
d<^ 2
;
/? 01
d*
- o2 ~
o
'
4.
i2_v"
#02
Re
'
",
e
Ye
'
and
Ye "
appear,
and
written
g
at
= s ih _
at
r #lat
sQ2
rfii
K 02
Ye
"
:x
-"oi
[89.22]
283
(*
-)*-*
<y
(i J
( \
-^K =
- 1*1
^02
[89-23]
s 2 -
V rd + s^-Y S2
e
n "
r^Y
<*1
'
-n-02
tC oi
s(q 2
V
%-Y
/t
2
,
')
'
- Aq, +
V
~Y
-ft
'\ '
= sQ" -
rQ'
[89-24]
01
where
fl'
fl, 1
P~ oi
Q2
Si'
p#02
Y e ".
Equation [89.24] shows that the stationary oscillation of the system occurs with two frequencies and Q" whose ratio is a rational number.
In addition to these frequencies there exists a frequency spectrum correspond-
m ing to the harmonics of the order ^Q', as can be shown from the study of the
s
90.
SYNCHRONIZATION
In the previous notation the total phases are
\Jj
</>
and
\p 2
(f>
whence
\p 2
rp
x
<f>
- -j<t>\
= &
d6/dt =
and
= constant,
which gives
s\p 2
rifri
constant
[90.1
Thus the total phases are "locked in synchronism" with each other.
The question now arises whether the condition [90.1] is stable.
Since the
Re
'
"
e
,
'
and
Ye "
variables a lt a 2
20
and
of these
Following
Aa
~dY
= Ra
x
(aio>
a 2o,0oM<*i
(aio,a 2 o^o^ a 2
+ ^VKo.^o^o)^ + R "{a
e
Aa
j
t
Ra {a 1Q ,a 20 ,e )Aa 1
l
R'a (a 10 ,a 2 o,9
2
)Aa 2
l0 ,
a 20
)A0
[90.2]
j~ AB
= F^ia^a^B^A^ + F (a 10 ,a 20 ,$ a2
)Aa 2
+ F (a lo ,a 2o ,6 o )Ae
284
where
*'
- -
4sfl 2
ih
rfij
R" - -
*(>
f
'
)*
and
[90.3]
F =
s^K" K
02
- r^-Y n 01
If all real parts of these roots are negative, the synchronized oscillations
are stable; if any one of these parts is positive, the synchronization is unstable. Assume that the oscillations are stable; then, for sufficiently small
Aa x> Aa 2
a 20
and
the
and A0 which
presence of such
a zone.
one can say that linear systems are characterized by a zone of synchronization
reduced to zero.
91
s,
fi
an d Wg/wj =
/uP,
For
co 2
=
<t>
a x cos(w
x
0j)
+
=
a 2 cos
{cot
<f> 2 )
[91
</>
-1
Putting
a sin
<f> ,
04 cos
a 2 cos
<t>
a cos
<f>
and
a x sin
a 2 sin
</>
one has
a2
a x2
a 22
2a 1 a 2
cos(<^>
2)
[91-2]
a cos
(cot
0)
[91 .3]
+ RAa) =
[91
A]
and
(a)
and
2 2 (a)
the decrements
d^a) and
d 2 (a)
285
and
and
Q2
=~ Jf = - ^(a)-; ^ -fl
da 2
*2<a)a 2
[91-5]
8 (a)- u
d 1 [a)
6 2 {a)
{a)
and
if
2 (a)
6
l
are ob-
It is apparent that,
If, however,
{0)
>
(0)
>
0,
at least one
Equations [91-5]
ax
awe
d,(a)t
;
a2
60(a)
a 20
e
[91.6]
01
1O
,
010
+ Q
(a) x
O) t;
02
02o
+ Q 2 i.a) -
cj
where a 10
[91 .2]
20
and
2O
From Equation
one has
a
9
2<Si(a)<
a 10
<J
a 20 e
<>2(a)]<
^)
+ 2a 10 a 20
If ^i(a) = 0,
k(a) +
cos([ 2 (a) if
(a)
^(a)]<
2o
1O )
[91 .7]
then a = a
10
= a1
<J
= 0,
then a = a 20 = a 2
In the
{a)
Q 2 {a) and
its harmonics.
>
0,
and
^
da
>
>
0,
(a)
and a 2
<f>\
02-
92.
they are applicable when, because of some couplings provided in the system,
286
Xi
xn
[92.1
where x
= a x cos (oj^t +
<j>
, x )
x n = a n cos
(u n t +
<f>
Substituting
F(x)
[92.2]
io n
and let y lt
Since any term y i will have the same frequency as the corresponding term x it we can write y i = Si x i so that Expression [92.2]
,
y n be these terms.
becomes
y
yi
yn
Xi
+.+
Sn xn
[9 2 -3]
definition of the
equivalent parameters S
Balance, Section 77-
1 ,
clear that by adopting this procedure the terms expressing combination fre-
quencies have been omitted inasmuch as the frequencies of these terms are
o) n
appearing in
Expression [92.3].
After the frequencies
co
ie
d ie
of the equivalent
linear system have been determined, the equations of the first approximation
so obtained can be written in the form
-
= -*ua
*>u
;
;
-^
JT
= -
6 ne
an
[92.4]
W=
r.
= " ne ~
[92 5]
'
there are regular forcing terms, they are added to the harmonics just mentioned.
so that for
ft
one
287
Let us apply these considerations to a linear system characterized by the impedance Z(ju); the relation between the current and voltage is then
e
= Z[ju)) i.
op)
tion becomes
Q(j(j)e
P{ju)i
[92.6]
ju>
&
'93.7]
transient state.
= -
F{i)
= -
juf(i)
P(>
When
fi
*)*
= ~ MQ(jr)m
[92.8
Thus, for example, instead of Equations [88.4] and [88.5], we can write
2
kb
ft
+ **)(* +
***4
+ Q?> = ~
^Mh w4
2
[92 9]
'
> fU) )
linear solution
0)
in the form
tp 2
i= a^osfa +
where
a2
,
a 2 cos
uzi{a u a 2t ^1,^2)
+ U 2 z 2 {a h a 2
ip
u ij> 2 )
"
'
'
[92.10]
z lt
l
,
ip
and
ip
= uXn {a h a
2)
12 (a lf
a2)
-jj
/uX21 {a h a 2 )
+ u 2X 22 (a
,a 2 )
[92.11]
-~ =
-3-2
?!
|iyu (a lf a 2 )
2
//
y12 (a 1(
o2)
+
)
= +
2
^l2i(ai,o 2 )
+ M 2 Y22 (a u a 2 +
288
of
.
.
Xn
X 12
X 21 X 22
,
>
J 11
V 12 x
...
>
>
anri
V 21' V22'
,
...
Carrying out these calculations for terms containing the first power
of n, that is, for
z lf
Xn
X 21
Yn
and
Y 21
of the first approximation so obtained are precisely those which were obtained
If, however,
one retains
the first three terms in Equation [92.10], the calculation results in what
equivalent linearization, which was more or less postulated in Part II, can
be justified on the basis of the preceding analysis and can be generalized for
CHAPTER XV
EXTERNAL PERIODIC EXCITATION OF QUASI- LI NEAR SYSTEMS
In preceding chapters we have been concerned with the phenomena of
self -excitation of quasi-linear systems of the autonomous type, that is,
t,
tion later in connection with the Mandelstam-Papalexi method based on the theory of Poincare.
93.
mi +
in which the time
t
kx
/uf(t,x,x)
[93 -1]
appears explicitly.
N
f(t,x,x)
= f ix,x) +
[f
n =
nl
(x,x) cos yn t
+ fn2 (x,x)
x.
sinyj]
[93-2]
where /
f nl
motive force
Z(ju
mju
+-J^.
When n =
is
0,
sin (<j
+
t
0)
[93.3]
aaj cos(cj
0)
where w
Vk/m and
<f>
When
fi
but suffi-
with the method of Poincare in Chapter VIII) with which we start the approximation. One may consider these expressions as an approximation of zero order.
Our purpose will be to establish an approximation of the first order which
290
nf[t, asm(co
<p),
ao) cos(tj
</>)]
[93-^]
fnl
fQ (a sin ip,
ato cos
ip)
^\ g
klO
N'
Hk (a) sin
kip]
fnl (asmip,
ato cosip)
= JT[g nl
N'
[93-5]
fn2 (asinip,
ao) cos
ip)
HO
Using these expressions and also the expansion [93-2], one can write the nonlinear term [93-^] as
e
^
k = o
N'
\
g^(a) cos
k(p n t +
<p)
h k (a)sink(tj Q t
0)1
Z Zi k,k^ +
h
n =
k =
n =
'
9 n2
(o)] sin[(/cw
+ yjt +
k<f>]
k -
tZZI^M + W^W^o n=
k =
yJ>t
H
k<j>]
i2727[W
n
=
A-
a)
a)
]
sin [<* w o
- yjt +
[93.6]
+ y n and ku
yn
appearing in
encountered. We shall limit our study in this section to systems in which none
of the combination frequencies approaches or is equal to the frequency
In other words
[kuj
co
.
yn]
a>
and
[ku
yn ]
a>
of Expression [93-6] one ascertains that the only harmonic having the fre-
quency
cj
is the harmonic
e\
= ug^a)
cos((j
<p)
4-
juh^a)
sin
(u Q t
4-
<p)
[93-7]
291
Using this expression together with the second equation [93-3], one
has
Z =
e
^
i
--[9ite) L
au
- Kia)j]
J
[93.8]
Hence, to the first order, the non-linear element of the system can be re-
placed by the equivalent linear one so that the symbolic equation of the
z - ze =
Written explicitly, this equation is
[93.9]
mp +
where p = -6 + ju,
lent linear system.
8
r =
[93.10
has
(-
Jtof
+ wg = r-JL+[g
(i$
- h.iri^i-
Jto)
[93-
H]
= -
9l {a);
to
co
- -rr-^h.ia)
[93-12]
oa
and
= u at
d</>
cor,
one obtains
-jt
dt
2mu>
ffxia); J
'
dt
ip
= =
to
t
2mto a
+
</>,
h^a)
93-13
^7 = 1T9M); dt 2mu)
i
%=
dt
x
fl
rJU,^h Ud 2mu a
1
co{a)
[93.14]
It is thus seen that the solution to the first approximation is still of the
form
a sin0
93 -15]
where a and
tp
[93-14]-
By analogy with the definition of the linear decrement, it is logical to introduce now a quantity X, the equivalent decrement, defined by the
equation
I = -
--gM
[93.16]
292
Since ^(a) Is the first coefficient of the Fourier expansion, its explicit value is
2"
g^a)
J
o
/
2n
[93.17]
whence
X
=
aco
TT J
fJa u
ip r
dip
[ 93- . L/ "
] J
one obtains
ip
dtp
[93 .19]
A
->
dt=-2^. a
da
dip
dI =
+ "=V-^
,/k
k'
ip
_.. [93.20]
.__
tions [93-20], and substitutes x and x into Equation [93-1], one finds that
the solution x - a sin
ip
mi +
~kx
+
2
//
(k
.
k' )x e
[93-21
the dependence of the "forcing function" on time does not appear explicitly.
This is due to the fact that, for the formation of these equations, only the
(x,x)
lim
f(r,x,x) dr
In other words, the first approximation deals with the average value
of the forcing function with respect to time, and the instantaneous behavior of that function is felt only in approximations of higher orders.
k(a n )
au n
fn (a o
[93-22]
(dX\
(4r) \da)a
>
= a
and unstable if
\daU
<
= a
Self -excitation may develop from rest if X(0) < 0; if, however,
X(0) > 0, no self -excitation takes place.
293
94.
FIRST-ORDER SOLUTION OF VAN DER POL'S EQUATION WITH FORCING TERM It has been shown that the operation of a self -excited thermionic
- Mil -
y )y
[94.1]
fi(l
y )y
= E smat
[94.2]
Usinat, where U =
+
-*
E _
=
iu[l
(x
+ U smat)
][x
4-
U a cos at]
[94.3]
In this equation
f(t,x,x)
[l
(x
Usinat)
~\[x
L^acoscy^J
[94.4]
one
E
(x,x)
W
1
x [94.5]
If we let
m=
]>
and k =
in Equa-
oj
o o o
o o o o
tion [93.1
since
Figure 94.1
] ,
one
/1
[94.6]
da
[94.7]
~dt
U <
2 there
begins to increase.
294
ox
= 2^1 -
^
Z
[94.8]
is
If
t/
>
.
2,
stable
The value
F U
(1
2
2
)
- a2
U2 <
2,
U >
2,
the external
95.
ance
= mjco+ -K,
f<>
-sin(* + mj Q + JL jQ
Likewise, for
e
2 m(w - p
y^
sin(Qt
)
0)
=
m(io
%$
- Qr
)
cos (Qt
0)
m/G +
cos (Qt
<t>)
~ f _
oJ
m( ^
wo)
cos(fl;
0)
w^ - w
2
J&S
52 o
)
sin(
0)
= asmip +
'
N'
yk
(1
- -
k 2)
lZ \
2 u n2
- +
,\
N'
ot
= o
^o
>
y y [MAw 2
ic
TV
TV'
h n2 ^)cos(kip
(A;w
+
2
yn t)
= o
yn )
JJ_
2m
+
M
y y (g
n ,
H i, k
h n2 Jco8(ty
- yj)
0/c=0
TV
Uw -
yn )
P rf> 2m n = 0/t =
& (g^
y n <)
TV
y* Y* (9m,k
W)
sin(fe ^
y n t)
[95.1]
it follows that resonance occurs whenever one of the divisors becomes either
295
fi
a sin
if/
a sin [u)[a)t +
<j>],
2.
The spectrum of combination oscillations with frequencies ku{a) t y n One could proceed with building approximations of higher orders in
u>
(k(o
^Aj
+ bn Xn
2
)
and
96.
HETEROPERIODIC AND AUTOPERIODIC STATES OF NON-LINEAR SYSTEMS; ASYNCHRONOUS EXCITATION AND QUENCHING
We shall now consider Equation [95 -l] when the autoperiodic oscilla-
In such a case
k
=
;
hk
0;
1, 2,
Snz,k
Ki,k
W
,
= h
>
for k > 0, which we have This follows from the fact that the numbers g k just asserted to be zero, are merely the non-constant terms in the Fourier
expansions of [93-5]-
Now for a =
their Fourier expansions, and hence all coefficients of the cosine and sine terms in these expansions are zero.
If we put g nly0 =
An
n
and
cosy n
h n20 =
t
B n Expression
,
[95
reduces to
JL
^A
+ B sinyj
[6j]
plied frequencies y lf y 2
yN
296
my +
where m, X 1) and
k x y + ky
6xy
= E
sina i t
+ E2
sin (ct 2
0)
d
1
[96.2]
k are
positive constants.
fi
Moreover, X : and
by writing
are small, so
mX-
6,
= m6
[96.3]
In order to make Equation [96.2] of the same form as [93.1], one can introduce
a new
sina
1
m(co
- a2
+
)
sin(a 2 t
+
ct
2
0)
m(u
[96.4]
)
+ kx
/u
kx
kE a
1
cosci 1 t
kE a
2
cos(a 2 t
+
2 )
0)
m{aj
m((J
+
a?
ct
Here k =
k +
E^sinot-^t
E
2
\
sin(q 2
0)
m w
m{u n
[96.5]
/uk
>
0.
By Equation [96.1]
= UkE a
1
cosct l
/ikE 2 ot 2
cos
2 n
{a 2 t
+
2\2
0)
M6E,
2k[m(a>
2
2\2
\cj,
m 2(\(J
+
2 - a 2\12 )}
[96.6]
UdE.;
HdE cos2a
x
2k[m(cj
- a 2 )] 2
,
2m[m(co*
2
ct
)]
(^ +
+
4a*)
fid
cos
2m[m(
2
)
)t
x
0]
/udE1 E2 cos[(a 2
.3/.
- a ;)(^ - a 2
)h
(a 2
af
'(cj
- a
)(u
cies
2ct lt
2a 2
+ a 2 , and Oj
a2
= f(E + Fcosat +
e)
[96. 7]
297
</>)]
[96.8]
where a and
<f>
linear conductor
ia
= f{e g
ia
= Se g
where
the equivalent parameter 5 must be so chosen that the fundamental harmonic of [96.8] is equal to the harmonic Sa cos
{oj
t
+</>).
>
5 =
f(Eo
"*"
Fcost +
1
dr2
[96.9]
It was shown in Chapter XIII that in the linearized scheme in which the non-
<J
jco
d Q (l
-)
[96.10]
where
co
The quan-
F cos
ctt
is absent,
In
S(a)
na
f(E +
()
a cos
cos
d<t>
In order to use this equation, it is sufficient to replace and to average it again over the period
2
77-.
by
cos at
oj
low (for example, audio frequency), the equivalent transconductance S(a) given by Equation [83.7] is a slowly varying function with frequency a.
If the
of the
298
2n of the heterautoperiodic state will also be periodic with the period TH = The reader will recognize the characteristic behavior operiodic oscillation.
of the so-called superregenerative circuit.
If, however,
than the autoperiodic one, its effect will be felt in the virtual modification
of the critical transconductance of the electron tube.
of the characteristic, this effect will sometimes manifest itself in the ap-
heteroperiodic period and the other dr 2 with respect to the autoperiodic one.
As an example (5), consider an electron-tube oscillator with a non-
fix)
= ax + px 2 +
yx
6x
ex
[96.11
= 0,
whereas for
a cos0
b cosi//
[96.12]
b
where a and
ones.
<j>
and
\p
the heteroperiodic
Replacing
S{a,b)
= -
2an
/(acos0 +
[96.13]
+ ^-eb 4
o
b
[96.14]
= 0,
= a + I ya 2 + | 4 o
ea
[96.15]
and
= 0,
= a - ||y|a 2 =
[96.16]
299
which gives
\lffj V o \y\
C96.17]
.*
periodic frequency is
5(0,0)
If, however,
= a >
S(a,0)
b
[96.18]
0,
the
= a + ~yb
[96.19]
b)
and a, 0, y, and
>
0)
in the
= a + ^-ya 4
|-|e|a
8
'
'
[96.20]
5|e|
5 max (a,0)
= a +
A^|
{b *' 0)
4
[96.21]
if
f ,6
,
>
[96.22]
that is, if
3
,2
s 15,
,,4
l .2
.
lrV<* <i/r 5
5 e
e
It should be observed that the notation here differs from that of Section 54.
300
summarized as follows:
a.
tions of self -excitation are influenced one way or the other according to the form of the non-linear characteristic.
CHAPTER XVI
NON-LINEAR EXTERNAL RESONANCE
97.
For simplicity, we confine our attention to systems having one degree of freedom and fractional-order resonance.
"0
In such systems
[97.I]
7"
where r and
that
s
>
1,
because
= j a + nQ
[97.2]
where n is small Let us consider an electron-tube oscillator whose non-linear element, the electron tube, has a characteristic
ia
= f(e).
'Let the
autoperiodic
oscillation be
e
a cos (u
<f>)
a cos f
<t>\
ia
= / E + F cos
a. t
4-
a cos
<f>
[97.3]
In this expression the quantity a is the amplitude of the autoperiodic oscillation, -a is its frequency, and
and
ce
writing
ia
=Se
a
[97.4]
see Section 77
^02
A].
By a Fourier series we
a
7TS
E +
Fcos olt
a cos
a t + 0)
cosf
ar +
rfr
0j cos
at +
<jA
sinf
ar +
0j
sin
4-
<f>
[97-5]
Sr a
where
cos
at +
<f>\
sin
( at +
0j
97.6]
Sr = S r
(a,<f>)
J /
-7ra
E + F cos 1st
<f>\
a cos
rr
cos
rr dr
.91.1]
S .=
Siia,*)
= -
K + Fcos
st
a cos rr
sin
rr dr
Sa
cos
(a
(/>)
(Sr
jSi) a cos f a
</>)
[97.8]
a
S is
complex quan-
tity
S = S r + jSi
whereas in a non-resonant system it is real.
[97.9]
remains the same as in Section 89, the only difference being that instead of
impedances we are now dealing with admittances since transconductance is an
admittance.
"
>(l-t)
o,.
[97.10]
where S
If we put
at
o
<j>
if/,
da
dt
1
'i
-^
=
dl.
Wn
a +
,
S>
d n -Fr
[97.11]
The variables in these equations cannot be separated since both S r and S { are
0.
a, 0)
303
The first case is characterized by the approach of a and tain fixed values a
1
<f>
to cer-
and
a
<f>
when
-> .
Thus we have
Si(a
<5
Sr(a
0o)
=0;
r
S
a +
,(f>o)
u)
- a =
s
m-7 ioi
Oq
to
[97-12]
jet;
this means that the frequencies u and a become "locked" in a certain rational
relation.
(cot
<t>)
a cosf at
<f>\
t)
heteroperlodic frequencies.
in Chapter XVIII.
and S r
]>
the resonant case degenerates into the non-resonant one, which has al-
f(E
f(E +
where Su S 2 gives
co
,
u)
= f(E
+ S,u + S
- S3 u
1
[97-13]
and
= 2, which
= a/2.
F cos
2
sr = s +
i
T &
=
S3 F 2 +
^S F cos A
2<t>
fs 4
2
;
Si
= -\S F
2
1
sin
20
[97-1^]
S-
S >- S -2 S *
d<f>
F +
a
S * Fc0s2 *) a
-J sf
d? =
grated.
<"
-2 -
1^-
S2 F
[97-15]
.
sin 2
The second equation [97-15] admits separation of variables and can be inte-
Putting
The proof of this proposition can be found in the Kryloff-Bogoliuboff text "Introduction to NonLinear Mechanics," Reference (1), pages 270 and 271.
304
Ad =
one has two cases:
In Case the expression
2 *
T
and
-25
0U )<., >
i
)
>
where
is given by
* Hi^
**"
&"]
One can select for 20 o the principal determination, that is, sin 20 o > 0;
cos 20 o > 0.
f r brevity, we put
dt
= Aa - 5a
[97.17]
When
A<
0,
negative and hence a cannot tend to any limit other than zero.
the only stable stationary amplitude.
If, however,
A >
0,
point a =
is unstable.
be shown to be stable,
a
(/gl"^-
S -
S3
F + |s Fcos20 o )
2 2
97
Since
<j>
->
O
when
> op,
periodic oscillation.
In Case 2, when
d(j)
Ad>X,
d<f>
(\
Ad
Asin20
Aojll
\
/,
Acj Aoj
X
(l
+ ^jsin20) =
,
dt
_,\ sin20)
.
Aoj
[97.19]
6
that is,
6.
2Ad
<j>
cos
20
= Aut
6
where
O
= Aut +
2Au
cos
20
is an arbitrary constant.
written as
{Aut
+ 1-cos2Uut + Ad
id
O)
[97-20]
<f>
305
The stability of
-S
F< +
-S
Fcos20
20
j
I
cos
20 dt
[97.21]
/cos2^^=-fcos20^=|-f-
\to
cos
20
sin
X.
20
d(20)
d{l
2Auk
k
2
fr
sintfi)
sin^
2Auk 2
log(l
k u)
u =
Si
S*F' = C
[
[97-22]
If
C <
0,
97
0,
"
state a =
C >
98.
FRACTIONAL-ORDER RESONANCE
We shall consider Equations
[
of fractional-order resonance.
the stability of the solution a = 0; for that purpose we develop the inte-
cos {st
-|-0)
in Equations [97-7].
E + F cos (st
(,
COS
VT dT
[98.1
2tt J
J"
+ F cos
st
sin
2rr dr
With r =
Sr
jf
Zn
[E
Fcosst][l
+ cos(2W +
20)] dt
S,
= -
2*
|7n
i
|X +
l
Fcoss*l sin(2r^ J
20) dt
^06
Since
jfa (E +
o
Fcosst)sin2rt dt
Sr =
J f
o
[E
+ Fcosst][l +
cos
2W
cos 20] dt
20 dt
If we set
2"
-jf [E
a
+ F cos st]dt =
N
[98.2]
and
jf
Equations [98-1] become
a [E
+ F cos st]cos2rt
dt
Sr = S
i
N
= -
cos2<f>
[98.3
A^
sin
20
dt= 6 (f
d<t>
(
/Nn -
l
)
X
t$
n JV,
acos2 *
[98.4]
dpNj
The nature of the solutions of these equations establishes the conditions for
the stability or instability of the autoperiodic oscillation.
If the only
possible.
sin
<f>.
da du =
dt
sin
d</> -1
dt
dt
and
dv
-77-
da
"77
sm
dt
-7^
dt
<f>
+
,
d<j>
a cos
-7 dt
[98.5]
507
(- p + A)
- B
(- p + C)
[98.6]
B
Hence the solution a =
tive real parts.
^=Af->UWS-(*--7i
negative, the roots will be real or conjugate complex.
value [98.2] gives
r a s
[98.7]
Replacing
iVj
by its
(J
--
>
60
2ttS
fa (E Q +
cos2rr dr
[98.8]
cj
--
r a s
<
2ttS
fa (E +
cos2rr dr
[98.9]
In Equation [98.8], the roots are conjugate complex, and self -excitation is
possible' if
> S
that is, if
2n
^jf (E
a
+ Fcosr)dr > S
[98.10]
In Equation [98.9
]>
(N - S
with
2
)
>
Sn
[98.n]
N
-
> S
2
)
(N
7(^0 -
<>o-or
do
[98.12]
This means that the autoperlodic state always sets in when the external fre-
W,
tf~
Z7T
L (E + F cos st
a n u
cos
2rr dr #
[98.13]
Since
308
systems where r/s is an integer have been eliminated because they do not possess subharmonic solutions, one can assert that the condition for the exist-
frequency a is
f-^fi
where
q =
[98.14!
0,
1,
2,
If the amplitude
N =
where faF [E
point
)
i!aM
Z7T
f cos
V
sr
cos 2 T
rdr
[98.15]
taken at the
for
s =
2r
[98.16]
Ni
_ FfaF {E
Li
Q)
fQr
2r
which means that in the zone of self -excitation defined by [98.12] only fundamental fractional resonance of the order one-half can exist, in which case
This fact was noted by Lord Rayleigh in his experiments with oscillating systems.
He observed that, if one of the parameters (L,ra) or ^/C,k), in the
99.
PARAMETRIC EXCITATION
Parametric excitation of a system is defined as the condition of
It consists of
L C
{i)
C = C
(l
psina)
509
where pC
capacity and p
If we assume that
is
#
o L o o
L,(i)
LAi)
L,
the problem
o
o o
linear theory.
3
a
2
AAAAAA/V[99-1]
(LC,
Flgure 99
This condition will result in fractional-
order resonance of the order one-half, as was just shown, and the steady-state
+
t
<f>\
[99-2]
and L, one
(l
non-linear parameter L
and write
LAa) =
(V(a
a7T
6
</>
2"
sin
0)
sin
d<j>
[99-3]
is the
has the
Figure 99-2
the inflection point near the origin in order to simplify the argument.
(1
e~
Kt
) ,
one has
L (!
(x-f
,
f'
X
3
-.-.)
[99
A]
dL(i)
di
is negative.
X'
2
^\
The argument is obviously valid for any point around which the
It is also apparent that the term
O
expansion is made.
X = L
represents the
310
constant coefficient of Inductance of the coll without an iron core; the re-
malnlng terms
p-;
0U).
(1
+ p sin at)
+
;
-r fidt psinlor?) /
:
= -
Cn (l + p sin at)
,
-e
f J
asinHj-*
\2
0J /
rf<
a cos
(?'
p
+
sin
) [99.5.
*!"
By the Principle of Harmonic Balance, Section 77. the fundamental harmonic of this expression must be equal to the voltage drop of the current i, given by
[99.2], across the impedance
Ze
= r e + jx e
sin
1
{^-t
f "J C
acos(0 ^ a
0)
- cos(0
0)a0;
|(l +
bx
1 fr =
/>sin20)
^C
acos(0 + s
0)
20)
sm(0
0) a0
.J
-(l + p
sin
one finds, after a few elementary transformations, the following values for
and x e
2*
sin(0
+
(l
0) cos(0
0)
-a
C
2
+ +
/o
dO
p sin 20)
aC
1
cos
20
[99.6]
cos (0
0)
20)
(l
d0
sin
aa
p sin 20
1
The
C.
=
1
C
sin
(l
20
f sin 20)
sln 2 0)
[99J
resistance R
capacity C
(l
+4
The equivalent
The decrement and
!
311
R
6
cos
aC.
20
pio
2(L
L,)
:99.8
"
U)
=
j/(L
w,
+ Lj)C
p ^- +T sm20
.
L,(a)
= (-^ C os20)a; 2L
0J '
tf0
wA
L {a) 2L
x
^o-f)+ ^f sin20
[99-9]
wnlcn
a PP ears
dt
wo
L(0)
u p
"
2L
2/
sin
20
2L
L[{0)
If we let
"oP o^
_ m;
2L
R =
n:
LAO) wo -
a
2
-2T"* -
= (m
cos 20 ~r
rf0
n)a;
dt
+ m sin 20
2L
L[{0)
[99-10]
The singular
0,
m sin 20 +
(sin 20 o > 0; cos 20 o > 0).
[99
-H
Let 20 o be one of the roots, which we will assume to be in the first quadrant
and the
by
77,
= (mcos20
dt
dt]
u o
n)$
[99-12]
Li'(O)
dJ
2L
The last constant term in the second equation clearly does not have any effect
on stability and amounts to a shift of the origin in the
(
rj)
-plane
The
n)
[99-13]
For self-excitation it is necessary that the free term as well as the coefficient of 5 be positive.
In this case the singularity is either an unstable
312
20 n >
m
[99.14;
cos20 o >
3m
On the
The first inequality is stronger than the second and should be used.
other hand, for real values of the argument, one must have
--M-<l m Lu p
inequality [99-1^] and the condition [99-H] we have
p
that is,
2
[99.15]
This merely imposes the additional condition that the index of modulation p should be below a certain critical value given by [99-15]. Prom the first
= w
sin
20 o
2
= m 2 (l -
cos
20 o
m
whence
cos
j 20 o =
m
2
<
Substituting the values of m, n, and p, we obtain the condition of selfexcitation which was derived by Kryloff and Bogoliuboff
2
(
"
ifd<f>/dt
[99.16]
reversed.
in
Equations [99-9]'
and the
2R
Loj p
[99-17]
L x {a)
For the stability of a stationary state one can again apply the
{a)
in a Taylor series
f
given by equations 6 = 6
and a = a x +
rj.
313
-P~
T]
(m
cos 20,
n)
2ma
(2m
cos
sin
20j)
[99.18]
dt
= v[(J
qL'^iaM +
20j
where
m
It is to be noted that
P
4
'
L/faJ
is negative,
decreases when a
[99.13] is
X
2
increases.
[~(mcos20i
n)2mcos20 + 2m5sin20
1
1]
where 5 = a q\L
l
[a l
)\
>
0.
clearly
3m cos
20!
n <
and
[99-19]
pm
cos 20!
(m cos 20i
n)
> Ssin20i
If the angle 20 : is in the first quadrant (sin 20 x > 0; cos 20j > 0), the pre-
n)
>
[99-20]
whence
cos 20!
>
71
Comparing this with the previous one, namely, cos 20! < -5
two conditions are not consistent.
-p.
<
n,
2<f> x
<
0,
sin 20! > 0, it is observed that both conditions, namely, ~}m cos 20j
n <
and
cos 20 :
100.
mx +
kx
juf(at,x,x)
t
[100.1]
with period
2n
In
314
w =
where w
gers.
^<?
[100.2]
s
equivalent linearization, Chapter XII, and for this purpose we seek a solution of the form
x
a sinf
at +
s
<t>\
[100.3]
F =
uf(at,x,x)
[100,4]
kx x
*!
[100.5]
F =
juf(at,x,x)
at
<f>\,
a cos
at +
<f>\
MOO
6]
F =
kl x
kl x
=
k
1
k}a
sinlat
<pj
a cos( at +
<p\
[100.7]
and k
k, 1
ixa
j
(
f\ST
V
<t>,
sin
rr, a
a cos rr
s
l
sin
rr dr
[100.8]
X,
\ral na
2*
flsr
J
\
<f>,
sin
rr,
a a cos rr)
s
>
cos
rr dr
Jfc,
1
na
\f\ST
J
\
<f>,
sin
rr,
aw n u
rr dr
[100.9]
2*
X,
=
7Taj
f[sr 6, a
\
sin
rr
I
cos
rr d
^; 2ra
l/iA
V
w u (i +
\
A)
2/c/
[100.10]
315
da
rf7
-2^ ai
Xj
d<f>
dt
k
l
= U
and k
l
r -^ a=
^-J a+
kx
__
^2k
[10 - 11]
two differential equations of the first order sufficient to determine the two
rod of length
sin at.
EI
d%
dx
q
yA^y
g
dt
+ Hsinat
p> =
dx
Q1
EI
y
is the weight of the rod per unit volume, is the cross-sectional area of the rod, and is the force of gravity, 32.2 feet per second squared.
A
g
y(l)
y xx (0)
y xx {l)
[100.13]
= z(t)smnj
[100.14]
one finds, upon the substitution of [100.14] into [100.12], the following dif-
ferential equation
yA
g
4
..
2
/
^ + EI ^jtl t
tt
r\
[100.15]
If we let
2
EIng
~ "M7^
;
"
'" ~
EIn 2
2
l
;
" F cr
[100.15] becomes
[100.16]
where
cj
rod and
Fcr
+ w
(l
- ps\nat)z =
a/2.
= asin(|f +
0)
[100.17]
316
where a and
da
dt
d<f>
pacor, cos
20
[100.18]
a
"
2
dt
pw n 4""'sin 20
Comparing these equations with Equations [99 9 of the circuit excited by a periodically varying capacity, we note that they appear as a particular case
]
{a) =
and
= 0.
We can therefore
] ,
- a
oj
<
f(x,x)
+ E
sin
at
[100.20]
We have w
a.
+ kx + f(x,x) = E
sinat
[100.21
sin (act
0)
[100.22]
da = - A_ 2m It
100.23]
d</>
dt
Jk + V m
fc
k1
f
J
nau
[f(a sinr,
aw
cos r)
E sin(r
0)] cost
dr
100.24]
k,
na
cos r)
sin(r
0)1 sinr
J
dr
If we put
2"
k.
=
7rawoi J
/(a
dr
[100.25]
k, e
na
1
2*
/(a sinr,
J
dr
317
It is noted that the coefficients k e (a) and k e {a) correspond to the absence
sin at.
we can write
2m
where
&e
--
2m
2m'
+
2:*nao>
sin0
^ 2m
fej
_^sin0 2maa
6e
2maa
sine*
[100.26]
Similarly,
,/*
+ m
*i
rv
2a
(k + m
1
\
- a
2
2 \
l
- (K -a 2a \m
1
/
#
ma
COS0
2a
where
a e
1/2 w.
- a c$
-^
ma
cos0
[100.27]
= V^/5.
The quantities
and
cj
odic force.
hl in Equations [100.23], and \ m one obtains the following expressions of the first approximation:
k +
da
-T-
= -
at
6e
".',
sin
4>
2ma
100.28]
2a
r,
d<f>
-t
or
ac
E a
'
'
ma
cos
sin
= 2maa<J
e
[100.29
E cosfi = ma(u
whence
a
[100.30;
m i/Y KU/
21 2
)
+
i
4<5
^i
order of approximation by exactly the same relation which gives the forced
amplitude of
will be shown now that the important difference between linear resonance and
More specifically,
w
318
If we set
R(a,(t>)
=
(w e
2
sin0 m
pi
2ccad e
t,
ip{a,<j))
ot
)a
cos0
m
pi
[100.31]
toff -,**#>
[100.32]
2aa-j- =
/7M
ip(a,<t>)
0;
0(a =
[100.33]
variational equations.
and
6<f>,
<f>
by da
2a
^
-jjip
- Rja +
R^<f>
[100.34]
2aa
= ^a <Ja + ^
6<t>
aS
- (aR a +
)S
+ (R a ^ -
R^
a
[100.35]
aR a + ^
<
0;
Ra
R^
>
[100.36]
aR a
4-
0,
= -
2aa^ da
=
mu>
tt
- 2a6 = e t
2a^p^ da
[100.371
2aa
5e
=
m-
f(a
[100.38]
^
W(a) =
2tt
au
f u
f(a
J
sinr,
awn u
cost) cost
dr
2>r
= -
/[a sin(w
i!
0),
aco
cos(w
0)] a w cos
/!
0) d
[100.39]
319
erage power dissipated by the non-linear force f{x,x) during the oscillation
x =
a sin {u
<f>)
with a so that
{a)
>
0.
fulfilled, as follows from Equation [100. 38] and from the condition
Wa {a)
>
0.
R(a,</>)
R a *L + R p. = _ R a
da.
da
d(f>
da
*da
whence
^rfoT
= "
*
[100.40]
KK
From Equation [100.31] we have
*a R4
^K - KK
fa
R = which gives
cos0;
^a #
R a = - 2ad
sin
iP
= - 2cm
[100.41]
- R a ip = 2a
(a cos m
a((o
6 e smcf>)
E
hence
(acos0 +
de
smd>)
)a
2a6 a
<t>
R*
~ R a ^ = 2aa 2 [K2 - a 2 )
2d e
2
]
so that
(K^-W^da.
= 2u*[(u?
-a )-26
2
2
e ]
j^ > j- <
Since the term
6
2
if
uj
> a
+ +
26 e
;i00.42]
if
a> e
< a
26 e
it can be neglected,
so
da
>
if
u,
> a
[100.43]
-77
dt
<
if
u)
< a
320
Let us
t2
2
(o>,
2\2
)
4<$V]
be the curve
Furthermore, let a =
{et)
{a)
= a.
(a)
the
F
,
(a)
creasing frequency a.
AB will
be tra-
At the point B, however, this stable zone ends and the amplitude
a
continuously increasing
At the
point D the stable region on this branch ends, so that the amplitude drops
through the same stages it traversed during the period when the frequency was
steadily increasing.
Thus, for example, if the frequency is decreased from
the value corresponding to the point L, the region D'H will be traversed in a
stable manner.
This region, as was just mentioned, was missed during the peIf the frequency continues
and a further change will occur along the branch H'E, and so on.
(ct)
321
in the energy input into the oscillating system supplied by the external peri-
x,
de
= 0,
to
d;
[100.44]
7T(l i
so that Equation
[100.30] gives
a = +
E
ma
r
u)Aa n )
= n;
.
[100.45]
<f>
for the
It is
0=0.
to e
noted that if
{a)
a,
co
(a
^ ao
0.032
when
a.
This
1
0.028
l 1
linear resonance.
Interesting examples
of jump phenomena have been
0.024
1
1
0.020
1
1/
<x
0.012
0.008
type
//
50
100
in
200
250
Frequency
Figure 100.2
322
-.
CHAPTER
XVII
We now propose
derived from the classical theory of Poincare, which was discussed in Chapter VIII.
Poincare'
s
and stability of solutions are treated in a relatively simple manner. Secondly, the description of the behavior of a system in terms of its characteristic
ized impedances and admittances used by Kryloff and Bogoliuboff is replaced here by the analytical method, which is probably a more familiar approach to
the subject.
It is noteworthy that the theory given in this chapter was found
102.
the form
x
nf{t,x,x)
[102.1
t, x, x)
t.
nf(x,x)
X sinnr
[102.2]
l/n.
tions having the same form as [102.2], but we shall not enter into these
generalizations here.
We shall consider the standard circuit shown in Figure 102.1 repre-
The exter-
and
been shown that, if the effects of the anode reaction and the grid current are
324
CL
dt *
+ CR
di
d7
i=
{<
+ c
=
dE
dt
)
[102.3]
is a non-linear funcia
.
where
ia
f(Vs
Vs
Vs
Mjr
where
dimensionless variables:
t<o
Figure 102.1
where I
=
-"o
L =
-'o
di
dr
di
u
,2
.
dt
,2
.
dr dt
_
'~~
dr n
i
i
''
d (di\
di
oj_
dt\dtl
JrVdti
dt
l?
(J
dE
dt
dE dr
dr dt
dE
dr n
[102.4
sin ut becomes
dE ^f dt
dE = Er,u u
cos
nr
cj
-r-f 1
dr
D + CR
oj
di
n dr
+
.
i a
n + CE,.co
cos
nr
becomes
CL^P, n dr
CRP n dr
R
I=
j+C&u
COS TIT
[102.5]
is in the neigh-
nu
325
CL
where
$
CO
CO'
[102.6]
CRco :co
n
CLRco 2 n
n Lco
2
co
Rn
Leo
u>Q
*>
CE.co
cos
Rn dl
If we divide this equation by
1
nr
(1
1
+ ),
noting that
+
1
we obtain
d I
2
Rn^ dj_
Leo
dr 2
Since /
=
-
dr
_J +
+
1
CE
/n (l
a>
cos
nr
[102J]
4s =/ tt^o where F
=
equation
MI
co/n,
n
CO
Vn
so that
'.
MLn
=
dl dr
#)
Setting
^
i?w
o
*/
/
[102.8]
dr 2
where
=
1
+ ^Adrl
28^- +
+ Q cos nr
Q =
Letting
CE
/ (1
0)
/ L(1
CLE (,co + )
Ec,n
Lw
^
'
dr.
F (j;)
[102.9]
one obtains
dr z
cos
nr
r,
F'(x)x
re
X sinnr
[102.10]
where A
= -Qn.
326
103.
quasi-
uf(x,x)
[103.1
with u =
0,
periodic solutions exist, but the phase trajectories are spirals winding onto
the closed trajectories, the limit cycles, which exist In the neighborhood of the generating solutions.
For Equation [102.2] the situation is similar, for when exists an infinity of such linear solutions of the form
x
fi
there
asinr
cost
+
1
s-
sin
nr
a(//)
[103.2]
and
b(/u)
which will yield periodic solutions for the non-linear case, that is, when
fi
but is small
If, when n
-*
0,
and
respectively,
that is,
a(M)u
+ o+
M^^
^6
[103-3]
for
/u
->
In
sinwr
[103-4]
is substituted into
juflz
\
An 5-sinWT, + l 1 n
+
1
nk
^t 7r
v
COSWT]
/
[103.5]
= =
cost
z sin
[103-6]
z sin
z cos
327
we find
V.
(Z
Z)C0ST
fit//
(u,V,T) COST
[103.7]
v
(z
+ z)sinr =
A
m//{u, v, r) sinr
where
ip{u,v,r)
= flu
sin
v cos
r +
g-sinwT,
wcost + vsinT +
2tt.
^-g-cosnrj
sin
v cos
r +
1
An
%i
-2
sin JIT
103,8]
= a
and
for
/j.
->
0.
b,
that
is,
U T=
"^
'
"T
= o
/?
[103.9]
where a and
/?
u T=0
+ Ujip{u,v, t)
cost
o?t
[103.10]
r
vT =
+ u
ip{u
v,
r) sin
rf-
v,
u
v
=
=
a
b
+ a + +
J3
fiC {T)
x
+ uaD^r) + uPE^t) +
{t) 2
(r)
[103.11]
2
fj
G 2 (r) +
3
fi
,
4
ft
,
Cj(t)
C 2 (t) =
ip{a,
b,r)
s\x\t
dr
[103-12]
and also
j(t) Z>,
dip
j Y~
r
7
~i
cosrrfr;
(t)
rf^
J dv
dip
J"
cost d-
[103.13]
sin
Z) 2
J |^
rdr;
(t)
dv
sinT q?t
328
dip
du = a = fi = 0.
and
dip
dv
u{0) =
and
v{2n)
{2tt)
=
[103.14]
C 2 (2tt) +
hence when
and
fi
+
b
J3E2 (2tt)
+ uG2 (2n) +
=
fi
is small and
which will satisfy Equations [103.14] and which will reduce to zero
.
when h =
fi
provided
C^tt) = jip(a,b,r)cosrdT =
0;
C 2 (2tt) =
ip(a,b,r) sinr
J
dr
[103-15]
b,
and
fi.
aD^rr) + fiE^n) + hG
{2tt)
=
[103.16]
aD
{27r) 2
fiE2 (2n)
juG 2 (2n)
/i.
=
if
fi
= D^n) D (2tt)
2
E^tt)
[103.17]
(2tt)
Hence the problem of determining the limit values of the coefficients a and
in Equation [103.2] when
fi
->
is solved by Equations
104.
r\
and
v = v
and v
are
and
and
we obtain
329
drj
= =
(/uip u
cos t)tj
(hi/> v
cost)$
[104.1]
dr
since the functions u
(/y^sinr)?/
(//^
sin
r)
and v
Equations [104.1]
rj
(t)
^(r) and
f 2 (0)
{r)
2 {r)
are two
conditions ^(O) =
1,
^(0)
arj^r)
Cfj^r)
and
?7
(0) = 0,
^(t
2tt)
(r
= =
6j? 2 (t);
^(t +
2 (r
2tt)
= a^(r) + = c^(t) +
b$ 2 (r)
[104.2]
t]
{t
2n)
drj 2 {r);
2tt)
d$ 2 (r)
?1
(2tt)
a;
^(2tt)
6;
77
(27r)
c;
2 (2tt)
[104.3]
rj
{r + 2n) =
^^(r)
The formation of
F(S)
- S
c
7i(2tt)
- 5
^(2tt)
rjJ27r)
U2tt) -
[104.4]
=
-
;io4.5]
with
p
= -
[^(27r)
fi
^ 2 (2tt)]
and
[^(2^)^(2^)
rf2
{2if)
iftr)\
[104.6]
The parameter
jt
^f-
dr
= 0,
so that
and
fi
= 0, ^p- =
v,(t)
e2 (r)
1;
r,
(r)
/u
= Ut) =
[104J]
Hence, for
/j.
= 0,
p = -2 and q = +1
For
and
>
and
0.
characteristic exponents h l and h 2 (see Section 27) are negative, which implies that
\e
<
e
|
<
1,
2;
+ p +
>
[104.8]
+ p +
(s l
- l)(S 2 -
1);
= -
(S 1
+ S2 )
[104.9]
330
fi
= 0, p = -2 and q =
+~\
as was shown.
Hence, the conditions of stability [104.8] can be satisfied for small posi-
tive values of
/j.
with respect to
fi
= 0.
77
and
in Equa-
tions [104.1] by
(27r) 1
rj
and
f2
and 2n.
2w
This gives
?7
/7
(0)
77j
(27r)
//J^T/jCOsr dr
+ ^J^^cosrrfr
[104.10]
2>r
2* u
r]
1
^(27r)
- ^(0) =
^(2tt)
mj
i^
sin
t dr + /yj^^sinrrfr
fi,
one obtains
\*
j^p
-^ cost
+ ^-^cosrjdr
[104.11
]
j(0J7i
siirr
+ I^smitHt +
^K^-^sinr +
-^
sin
r)dr
//
:<o
=
-> =
J
M cos
rrfr
= D
(2n-);
J // =
{ip
smrdr = D 2 {2n)
[104.12]
=
J
/i
^ cos r dr
' (2n-); 1
ja =
^v
sin
rfr
= E 2 (2n)
_ =-[A^)
r
d(p
+
d/u
9)
//
*.<2*)];
==
0;
[104.13]
d Qo
2
+
2
y)
.
Z) 1 (2jtt)
u =
> (27r) 2
tf,u
(2tt)
(2tt)
D^2tt)
+ E 2 {2tt) <
Dftn)
and
Ep.Tr)
D E
{2tt)
>
(2n)
[104.14]
Conditions [104..14-] have been formulated by Mandelstam and Papalexi (8). developed by Professor W. Hurevicz.
331
105-
\ = M.) = i % + a[V +
s
a2
a3
V + atf +
a4
'
a 5'C
[105-1
a5
'
=0, whereas
Vs
,
for a
We have also
seen that the important terms are those containing the odd powers of
but
Using the
V V
dl
dr
Ia
= f {x) =
x
Ia
axx
a2 x
a3 x
a4 x
a5 x
[105.2]
Using Expression [102.8], one obtains the following expression for fif(x,x) in
[102.2]:
Mf(x,i)
YT-f
+ 2a
2
([
ai
2e ^
2
>
+
3
3a 3 x
+ 4a
5a 5 x ]x
x)
[105.3]
+
;
a1
',
20(1
a2
$)
3og
;
A.
4q
4
;
Ar
5a 5
z
a2
a2
a2
= (k+ 2x + A 3 x
+ A
+ A
x )x
-^-x a2
[105-4]
<f>)
--
sin2r
[1
05 5
In this section we shall consider systems having soft self -excitation, that
is,
systems in which
b,
A4
= 0.
i//{a,b,r)
cost dr
and
\\p{a,b,T) sinr
o
dr
Since the function &[a,b,T) is, by definition, /(x,x), in which x has the
332
2*
2*
ip{a,
I
b,r) cost
dr =
(k
+ 2x + Ax
2
)
x cost
dr +
a2 a2
2
\
x cost
dr
[105-6]
2n
2.1T
ip{a,
(I
6,t)
sin
r dr
(k
+ 2x + A~x 2
x sin
t dr +
2jt
x
J
sin
rfr
Carrying out this substitution and the Integrations, one finally obtains
k
4Ma 4
V
+
+
+
+
2V
Jl-(
a(
+ )
a,/
[105-7]
(e
.
^) __
X
of the prin-
*' =
and the phase
a'
b'
= -
3 L
105.8]
9
t
^
d>
tan
=
a
tan
a
2
[105.9]
such that
>
0.
qX
is apparent that this condition is fulfilled if the second term on the right
qX
2
.
Hence, if
A3
and
If, how-
ever, Ao <
and k
>
0,
the condition
>
of the radical.
k,
and
>
0,
principal solution exists in practice. Its existence implies that the oscillation be stable, which requires that the conditions [104.14] be satisfied.
In
D^t)
{r)
{t)
and
EAt)
du
where
x
df dx dx du
du df dx dx du
and
dip
dv
df dx
dv
.
df_
dx_
[105.10]
dx
dv
dx
dv
sin
v cos
r -
sin
2r
cos
v sin
r ^o
cos
2r
Hence
dx =
du
sin
dx =
du
cos
dx =
dv
cost;
dx dv
sin
333
This gives
(2tt)
sin
t cost dr +
-=pr cos
r dr
EA2tt)
cos
rdr +
5/
sin
r cost dr
[105.11]
, (Z7r)
=
J
-r
OI
sin
t dr +
J
-r
ox
sin
t cost
<xt
E, ,(2rr)
= - j df dx
2*
2ir
sin
cos
t dr +
a/
5 a;
Sin
(XT
a/
<9x
2(1
+ A~x)x +
^V =
Jfc
2a;
+ A.x 2
[105.12]
The criteria of stability [104.1 4] can be applied now both to the principal
= 0.
Carrying
pressions [105-11] and [105.12] for both the principal and heteroperiodic
solutions, one finds that the conditions for stability, and hence for the
,,2 X +
c
<
Aj
0;
A,
/,
^(Z +
2Ao
9
-1
>
[105.13]
[h
+ AS
a;
^>0
>
H
J
105-14]
if-
<
0;
Tl/f
[105.15]
A <
3
becomes
k
X = +
Since
k =
/*
-1
+ A3
[105.16]
18
a i z 2e
the
^k
energy input from the electron tube outweighs the dissipation of energy in the
circuit, whereas when k <
the dissipation of energy exceeds the input.
a
From
<
* 0). if x
= 0,
= j-pr, which
coincides with the equation previously obtained from the theory of Poincare
I'si-'K-T)
On the other hand, from [105-15] one finds
2
[105 17]
-
$
same time
.
>
(*
lA
^-^
It is thus seen that one solution appears at the point where the other disap-
pears and vice versa, so that there is no interval in which both exist at the
--(*-W
"2
'- 5 , 91
where
ax
Un
20
20|
(I/O
2$l
wo
one can obtain two values ' and f" determining the limits of stability. These limits determine the zone of discrepancy A$ = " - ' between cj and nu within
which
A)
=
a
r
2
*'
2a 2 lAf(l
9
V
,
k
l
A3 s
-^)
36
[105.20]
Considering A$ as
function of X = A2 3
-
*o^3
'
2k
A3
[105-21]
exists.
given by the equality sign in [105-21], the real values for A$ appear and Af
increases up to a maximum value
4
fr"
w-
2a,
~- 2k A *
[105,22]
335
-y =
For a further increase of X0!
Xn
9
,
2(1
k A.s ) 3
[105-23]
Ail
X
fCfi f\.o
[105.24]
Summing up, one can say that the external periodic excitation having
a frequency
oj
oscillating circuit is tuned is capable of producing oscillations in the circuit provided the coefficient
Z
X
2
0J
2
sponding to values of X
_2_
in the interval
A3 -
Vl
2k Q A 3
i
\
-f
g -p.[l :
A +
3
Vl
- 2k Q A 3
106.
bility of the principal solution were established without specifying the sign
of the quantity
ax
20(1
a2
f)
u
291
a2
As has been mentioned, this quantity characterizes the stability of the system in the neighborhood of equilibrium, when the effect of the terms containing
the small quantity x (see Equation [105-3]) is negligible.
>
it
<
0)
or unstable
>
0)
We call such
sys-
differing from
w
If
the principal
.
oscillation will suddenly appear and will have exactly the frequency u/2
the parameter
|
is varied,
3^6
p
~
2^
<o
can
o>
.
or
is varied,
cj,
variations of
principal oscillation exists its frequency is always 1/2 (or l/w for subharmonic resonance of the n
order) of the externally applied frequency u, but
,
<j
and w,
,
simultaneously.
as
more generally, of the order l/w) presents features radically different from
those of ordinary linear resonance.
It is sufficient to investigate the be-
of this solu-
co
u as
is customary for
line'ar systems.
of the
Figure 106.1
ferent when
>
0.
107.
A4 *
and
A *
5
0.
337
2 r
4
t(
(
5
27
2
5X
2
4 *])'
-^ 3
z
--4iJ
a
4
\
w
3
V
,
+
6 9
2&
)
/.
[107.1]
k
(X
+ -9-) +
2Xq
^\ x
4
v 4 + ^0
2
Jl
5X'
\a
^^M*
3
\
+ " + 6 A
26
j '.
of asymmetry is very
that the square of the amplitude of the principal solution is much larger than
the forcing term, that is,
g9
the expression
4^X + 8
4
^X
4
[107.2]
Applying the same procedure as that followed in Section 105, one finds the
inequalities
(A R X
2
A,)
]/W~&
<<>
^
2
x2 + A
A+^-&)>
A X2
5
M07.3]
and the
A <
3
X
2
X =
2
-^ A
5
'
'
[107.4]
4>
=
8fe
and
= 0)
X' =
The condition
A,
A,
[107-5]
A X
5
A <
3
and
are
<
0,
<
0;
2.
>
0,
A3 <
0;
<
0,
A 3 >0
X =
A3
A,
8k
or
X' =
8k
[107.6]
538
A;
>
0,
These two cases therefore characterize soft self-excitation in that the amplitude increases with k beginning when k = 0.
It is noted that,
>
and
electron tube, exhibits an inflection point for a certain value of the amplitude (see Section 51)In this case
X = + A
,+
5
W(t)
tiOT.7]
iS[
It is observed that in this case one can also have k < 0, which means that a
In order that no
e& > w
that
\k\
>
1107 9)
-
fr
Thus in the interval [107-91 This implies
A3
order one-half.
[107.10]
Hence there
intervals do not overlap, so that the oscillation of one type appears at the
point where that of the other type disappears.
quantity
(-4M
curve
element.
Hence, if
($)
X
2
is considered as a function of ,
interval in which
339
value (-4/-
Section
51
This is illustrated by
,
Figure 107.1
applied amplitude X
For
remains constant
<
and for
||
sufficiently large,
and only
= x
Figure 107-1
see Point
in Figure 107-1.
With
further increase of
through
a
the amplitude
X
=
For
^"
suddenly disappear.
= x"
,
If, however,
and decreases them gradually, the subharmonic oscillation will start at the
point
= fx
'
.
creasing
at D;
for decreasing
ner.
The sudden jump during both the appearance and disappearance of the
(-4^-)-
108.
r +
b cos
sin
3r
[108.1]
Proceeding in the manner explained in Section 105, one obtains the following
t(*
+
= a
{a'
2^
^pA 3
32
[108.2]
k
A '* 2
&
2
-i
= + lab
32
X'
wx + i)
X=
0,
+ r
.2.2
^0 -^3
32 2
v X
we have
{
/
(*
I
32
^0^3
32
\ t
/J
a ?2
X'
108.3
j>ho
If we let
k
X' +
32/
Y =
Ad
A-.
7\
S 2
32
16
32 2
16
32
108.4]
It is seen that the quantity under the radical can be positive only if
A;
>
0,
externally applied frequency can exist only if the system is self -excited.
109.
EXPERIMENTAL RESULTS
The preceding theoretical considerations can be verified by the fol-
mately
self-excitation, that is, self -excitation starts from zero at a critical value
of the coupling k if one gradually changes the feed-back coupling.
When the
13
critical condition is thus established, the coupling is decreased below that critical value
underexcited.
(
k < 0).
is now introduced,
their appearance.
If the discrepancy
As soon
';
'
f rl
= x
in the Interval
A^
and disappears at
$[',
as shown
in Figure 109.1.
of the ampli-
CHAPTER XVHI
ENTRAINMENT OF FREQUENCY
110.
INTRODUCTORY REMARKS
If a periodic electromotive force of frequency
is applied to an
cj
frequencies decreases, the pitch of the sound decreases, and from linear theory one may expect that the beat frequency should decrease indefinitely as
\io
>
0.
tor frequency
frequency
10
within
entrainment of frequency
the difference
Aoj is
occurs is called the band or the zone of entrainment. Figure 110.1 represents
\u cj
\
u>\
the interval
- u>
\
lines.
The phenomenon of entrainment of frequency is a manifestation of the non-linearity of the system and cannot be accounted for by linear theory.
This effect was apparently recognized long ago, but its theory was not developed until recently.
Thus, for example, Van der Pol, who developed the the-
ory of the phenomenon (9), observes that "the synchronous timekeeping of two
"
oscillations.
In recent
Figure 110.1
3^2
but it is pref-
erable to treat this matter in more detail, starting from the discussion of
Van der Pol and concluding the analysis with the topological method of Andronow and Witt.
Electron-tube circuits, as usual, offer a simple way of obtaining
the differential equations from which conclusions regarding the phenomenon of
present
As far as is known, no special studies of mechanical entrainment
have been made so far, but the following example is worth mentioning.
If one
envelope suddenly becomes a straight line parallel to the motion of the re-
cording paper, and beyond this point no further beats are observed.
The non-
linearity of the torque in this case is generally due to the kinematics of the
in.
which
cj 1
inserted
irji
Lj-+Ri
+ jjfidt - Mjf- = EQ
o
smu) x t
[111,1]
ia
considered as
ia
f(e g )
= Seg (l - 3^)
[111.2]
3^3
<
Figure
where
IT
.1
and S
K
we obtain
j idt
CV.
'
a =
MS
LC
R
''
= I
3
MA
LC
;
B -
'
- LC
av +
,
yv
-3
+
,
2
6l>
Bui
sint
[m -3]
It is noted that the left side of this equation is of the same general type
as Equation [44.1
].
Van der Pol assumes as the solution of this equation the expression
v
ftjsinwji
b 2 cos
u^
[111-4]
where
and
b2
Substituting
this expression into [111.3], equating like coefficients, and neglecting second derivatives, we obtain
2b\
zb 2
- abJl
[m.5]
2b 2
zb 1
ccb 2 il
g-]
= - Bui
where
z
2( 0)n
uj
and
6j
b2
[ni.6]
It is apparent that, if
b2
3>44
oj 1
In the first
In the second
o,
and
sin
o> 1 t
0,
we
shall follow the presentation of Andronow and Witt (14) which will permit es-
tablishing a more definite connection with the representation of the phenomenon in the phase plane.
The results which will be so obtained coincide with
f -PCM,,);
if
b
x
f.
b
1
= <3<M^
and
b2
are taken as the variables of the phase plane; thus the condi= constant and
b2
= constant, reduce to
Q{b lt
b2)
= 0.
precisely the condition for the heteroperlodic state with a single frequency
u)
l
.
and
b2
quantities.
and
112.
*-^; a
Equations
y
[11
= ^;
a
= ^;
a
A =
-^; a
a
^
2
[112.1]
1.5] become
dx
ch7
x (1
2
)
ay
[112.2]
2 ')
%L =
ax
2/(1
+ A
dy dx
A +
ax + y{\ r ay + x(l - r 2 )
[112.3]
3^5
If we let
Xq
f.
yQ
fj
[112.4]
UT
J
P($,rj)
= m$ =
4-
nr]
+ terms with
2
,
rj
[112.5]
-p-
dr
Q(S,V)
Pi + QV
,
+ terms
)
with
2
,
rj
- (m +
q)S
+ (mq -
np)
[112.6]
For a stable singular point it is necessary to ascertain first that the singularity is not a saddle point, which implies that {mq
If this necessary condition is fulfilled,
-
np)
should be positive.
the roots should be negative; thus one has either a stable nodal point when
jugate complex.
since both
and
tend
* <.
and
77,
The quantities
bx
and
b2
are also periodic, which means that the Van der Pol solution [111.4] for
oscillations.
We know from the theorem of Bendixson, Chapter IV, that any non-
closed trajectory which neither goes to infinity nor approaches the singular
points winds itself on a limit cycle.
This limit cycle Is stable for
t
*
+00
* -<.
cycles are of interest; we know from Section 25 that in the interior of such cycles there exist, generally, 2n +
1
always +1.
The "coordinates"
{x ,y
equations
x
o= ~
ap
^ =
-^V^
p(l
p)
[112-7]
where p =
346
a p
p(l
= A
[112.8]
third degree which gives the "amplitudes" Yp = r of the singular points for any a = 2(<j - uj/a. The quantity A is the parameter of the family of the
curves [112.8].
nature of the singular points in the (p, a) -plane depends on the nature of the
roots of Equation [112.6].
113.
in Figure 113-1.
the curve
and
M/;
= 0.1
For an increasing
the branch
M
2
M-l
is further
shown for
It is noted that
M'.;
=0.1
Figure 113-1
the curves
of the family exist only above the a-axis and are symmetrical
If one substitutes Expression [112.4] into Equa-
tions [112.2], one obtains Equations [112.5] with the following values of the
=
dt
drj
^[(1
p)
2x
2
]
rj[- (a
+ 2x -
)]
+ terms
in
r)\
[113-1]
[a
Tt
2x
?/
T)
(1
p)
2?/
+ terms
2
,
in
rj
3^7
where x
and y
2(1
- 2p)S +
[(1
p)(l
3p)
2
]
[113.2]
"can
various types as was shown in Chapter III. These zones, when drawn in the same
(p.a)-plane as Figure 113-1 in which the curves [112.8] representing the loci
of singular points are drawn, will indicate the nature of singularities in
that plane
inequality
(1
/>)(1
3p)
<
[113-3]
for, with this condition, the roots are real and of opposite sign.
(1
The curve
p)(l
3p)
is an ellipse E with its center situated on the p-axis, and Condition [113-3]
means that the region of saddle points is situated inside this ellipse.
2
.
The
quantity under the radical sign in the expression for the roots S 1 and S 2 of
2 [113-2] is p
and
which bisect the first and the second quadrants represent the divides between
the real roots and the complex ones.
at p = 1/2.
Inside the angle BOBj formed by these lines lies the zone of nodal
The area in-
points and saddle points; outside it, the zone of focal points.
side the ellipse, as was shown, is the zone of distribution of saddle points.
The condition for negative real parts and hence for stability is
clearly
(1
2p) <
Hence the line PP' of the equation p = l/2 is the divide separating the roots
with negative real parts (stable singularities) from those with positive real
parts (unstable singularities).
The former lie above that line; the latter,
below it.
\T
p
(1
"
p)2
For a =
- 2p 2 +
- A2 =
[113.^]
A2
A2
> 4/27,
348
there exists one real root, and two conjugate complex roots which are of no
interest here.
For
< a2 < a
2
x
,
where a
p)(1
3p) + a
= 0, Equa-
tion [112.8] has three roots of which only one is stable as can be seen from
a!
< a
< + < a
2
<,
Hence,
< a
since
ficients
as
t
&!
and
-> oo.
stable singularity exists, we can assert by the Bendixson theorem that a limit
o
nal frequency u x
Hence, whenever a
> a
2
,
of a single unstable singularity, the solution [111.4] of Van der Pol has
and
and
o)
If, however,
< ax
2
,
case, and the stable singularity gives rise to a stationary heteroperiodic os-
to
family of curves [112.8] for different values of the parameters with superimposed regions of distribution of the various singular points, as shown in Figure 113-1. A topological analysis of this kind was carried out by Gaponow (15) on the basis of the general considerations of Chapter IV, where singularities
conditions.
of this analysis,
Figure 113.2
3^9
Figure 113-3
Thus, for in-
three singular points, namely, a stable nodal point, a saddle point, and a
stable focal point. The flow of trajectories for this situation is shown in
Figure 113.3-
K formed in
S.
SN issuing from
At
large distance from the singularities the trajectories are inwardly directed
as shown.
Depending on the form of the separatrix, the trajectories may apTheir approach to the nodal
point will be aperiodic from a definite direction; their approach to the focal
point will be in the manner of a spiral, which indicates an oscillatory damped
motion.
Figure 113.4.
The separatrix forms a closed loop with the unstable focal The trajectories arriving from distant points of the
singularity of
higher order.
Here
This
550
When en-
becomes
straight line.
114.
Unstable Foca
Point
During entrainment the heteroFigure 113.4 periodic and the autoperiodic oscillations
become "locked," and the former imposes
ties
and
b2
b 1 sinu) l
b2
cos(J 1
Vb 1
b2
si
nw i^
cos
coswjisin^j
[114.1
]
= 6sin(<V +
where
bi
<f>)
A
and where
+ K
cos0;
=
Vbl
sin
\/bl
bi
<f>
One has
Mo.
<j>
-
b\
(1
r )a
2( u.
[114.2]
tjj)
where a and p are the coordinates of the stable singular point In the (p,a)
plane.
The amplitude
6
of the oscillation is
'
where k =
cies
u)
^Va
x
(1
2
)
- oj
lf?
[114.4]
[54.5].
is affected by the
entrainment factor
351
115-
More spe-
effect was made by K. Theodorchik and E. Chaikin (17) at the sugtion of Mandelstam and Papalexi.
Figure 115.1
and
A2
The mechanical system
A^Ag
is described by a
cj.
The
co
on the limit
- to
and
to.
If the
value of this difference is decreased, one finds that both frequencies coalesce into a single frequency
to
is thus entrained by
top
to
352
and determining
1,1
6.
ization of frequencies.
an electric motor with a high degree of accuracy can be solved by synchronizing the motor's frequency with the standard frequency of a quartz oscillator,
obtained, the motor speed can be maintained with the same accuracy.
particular example, the frequency fq of the quartz oscillator is generally many times greater than the rotational frequency fm of the motor. This difficulty is eliminated, however, by a frequency-demultiplication network, which
permits obtaining
a
The
problem then consists of "locking" the two frequencies fg /n and fm by some kind of entralnment phenomenon.
In the preceding sections of this chapter we have investigated the
of the tube which was approximated by retaining the cubic term in the repre-
ia
= f{
small zone of entrainment and also because it is difficult to modify the char-
in which the zone of entrainment is artificially made large by suitable circuits. In this manner one obtains a kind of artificially produced entrainment
which is more adequate for practical purposes than the simple type investigated by Van der Pol
As an illustration we shall investigate one such scheme suggested by
operates as an oscillator with frequency u^ having C : and L 1 as its oscillating circuit. The coefficient of inductance L x can be varied within certain
limits because the coil L^ is wound on an iron core whose state of magnetic
There are
353
Figure
two other coils:
11 6.1
K lf which
which merely maintains the oscillation with frequency u lf and the coil
as an amplifier.
K2
V 2 through
AB branch CD
branch
The network
lo 2
K also has
a second
is inductively transferred.
BACDis
closed on a bridge
formed by rectifying
BACD there are two induced voltages: E with frequency D induced in the AB-branch, and E 2 with frequency u 2 induced in the CD-branch. We shall consider the case when the difference u 2 - <o = Au is
In the circuit
1 1
small.
11 6. 2;
assumed to be fixed;
Aoj
in one
- io 1
Er
is the voltage
The frequency
L^i)
is a
i.
non-linear function of
It is apparent that, if
d<f>
w2
ux
oj
1
dt
= u2
= Au
[n6.l]
of the oscillator.
In the vector
Figure
11 6.
354
E2
as center.
The resultant
Er
ini
Er
reduced.
un-
'|f
is stable.
= " "
2
"i
[116.2]
(*>i
provided
<f>
and n.
of the circuits,
(O <
u> 1
</>
< n)
for which
[11
by the ex-
and
cj 2
E2
rotates
constants of the circuits the magnetic saturation control may not be suffi(v
1
so as to "lock" it in synchronism
co 2
and
If the parameter
n
be impossible to obtain the synchronization of the two frequencies
1
were constant, that is, in the absence of the saturable Iron core, it would
u and
cj 2
CHAPTER XIX
PARAMETRIC EXCITATION
117.
parameter
Reference (10) an old experiment of Melde (19) which he reproduces and analyzes.
In this experiment a stretched string is attached to a prong of a
appearance of transverse vibrations of the string with a frequency of f/2. Later, M. Brillouin (20) and H. Polncare (21) investigated a similar effect
in electric circuits.
of time.
experiment are
here with a frequency equal to one -half the external frequency of the tuning
fork.
excitation of electron-tube circuits can be traced to the fluctuating transconductance of the tube caused by the oscillatory process itself.
In all
electron-tube circuits the periodic variation of the parameter, the transconductance, appears as an explicit function of the physical quantity which
characterizes the process, for instance, the grid voltage, and depends only
implicitly on time.
Self -excitation of electron-tube circuits therefore beThe concept of autoparametric excitation
356
With heteroparametric excitation, however, the situation is different. Since one or several parameters of the system appear as explicit periodic functions of time,
iation of a parameter.
It must be noted that, although the theory of the Mathieu-Hill equa-
that the known types of Mathieu-Hill equations are linear equations and, as
such, possess unstable solutions increasing indefinitely in their unstable
region.
proaching a definite steady state, one should apply some kind of non-linear
differential equation with periodic coefficients.
Unfortunately, no theory
Section
24
punctured.
the voltage builds up to a finite value, and the generator functions in a sta-
ble manner.
118.
2p(t)z
q(t)z
.
[118.1]
xe-
Jpdt
[118.2]
Equation
[11
8.1] becomes
x
+ M{t)x =
[118.3]
where M(
t)
q -
p is a periodic function.
357
M(t):
1.
t)
o)
+ a
cos kt
M(t) Is
a Fourier series,
cos kt +
A2
cos 2kt + +
sin kt +
B2
sin 2kt +
in which case
equation.
The Mathieu
(u) Q
+ a
cos
t
kr)x
=
[11 8. 4]
{cj
+ a2
cos
Ox =
=
where
10k = u>
and ak = a
u2
2
ot
cos
Is periodic,
its
solutions are not necessarily periodic although under certain conditions they
may be periodic.
Since Equation
[11 8. 5]
1
is linear, one
a
f and f 2 forming
F = AJ + A
X
f2
(
[118.6]
t
where
A
+
and
A2
Moreover, since f1
+ 2n) and
1 (
f2
f2
2tx)
t)
and
(t)
A(t +
2tt)
af.it)
+ bf2 (t)
[118.7]
f2
From
[11 8. 6]
(t
2tt)
cf
{t)
+ df2
(t)
F{t +
2?r)
= A1 f
(i
2tt)
+ A 2 f2 (t +
2tt)
[118.8]
solution
aF(t)
such that
[118.9]
F{t +
2rr)
- A f
x
(t
2tt)
+ A f
z
(t
2tt)
0;
/;(0)
1;
/2 (0)
1;
/;(0)
[118.10]
/2 (0)
#
/,'(0)
/2'(0)
358
f2
is fundamental.
From Equations
=
6;
/ (2tt) =
2
d;
/*(2tt)
a;
/2'(2tt) =
[11 8. 11
2tt)
= A
{afx
[118.12]
[ 1 1
8 5
.
] >
clearly
f" + Mit)f =
x
0;
/2"+ M()/2 =
-
therefore, //V2
f2 "fi
tnat ls
/g'/i
constant
[1-18.13]
is
h
/;(0)/ (0) 2
-f^f.iO) =
f[{27T)f {27r) 2
- /2 '(27r)/ (27r)
1
[1l8.l4]
ad
be
[118.15]
[11 8. 11
A A
system [1l8.l6] is
a
2
{a
x
a)
+ A2c =
o)
+ AAd -
[118.16]
Thus, in view of [118.15], the condition for the non-trivial solution of the
(a
d)c
(ad
be)
(a
d)a
[118.17]
,,,,
il
j/(*)' -
Ai
-4
j/i-li+J)!
[1l8 l8
.
If we put ^-4
= cos 2tth,
a,
9
cos 277-^
;'sin27r^
J27TM
[11 8.
19]
If
**
cos
27r^
ls real; hence,
//
is imaginary; hence, a is real, and there is a root greater /V -L W ~ = 1 = 0, and hence a = 1; than one which indicates instability. If ju
Equation
[11
8.9].
> 1,
/i
Mathleu equation.
a
359
= y(a +
d)
[118.20]
119.
in the Interval
/'(0) = 0,
0(0) =
0,
0(0
= Csin + D cos
a2
6t;
h(t)
sin
eJf
Fcos
t
<
27r,
[119-1]
where y =
t
Vw 2 + a 2 and
obtain
y/u 2 -
= n, we
/(jr)
gin)
g '(n)
cosy7r
Csin<57r
+ Dcosdn
<?7r
(tt)
=
= =
y sin y7r
C<J cos
Dd
sin
<$7r
4>{tt)
/z
(7r)
or
[119-2]
y
0'(7r)
h'in)
or
cos y7r
= Edcosdn F6
sin
dn
C,
2?^
f(2n)
0'(2tt)
-2
0(2tt)
A'(2tt)
hin [1 9. 3]
1
w2 > a2 >
0,
-7-
+
y
[119-^]
and for
io
< a
cos2nju
cos
ny
cosh
my
[119-5]
where
77
= Vet 2
- cj 2
2
, )
{u>
-plane which
are the boundaries between the regions of unstable motion (shown in white in
This discussion is
the character of the stable and unstable regions for various values of the two
The unstable regions cover a larger area than the stable ones.
2.
Below the ^5-degree line in the first quadrant the motion is, in gen2 Here a < 2
eral, stable.
u>
360
Figure 119.1
!9
does not touch the zero line, and the coefficient of x in the Mathieu equation remains positive. Thus, without the ripple, one would always have stable
motion.
3.
Above and to the left of the ^5-degree line the motion is generally
Without the
ripple the motion is unstable in this region so that the ripple under certain
It is found that,
for a certain band of frequencies and for a certain amplitude of the motion
of the support in the vertical direction, the unstable pendulum exhibits
stability.
In what follows we shall be interested particularly in the unstable
parametric excitation.
361
120.
These regions of instability have not as yet been explored to any extent because the aim of previous analytical studies has been the establishment of
in which we are interested here, the unstable regions present greater interest.
already mentioned, is the fact that since the Mathieu-Hill equation is linear, there is no indication whatever as to how the gradually increasing os-
To determine this it
happens in the unstable region of solutions of this equation by the following argument of Mandelstam.
capacity which varies periodically between the two limits C max and C min
the capacitor have initially, that is, when
t
Let
= 0,
certain charge
q;
the
solutions of the Mathieu equation and those of the Hill equation with the
function
Mit)
(j
a
it
(cost
V
-^cos3
6
-^cos5
5
)
'
[120.1
representing
of Van der Pol and Strutt and consider abrupt variations of capacity from
C m ax
by
to ^min
AC
C max
C min
electrostatic, it is apparent that the impulsive work done during this sudden
decrease of capacity is
w>
AC
'
1120 2!
-
This amount of energy is thus added to the initial weak electrostatic energy
= 0.
Assume now
362
that one-quarter period later, when the capacitor is totally discharged and
the energy is entirely electromagnetic
[Li /2)
inal value of the capacity C max by giving the increment + AC to the capacity.
In so doing no work will be performed since the electrostatic energy is zero
at this instant.
the electrostatic and, hence, the total energy of the system has received an
is apparent that it is still present in the system which increment -^rl 2 has been assumed to be conservative. If one-quarter period later we repeat
<1
>
= 0,
ity is one-half the period of the free oscillatory phenomenon. The argument remains the same if, instead of capacity variations,
variations is exactly the same as for capacity variations, namely, the coefficient of the inductance is decreased (- AL) at the instants 0, T/2,
and
J>T/k
ever, there will correspond a diametrically opposite effect, that is, at the
J>T/k
121.
rods of electric locomotives (26) and was found useful by other investigators
(25).
as the Hill-Meissner equation; we will inquire further into the nature of its
solutions.
lar form of its periodic coefficient, the ripple, which permits a simple dis-
365
is to be understood, however,
unstable solutions of equations of both types has been attempted so far, and
the above assumption seems to be a plausible hypothesis convenient for a qual-
With these remarks in mind, we shall write the Hill-Meissner equation in the form
x
(a
)x
[121
.1 ]
(a
)x
0;
(a
)x
[121.2]
during each half period n of the ripple, with the understanding that the solutions have to be continuous on physical grounds although not necessarily analytic at the points at which the changes from (a 2 +
versa, occur.
b
2
)
to (a
),
or vice
>
citation discussed in the preceding section and write the differential equation of the non-dissipative circuit in the form
L^
where L
is the inductance,
^=0
q
[1
21 .3]
C varies
The preceding
AC and C min
AC in
stepwise manner.
where y c = AC/C
becomes
If l/L
C = w o 2 an d
if we assume that y c
(1
yc )u
[121.5]
(1
yc )(J q
=0;
yc )u q
[121 .6]
t/u
-^
where a
2
d 2(l
2 + cgq =
,.
0;
q _2. j^+aJq^O
,
[121.7]
+ yc
=1 and 2 2 = i a
yc
364
by the integral curves, or phase trajectories, of Equations [121.7], and the dynamical
I.
Figure 121
.1
For y c =
and
a*
2
ce 2
the
If y c * 0,
and
r2
of concentric homothetic
1
The family
corresponding to a x2 >
has a
a*
r2
has a ratio
=1
ye
The family
AC
r2
C +
AC.
The origin
is the
r2
thus serve
Is*
For
certain initial conditions, say {q ,0), are given and the value of C is prescribed, for example, C = C n - AC, the process is depicted bj
example, if for
t
A corresponding
instant
lt
r2
(C= C
AC)
and
This representation of the solutions q(r) of Equations [121.7] by phase trajectories is a convenient way of ascertaining the various circum-
Let us start from a point A(q ,0), see Figure 121.2, after the capacity has
been reduced {C = C
AC).
AB
+ AC)
r2
is
followed.
CD
is of
365
the family
and so on.
After
corresponding to
q 2n
> g
which
Figure 121
.2
variations.
that, instead of injecting energy into the system by providing external impul-
sive work which will overcome the electrostatic forces, energy will be with-
drawn because the electrostatic forces will do the impulsive work and will
thus diminish the energy content of the system.
the trajectory AB'C'
in Figure 121.2.
,
If,
at B' decreased, at
Increased,
energy
In this example the trajectories are spirals made up of elliptic
arcs; these spirals have continuous tangents at every point, although there
at which
Meissner equation with which we are concerned here are piecewise analytic
curves, possessing continuous first derivatives but discontinuous second de-
In a more general
122.
DEPENDENCE OF HETERO PARAMETRIC EXCITATION ON FREQUENCY AND PHASE OF THE PARAMETER VARIATION
In the preceding section we studied a special case in which the dis-
continuous changes in the rectangular ripple occurred at the instants when the
representative point crossed the coordinate axes of the (x,x) -plane and the
frequency of the ripple was twice that of the circuit.
This case, which is
the one studied by the early investigators, is also the one most frequently
366
Figure 122
.1
=
cos
[122.1
sin
a,
where
axis; and
a*
,
+ y c as before.
and y 1 =
sin 0j
where r x
=
<f>
1 ,
cor-
responding to C
point N(r 2 ,0 2
and a new arc
)
AC will begin
at the point
M
C
AC to C
AC occurs
"
NP
of the family
will be traversed.
0i<
,
'"
>
It can
ke
rence procedure.
A in Figure
"N
2v
= =
a,
(
f. 2v
4
[122.2]
J 2 J
/2
'
<*1
N
where
TV +
a.
/l/s
J Zv +
' * '
[122.3]
4/4
2
Ai/
./(0,)
a2 +
a,
2
tan *i
,
+
,
tan v-
[122.4]
f.(-j>.)
/.(0.
77)
= /\(-0 +
2
7T)
[122.5]
The only case of practical interest is that in which all intervals are equal
If
(2tt)
N=
4 will be
367
Figure 122.2
<t>
>
+ fl
+ ~T> 03 =
2
iv"
N=
4 and
system.
It is to be noted that in a more detailed investigation of this
phenomenon one has to take into account the fact that the equiphase intervals
are not equltlme intervals, that is, intervals of equal time, because of the
similar to the functions fi [4> i ) just introduced, we certain functions g {<l> shall not elaborate on this subject here but will investigate the principal
tant practical cases, when the changes of capacity occur on the coordinate
axes of the phase diagram, both types of intervals coincide; when they do
not, the introduction of equitime intervals, while complicating the calcula-
tions somewhat, does not change the qualitative aspect of the phenomenon of
heteroparametric excitation.
It is convenient to consider the following four groups of numbers N.
1.
v
N
= 1,
= kv;
2, 3,
2.
N=
{2v + 1)2;
q
3.
= 2v +
4.
= p/q, where
and p and
groups
This
First group:
N=
4, 8,
12,
N=
4,
{<f> i
).
4
<f> x
368
of the functions
B
fi{<t>i)
it is apparent that
*
1
\f
^^V
\A
M)
"
Jz
4
[122.6]
i/s
^Hf
^/
D
f2 f J
\^
that is,
f
Figure 122.3
If we substitute for f 1 and
Equation [122. k], it is easy to discuss the conditions for self -excitation of
the heteroparametric oscillations.
in which
self -excitation occurs are located within the shaded sectors shown in Figure
122.3.
In the non-shaded sectors, self -excitation does not occur.
O
The lines
and
<f>
1
4
+
-
y
7
a,
10,
[122.7]
Second group:
N=
6,
14,
18,
N=
By Equation [122.2]
[122.8]
J J f2 f J 6 f
A
/p
<f>
=
.
f3 = /6
fx = /4> and f2
f5
which
shows that a 6 = a
Third group:
N=
3,
5,
1,
9,
Since
= 3.
a,
d "2
/ /
J 2
From the form of the subdivisions and from Equation [122.5], f3 = f the condition for self -excitation will be
Hence
^
Of
[122.9]
/.
^69
By introducing for
in this case
4.
f1
carrying out the calculations, one finds that no self -excitation is possible
Fourth group:
N=
p/q
most fractions p/q are either of no Interest or fall within the scope of the The only cases in which heteroparametric exgroups previously investigated.
N=
4/3
4/5
4/7
* ,
"'
8/3, 8/5,
A<f>
- 37r/2,
5^/2, 7^/2,
123.
energy communicated by the variation of a parameter must, on the average, exceed the energy dissipated by the system.
to an additional condition.
+ R
[123.1]
C is
modulated by a ripple
If we
AC
so that
and
X.)
L C
Equation [123.1] becomes
q
(1
=
yc )
*(!
+ 2p
(1
y)q
[123.2]
= Qe-fp" dt
[123.3]
Q + ^;
(1
+ 6)Q =
t
[123.4]
u)
x
to
*)Q
r =
t,
one obtains
[123.5]
^4
(1
nate sequence of the following two equations replacing each other at each
370
J
d
o,
^al
,'
[123.6]
where
2 2
-i+^-i+^r..
(j
1
i
[123J]
-
4C. Equa-
tions [123.6] have the same form as Equations [121.7], so that the conclusior
reached for those equations are applicable here except that Equations [123.6]
contain the dependent variable Q whereas Equations [121.7] contain the variable
q
;
The trajectories
decreases monotonically.
q,-r~)
Q = Q
where p 1 ^ p
.
+fpidt
[123.8]
divergent spirals with the absolute value of the average negative decrement greater than, or at least equal to, the positive decrement p = R /2L Physically this means precisely the of the dissipative circuit {R ,L ,C
\p 1
\
condition stated at the beginning of this section, namely, the energy injections into the circuit by the ripple t AC must, on the average, be greater
parameters [R
L ,C
the decrement is p
R /2L
S^L
= e ~^n
[123.91
(N=
+p
4,
=
which defines the increment
Pi
1
=
~
log
2tt
Of "2
[123.10]
371
Expressing the condition for parametric excitation, namely, p a 2 and p their values, one gets stituting for a 2
, ,
^ pQ
and sub-
to
2
e"<>
where
fi
""V^o^o*
a* > a
2 2
,
one
This gives
2 9 *
ye
[123.12]
^ yc ^ y" c
[123.13]
When
= 0,
= 0, we have n
and p
YL
cj
= It is
/C
obtained by
For a
ation of capacity and inductance arises from the fact that both these factors
enter symmetrically into the expression
cj
= l/L
(through
2
(jj
R /2L
The
inductance L
decrement.
appears both in the expressions for the frequency and for the
Hence, a priori, one may expect different results in both cases.
The
only difference lies in the fact that for the inductance ripple instead of an
interval in which the index of modulation y c must remain in order to obtain
self -excitation, the condition for self -excitation is given by an Inequality.
It is also noteworthy that in the preceding discussions it was assumed that
y
y.
cated but the qualitative picture of the phenomenon remains substantially the
same
372
124.
R{t)x
C(t)x
[124.1
C has
As to the parameter R, it
must be noted that only those parametric variations which extend alternately
into the region of negative resistance are capable of producing parametric
excitation.
physical grounds negative resistance means the supply of energy from an outside source.
rangement consisted of a series of coils located on the periphery of a stationary disk; the inductance of these coils was varied by the periodic passage
of teeth and slots on a rotating disk parallel to the stationary disk.
The
frequency of the Inductance variation thus obtained was of the order of 2000
cycles per second.
In a circuit of this kind devoid of any source of energy other than
the kinetic energy of the wheel, electrical oscillations of exactly half the
The non-linearity by
a
saturated-
core reactor; an auxiliary d-c winding served the purpose of displacing the
desired value.
Similar experiments were produced with a periodically varying capacity.
by neon tubes which permitted maintaining the voltage at about 600 to 700 volts.
equation, and the voltage rises rapidly to between 2000 and 3000 volts and the
insulation Is punctured.
373
circuit so as to deviate from the condition of exact fractional-order resonance, the amplitudes of the parametrlcally excited oscillations decrease, all
other conditions being equal, until finally the excitation suddenly disappears
at a certain critical threshold as was analyzed theoretically in Section 123.
125.
shall now show that the differential equation [102.2] of subharmonic external
resonance which we have discussed in Chapter XVII on the basis of the theory
of Polncare can be reduced to an equation of the Mathieu-Hill type.
For that
turbations a and
/?,
Equation
03 -9
Putting
x
x (r)
[125-1
and substituting it into Equation [102.2], we get, after expanding f[x,x) in a Taylor series,
x
/i/(x ,x
sin
nr +
pp.f {x Q ,x Q ) x
+ pufx
(x ,x
[125.2]
Since x
=
)
pftfXg (x ,X
yd///;
o
(x ,X
[125.3]
where fx (x
and fx (x ,x
and x
and, hence,
ze
[125
"
1- Hh
+ ^T-(fi) ~ x 2 dr
o
%-ifi x
2
z
o
[125-5]
Mk
e
%(r) +
[125.6]
2
where
<j>
and
<f>
and 119 if
"
374
126.
excitation because in
well as the steady state of non-linear oscillations can be discussed more con-
veniently on the basis of the theory of Polncare than by treating it as autoparametric excitation.
In some special problems, however, the concept
oi
autoparametric excitation of oscillations may be convenient. In this section we propose to apply this method to
motion of
26
gravity.
Figure 126.1
gation
of the spring.
degrees of freedom.
T =
and its potential energy is
m
2
(r
<{>
[126.1]
V =
where
1
(r
mgril
[126.2]
to the elasticity of
<?
cos 0.
V corresponds
mg
k
1=
and a new variable
z
+ -
[126.3]
[126.4]
T =
ml'
(z
+
,
/2
<f>'
+
,
2zd>
[126.5]
mV
k 9 {- z + 0 + \m m
2 2
,
_ ~rz
I
[126.6]
375
where
and
-z
MV 2i0 +
2 )
=
[126.7]
+ -0 + (20 +
It is seen that,
2z0")
de-
gree of freedom with frequency u z = Vk/m, and the second equation will give a
= yfg /I
These terms represent a non-linear coupling between the two degrees of freedom; we will now investigate this condition.
It is to be noted that In the general case when
oj z
v
Q^<t>
the non-
o> 2
2w
so that
0=0
z n cos U)
[126.8]
22 o cosw z )0
(2(J
z
sinw<)0
+ u*(l +
cos
u) t)</>
z
[126. 9]
which is
w2
2w
oscillation in the
degree of freedom.
build up.
This curious phenomenon of autoparametric self -excitation was actu-
degree of freedom.
z
degree
376
One could, of course, start from the ing the pendulum from an angle
</>
<f>
for
tion
<j>
cos u^t
[126.10]
Substituting this expression into the first equation [126.7], one has
2
cj
= ^P-{\ 4
3cos2w.)
[126.11]
cj z
Hence, in the
self in the fact that the energy appearing Initially in one degree of freedom
is eventually transferred into the other degree of freedom.
There exists,
fests itself in classical linear resonance with which the z-osclllation builds
up.
metric excitation is still present in the form of the centrifugal force whose
degree of freedom.
It is noteworthy that in both cases the frequency with which the pa-
rameter varies is double that with which the self -excited oscillation occurs.
If one starts the oscillation in the
z
cause
u) z
2w
If, however,
degree of
the effect of the centrifugal force so that in both cases the condition of
recently by Sekerska (22), who passed an alternating current of 50-cyde frequency through a stretched metallic wire capable of oscillating laterally
with
frequency of 50 cycles.
377
autoparametric excitation, produces self-excitation of lateral oscillations with half the frequency of the parameter variation.
These phenomena occur not only for the ratio
is the frequency at which a parameter varies and
co r /co
= 2/1
where
(j r
excited autoparametric oscillation, but also for the ratios 2/2, 2/3,
Migulin (30) investigated, both theoretically and experimentally, these phenomena when this ratio has the value 2/3 and found that the resonance curves then resemble those obtained by Mandelstam and Papalexi, Chapter XVII, in
their studies of subharmonic resonance of the n
order.
As a matter of fact,
are closely related to each other and merely represent different aspects of
the same physical phenomenon.
378
REFERENCES
(1)
931-1937
"Mechanical Vibrations," by J.
1940, p. 403
P.
by B. Van der Pol, Philosophical Magazine, Vol. 3, Number 13, January 1927(10)
Philosophical
November 1922.
(14)
"Zur Theorie des Mitnehmens von Van der Pol" (On Van der Pol's
,
Theory of Entrainment)
Vol. 24, 1930.
flir
Elektrotechnik
379
(15)
935
USSR,
Poggendorf f
'
(20)
(On
Poincare, Eclairage
line"alres"
January- June
(26)
Mathleu and Related Functions), by M.J.O. Strutt, J. Springer, Berlin, Germany, 1932.
380
(28)
July
9^5
fiir
zwel
Number 2, 1937-
PART
IV
RELAXATION OSCILLATIONS*
INTRODUCTORY REMARKS
The term relaxation oscillations
,
127.
introduced by Van der Pol (1) + (2) and commonly used at present, generally designates self -excited oscillations
Ph.
maining on each level alternately for a relatively long time but passing from
one level to the other so rapidly that in the idealized representation the
Figure
27
W sufficient
to
S.
and at the instant when the moment due to the weight of the water becomes
greater than the moment due to the weight W, the system tumbles over against another stop S'.
The container then
W brings
S,
the
af-
Figure 127.1
The text of Part IV follows the presentation in "Theory of Oscillations," by Andronow and Chaikin, Moscow, (Russian), 1937. A complete bibliography on the subject of relaxation oscillations appears in this volume
*
382
S and S' respectively, and the representation of this system in the {6, t)-
plane appears as a periodic rectangular ripple with the length of its hori-
conductor, the neon lamp flashes and the capacitor is suddenly discharged,
whereupon the gaseous conduction ceases abruptly and the charging period begins anew. Here, again, there are two levels, the voltage V immediately befor
x
V2 immediately after the extinction of transition from V2 to V is gradual, but the inverse tran
x
sition from
V2
is quasi-discontinuous.
represented in the
(V,
for large values of the parameter n in the Van der Pol equation. Figure 37 -1c
shows quasi-discontinuous changes in the variable x{t) between the two region;
in which it changes but little.
ions of very large curvature, such as the curves shown in Figures 37 -2c and
37 -3c. By idealizing these very rapid changes as discontinuous changes, closec
discontinuous stretches.
Available analytical methods are inadequate for a rigorous treatment
of these phenomena.
eter
fi
nf{x,x)
[127.1]
Is very small.
/u
is 10.
Attempts have been made to extend the analytical methods to oscillations in which
/j.
is large.
phase trajectories when n was very large. N. Levinson (11) extended the prooi of the existence of closed trajectories to cover oscillations in which n is
not small.
In a recent publication (12) J. A. Shohat has indicated a form of
fi
series expansion formally satisfying the Van der Pol equation when
is large.
These various attempts, however, did not result in any complete analytical
383
theory comparable to the one which has been studied in Part II in connection
with oscillations in which n is small. Moreover, as will appear below, not all known relaxation oscillations seem to belong to the group of equations [127.1] of which the Van der
Pol equation is a particular example. More specifically, it will be shown in
evolve
oscillations
of Part IV.
minal" conditions are correlated on the basis of certain additional information not contained in the differential equations themselves.
This permits
obtaining the correct overall effect of the impact without knowledge of its
details.
the theorems of momentum and kinetic energy; to this information is added, for
o2a
are composed.
'(a)
xl
perfectly elastic ball rolls without friction on a horizontal plane and strikes
X
'
elastic walls
WW
(b)
384
AB
and
CD
velocities x = v
it is discontinuous.
continuities in the two systems is, however, different. For the ball striking
the walls there exists a definite external actuation, the reaction of the con-
straint, the wall, applied to the dynamical system, the ball; this actuation
is properly "timed" by the distance 2a between the walls which determines the
Mw of
Mc
Mc = Mw
position from
to 6'
is not instantaneous,
the long periods of filling and evacuation of the container it may be consid-
varies still
shorter, which, in turn, makes the relative time intervals of filling and of
those
Impulse-excited oscillations do not require any particular additional information for their treatment, as will be seen in Chapter XXIV.
For
jvi
385
Section 130 provides a criterion sufficiently broad to cover all known types
of relaxation oscillations.
Finally, inasmuch as the representation of a rapidly changing process by a mathematical discontinuity is always an idealization, it becomes
we have noted that the change from the angle 9 to the angle
may be consid-
ered as quasi-discontinuous.
such as the
In both
employed.
It is useful
ALRB has
no ca-
In other words, we
L and resistance
Lnterval
{t
0,
+ 0).
1"
i
=: e
quency.
Figure 127.3
386
r)
describable by
dif-
ferential equation of the second order whose phase trajectories are spirals
converging toward
The closing or
certain transient.
<
It is noted that,
in parallel
+ Ri = E-
dV rC^~ + V = E dt
t
[127-2]
o;
where
for
and
= constant for
finds that
- |(i - .-*'
V = E(l It is seen that for
t
di
-
= E
L
~dl
[127.3]
"*>');
dV
dt
E
rC
i(t) and V(t)
are continuous but not analytic in the sense that their first derivatives
the variable i jp j? di and i o discontinuities j- and respective izations the functions ^r dt (di \ ly. If one takes the plane of the variables [^j> l c)> the process occurring
dV 4f dt
= f; L
'
Noting that i n = C-%? we| (4r) -'S*' dt \dt lo and state that under the assumed idea
fij
A whose coordinates
are
and
y-;
see Figure
~\2~(
A
ouj
AB
will eventually approach a damped oscillatory motion represented by a conver gent spiral.
di+
dt
tl
the switch is opened, another jump will occur, but this jump will generally not bring
s'
It
i\ the history of
Figure 127.4
387
representation
certain variables is possible only because we have introduced certain idealizations into the problem.
a.
-j-
di
and
ic
which
As for Assump-
tion c, it is clear that if one selected some other variables, for example,
i
and
V instead
of
and
di
dt
i c
formulating some
CHAPTER XX
FUNDAMENTALS OF THE DISCONTINUOUS THEORY OF RELAXATION OSCILLATIONS
128.
ferential equation of the second order with constant coefficients, for example,
ax
bx
+ kx =
[128.1
= R, and k = l/C;
in a mechanical one, a =
(mass),
is the coefficient
First, if
approaches zero.
For
6=0,
It is
thus seen that there is a definite difference between the qualitative aspect
of trajectories when
b
From
or
0.
->
0.
a =
Equation [128.1] becomes an equation of the first order and its solution
e"b
[128.2]
where x
is that constant.
x
E(y
= - \x
= - jx
[128.3]
It is seen that the coordinate x and the velocity x are not independent but
trajectories of Equation
v [
and
that the phase space of a differential equation of the first order is uni-
389
itself.
The solution of [128.1] is
x
= C
rit
e
x
+ C2 e V
[128.4]
where
C and C 2
x
ar
If the initial conditions
X
t
br
,
=
are given, one obtains
^2 r2
= 0, x = x
;
and x = x
= C + C2
x
^I r i
[128.5]
C2
_ =
Xr.
r2
"
and
ri
[128.6]
where
ri
>
2a
4a 2
Mb m
~
k_
b_
n28:ri
In this expression only one term is retained in the expansion of the square
This gives
b_
and
[128.8]
mate expressions for the roots r 1 and r 2 are substituted in Equation [128.4],
the approximate solution x {t) of [128.1] is in the form
x
x x {t)
ak "ty e
b'<
An
e
[128.9
is an approximate one
because the expansion of the square root has been limited to the first two
terms; this is justified by the assumed smallness of a.
the second order becomes the same as that of the equation of the first order
given by [128.2].
as that of Equation [128.1] when complete degeneration occurs, that is, when
a = 0, we will write it as
k
.
x it)
xn
[128.2]
<t>(a,t)
x x {a,t)
x(a,t)
= -
x ~r2-e
-r-x
[128.10]
390
This function represents the difference between the approximate solution a;^*
of [128.1] in the neighborhood of the point of degeneration, where a is very
~x[
The function
*- 0.
<
<
when a
(x
j +
i )e~*'
^i
e~*'
[128.11]
the function
x j- + x
and it is im-
ciently large
t,
value
number
<j>{a,t),
con
t
sidered as a function of
a,
<
<
when a
of
t
>
0,
For
x^t)
x
to the function
when a
>0
value of
+ x
such that
</>
x j- +
tion x 1 (t) of a quasi -degenerate system when a is very small and the corre-
approache
<
= 0;
<
when a
->
t)
>
0,
ax
+ +
bx
[128.12]
M
Mis
[128.13]
where
M is
The value of
determined by the
ax
bx Q
[128.14]
391
^L
+ Ce
a'
[128.15]
x(t)
If,
T-(l
b
a ')
[128.16
in
solution is
x,{t)
<B
e~*
yi
(l
"')
[128.17]
x x (h,t)
x(k,t)
and
&(k,t)
% x (k,t)
x(k,t) [128.18]
functions
<f>{a,t)
when k
> >
<
<
t
whereas
=,
except when
>
for
129.
*
where
i
i,=o
(L = 0).
is
[129.1]
The corresponding
L very small,
[t),
[t)
For
A and
decays
very small, the current starts from zero, increases very rapidly, and follows
the curve i'xU) which becomes practically identical with the curve J(t) after
a very short time;
According to the
392
0,
mathematical discontinuity OA at
.2
] ,
[t)
of
with
L^O,
has a quasl-dlscontlnulty
OA when L approaches
zero.
and
two differential equations of the first order. Since the solutions of each of
these equations are determined by one constant of integration, there appears di of the phase plane which does not a relation between the variables -r- and i
at
with Figure 127-3 not only permits a discontinuous treatment of the problem
but also indicates the direction of the jumps in the phase plane.
It is impossible, however,
relaxation oscillations.
mechanism by which the phenomenon of "closing" or "opening" the switch is produced spontaneously by the internal reactions of the circuit itself.
In order to be able to formulate this condition and thus to complete
the discontinuous theory it will be necessary to introduce an a priori propo-
sition whose validity is justified only by its agreement with the observed
facts.
This emphasizes once more the physical nature of this theory as disI
130.
CRITICAL POINTS OF DIFFERENTIAL EQUATIONS; BASIC ASSUMPTION It was shown that, as a result of certain idealizations, discontinu-
393
matical discontinuities.
the basis of some kind of Impact is difficult because the energy delivered by
the external source, a battery, remains constant, and one cannot very well
correlate the apparent continuity of the energy input into the system with the
ment.
what actually happens and, still less, to predict theoretically the existence,
or non-existence, of such effects.
basis as was possible for the quasi-linear oscillations with which we were
concerned in Parts
I,
Critical points are the points at which the differential equation describing
in a certain
phenomenon
domain ceases
the
to describe
it.
Basic Assumption:
Whenever
phenomenon reaches a
occurs
in
some variable
of the
system.
dx
dt
P(x,y) R{x,y)'
dy
dt
Q(x,y) R(x,y)
lU
'
59^
describe
for which R{
y{
= 0.
i ,
yi
= 0.
trajectory is concerned, the passage through a critical point does not in any
R cancels
-r- = -.
ax
neighborhood of
critical point.
critical point are opposite to those of a singular point where the trajectory
is indeterminate but the motion is determinate.
2.
ure 130.1.
is a
and
belong to two
Let us as-
Fi
,,
kind of analytical impasse from which there is no normal issue, that is, along
the integral curves. In fact
T' passing
through P nor can it turn back on T since, in both cases, this would be inconsistent with the differential equations prescribing a definite direction on
the trajectories of the two regions
and N.
necessarily occur once the representative point has reached some point on
tigation of relaxation oscillations in relatively complicated circuits in
L.
395
of the basic assumption to the simple example given previously, Figures 127.3
and 127.4.
opened, the differential equation of the second order ceases to describe the
phenomenon since the right and the left portions of the circuit, Instead of
being in series, become in parallel.
Hence this instant corresponds to a
type, and the application of the basic assumption does not yield anything of
interest
We shall see that in connection with relaxation oscillations proper
the basic assumption will be a useful tool by which the possibility of relaxa-
131.
CONDITIONS OF MANDELSTAM
At the end of Section 127 certain idealizations and a choice of var-
~ and 4r are
at at
selected.
we have formulated a sufficient condition for the occurrence of a discontinuity on the basis of a certain basic assumption.
R has reached a critical point k{x ,y ). We may question into which other point B{x 2 ,y 2 the representative point will jump
1
)
we have already touched this subject and found that in the very special case
considered there the jump is from A(0,0) to B( ,j\. L. Mandelstam formulated the conditions of a jump on the basis of
certain plausible assumptions regarding the continuity of energy during the
It is to be noted that
these conditions of Mandelstam are useful for relaxation oscillations and not
for impulse-excited oscillations for reasons which will appear later.
The
Since
396
i[t) and V(
t)
~-
-2
stored in
an inductance and the electrostatic energy stored in a capacitor are also con-
+
-
Ai
t
=
n
0;
-0
AV
tn
[131.1]
where
(t
0,
not hold.
CV ^-
totally converted into heat. But in order that this may occur, a finite dissipative parameter must be present.
If, however,
exists in the circuit, there exists also a finite time constant so that the
disappearance of the charge, and, hence, of the energy, cannot be instantaneous, and it is sufficient to define a small time interval consistent with the
Mandelstam.
The usefulness of the conditions of Mandelstam is limited only to
taneously.
can be stated that the basic assumption and the conditions of Mandelstam are
excited oscillations.
132.
second order can always be represented by a system of two differential equaLikewise, a tions of the first order if a new variable y = 4r Is introduced. dt th order can be reduced to a system of n differdifferential equation of the n
-rr = y x
<fc-, "'
dt
dt
...
397
phase space, instead of being two-dimensional, that is, a phase plane, becomes
original differential equations of the second order degenerates into one equation of the first order, the system of the fourth order reduces to one equation of the second order, and its solutions can be represented by trajectories
in a phase plane.
express this by saying that we have a doubly degenerate system. Since each of
the two differential equations of the first order admits discontinuous solu-
tions, the doubly degenerate system of the second order will also possess cer-
tain discontinuous stretches in its phase plane so that its trajectories, in general, will be composed of certain analytic arcs joined by these stretches.
Under certain conditions a doubly degenerate system of the second order may
degenerate into a single differential equation of the first order; we can call
such a case a triply degenerate system.
We shall encounter one such system in
what follows.
first order represents the result of the degeneration of the system of the
fourth order
exist inside such "closed" trajectories; in some other systems such trajecto-
They reflect to some extent the fact that the discontinuous theory
this fact makes
lytic functions.
398
of important practical problems whose solutions are beyond the reach of exist-
CHAPTER XXI
DEGENERATE SYSTEMS WITH ONE DEGREE OF FREEDOM
33
[133.1]
Moreover,
one can assert that if the function f(x) is single-valued, no continuous, alIn fact,
in order that some sort of periodicity may exist, it is necessary that the
system traverse the same line x = x x with two oppositely directed velocities;
this, however, is impossible if f[x) is single-valued.
In the example illus-
trated by Figure 127.2b, we saw that the "closed" periodic trajectory ABCDA = is actually doubleis characterized by the fact that the function f{x) - +
valued.
a
It so happens that,
constant.
It is easy, however,
the plane on which the ball rolls, Instead of being horizontal, rises toward the right wall.
Figure 133.1.
instead of rolling on
plane, the
In this
by stretches
AB and CD
of analytic
Figure 133.1
Very frequently
the phenomenon is described by one differential equation and during the other
the function f(x) happens to have branches which are symmetrical with
400
134.
omit the familiar details and will endeavor primarily to show the application
of the discontinuous theory in order to prepare the groundwork for more com-
A/WWW
R
'
''
i 1
Figure 134.1
E - V R
^-
Figure 134.2
Figure 134.1 shows a circuit with the usual notations and with the
positive directions as indicated;
tube
is a gaseous conductor,
such as a neon
ABD
R(i +
i,)
+ V = E;
dV = Cjj-
[134.1]
These equations reduce to the following differential equation of the first order
dV
dt
RC
E - V -
R<j>{V)
[134.2]
This equation is valid only when the discharge exists; during its extinction
il
<f>{v)
= 0, and we have
v
dV
t
RC
(E
V)
[134.3]
401
Since these differential equations are of the first order, only one constant
variables
<j>{V)
relation between these variables, as has been mentioned in connection with Equation [128.3].
During the time intervals, when the discharge exists, this
parent that during the intervals when the discharge exists the characteristic
appears as the phase line and during the intervals of extinction the phase
line is the F-axis.
As is well known from elementary theory, the points of equilibrium are given by the points of intersection of the characteristic and the straight
line
i
v
.
p K
AB
BD
an elementary procedure.
cuts the
upper branch of the characteristic is obviously of no interest from the standpoint of oscillations.
We shall* confine our attention, therefore, to the case
and the characteristic intersect at some point S situated on the lower unstable branch BD.
Assume that we start the investigation of the phenomenon at the instant when the discharge has just appeared; this instant is represented by the
point
in Figure 134.2.
Since
y-< at
A
the characteristic, which, as was just explained, is also the phase line, in-
It
follows, therefore, that having reached the point B, the representative point
this point.
static energy
TV 2 ^-~-
+o
AV
which means that "during" the jump the voltage V across the capacitor remains
constant, that is, the jump occurs parallel to the
BC.
ix
tinuous theory.
402
representative point moves along the V-axis with a finite velocity until the
point D is reached.
Here the discharge strikes again, and the representative
repeated.
AB and CD
"closed" by t|
I.
ABCDA
is
This cycle
resembles in all respects a similar cycle shown in Figures 127.2b and 133-1
describing the idealized behavior of a ball undergoing reflections from constraining walls.
So far,
ar:
altogether impossible.
135.
RC -MULTIVIBRATOR
As a second example of a system which can be described by one dif-
V2
is a nonj
). g
The tube
appear
here merely as a linear amplifier amplifying the potential difference ri, the
feed-back voltage, between the points B and D and applying the amplified volt
age
e
g
V2
The
circuit is idealized in the customary manner, that is, the effects of the pal
asitic inductance, the grid current, and the anode reaction are neglected.
The differential equations of the circuit with the positive direc-
kOj
Figure 135-1
(R
r)i
+ V =
R<t>{kri);
r dV
at
[135.1]
(R
di
r)
IdJ
[135.2]
ix
f{i x )
[krR<t>'{kri
- (R +
r)]
[135-3]
i
where
</>'
{kri x
One can also apply the argument given in connection with Figure 130.1.
For
this purpose the graphical procedure shown in Figures 135 -2a and b will be
useful.
<f>[kri)
of
V multiplied
2
by a constant factor R.
V2
in
when
>
di
<f>'
mm
oo.
assume that
sible.
RS > R
+ r, without which,
(R + r)i.
R(f>{kri)
The curve
multiplied
It is apparent that
{R + r) from the origin 0, the ordinates f{i) of this curve represent the
ical points, and by transferring these points on the diagram of Figure 135>2a
D situated on
metrical with respect to the origin under the assumed idealization of the
404
characteristic.
(-i 1 <
It Is apparent
< +i
1 )
is unstable since
f[i) >
in that interval.
On the
> +i
1 ;
< -i 1
is stable since
f{i) <
in that interval, as is
The argu-
therefore applicable.
Thus, for
Figure 135-2
At the
D another jump occurs along the stretch DA, so that a closed piecewise analytic cycle ABCDA results. The phenomenon therefore follows a pattern sub
stantially the same as that which has previously been investigated in connection with the ball striking the walls and also with the neon-tube oscillator.
The difference between the neon-tube circuit and the multivibrator circuit is
that for the former the phenomenon is described alternately by two separate
[1
j4 .3
] ,
Because of the
405
136.
SYSTEM WITH ONE DEGREE OF FREEDOM DESCRIBABLE BY TWO DIFFERENTIAL EQUATIONS OF THE FIRST ORDER
The systems with one degree of freedom so far considered could be
Chaikin and
Lochakow
("1
second order will account for the representation of the phenomenon in the
establishing the principal features of the phenomenon, although a direct intuitive argument, such as is applicable in the simple systems of Sections
134, and 135, Is insufficient here.
1
33
shown in Figure 135-1 Is that here the inductance L replaces the resistance
of the circuit of Figure 135-1
this circuit exhibits features entirely different from those of the circuit
of Figure 135-1
<t>{kri)
dl
ri dt
i;
Jidt =
[136.1]
Figure 136.1
koe
and differentiating
J
T
these equa-
dt
y
ip(x)
dx It
dy
^(x)'
dt
krLC
0'(x)
kL
[136.2]
where
ip(x)
[136.3]
krLCy
kL
[136.4]
of the characteristic of electron tubes that the function ^(x) decreases mon-
-j^-
- -i,
when
a;
*> o.
la;,,
^ and dV
(Jy
at
^f-
at
become infinite
critical points
tx 1
determine two critical thresholds; that is, they are loci of the critical
points of this system. It is to be noted that at these thresholds the tangent
to the phase trajectories
ldy_\
1
\dx)
kL
[136. 5]
is determinate.
problem slightly without, however, introducing any qualitative changes. Figure 136.2 represents the characteristic
of the electron tube
136.1.
V2
shown in Figure
Figure 136.2
407
I,
<
2#)
or
(if r >
2^)
1 ;
in the
>+xJ.
A continuum of hyperbolic
Figure 136.3
-^IX JO
in Equation
the positive directions on the trajectories are oriented as shown by the ar-
rows.
tx
tx
.t
tn
AI
t
ydt =
'<
0;
AV
tn-0
<
-0
^V krCJ
_
xdt =
[136.6]
_o
tW
'
fir kr
= ^2) ~ Co tZ'> Ir
;
X, *i
Xt
kL
[136.7]
where x 2 and y 2 are the coordinates of the representative point immediately after the discontinuity. Since the function <t>{x) is an empirical curve, it is
<f>{x)
- --
One obtains for 6[x) a curve similar to the curve V(i) of Figure
408
135 -2a.
tion 0(x
x 2 if x 1
is given.
We omit this
-x 2 (with
-x : to +x 2
\x 2
.
>
\x 1 \)
and from
Figure 136.4
of coordinates {x 1 ,y 1
).
AB in
the phase
plane.
although we omit the proof, that after one turn of the radius vector the
piecewise analytic spiral ABCDE will approach the origin, which means that the
point E is below the point
origin 0.
near the
x-axis on the critical line x = t%i* one finds, on the contrary, that E' is above A'. Thus the large spirals shrink and the small ones grow with each
turn of the radius vector.
A and E coincide
comes closed.
jectory and the piecewise analytic cycles described in Sections 133. 134, and
135. which are not or the limit-cycle type.
409
in Figure 136.5.
Figure 136.5
BC and DE corresponding to
val.
relatively slow
motion of the electronic beam on the analytic trajectories of the outer interThe quasi-discontinuous jumps EB and
of the
much higher speed of the beam in the region of the inner interval
It seems
logical to assume that when the phenomenon starts from rest, one hyperbolic
trajectory of the inner interval is actually traversed, but once the first
critical point has been reached, the phenomenon begins to "skip" the inner interval and continues to do so thereafter.
Once the essential features of this phenomenon have been ascertained
the current
C
at
remain continuous.
first derivatives.
but are not analytic at the jumps in the sense that they have discontinuous
compli-
'
CHAPTER
XXII
MULTIVIBRATOR OF ABRAHAM-BLOCH*
As an example of a system with two degrees of freedom describable
by two differential equations of the first order we shall investigate the be-
current and anode reaction and using the notations and positive directions
h =
RI
X
hi +
ri 1
h ~
Rio 2
'
'aZ
~*~
+ -w\i
hi =
dt
E;
+ kfa Q ) -2
dt
ri 2
= E
[137-1]
^Ki) =
0(ri 2 );
'a2
0(O
4( r h)
where Ia =
<t>{e
Y and
x
V2
the other two groups, we obtain the following system of differential equations:
(R
+ r)j + 1*, +
(R
Rr<fi'(ri 2
)^
=
[137.2]
Rrfirijf +
r)
di<
+ i 2 =
Figure 137-1
*
This device is described in Reference (15)'
411
^
dt
(R
+ r)| -
Rrf(ri^
- (R +
r)'
r <j>'(ri 1 )<j>'(ri 2 )
<& _
d
g
R
2
2
4
'
[137.3]
- ^-(r,-,)^
r <t>'(ri l )4>'(ri 2 )
- (R +
r)
PUi,i 2 )
U(ilf i z )
di 2
dt
dt
Q(i lf i 2 ) U(ix ,i z )
, 7 kl L^f.'+J
r
One notes the symmetry of Equations [137 3] with respect to the variables
ix
and
i 2
i
x
i2
equation
di 2
Q(ii,i 2 ) f(ii,t 2 )
[137-5]
From the explicit values of the functions P and Q, Equations [137-3], one observes that the origin i 1 = i 2 = is a singular point, that is, an equilibrium point of the circuit.
We first inquire whether closed analytic trajectories are possible
in the system.
dP
Jil
dQ
2(R +
r)
dT =
2
constant
r
'
[137 6]
i2
=
) i
If we let 0'( ri 1
<f>'
ri 2
RrS > R
+ r
Equations [137
rftx
-3]
become
r
.
.
dt
g +
:
RrS
di^
dt
RrS
M
M =
h;
~
r)
2
]
M
>
h +
R +
[137 7]
-
where
C[R 2 r 2 S 2 - (R +
2(R +
r)
[Cg
r)
+ grgjjXg +
r)
ftrg]
_
r
'
fl [137.8]
The origin [i 1 =
that, when
i2
=0)
r,
RrS <
R+
RrS
>
+ r
412
ix
and
i2
begin to increase.
<f>{ri)
,
when
i -* o.
The function
U{i x ,i 2 )
= C[R 2 r 2 <f>'(ri
)j>*&i 2 )
- (R +
r)
2
]
[137-9]
ix
and
i2
increase
and
i
x
i2
and
for which
U=
0.
tem [137-3] has critical points so that by virtue of the basic assumption,
The expression
i
1
U{i
i2
.
defines
certain curve
and
ues of
and
R 2 r 2 <f>''(rx
)4>'(ry x )
- (R +
r)
[137-10]
<fi'(rx x
)
of
the curve
is a symmetrical closed
reached a point (x ,y on the curve r a jump must occur, and the point
1 1 ) x
,
(x 2 ,y 2
Figure 137-2
differences
continuity.
and V" across the capacitors remain constant during the dis-
= E -
- (R +
r)i x
V" =
E -
R<j>(ri x )
- (R +
r)i 2
[137-H]
the con-
413
RtiryJ - {R +
r)x
l
=
x
R<fi(ry 2 )
)
- (R + r)x 2
[137-12] (R
R&irxt) - (R + r)y
In these notations (x 1 ,y 1
= R&{r% 2 -
r)y 2
curve
Since there is
)
generally a one-to-one correspondence between the points {x ,y i and (x 2 ,y 2 ), we conclude that the locus of the points {x 2 ,y 2 is another curve r2 which is
)
and
i2
since
A on
the curve
At the point
A begins
stretch
By
and
fiQ,
= -i 2 and
<^>'(ri 1
= -0'(rt' 2 ).
With this
(R + r) + Rr<f>'{ri) i_ ~ R 2 r 2 [4>'(ri)] 2 - (R + r) 2 C
L^MJJ
- (R+ r)]^ =
[137-^]
laxation oscillation.
414
138.
gate the so-called Heegner circuit (16) shown in Figure 138.1, which is the
same as the circuit of an asymmetrical RC relaxation oscillation shown in Figure 135-1
>
Figure 138.1
'.
'
+ A +
,
{.;
h = C ^^ RI ~
^i);
h =
C2
/,
[138.1]
Putting I a = ^{krl^
we obtain the
W
dt
x "'"
+ rC A rC
l
dh
dt
R - RrWikrlJ ~ RrC
1
R +
l
[138.2]
Rrktfi'jkrlJ
RrC2
The only singular
point
tion
|
is,
'/j
= I = 0. 2
|
0' {/cr/j
>
when l^l
415
[138.2] is
X
2
Cz_
R -
RrkS
rC 2
a
+ RrC D A/> C2
x
[138.3
saddle point.
Hence, it is
either
R +
If the singularity is unstable,
- RrkS
C^
[138.4]
By
by a polynomial and by
139-
now connected by an adjustable sliding contact to some point E along the resistor
r2
r,
Let
r.
be that between
E and D, where
+ r 2 = r and
r x /r =
fi.
Proceeding as we
fi)rC x
ll
(1
fi)rC 2
[139-1]
[fir
+ R -
rR<t>'kr(I x
(1
=
dt
+ 0I 2
~
)]jf
fir
[r
- 0)[R +
prR<}>'kr{l
+ 0I2 )]
416
Figure 139.1
& =
that is,
r x = r,
For
/?
(1
- /S)
in the denominator
IS
IK
of the second equation [139-1] may vanish, which means that the otherwise con-
dis-
139-2.
system
)
39
"
is characterized by a
I,
We shall not enter into a further study of these complicated and rel-
Figure 139-2
CHAPTER XXni
MECHANICAL RELAXATION OSCILLATIONS
140.
INTRODUCTORY REMARKS
To date, mechanical relaxation oscillations have been studied less As was mentioned in the Introduction to this report,
First, the
determination of the parameters of a mechanical system is generally more complicated than the determination of the parameters of an electric circuit.
Secondly, mechanical relaxation oscillations always appear as undesirable par-
asitic phenomena, and the endeavor of engineers is directed toward their elim-
similar factors. Whether any useful applications of these effects can be found
is difficult to say. Very likely this state of affairs will continue until the
laws of friction, which usually account for the appearance of such effects,
are better understood and can be controlled with a prescribed degree of relia-
bility.
For the time being, the whole subject of mechanical relaxation oscil-
we will indicate examples in Sections 141 and 142, there exists a class of
These oscillations
trajectories.
mi + kx = - F{x) = - uf(x)
where F{x) is
a
[140.1]
is of a quasi-linear type
More specifically,
for particular forms of the function f{x) self -excitation may occur, and the
in
418
When n is large, Equation [1*10.1] cannot be solved by existing analytical methods, and at present the only guide to Its solution Is the quallta
tive analysis of Llenard;* see Section 36. We know from his analysis that the
periodic trajectory of the differential equation when n is large generally consists of two pairs of branches. On one pair of branches the motion of the
representative point is slow and the displacements are large, so the system
remains for a relatively large fraction of its relaxation period In that region.
On the other pair of branches, however, the motion is very rapid. This
phase of the motion is of very short duration so that, in spite of the large
0:
Llenard
analy
a:
when the slow motion becomes a state of rest, followed by an interval of exceedingly rapid motion, followed by another period of rest, and so on. We are
thus confronted with a special type of relaxation oscillation appearing as a
l4l
by the balance between the restoring force kx and the friction force nf[x).
When this balance is momentarily lost at a certain Instant of the cycle, the
acceleration and the velocity of the system may suddenly reach very large val
ues since, by our assumption, the mass is very small.
When
m=
0,
= - kx
[141
.1
which describes the pair of branches of the Llenard cycle on which the system
remains
a
The
balance between the restoring force and the friction force takes place on
these branches.
Recent research (12) of J. A. Shohat gave hope of bridging this gap, but these efforts have been interrupted by his untimely death.
419
F'(x)'x
= - kx
[141 .2]
region F'(x)
a
0,
mx will play
a role in
that this condition characterizes the second pair of branches of the Lienard
cycle the acceleration is very large and the velocity varies in a quasi-dis-
continuous manner by a finite quantity, but the coordinate does not change
respects to that which occurs when a ball strikes a wall. However, the underlying dynamical facts are different.
and
!=-'*%)
and, hence, ceases to describe the phenomenon at that point.
[I*.*]
In order to de-
physical problem, we must again apply the condition of Mandelstam. idealized case,
ra
In the
From
resembles the results obtained in the theory of mechanical impacts, with the
difference, however, that the quasi-discontinuity here arises from the existence of a critical point in the differential equation and is not due to exter-
420
142.
clude that relaxation oscillations are possible in a system of this kind when-
that is, that F[x) should be a two-valued function of x, must also be ful-
filled.
In practice, these theoretical conditions are frequently encountered. Thus, when a shaft rotates in a bearing, Sommerfeld (17) has indicated
that the friction force goes through a minimum for a speed v 1 given by the
equation
1
d2p
nko
where
is the pressure,
///////////////////////,
S.
A definite friction
Figure 142.1
order), is
r(F[{Q - 0)r]) =
where F{v) = F[{Q
- <j>)r]
c<t>
[142.2]
isv the
Q
<j>
is the angular velocity of the shaft K, is the angular velocity of a for small departures
is a constant depending on the spring's strength.
It is
<f>,
and
When
v = 0,
<f>
0=0
and v = v
421
the shaft
K move
they
<j>
= 0. As long
= Q, the spring
is a static force
Figure 142.2
Flf
where F\ is the
tj>
x%
for which
c<t> x
r,
until it
F'[(G -
j>)r\4>
c<j>
[142.3]
<j>
dF
It is seen that
<
when
and
<f>
Moreover, for v = v
quantity
<j>
of the shaft
BD
acteristic of friction which may be considered here as the phase line of the
process, see Section 13^.
<j>
On the part
BD
and
have the same sign; this part of the phase line is therefore unstable,
It follows,
D during
CD
of the
DBA.
We conclude
is a critical point.
<j>
= y in Equation [142.3],
which gives
dy
dt
r
2
cy
F'[(Q - y)r]
-%- =
<f>
oo
at the point
0)r] =
at this point.
A where
the
422
where F' = 0.
DA
occurs, after
which the phase CD of static friction begins anew; during that phase of the
motion the representative point again moves continuously along the branch CD
of the cycle DABCD.
oscillation and the example of the relaxation oscillation of a neon tube, Section 134.
In both examples, relaxation oscillations are possible if the pheIn the neon-
interp sects the characteristic on its upper stable branch, see Figure 13^.2; like-
corresponding to
teristic
.
'
CD
ABCDA in both Figures 134.2 and 142.2 corresponding letters. Thus, for example, the branch CD
corresponds to the period when the capacitor is charged (Figure 13^.2) and to
the period when the static friction force
Plotted in
Figure 142.3
CHAPTER XXIV
OSCILLATIONS MAINTAINED BY PERIODIC IMPULSES
143.
INTRODUCTORY REMARKS
We have defined "relaxation oscillations" as stationary self -excited
ary state does not depend on the initial conditions of the system.
In that
continuous oscillations.
As a somewhat different type of oscillation appear the so-called
For both
types, also, the stationary state does not depend on initial conditions;
tions with which we have been concerned in preceding chapters of Part IV, the
energy stored in the system during the quasi-discontinuous interval does not
change appreciably.
with an assumption that the energy does not change in the interval
t
0,
+0).
424
144.
elements:
1.
elements.
The purpose of the escapement is to release periodically the energy stored in
the second element in the form of an impulse and to apply this impulse to the
ment is uniquely determined by the motion of the system, the system is auton-
3,
of the mechanism
it is possible to arrange the operation of an electron-tube circuit so that it resembles the performance of a clock, but in general the mode of operation
commonly used.
The escapement communicates to the oscillatory system periodically
timed impulses which we shall first idealize as instantaneous and of a constant value a.
Let v x be the velocity of the system immediately after the
If we assume
that the system is a linear dissipative one with a decrement d per cycle, the
velocity
v2
v2
= v 1 e~ d
Immedi-
= v2
d + a = v 1 e~ + a,
number n of impulses, v n
vn +
vn
which gives
y^r
W,i]
As
was shown in Section 5> these spirals form a continuous family so that through every ordinary point of the phase plane passes one and only one such spiral.
Let us designate as u the segment
AB
of the
a;
425
2n.
From the
sidered as a function of
r
.
= OA,
is a
monotonic function u(
Since the
On
constant and hence are represented by constant jumps a along the x-axis.
It
Figure 144.1
follows, therefore, that a steady-state condition will be attained when a radius vector r
=
OA
BCAB
BCA
(since, by our
AB
is of
BCAB
limit-cycle type in the sense that, if one starts from the relatively
= OA, the value of a is initially
)
in this region of
impulse is greater than the amount of energy which the system is able to dissipate during one cycle
27r,
radius vectors r
when the impulsive increments a are just equal to the intercepts u(r
particular region.
the function u(r
),
in a
BCAB
This elementary
discussion accounts for the fact that the clock's performance is of a limitcycle type and as such does not depend on initial conditions, see Chapter IV;
in other words, the ultimate performance of a clock does not depend on how
it has been started.
426
entirely on the parameters of the system and has nothing to do with the Initial conditions.
The converse Is true for a mathematical pendulum.
In one respect the elementary theory discussed above falls to ac-
plified theory Indicates that the clock should start by Itself even if the
initial disturbance Is infinitely small.
self -excitation of the clock.
In other words, it predicts a soft
This threshold
Is partially due to the existence of Coulomb friction not taken into consider-
not complete, namely, our assumption that the jump a in velocity caused by the
Andronow and
Chaikin (13) have shown that by a slight refinement of the preceding theory it
is possible to explain better the observed facts.
Let us assume that the change of kinetic energy during the Impact
mv n
2
constant
yi
y<>
h'
[144.2]
where
Us
y 1 are the values of the radius vector immediately before and immediately after
the impact. It follows, therefore, that the jump a in the phase plane, Instead
of being constant as was originally assumed, is now given by the equation
= yV +
2
[144.3]
according to a hyperbolic
This refinement of the theory,
\
\ \
V
although it leads to conclusions concerning the existence of a stable piecewise analytic limit cycle which are similar to those obtained by the original
Figure 144.2
by phase trajectories.
427
145.
PHASE TRAJECTORIES IN THE PRESENCE OP COULOMB FRICTION The customary idealization of Coulomb, or dry, friction consists in
assuming that the friction force f remains constant during the motion and changes its sign with a change of the direction of motion, as shown in Figure
14$.1
+ kx = + f
w
x
mi +
au
2
2
kx
= f
[1^5-1]
Putting
and |/
+
+
,
cj
q
2
x x
= + au Q = au
-
when when
<
[145.2]
oj
>
+
+
ut
Xj
=
=
when
when
x x
<
[145.3
X,
Wn Xo
>
These equations are obviously identical, with the difference, however, that
the center of oscillation is displaced from +a to -a, and vice versa, each
the equations of motion. The "change of equations" occurs at the instant when
x = 0. In the
(x,0-plane such
in Figure 145.2.
=0.
The sys-
fr
Figure 145.1
Figure 145.2
428
the t-axls.
is
|x 02
|
= |x 01
2a|
with respect
to
the
t-axis.
displaced a
sinusoidal, so that the point C (at which x = 0) is at the same distance from
the
t
2a.
The time interval between two consecutive maximums (A and C in Figure 145.2) is
the same as for a harmonic oscillator, as follows from the preceding discussion.
It is to be noted again that the curve ABC
[145.3] become
a)
dy _ dx dy dx
o)
(x
when y <
[H6..41
n
io
(x
+
y
a)
when y >
u K+
/
%2
a)
p~2
\2
+ P
,
y
2,,2
2
when y <
{x
[145.5]
.
a)
^~2
2
+ P 2 /|2 = K 2 U)
when
>
where
and
R2
and
A representing
of the trajectories
y = 0,
r2 r
x
the change of equations takes place, and the representative point goes
passing through B.
trajectory
r2
jectory would be confined within the zone limited by the broken lines x = +a
jectory is reached which does not emerge from that zone, the motion ceases.
429
'
families
and
Figure 1^5.3
equilibrium C^Og
O^
friction
dr 2
xx
= =
when when
<
[145.6]
1^
d x2
x2
>
circles whose centers are displaced on the x-axis on both sides of the origin
by a fixed quantity a. By reproducing the preceding argument and by expressing
the condition that the ultimate trajectory always emerges from the "dead zone" of static friction, these authors show that, if the trajectory is to emerge
ultimately from the dead zone, the impulsive change of kinetic energy h 2 must
be equal to or greater than
lo/
where f
This means
2
and l6/
fied, the clock will start and will approach its piecewise analytic limit
430
146.
purely formal, but it will be helpful in approaching from a somewhat new viewpoint the difficult subject of self -excited oscillations.
The interesting
feature of the analysis made by Andronow and Chaikin is that it permits inves-
inductance L.
Ia )dt
[146.1]
Lx + Rx +
where I a = f(e g
)
f(e
g)
[146.2]
We
will consider a peculiar performance of this circuit, namely, the one which
M is
Ia
As a
large; this will cause the electron tube to act more or less as a switch oper-
0,
and B, when
= Is
where I s is the
A B
1
corre-
current to vary, is small in comparison with the total interval AB, and it is
Ink
x-I a
^W
Figure 146.1
&
A,
B,
Figure 146.2
431
+ 2hx
4-
+ 2hx +
u)
x=(J
Is
[146.3]
2 2 = y~, and u> I s = y-~f{ + e g ). It is obvious that the change where 2h = y, u> from one equation to the other occurs at the point where x changes its sign.
> 0,
eg
>
when x <
later.
0,
and
eg
<
when x >
0. We
0.
through the capacitor branch of the oscillating circuit since in the inductive
L^t in
at
remains continuous.
V across
the
other, mentioned above, resembles the situation which we have already encoun-
We can write
+ 2hx + u*x =
[146.4]
where x = x
Is
are two linear differential equations possessing focal points on the -axis
at a distance I s
stant when the amplitude is x x and the electron-tube switch is off, which
original amplitude x lt see Figure 146.3, will be reduced by damping and will
become x 2
.
At the point
lt
if
432
7 S is large enough.
Beginning with
'T" _
A
A, and so on.
3
*
^
Is
11
,
.
T \ 2 \
X
V
It
in the oscillating circuit may grow, thus outweighing the effect of the
*2
Vy
;
/ /
dissipation of energy.
It is easy to
'
ly.
Figure 146.3
x2
hi
2
has
=
+
hi x1
e
2
;
I.
2hT
hi
I.e
2
2hT
xxe
X3
X3
+ h =
Xi e
hi
I.e
2
i:
which gives
xo
+ -
hi
2
)
e~
e
hT
hl
e
2
[1^6.5]
namely,
hT
hT
xx
7 (l g
[146.6]
In the
(asj,
x 3 ) -plane,
e
,
a slope tan a =
and an intercept l s
= x3
['\
= A.
The point
.
of intersection of these
'
plitude will be x 3
will
' ,
correspond on the x 3 -axis, and its corrected value on the Zj-axis will be x 5
433
amplitude x
amplitude is stable.
will occur in
direction opposite
Figure 146.4
standpoint, the circuit behaves as if it had a kind of negative Coulomb damping when
M>
M<
0.
damping.
147.
in the preceding section becomes still more striking if we analyze the phase
+ 2hx +
+ 2hx +
x
u> n
x
Xj
cj
=0
where
Xj
= x
- I s
Figure 147-1
and
that of the
second equation.
resenting the first equation [147-1]. At the point 2 the "change of equations"
Kjh
and so on.
ultimately approaches
a closed curve
two
Figure 147.1
its
ABCDA formed by spiral arcs, ABC with as focal point and CDA with
One obtains
similar conclusion if, instead of starting with a point 1, one starts with
point
ABCDA
We
shall designate by r the radius vectors with respect to the focal point
the representative point moves on the upper spiral arcs, and by
when
the radius
For the subsequent radius vectors along the x-axis, we have the
r2
r e
1
nh
nh_
_ nh
nh
r2 e
_ 2nh
rk
"
r
a
_ nh
+ +
u
(k-2)
e
nh
(k-l)
nh
r1 e
rk
rk
[1^7.2]
+
s
_2)^-"i
r1 e
U>
r.
increase indef-
initely, that is, if the number of turns of the plecewise analytic spiral
lm2njipk
when k
>
and
oo,
namely,
lim
(r.)
1
nh
lim {rs )
[147.3]
4^5
The conclusions remain exactly the same if one repeats the argument for a
'
starting at
relatively distant
ABCDA
oscillation approaches as
-> .
It is
useful to emphasize once more that this limit cycle characterizing stationary
Is
and
cj
ditions.
Such a pro-
cedure becomes possible only after we have idealized the relatively compli-
step-function like that shown in Figure 145.1. In connection with this it may
be useful to recall the statement made in Section 3 that the really important
This
argument is
CHAPTER XXV
EFFECT OF PARASITIC PARAMETERS ON STATIONARY STATES OF DYNAMICAL SYSTEMS
148.
PARASITIC PARAMETERS
The study of relaxation oscillations in the preceding sections was
that such cases appear generally as rare exceptions, it is important to analyze this matter in greater detail now that we are acquainted with the general
Thus,
conditions."
which some initial values can be assigned, these factors cannot have entirely
The discontin-
from the solutions of one form to those of the other form, just as the classical theory of mechanical impacts ignores the unknown dynamics of the collision
process. Both theories follow a somewhat similar argument, namely, the rapidly
0,
which permits solution of the problem in the large although certain local details are inevitably lost in such a procedure.
In the theory of mechanical
4^7
during the Impact; likewise, in the discontinuous theory of relaxation oscillations this additional information appears in the form of the conditions of
Mandelstam.
The use of the concepts "infinitely short time interval," "infinitely large acceleration," and so on, instead of the more correct terms "very
short time interval," "very large acceleration," and so on, while convenient
for the description of a phenomenon in the large, is sometimes capable of in-
troducing serious errors in a theoretical argument and of leading to conclusions at variance with experimental facts.
Thus, for instance, in dealing with an idealized {L,R) -circuit, we
L~ at
+ Ri = E(t)
[148.1]
This equation describes the phenomenon with adequate accuracy in a range sufficient for practical purposes (on the scale, say, of milliseconds).
If, how-
ever, the behavior of the same circuit is studied in a different range (say,
on the scale of microseconds), this equation may not give a correct answer. In
fact, any resistor or inductance coil inevitably has a small parasitic capacity
be
Cp
and if we take this capacity into account a more correct equation will
d
_ di + L^-i +
l
RLC, '"
ix
dt 2
Ri,
E(t)
at
[148.2]
It was shown in Section 128 that under certain conditions the solutions i{t)
and
i-iit)
of Equations
tures locally, although in the large these solutions are practically indis-
tinguishable.
simpler equation [148.1] rather than the more complicated one [148.2],
facts are too well known to need further emphasis here. However, they are fre-
quently very troublesome. Thus, for instance, a student attempting to investigate the behavior of a simple circuit by means of a cathode-ray oscillograph,
arise which lead to paradoxes already noted by certain authors (19) (20).
438
Lq + Rq +
and assume that L is very small.
If
q =
[148.3]
range in which the term Lq is negligible compared with the other two terms.
Still more confusing is the effect of the degeneration when leads to a meaningless result.
If, however, we make
*- 0,
which
C > <,
at the limit we
obtain
L^- +
at
Ri
[148.4]
- as
a parameter.
sponding terms in the differential equation are small. The first condition is
merely a definition, whereas the second condition specifies the range in which
a small parameter can be neglected.
149.
differential equation
{n)
(n_1)
a 1 a;
+'+
an
_2
an
an x
[149.1
where x
_ ^, = d^x
{n}
dV
and so on.
n
a X
ay
+..'+
a_
+ a.^X +
an
[149.2]
The equilibrium is stable if^the real parts of the roots of [149.2] are negative, which can be ascertained by the well-known criteria of Routh-Hurwitz
As was previously stated, we never know the exact form of the differential
^9
If it is of the inductance type, the non-degenerate equation will be
ax
a x
a x
x
"
an
x
l
an x
[1^9-3]
a1 x
"
an
an x
+ aj x
dt
[149 .4]
aQ x
a x
x
"
an
x
1
an x
ctx
[149-5]
For both types of parasitic parameter the order of the original differential
aX
n +
+
+
1
a y x"
ajX""
a n _jX
an
[149.6]
a X"
fl^X"
a 2 X""'
an X
+ a =
[1^9-7]
1)
roots change but little owing to the appearance of the new root, and we may
The matter hinges, therefore, on the new root X w + We can write Equation [149.2], in terms of its roots, as
a (X
X,)(X
X2 )
(X
X)
[149.8]
a(X - X
)(X
- X2
(X
X)(X
w
'
- Xn
)
]
or
a X
+ a,X
a w _,X
ar
(x
xn +
[149.9]
Expanding the right-hand side of this equation and identifying it with Equation [149.6], we get
*n
+
-7"^
[14 9 .10]
- - T~
[1^9.11]
librium.
= C,/
'
+ C 2 / 2< +
+ C n e kn< + Cn + /" +1
1
'
[149.12]
440
are fulfilled, the stability of the new equation [149.3] or [149.5] will de-
pend solely on the new root X n + 1 If this root is stable, that is, if its real part is negative, the equilibrium will still be stable. If, however, the
.
For
parameters
Conclusions obtained from the simple discussion given above may appear to some extent paradoxical.
Thus, for instance, if the parasitic param-
eter is of the "inductance" type, see Equation [14-9.10], the new root X n + 1 is larger as the parasitic parameter a is smaller. If, however, it is of the
"elastance" type, see Equation [149.11], the magnitude of the root X n+1 is proportional to a.
The rather erratic "floating" of a system around its theoretical
150.
21
Investigated in detail by means of Liapounoff's equations of the first approximation on the basis of a differential equation with finite parameters L, C,
and p.
The behavior of the arc is far more complex if some of these param-
eters are small, a condition which results in a degenerate form of the dif-
equations are
ty-V
di
-#(.);
,/.v
dV C^- = E
- V - Ri
[150.1]
ricAll
where
Va
150.1.
Hi
A/VWW
^WfftfSTT
If-
Figure 150.1
points (Points
1,
2,
Figure 150.2
ip{i).
J_
~RC
dt
[150.2]
where p =
the point
\p'{i
i
at
+ (J- + 11 \RC L,
U'
+ 1) -
[150.3]
1
corre-
>
0.
therefore, either a
stable nodal, or a stable focal, point according to whether the roots are real
or conjugate complex.
and, moreover,
\p\
> R.
is
< R.
21
we propose to in-
parameters L and
C approaches
C*
-*-
0.
(A + cp\ k + 1 Ll \R R L TR
'
{R+P)
[150.4]
442
lx
(^
P )x
+ -^(R +
p)
[150.5]
Putting
^-^
CR~p~
[150.6]
Putting L =
= ~
iR
p)
[150.7]
It is
In the L-
The following explanation was suggested by S. Chaikin (20) Let us consider the characteristic equation
(21).
aX
with a >
and
c
+ bX +
=
and as negative when
[150.8]
We
>
< 0.
Since
If, however,
it follows the
phase plane.
In this case,
therefore, a threshold exists similar to that which exists for the asymptotic
Case
3;
Only in a purely theoretical case when the motion of the representative point takes place along the stable separatrix or asymptote of a saddle point, may the latter be considered as a stable singularity. If the actual motion of the representative point occurs in the neighborhood of this theoretical motion, In view of it may appear that the saddle point is a stable singularity, at least for a limited time. this, one could characterize the saddle point as an almost unstable singularity.
hki
For practical purposes, if precautions are made to reduce the parasitic parameter to a very low value, the representative point will follow a
trajectory very closely to the stable separatrix so that one has the impression that the phenomenon would ultimately "settle" at Point
point of stable equilibrium.
2 as if it
were a
By
for a
of stable equilibrium
< R.
Point
systems.
151.
ior of the circuit shown in Figure 1351 taking into account the presence of the small parasitic inductances L and L' shown in Figure 151.1.
Kirchhoff's
*ff + J
Ia
ri
- Lx
~
dt
- i/itfc =
C
/
[151.1]
Moreover,
= <f>[kri) as before.
Elimination of I a and
between
L +
- (R + r)i - V - Lkr0'(kri)
i
dV
dt
i_
[151.2]
-
and #0(0)
= 0.
On the
Figure 151
.1
444
0(0)
kri<t>'{0)
Limiting the expansion to the first term and noting that #'(0) =
Liapounoff equations of the first approximation are
di
S,
the
V
y
dV
dt
d7
where
6
[151.3]
+ R(l - Skr)
and
L'
+ L(l - Skr)
[151.4]
yC
[151.5]
^
\
Ay
C
[151.6]
2y
For y <
For
nodal point or a focal point, depending on whether the roots are real or
conjugate complex.
jugate complex, with
< 0, it is
with
Figure 151
.2
equation [151-3] by
tlon gives
y,
one has di +
V=
0,
that is,
-*-,
dV
dt
V
6C
[151-7]
R<f>{0)
>
< 0.
The points
445
of ordinates.
in the region of "negative" saddle points and that of unstable equilibrium in the region of "positive" ones.
It was shown, however, in the preceding sec-
tion that such stability is of theoretical interest only, because it is impossible in practice to prescribe the initial conditions accurately enough to
make the representative point follow the stable separatrlx, just as it is im-
possible to Impart to a pendulum a velocity just sufficient to produce an asymptotic motion. Moreover, no physical system is completely degenerate, but
Is only quasi-degenerate.
ity conditions are influenced not only by the existence of parasitic param-
Skr <
but
is sufficiently small to
render
= r + #(1
for any value of L, however small, because the real part of the roots is positive, that is,
<J
>
equilibrium is stable, however, as follows from Equation [151 -7]unstable in this case.
Skr <
0.
-
It is obvious that,
= r +
R{~\
Hence, the
0,
=0,
appreciable "effect" on the system, and its behavior then becomes entirely
unpredictable.
446
REFERENCES
"On Relaxation Oscillations," by B. Van der Pol, Philosophical
1
(1)
926
by B. Van der Pol, Zeitschrift fUr Hochfrequenztechnik, Vol. 28, 1926, and
Vol. 29, 1927.
(3)
18, No.
October 1926.
bei welchem die KapazitSt von untergeordneter Bedeutung 1st" (On the Excita-
I.
1935-
tion"
447
(12)
cow,
(Russian),
(14)
937
S.
Vol.
1935(15)
Electric Oscillations), by
Vol.
12,
H.
No.
9,
1919.
(16)
(On
(19)
"Warum kehren sich die fUr den Lichtbogen giiltigen Stabilibei Elektronenrohren um?"
(Why Do the Conditions of Stability
ty sbedingungen
(21)
Dynamical System," by
INDEX
Analytical definition of stability, 41-42 Analytical impasse, 39^ Analytical method, 11,111,135-245,323,382 of Kryloff and Bogolluboff, 4 of Poincare', 4,87,99,104-105,138-166 of topology of phase trajectories, 26-29 of Van der Pol, 4 Analytical periodic solutions, 383
Analytical trajectories, 383,4ll,4l3-4l4
Andronov, A., 2-4,63,75,135,137,148,381,383,426, 429-430
Autoperiodic excitation, 297-298,305 Autoperiodic frequency, 297-299,325,342,349 Autoperiodic oscillation, 301,345,348,350 amplitude of, 350 conditions for stability of, 306 definition of, 295 existence of, 295-300 self-excitation of, 300 stability of, 306 stability of stationary, 3 04 synchronized with heteroperiodic oscillation,
303 synchronous, 304
Andronow and Witt presentation of, 344 topological method of, 342
Aperiodic damped motion, 106 equation of, 14 phase trajectories of, 18-19 Aperiodic damping, 262
Aperiodic motion, 3,36,48
Appleton, 34l
Barkhausen, 260
Barrier, 177,203
Beats, 303, 341-342, 351, 35
1 *
quantitative method of, 2-3,135-136 advantage of, 3 second, 210 . example of, 228 theory of first, 137,171,183-209,233,241,258 applied to non-linear conservative system, 189-190 third, 210
Bendixson, 4 application of first theorem of, 79-82 criteria of, 75,77-79,105,345,411 first theorem of, 75,77-78,3^5 negative criterion of, 75,77-78,345 second theorem of, 75,78-79,348-349
Critical point)
Bifurcation theory, 415 for quasi-linear systems, 160-161 of Poincare, 87-104 Bifurcation values of parameter, 87
Biot-Savart law, 34
Birkhoff, G. D., 4,65
449
450
Bogoliuboff, 2-3,67,135,157,181,210,234 (see also Kryloff and Bogoliuboff) method of equivalent linearization of, 137,209,
233
notation of, 254,263 quasi-linear theory of, 4,234,244,249,258,331, 342 theory of first approximation of, 183,186-188, 209,258,289,323 treatise of, 251,288,303
Boltzmann, theory of, Bowshewerow, 73
5
Constant linear parameter, 317 Constant parameter, 201,252,370,435 linear dissipative system with, 258
Continuum
Canonical equation, 45 Canonical form, 329 of linear equations, 42-44 of non-linear system, 52
Car tan, E. and H., 105,181
of of of of
circles, 326 closed curves, 62,119 closed trajectories, 12-13,62, 140-141 phase trajectories, 62,128,407
Coulomb friction, 426-427,429-431,433 existence of, 426 Cauchy s theorem of existence for solution of differential negative, 430 equation, 2,9 phase trajectories in presence of, 427-429 of existence of linear differentical equation, 2 Coupled electronic oscillators, stability of, I58 of phase trajectories, 11, 14 of uniqueness, 24 Coupled frequencies, 265,278
Chaikin, E., 351 Chaikin, S., 3-4,75,135,137,381,383,405,409,420, 422,426,429-430,442
Coupling autoparametric non-linear, 375 between degrees of freedom, 253,270 critical value of, 340 inductive, 260
lack of, 254 non- linear, 375
Critical points, 31,38,102,395,399,401-4o6,4o8-409, 412-415,419-423 criteria for existence of, 393-394 locus of, 394,406-407 of differential equations, 392-395 theory of, 130
451
Critical threshold, 5, 11,294, 373,406 (see also Divide) Critical value of amplication factor, 102 of amplitude, 203 of coupling, 340 of damping, 123 of equivalent parameter, 234,278 of index of modulation, 312 of negative resistance, 263 of parameter, 31-35,39,87,97, 104, 160-161, 165, 203-204 of resistance, 266,269 of self-excitation, 297 of trans conductance, 302 of variable, 197
Curve of contacts of Poincare', 75
Differential equation, 1-6,10-11,14,22,24,31-32, 35,37,42-43,46,49,51,62,64,66-67,72,80,82,87, 90,95-96,101,106,112,117,138,192,215,241,323 critical points of, 392-395 degeneration of, 5,395-397
linear,
1
'
Cauchy s theorem for existence of, 2 with constant coefficients, 1 linearized, 67,87 non-linear; see Non-linear differential equation of damped oscillator, 16 of dynamical system, 20 of electromechanical system, 117-118 of first order, 17, 388-389,392,408, 413, 438 degenerate, 385,387,395,397 describing system with one degree of freedom, 405-409 discontinuous solution of, 397 of harmonic oscillator, 7 of mechanical system, 257 of motion, 11 of synchronous motor, 126 of non-disslpative circuit, 363
of n tn order, 9 of pendulum, 19 of second order, full, 385
Damping aperiodic, 262 Coulomb, 195 negative, 433 positive, 433 critical, 371 critical value of, 123 linear, 193-196 negative, 105,178-179,197,199 definition of, 21 examples of, 21-22
non- linear, 5 non-linear dissipative, 192-196,240-241 non- linear variable, 105,178-179,197-199 positive, 18,105,178,197 quadratic, 194-196,240 variable, 105,181, 216-217
10,17,24,386-388,392,395,^36
of, 4 18
quasi-discontinuous solution
of synchronous motor, 126 of Van der Pol, 342-344 quasi-linear system of, 148
deForest, Lee, 5
singular points of, 17,344 singularities of, 3,11 solutions of Cauchy 's theorem of existence for, 2,9 in neighborhood of point of degeneration, 388-391 represented by phase trajectories, 250 systems of degenerate, 396-398 theory of, 2,4 with periodic coefficients, 250,304,322,356-360, 372-373,375 Discontinuities, 393-395, 4l2-4l4, 421, 423, 428,431 occurrence of, 393,395-396,407
Degenerate differential equation, 396-397 systems of, 396-398 Degenerate equation, 390,436,438,440 absolutely, 391-392,438 of first order, 112,385,387,395,397 Degenerate limit cycle, 70 Degenerate phase trajectory, 12 Degenerate system, 148,440,442,445 absolutely, 391,442 completely, 390 conditions for stability of, 397 of first order, periodic solution of, 399 transition between continuous and discontinuous solutions of, 415-416 with one degree of freedom, 399-409
Discontinuous jump, 166,386,392,395,^01-402,404, 408, 412-413, 416,421-422, 425-426 Discontinuous periodic motion, 383
Discontinuous solutions of degenerate system, transition between continuous and, 415-416 of differential equations of first order, 397 Discontinuous stretches, 382,422-423
Discontinuous theory of relaxation oscillations, 383, 388-398, 400-402, 405, 409, 436-437
Degeneration absolute, 444 complete, 389-390 condition of, 4 18 of equations, 4-5,395,420 of system, 44l point of, 4-5,388,390
Dissipative parameter, 269,396 Dissipative system, 201,207,425 heteroparametric excitation of, 369-371
Divides, 3,11,13 examples of, 72 of common topography, 32
452
Dreyfus s, 126
Dry friction, 22,427
Durand, Dr. W. F., 195
Dynamical system, 10-11,32,51,69,82,87-88,107 differential equation of, 20 equilibrium of, 438-440 stationary states of, 40,436-445
Electrical oscillations in circuit containing iron core, 192 An circuit containing saturated core, 242
Electrodynamical system, 87
Electromechanical system differential equation of, 117-118 self-excitation of, 99,102-104 self -excited oscillations in, 99,102-104
Entrainment of frequency, 4,249,341-354 acoustic, 342,351-352 artificially produced, 352 band of, 352 definition of, 34l mechanical, 342 other forms of, 352-354 pure,, 348 state of, 350 steady state of, 350 transient state of, 346,350 zone of, 341,348,351-352,354
Equations absolutely degenerate, 391-392,438 autonomous 155 canonical, 45 change of, 427-428,433 characteristic; see Characteristic equation degenerate, 390,436,438,440 of first order, 112,385,387,395,397 degeneration of, 4-5,395,420 differential; see Differential equation Lagrangian, 37^-375 linear, 80, 136-137, 139, 179, 186,407,431 approximate, 100 canonical form of, 42-44 equivalent, 235-236 harmonic solution of, 167 Mathieu, 372 Mathieu-Hill, 356 of first approximation, 49 system of, 18, 23,80,83 with constant coefficients, 1 > linearized, 67,87,100 equivalent, 244 non-degenerate, 439 non-linear, 1, 118, 139, 181-183,242 of oscillation, 192 non-linear differential; see Non-linear differential equation of aperiodic damped motion, 14 of damped oscillatory motion, 13 of first approximation, 3,49-52,55-56,65,81,87, 9^,97,99,101,103,114,120,127,167-168,173, 176,179,186-188,193,217,224,229,236,244,259, 271,275,277,279,281,285-286,288-293,301-305, 311,315-317,322,411,440-441,444
,
Equations - continued of Proude for rolling of ship, 195,241 of Hill, 361-362 definition of, 357 of Hlll-Melssner, 362-363,365 topology of, 362-365 trajectories of, 362,365 of improved first approximation, 288 of Isoclines, 121 of Lienard, 107,109-110 of limit cycle, 180 of Mathieu, 357-358,360-362,372,375-376 definition of, 356 linear, 372 parameter of, 375 stable and unstable regions of, 359-360 of Mathleu-Hill, 356-357,361-362,372-373 linear, 356 stability of, 361 subharmonic resonance on basis of, 373,377 of n th order, 396 of phase trajectory, 66,88,128 of Poincare', 187 of Raylelgh, 178-179, 18 1-182, 199 of second order, 396-397,^18 of separatrix, 35-36,38 of unstable motion, 13 of Van der Pol, 79,85,97,103,105,108,111,113, 161,175,181,187,197-198,200,217,219,293-294, 348,382-383 quasi-degenerate, 391-392 quasi-linear; see Quasi-linear equation variational, 83,153,283-284,311-312,318,328 with both linear and quadratic damping, 195
Equilibrium branch point of, 3,42,48,55,60,82,165,445 conditions for stable, 161 conditions of, 11,29,60,80,94,205,443 dynamical, 282 half-stable, 51 Kaufmann's criterion for stability of, 61
line of, 429 loss of, 384 motion in neighborhood of, 40 neighborhood of, 335 of circuit containing non-linear conductor, 55-61 of conservative system, 40 of dynamical system, 438-440 of electromotive forces in circuit, 192 of point, 32 point of, 11,24-26,29,40,51,56,60,66,7^,99,125,
170-171,335,35^,401,411,429,440-441,443-444 coalescence of, 33 disappearance of, 3^ in conservative system, 34 motion in vicinity of, 29 stability of, 35^ stability of motion in vicinity of, 24,26 position of, 18,22,32,38,49,185,302,39^21,440 change of, 384 stability of, 9 stability of, 33-35,^0-61,71,85,87,93,127,169170,202,35^,358,401,438-445 according to Liapounoff, 48-61 criteria for, 60-61 theorems for, 29 state of, 22 types of, 58
453
Equitlme interval, 367
Equivalent linearization for multiperiodic systems, 274-278 method of, 137, 233-236, 269, 279, 288, 297, 3l4 applied to steady state of quasi-linear system, 258-262 applied to transient state of quasi-linear system, 262-267 examples of application of, 239-245 in quasi-linear systems with several frequencies, 285-288 principle of, 239,258,275,277,309 Equivalent linearized equation, 244 Equivalent linearized system, 314 self -excitation of, 278 Equivalent parameter, 236-239,241,281-283,297, 310,314,316-317 critical value of, 234,278 definition of, 234-235,286 determination of, 233-236 non-linear, 277 of multiperiodic system, 275
Euler's identity, 251
Existence - continued of nodal point, criteria for, 13 of periodic solution, 105-115,121,144-145,220,224 condition for, 136,152 geometrical analysis of, 105-115 In non-conservative system, 224-229 proof of, 105 of principal solution, 333 conditions for, 335 of saddle point, criterion for, 13 of self-excited oscillations, l8l
Expansions of Polncare, l4l
External excitation, 289,335-336,340 absence of, 337-344 non-resonant, 294-295 system with, 289
Excitation asynchronous, 234,295,298 autoparametric, 355-356,374-377 condition for, 374 autoperiodic, 297-298,305 external, 289,335-336,340 absence of, 337-344 system with, 289 Fractional-order resonance, 301,303,305-309,373 external periodic, 249,289,294,313,317,320,335, Franck, A., 6 342,376 of quasi-linear systems, 289-300 Frequency system with, 323 acoustic entrainment of, 342,351-352 heteroparametrlc, 355-356,360,363,373-374 autoperiodic, 297-299,325,342,349 conditions of, 356,371 coalescence of, 279,3^1,351 dependence on frequency and phase of parameter combination, 273-274,278-279,286,290,296,301 variation, 365-369 correction, 174,193,217-219,240 of dissipative system, 369-371 coupled, 265,278 principal features of, 367-369 demultiplication, 272-273 heteroperiodic, 298,355-356 demultiplication network, 352 absence of, 299 entrainmeni of, see Entrainment of frequency non-resonant external, of quasi-linear system, enternal, 307,341,348,351,353-35^ 294-295 externally applied, 336,340 parametric, 4,249,308-313,355-377 harmonics of, 296 definition of, 308 heteroperiodic, 297-301,348-349 of critically damped or overdamped circuit, of pendulum, 342 of quasi-linear oscillation, 217 371 periodic non-resonant, 289-292 of ripple, 363,365,367-368 of self-excited non-linear system, 265 Existence of self-excited oscillations, 261 of autoperiodic oscillation, 295-300 of subharmonlc oscillation, 336 of closed trajectories, 21,108-113,129 of thermionic generator, 158-159 of critical point, criteria for, 393-39^ synchronization of, 352,354 of focal point, criteria for, 13 of limit cycles, 99, 181-182, 198,200-203, 224Friction Coulomb, 426-427,429-431,433 229,3^8 existence of, 426 condition for, 169,173,176,179,199,207,227 negative, 430 criteria for, 75-77 proof of, 79 phase trajectories in presence of, 427-429
Focal point, 16, 18,22,69-70,75-77,79, 101-103, 114, 120,123,125,127,347,349,386,388,397,407, 415-416,431,433-43^,441,444 criterion of existence of, 13 definition of, 13 motion in neighborhood of, 13 stability of, 47-48,57-60,88-89,91-92,97,99, 165-166,175-177,202 stable, 13,202 unstable, 13,311
454
Friction - continued dry, 22,427 non- linear, 420 static, 428-429 zone of, 429
Froude
'
Gauss's theorem, 77
Generating solution, 86, 141-144, 146-147, 149, 152-153,158,160-161,172,181-182,184,190,199, 223,236,238,289,293,326 amplitude of, 146 definition of, 141 of Poincare, 350
Geometrical analysis of existence of periodic solutions, 105-115
Heteroperiodic oscillations, 301,335,338-340, 345,350 definition of, 295 existence of, 295-300 possibility of, 306 stationary, 348 synchronized with autoperiodic oscillation, 303 Heteroperiodic solution, 333-334,344
Heteroperiodic states, 344 condition of, 305 of non-linear systems, 305 stability of, 305 Heteroperiodic variables, 298
Hill equation, 361-362 definition of, 357
Half-stable equilibrium, 51
Hard self-excitation, 74,99,166,175,178,203,320, 331,339,^26,429 condition for, I65-I66 definition of, 71 example of, 87 subharmonic resonance of order one-half for, 336-339 Harmonic Balance of Energy, Principle of, 233, 238-239,286,297,301,310
Harmonic motion, 106
Improved first approximation, 210-214,224,232, 294-295 application of, 214-219 equations of, 288
Index of modulation, 371 critical value of, 312 of singularities, 76-77,117,345 of stepwise modulation, 363 of trajectory, 76
Instability, regions of, 3
Heteroparametric excitation, 355-356,360,363, 373-374 conditions of, 356,371 dependence on frequency and phase of parameter variation, 365-369 of dissipative system, 369-371 principal features of, 367-369
455
Isochronous oscillations, 189-190,202,239 Isochronous system, 171,265
Isoclines application of method of, 108 definition of, 20 equation of, 121 method of, 20-21,105,113
Lle'nard
continued cycle of, 418-419 equation of, 107,109-110 method of, 106-108 plane of, 107-115 limit cycles in, 113-115 qualitative analysis of, 4l8 theory of, 107
Jacobian, 140
Jump, 313,320-321,339 discontinuous, 166,386, 592, 395, 401-402,404,408,
Kaden, 352
Kaidanowskl, 420,422
Kennard, Prof. E. H.
Kryloff and Bogoliuboff, 2-3,67,135,137,181, 209-210,234 analytical method of, 4 argument of, 308 condition of self-excitation, 312 method of equivalent linearization of, 137, 2Q9,233 notation of, 254,263 quasi-linear method of, 331 quasi-linear theory of, 4,234,244,249,258,331, 342 theory of first approximation of, 183,186-188, 209,258,289,323 treatise of, 251,288,303
theorem of, 29
Levenson, M.
Limit cycle, 90-92,169,172,177,182,199,228-229, 234,302,326,344-345 amplitude of, 166 analytical examples of, 62-66 bifurcation of, 87 branch points of, 88 coalescence of, 72-73,166 condition for stationary oscillation on, 200 definition of, 23,62 degenerate, 70 disintegration, 87 equation of, 180 existence of, 99,181-182,198,200-203,224-229, 348 condition for, 169-173,176,179,199,207,227 criteria for, 75-77 proof of, 79 geometrical theory of, 63-86 half-stable, 70 in case of polynomial characteristic, 206-208 in clock, 67-68 in Lienard plane, 113-115 in Van der Pol plane, 113-115 inwardly stable, 70 motion on, 171 nature of, 99 of first kind, 125,130 definition of, 116-117 of Poincare', 62-86 of second kind, 123,125,129-130 definition of, 116-117 of thermionic generator, 158-159 outwardly stable, 70 physical examples of, 66-88 piecewlse analytic, 408,425-426,429,435 possibility of, 197 properties of, 75,105 representation of, 171 stability of, 62,64-66,68-73,75,85,93-94,98-99, 102, 161, 165-166, 169-171, 173, 175, 177-179, 200, 203-206,345,349 stable, 415,417,435 systems with several, 200-203 theorem of, 72-75,205-206 theorem of stability of, 70 topology of trajectories in presence of, 68-75
Lindstedt, 137 method of, 5,210 method of Gylden and, 209
Liapounoff, 4,173 criteria of stability of, 48-49,311-312 application of, 4g-55 equation of first approximation of, 114,176, 179,440,444 method of, 49,56,105 stability in sense of, 40-42,48-61,155,326 theorem of, 3,29,49,51-55,61 advantage of, 55 proof of, 49,51-55 treatise of, 161
Lle'nard,
Linear approximation, 1,22,100 Linear circuit characteristic equation of, 269 theory of, 252-255,259
105,181,382 analysis of, 4 approximation method of, 111 conditions of Car tan-, 182
456
Linear dissipative system with constant parameters, 258
Linear equation, 80, 136-137, 139, 179, 186,407,431 approximate, 100 canonical form of, 42-44 equivalent, 235-236 harmonic solution of, 167 Mathieu, 372 Mathieu-Hill, 356 of first approximation, 49 system of, 83 with constant coefficients, 1 Linear parameter, 265 Linear resonance, 249,317
Mo Her, 34
Linear system, 23,28,67,138,275,284 phase trajectories of, 7-23 Routh-Hurwltz theorem for, 3 with several degrees of freedom, 278
Motion aperiodic, 3,36,48 aperiodic damped, 106 equation of, 14 phase trajectories of, 18-19 asymptotic, 3,28,30-31,36 definition of, 16 of pendulum, 16,442 asynchronous, 125 damped oscillatory, equation of, 13 differential equation of, 11 discontinuous periodic, 383 harmonic, 106 in neighborhood of equilibrium, 40 in neighborhood of focal point, 13 in vicinity of equilibrium point, 29 in vicinity of saddle point, 26 isochronous, 42,l46
laws of, 2-3 linearized, 27 Method of Small,
Linearized differential equations, 67,87 Linearized equation, 67,87,100 equivalent, 244 Linearized motion, 27
Lochakow, 4,405,409
Ludeke, 321
1,66
non-isochronous, 172 non-stationary, 40 of conservative system, 42 of elastlcally constrained current-carrying conductor, 34-37 of harmonic oscillator, 86 of pendulum, 14-15,29-30^100,118,123 of representative point, 10 of rotating pendulum, 37-39 of synchronous motor, differential equation
of, 126 of system with one degree of freedom, 82 on limit cycle, 171
Mandelstam and Papalexi, 351,355-356,372,377 conditions of stability of, 330 discontinuous theory of relaxation oscillations
of, 383
heteroparametric machine of, 372 method of, 289,323 school of, 4,249,323,336 theory of, 249
Marconi, 5
Mathleu-Hill equation, 356-357,361-362,372-373 linear, 356 stability of, 361 subharmonic resonance on basis of, 373,377
Maxwell, theory of,
5
v
Mechanical system, 99,124,255-258,417 differential equation of, 257 parameter of, 417 self-excited oscillations of, 99-102 stability of, 102
Meissner, 362 (see also
Melde, 376 experiment of, 355
stability of, 82-86 periodic stationary, 63 quasi-isochronous, frequency of, 217 stable, 155 stability of, 24,41-42,85,155,359-360,373 in neighborhood of singular point, 4o-4l in vicinity of critical value of parameter, 33-34 in vicinity of equilibrium point, 24-26 of harmonic oscillator, 86 stationary, 23, 40-41, 148, 158,302 stability of, 161,358 stationary periodic, 61,435 uni-dimensional real, 8 unstable, 28-29,37 equation of, 13 phase trajectories of, 14-16
Multiperiodlc system equivalent linearization for, 274-278 equivalent parameters of, 275 resonance in, 271
Multiply degenerate system, 4lO-4l6
Hlll-Meissner equation)
Migulin, 377
Modulation critical value of index of, 312 index of, 371 index of stepwise, 363
457
Negative -continued definition of, 21 examples of, 21-22
Nodal point, 18-19,22,45-46,59,69,76-77,81-82, 101-103,114,127,345,547,349,397,407,411, 415,441,444 definition of, 13 criterion for existence of, 13 stability of, 13,45,47-48,58,60-61, 202,311 theorem for occurrence of, 45
Non-conservative system, 17,23,106 closed trajectories of second kind in, 121 cylindrical phase trajectories of, 119-121 existence of periodic solutions in, 224-229 with non-linear variable damping, 197-199
363
Non-linear system, 23,42,55,284,287 autoperiodic state of, 295 canonical form of, 52 characteristic equation for, 268 criteria of stability for, 3,55 frequency of self-excited, 265 heteroperiodic and autoperiodic states of, 295-300,305 periodic solution of, 138, 140 stability of, 49,52 with one degree of freedom, 278 with several degrees of freedom, 278-279
Non- linear variable damping, 105,178-179,197-199 non-conservative systems with, 197-199
Non-resonance, 271,290
1
Non-resonant oscillations, 275 Non-resonant self-excitation of quasi-linear system, 267-269 Non-resonant system, 277,281,303
Non-stationary motion, 40
Non-linear differenatial equation, 1,24,63,66, 75,85,135,137,141,201,216,234,272,417,431 conditions for periodicity of solutions of, 138-153 exact solutions of, 135 in dimensionless form, 323-325 of dissipative type, 179 of oscillating circuit, 162 of second order, 158 of self-excited oscillations, 66 of torsinal oscillation of shaft, 191
Non-linear dissipative damping, 192-196,240-241
Non-linear^ dissipative system, 182
Non-linear equation, 1, 118, 139, 181-182,242 of oscillation, 192 stability of, 50,52 Non-linear equivalent parameter, 277
458
Oscillations - continued isochronous, 189-190,202,239 maintained by periodic impulses, 423-435 non-isochronous, 191-192 non-linear, 100,137,233 self-excited, 197 non-linear equation of, 192 non-resonant, 275 of conservative system, 214 of limit-cycle type, 423 of pendulum, 1 of synchronous motor, 124-130 quasi-discontinuous, 408,417 quasi-harmonic, 183 quasi-linear, 128,158,393 frequency of, 217 relaxation; see Relaxation oscillations self-excitation of, 58-59,66 self-excited; see Self-excited oscillations stability of, 284-285,332,372,432 stationary, 66-67,71,263,269,283,297,303,344 on limit cycle, 200 subharmonic, 315,338-340 frequency of, 336 self -excitation of, 355 synchronized, stable condition of, 284
Parameter
continued 271,426 ordinary value of, 32 oscillatory, 308,385,436,441 parasitic, 387 effect on stationary states of dynamical systems, 436-445 existence of, 440 influence on equilibrium of dynamical systems, 438-440 periodic variation of, 308,322,356 small values of, 85,128,172,184,187,286,288,296, 327,382,438 variation of, 91-92,284,310,355,365,369,372-373, 376-377,415
-
of system,
Parametric excitation, 4,249,308-313,355-377 definition of, 308 of critically damped or overdamped circuit, 371 Parametric self-excitation, 36l
Parasitic capacity, 437 Parasitic parameter, 387 effect on stationary states of dynamical systems, 436-445 existence of, 440 influence on equilibrium of dynamical systems, 438-440
Pendulum, 360,445 as example of non-linear conservative system, 190-191 asymptotic motion of, 16,442 differential equation of, _190 elastic, 374 energy of, 26 frequency of, 3j2 Froude's, 21-22, 67, 178-180,417 mathematical, 426 mechanical, 342 motion of, 29-30,100,118,123 in neighborhood of unstable equilibrium, 14-15 of clock, 424 oscillations of, 1
Oscillator coupled electronic, stability of, 158 damped, differential equation of, 16 electron-tube, 296 harmonic, 86, 376,428 differential equations of, 7 motion of, 86 non-dissipative, 23 impulse-excited, phase trajectories of, 433-435
Oscillatory damped motion, 48,106 phase trajectories of, 16-18 Oscillatory parameter, 308,385,436,441 Oscillatory system, 68,271 non- linear, 272 resonance of, 271
Periodic solution, 66,79,82,84,86,122-123,135-136, 141,146,152-153, 159-160, 229,373 analytical, 383 condition for stability of, 328-330 existence of, 105-115, 121, 144-145, 220, 224 condition for, 136,152 geometrical analysis of, 105-115 in non-conservative system, 224-229 proof of, 105 of degenerate system of first order, 399 of non-linear problem, 140 of non-linear system, 138, l4o of quasi-linear equation, 221,326-328 stability of, 153-158
459
Periodicity, 116,156,147 condition of, 122,129,138-153
Phase trajectories - continued source of, 69,78 topology of, 171,348,406 analytical method of, 26-29 graphical method of, 25-26 in neighborhood of singular points, 25-29 in phase plane, 29-32 in presence of singularities and limit cycles, 68-75
Phase velocity, 9,16-17,24 definition of, 9
188,233-234,275,277,279,281,283,291
Phase angle, 7,138,159,172,216,310-313,368 critical, 368 equilibrium, 354 of ripple, 366-367 perturbation in, 311 Phase diagram, 26,30,367,433
36,43,45-47,51,62-63,68-69,71-72,74-75,77, 80,82,87, 105, 116-117, 119, 140-141, 148, 167, 169-170,174-175,198,326,344,348-349,362, 364,366,370,382,385-388,392,394-395,397, 405-406, 408-409, 413, 423-426, 433, 442 definition of, 7-8 representation of phenomenon in, 344-346
Phase space, 2,8,10,26,388,397,401 cylindrical, 116-130 uni-dimensional, 388,401
Phase trajectories, 2-3,7-23,40-41,43,45-46,62, 64-66,68-71,73-78,87,90,94,97,102,105-107, 109-110,113-114,117-118,125,128,167,169-170, 198, 326, 344, 349.,360, 364-366, 370, 382, 384, 386, 388, 392-394, 397, 399-400, 402, 406-407, 409, 411, 422-424,426,428-429,435,442-443 analytical, 383,4ll,4l3-4l4 analytical method of topology of, 26-29 behavior of, in neighborhood of singularities, 12-14 Cauchy's theorem of, 11, 14 continuum of, 62,128,407 continuum of closed, 12-13,62, 140-141 cylindrical of conservative system, 118-119 of non-conservative system, 119-121 definition of, 8 degenerate, 12 equations of, 66,88,128 existence of closed, 21,108-113,129 graphical method of topology of, 25-26 index of, 76 in presence of Coulomb friction, 427-429 in presence of singularities and limit cycles, 68-75 islands of, 30-32,62,119,128-129 of aperiodic damped motion, 18-19 of Hill-Meissner equation, 362,365 of impulse-excited oscillator, 433-435 of limit-cycle type, 402 of linear system, 7-23 of non-linear conservative system, 24-39 of oscillatory damped motion, 16-18 of second kind, 121-123,129 of unstable motion, 14-16 of Van der Pol equation, 348 proper, 12 sink for, 69,79 solution of differential equations represented by, 250
4,34,38,135,149,249,302,327,344, 355-356 analytical method of, 4,87,99,104-105,138-166 bifurcation theory of, 87-104 classification of singularities according to, 42,47 condition of, 158 criteria of, 105,165 criteria of stability of Liapounoff and, 311-312 curve of contacts of, 75 equations of, 187 expansions of, l4l functions of, 147,169 generating solution of, 350 indices of, 75-77 limit cycles of, 62-86 method of, 5,135-166,172,184,230,289,299,373 applied to systems with several degrees of freedom, 148 notation of, 154 quantitative method of approximation, 2 research of, 2 rule for ascertaining stability of motion in vicinity of critical value of parameter, 33-34 theorems of Indices, 76-77 theory of, 32, 63, 93-9 *, 117, 136-137, l4l, 162, 169, 175,178,200,209-226,247,289,292,373-374 topological methods of, 2,203 variational equations of, 328
,
1
Poisson, 183
460
Quasi-degenerate system, 390-391,445
Quasi-discontlnulty, 419,423
Relaxation oscillations - continued mechanical, 417-422 quasi-discontinuous, 415 quasi-discontinuous stationary, 392 stability of, 443-445 theory of, 4
Representative point, 8,10-16,36,40,41,45,47,62, 69,73,94,112,115,125,170,172,174,364-365,
367, 386, 393-395, 401-402, 404, 406-408, 412-413, 4 18, 421-422, 428, 434, 442-443, 445
motion of, 10
Resonance external, 3OI-322 definition of, 249 fractional- order, 301,303,305-309,373 in multiperiodic system, 271 in quasi-linear system, 271 internal, 281,284,301 definition of, 249,279 of order one, 284-285 linear, 249,317 non- linear, 4,249-377 external, 3OI-322 internal, 284 undamped, 321 of order n 323 of oscillatory system, 271 of quasi-linear system, condition of, 271 subharmonicj see Subharmonic resonance
,
Quasi-linear equation, 167,183, 186,189, 192, 200, 214,219,223,230,233-236,239,313,316,382,417 definition of, 136 of system with external excitation, 289 periodic solution of, 221,326-328 with forcing term, periodic solutions of, 326-328
Quasi-linear method of Kryloff and Bogoliuboff
331
Quasi-linear oscillations, 128,158,393 frequency of, 217 Quasi-linear system, 160-161,229,233,237,275, 278,301 aut onomous , 28 9 "bifurcation theory for, 160-161 condition of resonance of, 271 external periodic excitation of, 289-3OO Kryloff -Bogoliuboff theory of, 289 method of equivalent linearization applied to steady state of, 258-262 applied to transient state of, 262-267 non-resonant external excitation of, 294-295 non-resonant self-excitation of, 267-269 of differential equations, 148 resonance in, 271 resonant self-excitation of, 269-271 self-excitation of, 289 with several degrees of freedom, 251-271,274 with several frequencies, 285-288
Raylelgh, Lord, 341,351,355 equation of, 178-179,181-182,199 experiments with oscillating systems, 3O8
Saddle point, 23,28-31,34-36,38-39,41,46-48,58-59, 61-62,76-77,81-82,119,121,123,127,345,3^7, 349, 397, 407, 411, 415-416, 442, 444 criterion for existence of, 13 definition of, 12 example of, 16 motion in vicinity of, 26 negative, 444-445 positive, 444-445 theorem for occurrence of, 46
Relaxation oscillations, 68,128,130,381-445 definition of, 381,384 discontinuous stationary, 392 discontinuous theory of, 383,388-398,400-402, 405,409,436-437 examples of, 97,115
Secular terms, 141, 147, 159,185, 221 appearance of, 147 condition for absence of, 222-223,226-227 condition for elimination of, 291 definition of, I36 effect of, I83 elimination of, 137,209-210,221-223,226-228, 231-232
461
Secular terms - continued in solutions by series expansion, I83-I86 presence of, 186
Sekerska, 376
Self -excitation, 58-59, 73, 159, 278-279, 308, 360, 374 asynchronous, 299 autoparametric, 375,377 autoperiodic, 298,300 condition of, 80,99,173,204-205,264,266-267, 269, 307, 311-312, 316, 356, 371 critical value of, 297 disappearance of, 74 existence of, 292,368-369 hard, 74,99,166,175,178,203,320,331,339,426,429 condition for, 165-I66 definition of, 71 example of, 87 subharmonic resonance of order one-half for,
A.
5,136,382,418
Shottky, 5
336-339 heteroparametric, 364 lack of, 285,291,294,338 non-resonant, of quasi-linear system, 267-269 occurrence of, 417 of autoperiodic oscillation, 3OO of autoperiodic state, 308 of circuit, 433 'of clock, 426,429 of electromechanical system, 99,102-104 of electronic circuits, 66 of electron- tube circuit, 355 of equivalent linearized system, 278 of heteroparametric oscillation, 368 of non-linear circuit, 1 of oscillation, 58-59,66 of quasi-linear systems, 289 of shunt generator, 205 of simple circuit, 262-267 of subharmonic oscillations, 355 of system, 203,334,338 of thermionic circuits, 87,172 of thermionic generators, 95-99,161-166,293 parametric, 361 point of, 340 possibility of, 307,370,403,415 prevention of, 177-178,299 resonant, of quasi-linear system, 269-271 soft, 72,99,166,175,331,333,338,340,426 condition for, 164-165,299-300 definition of, 71 example of, 92,102 subharmonic resonance of order one-half for, 331-335 zone of, 297
Singular point, 9,11-28,35-38,40,49,51,64,69-70, 72,97,116-117,119-120,123,127,173,302,344346, 394,406, 411, 414 classification of, 42,47 coalescence of, 127,349-350 coalescence with limit cycles, 73 definition of, 11 nature of, 345 number of, 31 occurrence of, 311 of differential equations, 17,344 stability of, 38,91,345,350 topology of trajectories in neighborhood of, 25-29,68-75
Singularity, 2, 41, 6l, 78-79,121,176.179-180,311, 335,383,397,444 almost unstable, 41,442 classification of, 42,47 coalescence of, 29,34,36-37,39 distribution of, 346-350,444 index of, 76-77,117,3^5 nature of, 346-350 of differential equation, 3,11 simple, 29 stability of, 38,41,69,87,91-92,103,166,173,175, 348-349,415,442,444 criteria for, 47 transition of, 58,88,91 zone of, 82
Sink, 87,348 for trajectories, 69,79
Sommerfeld, 420
Source, 87,348 of trajectories, 69-78
Self-excited oscillations, 66-68,97,103,175,178179,182,376,381,415,435 amplitude of, 104 existence of, l8l frequency of, 26l non- linear, 197 non-linear differential equation of, 66 of electromechanical system, 99,102-104 of mechanical system, 99-102 of non- linear circuit, 1 stationary, 261-262,423 Self-excited state, 260 Self-excited system, 294,340 asynchronous action on, 300 non- linear, 351 frequency of, 265
Stability, 3,125,127 absolute, zone of, 6l analytical definition of, 41-42 conditions of, 55,99,148,156-158,317-320,330, 333-335,337,3^5,3^7,361,440 conditional, 61 criteria of, 22-23,33,333 applied to non-linear system, 3,55 of equilibrium, 60-61 of singularities, 47 defined for motion in neighborhood of singularities, 40 definition of, 40-41 exchange of, 34,38 geometrical definition of, 4l in sense of Llapounoff, 40-42,48-61,155,326 Kaufmann's criteria of, 61 Llapounoff s criteria of, 48-49,311-312 application of, 49-55 of autoperiodic oscillation, 306
462
Stability- continued
conditions for, 306 of coupled electronic oscillators, 158 of degenerate system, conditions for, 397 of electric arc, 440-443 of equilibrium, 33-55,40-61,71,85,87,93,127, 169-170, 202, 354, 558, 401, 438-445 according to Llapounoff , 48-61 Kaufmann's criteria for, 6l theorems for, 29 of focal point, 47-48,57-60,88-89,91-92,97,99, 165-166,175-177,202 of hetercperiodlc state, 305 of limit cycle, 62,64-66,68-75,75,85,95-94, 98-99, 102, 161, 165-166, 169-171, 175, 175, 177179,200,203-206,545,349 theorem of, 70 of Mathieu-Hlll equation, 36I of mechanical system, 102 of motion, 24,41-42,85,155,559-560,573 in neighborhood of singular point, 40- 41 in vicinity of critical value of parameter, 55-54 in vicinity of equilibrium points, 24,26 of harmonic oscillator, 86 of nodal point, 13,45,47-48,58,60-61,202,511 of non-linear equation, 50,52 of non-linear external resonance, 515,520 of non-linear system, 49,52 of oscillations, 284-285,552,572,452 of periodic motion, 82-86 of periodic solution, 155-I58 conditions for, 528-55O of point of equilibrium, 554 of position of equilibrium, 9 of relaxation oscillations, 445-445 of singular point, 38,91,345,350 of singularity, 38,41,69,87,91-92,105,166,175, 175,348-349,415,442,444 of stationary autoperiodic oscillation, 504 of stationary motion, 161,558 of stationary state, 292,518-319 Poincare-Liapounoff criteria of, 3II-5I2 regions of, 5 Routh-Hurwitz criteria for, 156,458,440 threshold of, 127 Static friction, 428-429 zone of, 429
Synchronization, 285-284 of autoperiodic with heteroperiodic oscillation, 305 of frequencies, 552,554 zone of 284
,
System absolutely degenerate, 591,442 autonomous, 424 definition of, 10 non-linear non-conservative, 62 quasi- linear, 289 with one degree of freedom, 105 completely degenerate, 590 conservative, 7, 25, 25, 58, 40, 42,106, 128, l6l, 185,189,224,267,562,375-376 cylindrical phase trajectories of, 118-119 equilibrium of, 40 motion of, 42 oscillations of, 214 periodic motion In, 62-63 points of equilibrium in, 34 degenerate, 148,440,442,445 conditions for stability of, 397 of first order, periodic solutions of, 599 transition between continuous and discontinuous solutions of, 415-416 with one degree of freedom, 599-409 degeneration of, 441 dissipative, 201,207,425 heteroparametric excitation of, 569-571 doubly degenerate, 597,415-414 dynamical, 10-11,52,51,69,82,87-88,107 differential equation of, 20 equilibrium of, 458-440 stationary states of, 40,456-445 electrical, 255-258 electrodynamlcal, 87 electromechanical differential equation of, 117-118 self-excitation of, 99,102-104 self -excited oscillations in, 99,102-104 equivalent linear, 256,281,286 equivalent linearized, 514
463
System - continued self-excitation of, 278 isochronous, 171,265 linear, 23,28,67,138,275,284 phase trajectories of, 7-23 Routh-Hurwitz theorem for, 3 with several degrees of freedom, 278 linear dissipative with constant parameters, 258 mechanical, 99,124,255-258,417 differential equation of, 257 parameter of, 4 17 self-excited oscillations of, 99-102 stability of, 102 multiperiodic equivalent linearization for, 274-278 equivalent parameters of, 275 resonance in, 271 multiply degenerate, 4lO-4l6 non-conservative, 17,23,106 closed trajectories of second kind in, 121 cylindrical phase trajectories of, 119-121 existence of periodic solutions in, 224-229 with non-linear variable damping, 197-199 non-degenerate, 443-444 non-dissipative, 268 non-isochronous, 240,271 non- linear see Non- linear system non-linear conservative; see Non-linear conserative system non- linear dissipative, 182 non- linear non-conservative, 23,40,63,105 autonomous, 62 higher approximations for, 223-229 non-resonant, 277,281,283 of degenerate differential equations, 396-398 of equations of first order, 9 of first order, 97 of linear equations, 18, 23,80,83 oscillating, experiments with, 308 oscillatory, 68,271 non-linear, 272 resonance of, 271 parameter of, 271,426 quasi-degenerate, 390-391,445 quasi-isochronous, 269 quasi-linear; see Quasi-linear system resonant, 281,301-305 self -excitation of, 203,334,338 self -excited, 294,340 asynchronous action on, 300 non-linear, 351 triply degenerate, 397,414 underexcl ted', 335 subharmonic resonance for, 335-336 with external excitation, 289 with external periodic excitation, 323 with internal resonance, 281 with more than one degree of freedom, 75 with non-linear variable damping, 197-199 with one degree of freedom, 10,82,427 conditions of periodicity for, 138-148 differential equation describing, 405-409 motion of, 82 with several degrees of freedom, 148 Poincare method applied to, 148 with several limit cycles, 200-203 with two degrees of freedom, 148-153,410 with variable damping, 105
;
Thermionic generator, 5,73,137,243-245,424 amplitude of oscillation in, 67 frequency of, 158-159 limit cycle of, 158-159 self-excitation of, 95-99, 161-166, 293 condition for hard, I65-I66 condition for soft, 164-165 self-excited, 293
Threshold, 48,370,394,407 critical, 3,11,294,373,406 of stability, 127
Threshold condition, 82
Topological methods, 7-130 of Andronow and Witt, 342 of Poincare, 203 of qualitative integration, 2-4 advantage of, 2 limitations of, 3 Topological representation, 148 Topological structure of trajectories, 4o6 Topological study of trajectories of Van der Pol's equation, 348
Topology of Hill-Meissner equation, 362-365 of phase trajectories, 171,348,406 analytical method of, 26-29 graphical method of, 25-26 in neighborhood of singular points, 25-29 in phase plane, 29-32 in presence of singularities and limit cycles, 68-75 of Van der Pol plane, 169-172,175
Trajectories; see
Phase trajectories
Transition of singularities, 58
theorem of, 24
464
Van der Pol - continued plane, 113-115,167,169-172,175
limit cycles in, 113-115 solution, 197,3^8-350 theory of, 1,178,200,292 theory of performance of heart, 68 variables of, 169,171-172
Velocity field, 11
Vincent, 34l
Wronsklan, 357
/I
Date
Due
War
/> 1
1932
1111 JUL
1 1QW 3 Wit
Introduction to
531M666i C