Documente Academic
Documente Profesional
Documente Cultură
1 JANUARY 2002
222
PAPER
SUMMARY Finding DC operating-points of nonlinear circuits is an important and dicult task. The Newton-Raphson
method employed in the SPICE-like simulators often fails to converge to a solution. To overcome this convergence problem, homotopy methods have been studied from various viewpoints. The
xed-point homotopy method is one of the excellent methods.
However, from the viewpoint of implementation, it is important
to study it further so that the method can be easily and widely
used by many circuit designers. This paper presents a practical
method to implement the xed-point homotopy method. A special circuit called the solution-tracing circuit for the xed-point
homotopy method is proposed. By using this circuit, the solution curves of homotopy equations can be traced by performing
the SPICE transient analysis. Therefore, no modication to the
existing programs is necessary. Moreover, it is proved that the
proposed method is globally convergent. Numerical examples
show that the proposed technique is eective and can be easily
implemented. By the proposed technique, many SPICE users can
easily implement the xed-point homotopy method.
homotopy method, xed-point homotopy, DC
key words:
1.
Introduction
The author is with the Faculty of Science and Engineering, Chuo University, Tokyo, 112-8551 Japan.
a) E-mail: inoue@po.cc.toua-u.ac.jp
The algorithms proposed in [15], [16], [18] were implemented on SPICE in SANYO Electric Co., Ltd. and they
are working very well in circuit simulation.
223
(1)
, Dg g(Dg T v) + DE i + J = 0
, DE T v E = 0
(2a)
(2b)
where f = (fg , f E )T , fg : Rn RN , f E : RN
RM , x = (v, i) Rn , and n = N + M . The variable
vector v RN denotes the node voltages to the datum
node and the variable vector i RM denotes the branch
currents of the independent voltage sources. Also, the
continuous function g : RK RK is a VCCS (voltagecontrolled current source) type. In addition, Dg is an
NK reduced incidence matrix for the g branches and
DE is an N M reduced incidence matrix for the independent voltage source branches. Moreover, J RN is
the current vector of the independent current sources
and E RM is the voltage vector of the independent
voltage sources.
Consider solving the MN equation (1) or (2) by
the FP homotopy method [4], [19], [21], [22]. The FP
homotopy equation is written as follows:
h(x, t) , tf (x) + (1 t)A(x x0 ) = 0
2=1
(x i (s)) + t(s)
(5a)
(5b)
iI
(3)
3.
Solution-Tracing Circuit
224
Fig. 1
Fig. 2
Otherwise, set xI = x0 .
(2) Execute the transient analysis using xI and t0
as the initial values.
Note that the arc-length s is regarded as the time in
the transient analysis. Also, by utilizing some command provided by the SPICE-like simulators, the transient analysis automatically stops when t reaches 1.
In Step 2, we begin to trace the solution curve of
Eq. (5) starting from (xI , t0 ), where xI is the solution
to h(x, t0 ) = 0 and close to x0 , and t0 is a positive value
suciently close to zero. By utilizing t0 > 0, we can
apply the solution-tracing circuit technique to the FP
homotopy method. Since the rst term in Eq. (3) does
not completely disappear at t = t0 , the initial solution
xI at t = t0 is not generally equal to x0 . However, it
is very easy to compute the initial transient solution,
since, in Eq. (3), the second term becomes major at
t = t0 (t0 0) and A(x x0 ) is linear.
If GFP and RFP are suciently large and t0 is close
enough to 0, then xI becomes very close to x0 , i.e.,
xI x0 . For instance, if x0 = 0 in practical circuits,
then we may regard xI as virtually equal to 0. Thus
the computation of the initial transient solution is not
always necessary. In this way, the solution curves of
homotopy equations can be traced by performing the
SPICE transient analysis. Therefore, no modications
to the existing programs are necessary.
Consider transforming Eq. (5) into circuits including some special circuit. Dividing Eq. (3) by t > 0 [19]
and considering Eqs. (2) and (4), we obtain
Dg g(Dg T v)+DE i+J +
DE T v E
(1t)
GFP (vv 0 ) = 0 (6a)
t
(1 t)
RFP (i i0 ) = 0.
t
(6b)
225
in the original circuit, respectively. The CCVS is connected in series with voltage source j in the original
circuit. In this way, we can make VCCSs and CCVSs
for all i {1, 2, , N } and j {1, 2, , M }, respectively.
In our approach proposed in this paper, A is extended in a more general form
g T
g GFP D
0
D
(7)
A =
0
RFP 1M
g is some N K
reduced incidence matrix
where D
(8)
+ M )n matrix represented as
where B is some (K
g T
D
0
B =
.
(9)
0
1M
T
JE
Jg
(13)
(14)
where uE RM , ug RN M , JE RM , and Jg
RN M .
Moreover, in order to make the discussion more
general, we introduce a positive semi-denite diagonal
dimatrix GFP = diag(GFPj ), j = 1, 2, ..., K, whose K
agonal elements are GFP s > 0 and the others are zeros.
Then we may rewrite the rst diagonal block in Eq. (7)
g T = Dg GFP D T , and Eq. (7) becomes
g GFP D
as D
g
0
Dg GFP DgT
A =
.
(15)
0
RFP 1M
In this case, GFP Dg T u = GFP vb represents the current
vector of the K branches of GFP whose branch voltage
vector is vb .
When we use the fundamental MC equation and
226
(16a)
(17a)
(17b)
Numerical Examples
Fig. 3
227
Fig. 5
Fig. 6
Fig. 7
Fig. 4
Diode-transistor circuit.
As shown above, by utilizing the solution-tracing circuit technique, the FP homotopy method can be implemented with no modications to the existing simulators. In this section, we further show that the FP
homotopy method can be easily implemented directly
on the existing simulators.
As discussed in Sect. 3, Eq. (6) is composed of the
original circuit to be solved and the other additional
terms. Hence, if we replace Eq. (5a) with Eq. (6), the
linearized equation of Eq. (5), used in the BDF curvetracing algorithm, can be written as
228
Jo
0
0
0
x
bo
=
+ bH (18)
+ JH
t
0
where J o and bo are the Jacobian matrix and the excitation vector of the original circuit, respectively. J H
and bH are the Jacobian matrix and the excitation vector of Eq. (5b) and the last terms in the LHS of Eq. (6),
respectively.
Equation (18) shows that the Jacobian matrix and
the excitation vector of the linearized equation can
be constructed by only adding J H and bH to those
of original circuit to be solved. This implies that we
need no modications to the existing programs but
only an additional subroutine which constructs J H and
bH . Therefore, by using our technique, the FP homotopy method can be easily implemented directly on the
SPICE-like simulators through a small modication to
the existing programs.
7.
Conclusions
[6] T. Ohtsuki, T. Fujisawa, and S. Kumagai, Existence theorems and a solution algorithm for piecewise-linear resistor
networks, SIAM J. Math. Anal., vol.8, no.1, pp.6999, Feb.
1977.
[7] A. Ushida and L.O. Chua, Tracing solution curves of nonlinear equations with sharp turning points, Int. J. Circuit
Theory & Applications, vol.12, pp.121, 1984.
[8] L.T. Watson, S.C. Billups, and A.P. Morgan, HOMPACK:
A suite of codes for globally convergent homotopy algorithms, ACM Trans. Math. Software, vol.13, no.3, pp.281
310, Sept. 1987.
[9] L.T. Watson, Globally convergent homotopy algorithm
for nonlinear systems of equations, Nonlinear Dynamics,
vol.1, pp.143191, Feb. 1990.
[10] K. Yamamura and K. Horiuchi, A globally and quadratically convergent algorithm for solving nonlinear resistive
networks, IEEE Trans. Comput.-Aided Des. Integrated
Circuits & Syst., vol.9, no.5, pp.487499, May 1990.
[11] L. Trajkovi
c, R.C. Melville, and S.-C. Fang, Finding
DC operating points of transistor circuits using homotopy
methods, Proc. 1991 Int. Symp. Circuits & Syst., pp.758
761, June 1991.
[12] Y. Inoue, DC analysis of nonlinear circuits using solutiontracing circuits, IEICE Trans., vol.J74-A, no.11, pp.1647
1655, Nov. 1991, also in Electronics and Communications
in Japan, Part 3, vol.75, no.7, pp.5263, July 1992.
[13] M.-C. Chang, J.-H. Chern, and P. Yang, Ecient and robust path tracing algorithm for DC convergence problem,
Proc. 1993 IEEE Int. Symp. Circuits & Syst., pp.16351638,
May 1993.
[14] R.C. Melville, L. Trajkovi
c, S.-C. Fang, and L.T. Watson,
Articial parameter homotopy methods for the DC operating point problem, IEEE Trans. Comput.-Aided Des.
Integrated Circuits & Syst., vol.12, no.6, pp.861877, June
1993.
[15] K. Yamamura, Simple algorithms for tracing solution
curves, IEEE Trans. Circuits & Syst., vol.40, no.8, pp.537
541, Aug. 1993.
[16] Y. Inoue, A practical algorithms for DC operating-point
analysis of large-scale circuits, IEICE Trans., vol.J77-A,
no.3, pp.388398, March 1994, also in Electronics and Communications in Japan, Part 3, vol.77, no.10, pp.4962, Oct.
1994.
[17] M.M. Green and R.C. Melville, Sucient conditions for
nding multiple operating points of DC circuits using continuation methods, Proc. 1995 IEEE Int. Symp. Circuits
& Syst., pp.117120, April 1995.
[18] Y. Inoue and K. Yamamura, Practical algorithms for DC
operating-point analysis of large-scale circuits, Proc. 1995
Int. Symp. Nonlinear Theory and its Applications, pp.1153
1158, Dec. 1995.
[19] L. Trajkovi
c and W. Mathis, Parameter embedding methods for nding DC operating points: Formulation and implementation, Proc. 1995 Int. Symp. Nonlinear Theory
and its Applications, pp.11591164, Dec. 1995.
[20] J.S. Roychowdhury and R.C. Melville, Homotopy techniques for obtaining a DC solution of large-scale MOS circuits, Proc. 33rd Design Automation Conf., pp.286291,
June 1996.
[21] K. Yamamura and S. Takahashi, Globally convergent algorithms using the xed-point homotopy for solving modied
nodal equations, IEICE Trans., vol.J81-A, no.7, pp.1094
1098, July 1998.
[22] K. Yamamura, S. Sekiguchi, and Y. Inoue, A xed-point
homotopy method for solving modied nodal equations,
IEEE Trans. Circuits & Syst.I, vol.46, no.6, pp.654665,
June 1999.
229
Appendix:
Proof of Theorem 1
uET i
and considering u DE i =
= i uE , we obtain
T
(u uI )
h(x, t0 )/t0
(i iI )
T
(u uI )
=
{h(x, t0 ) h(xI , t0 )}/t0
(i iI )
= (u uI )T {Dg g(DgT u) + DE i + J
+
(i iI )T {uE E 0 RFP (i i0 )
uEI + E + 0 RFP (iI i0 )}
(A 1)
Since g(vb ) + 0 GFP vb is uniformly passive on vbI =
DgT uI , there exists a gG > 0 such that
Next, we prove the above item 2 in the following lemmas. First of all, we have the following lemma as preliminaries for the subsequent lemmas.
Lemma 2: There exist a > 0, a 1 > 0 and a 2 > 0
such that for all x Rn and t [t0 , 1),
(u u0 )T hg (x, t)/t
(DgT u vbp 1 )2 2
+ (uE uE0 )T (i + JE ).
(A 2)
230
(A 3)
g(vbp )
LDgT u
vbp
{g(DgTu)g(vbp )+g(vbp )}
(DgT u vbp )T {g(DgT u) g(vbp )}
+ (DgT u vbp )T g(vbp )
(DgT u0 vbp )T {g(DgT u) g(vbp )}
(DgT u0 vbp )T g(vbp )
DgT u vbp 2 (3 + 4 L)DgT u vbp
3 4 .
1
i = 1 RFP
(uE E) + i0 .
(A 5)
+ i0 + JE }.
(A 4)
(A 6)
1
{uET uE (uE0 + E)T uE + uE0 E}
= 1 RFP
T
In order to prove the above item 2, we divide the discussion into three steps by dividing the variable vector
x into the three components, uE , ug , and i. In the rst
step, regarding uE , we obtain the following lemma.
231
(A 9)
(ug ug0 )T hgg (u, t)/t
=
{DgT (u
ug Jg
)} g(DgT u)
ug0 Jg
(A 10)
(A 11)
Jg = 8
T
(uE uE0 )g(DgT u) g(0)
DgE
T
DgE
(uE uE0 )g(0)
T
DgE
(uE uE0 ) GFP DgT (uu0 ).
{DgT (u
ug Jg
T
DgE
(uE
T
DgE
(uE
)} g(DgT u)
ug0 Jg
T
uE0 ) L DgT u
uE0 )g(0)
T
(uE uE0 ) GFP DgT (uu0 ).
DgE
232
Yasuaki Inoue
was born in Niigata,
Japan, on September 6, 1945. He received a diploma from the Department
of Electronics, Nagaoka Technical High
School, Niigata, Japan, in 1964 and the
D.E. degree in electronics and communication engineering from Waseda University, Tokyo, Japan, in 1996. From 1964
to 2000, he was with Sanyo Electric Co.,
Ltd., Gunma, Japan, where he was engaged in research and development in analog integrated circuits and analog/digital CAD systems. In Sanyo
Semiconductor Company, he was General Manager of the CAD
Engineering Department from 1993 to 1998 and the Memory Development Department from 1998 to 2000. He holds over forty
patents. Since 2000, he has been a Professor with the Department of Integrated Cultures and Humanities, also a Professor
with the Graduate School of Integrated Science and Art (Multimedia Masters Course), University of East Asia, Shimonoseki,
Japan. His research interests include numerical analysis of nonlinear circuits and systems, analog circuits, and CAD systems.
He was an Associate Editor of the IEEE Transactions on Circuits and Systems Part II from 1997 to 1999. He received the
Ishikawa Award from the Union of Japanese Scientists and Engineers in 1988 and the Distinguished Service Award from the
Science and Technology Agency, the Japanese Government, in
1999. Dr. Inoue is a member of IEEE.
Saeko Kusanobu
was born in
Okayama, Japan. She received the B.S.
degree in mathematics from Shimane University, Matsue, Japan, in 1994 and the
M.S. and D.E. degrees in information science from Hiroshima University, Hiroshima, Japan, in 1996 and 1999, respectively. Since 2000, she has been an Assistant Professor with the Department of
Integrated Cultures and Humanities, University of East Asia, Shimonoseki, Japan.
Her research interests include numerical analysis of nonlinear circuits and systems, and the applications of the statistical technique to image analysis. Dr. Kusanobu is a member of the
Japanese Society of Computational Statistics and the Japan Statistical Society.
233
Kiyotaka Yamamura
was born in
Tokyo, Japan, on December 19, 1959. He
received the B.E., M.E., and D.E. degrees,
all in electronics and communication engineering, from Waseda University, Tokyo,
Japan, in 1982, 1984, and 1987, respectively. From 1985 to 1987, he was a Research Assistant with the School of Science and Engineering, Waseda University.
From 1988 to 1998, he was an Associate
Professor with the Department of Computer Science, Gunma University, Kiryu, Japan. Since 1999, he
has been a Professor with the Department of Electrical, Electronic, and Communication Engineering, Chuo University, Tokyo, Japan. His research interests include numerical analysis of
nonlinear circuits and systems. From 1994 to 1998, he was an
Associate Editor of the IEICE Transactions on Fundamentals.
He received the Niwa Memorial Award from the Niwa Memorial Foundation in 1986, the Shinohara Memorial Young Engineer Award from IEICE in 1986, the Inoue Research Award
for Young Scientists from the Inoue Foundation in 1989, the
Telecom-System Technology Awards from the Telecommunications Advancement Foundation in 1990 and 1999, the Best Paper
Award from IEICE in 1999, the IBM Japan Science Prize in 1999,
and the Ohm Technology Award from the Promotion Foundation
for Electrical Science and Engineering in 2000. Dr. Yamamura is
a Senior Member of IEEE.