Documente Academic
Documente Profesional
Documente Cultură
6, DECEMBER 1972
tu
Abstract—A high-performance low-cost IC time-delay generator
has been developed to be used as a building block for automotive
electronic fuel injection systems. Time-delay accuracies are achieved
by employing the IC to precisely control both the reset time and ‘K V02
the initial ramp voltage of an external RC voltage sweep circuit Ic o~
“+ TIME DELAY ..-
by means of a gated voltage regulator. Reference information, GENERATOR L
tz t
derived from the rotating distributor shaft, triggers an input flip-flop REFERENCE V03
POINT /
that controls a reset generator to produce the required gating pulses. OS
-..
The resulting exponential voltage sweep is sensed at the input of L
three externally programmable voltage comparators that provide *3 ‘
VOLTAGE SWEEP
CIRCUIT
T
/-:-’ ~
V3
[(V, + @) – 1!?]is a minimum. This minimum value will
be 240 mV for the 3-ins time delay, since the linear slope
cOMPARATORS K, can now be defined as 12.5 ins/V. From (2), the worst
Fig. 2. Basic time-delaygenerator. case variation of &3 mV in the comparator offset voltage
(Darlington differential inputs) would produce a +l.5-
percent variation in the 3-ins time delay [1]. Also, for
a maximum variation of + 1 percent in the initial sweep
-4 b
voltage, a +4-percent change would occur in this same
1’ minimum delay interval, from (3). Thus, the required
E“-- --- VI -—- *l-percent time delay accuracy over the defined tem-
s v+
‘:2 c
perature range is difficult to achieve using this linear
1’
=
E
tl
t
voltage sweep technique not only because of the required
+ l-percent linear current source accuracy, but also
I
since the small variations in the comparator offset voltage
and the initial sweep voltage produce large errors in
(a) (b) the smallest time delay interval.
Fig. 3. ‘1’ime-delaygenerationusinga linearvoltage sweep genera- To reduce these IC parameter variations to a satis-
tor. (a) circuitry. (b) Resnlting waveform. factory level for this application, additional circuit
complexity would be required and a substantially larger
t, = At + K[(V, + r$) – E] (1) die area would result. This makes this linear voltage
sweep approach uneconomical, especially in view of the
where tight tolerance specifications and thus, other approaches
At reset switch closure time, were considered.
K slope of the linear voltage ramp (K = I/C),
B. Exponential Voltage Sweep
VL comparator reference voltage,
E initial sweep voltage, By replacing the precision current source with an
@ comparator offset voltage. external precision resistor, the external resistor–capacitor
sweep circuit that results can produce a precision expo-
The reset time At must be considerably smaller than the nential voltage sweep to again produce the desired time
second term of (1) to ensure that the maximum At variation delays, Fig. 4 [2]. The general expression for the time
will always produce less than a &l-percent change in the delay t, is
smallest time delay interval. In addition, variations in the
linear slope K, and thus the current source I, must always V.. – E
(4)
be within +1 percent, since the time delay is directly ‘L= At+ Rcln [ V.C–(V, *4) 1
proportional to these quantities. The normalized change where
in this time delay due to a change in the comparator
offset voltage is described by R external resistor value,
C external capacitor value,
(2) Voc positive supply voltage,
and the remaining parameters are as previously defined.
The normalized change due to variations in the initial The reset time At is again assumed negligible with respect
sweep voltage can also be written as to the second term of (4). The time delay, from (4), is
&w – AE essentially independent of the supply voltage V., provided
(3) the initial sweep voltage and the comparator reference
t, ‘( V,+ 1$) -E”
voltage are both referenced to V... The normalized
The term associated with time At has been omitted since change in this time delay due to a change in the com-
it is dominated by the linear voltage sweep term. parator offset voltage is described by
To evaluate (2) and (3), the magnitude of the pa-
rameters must be determined from the constraints of ~- Ad
t, =
. (5)
the fuel injection system. In general, the shortest required V,. – E
crankshaft referenced time-delay interval for most fuel ‘v” – ‘v’ * 4)] ln [ v,. – (v, * 4)1
464
+v~c
?+
v
+
IEEE JOURNAL OF SOLID-STATECIRCUITS,DECEMBER 1972
+ v~~
GATING
Vcc
WJ
TERMINAL
R INPUT
REFERENCE
OUTPUT
..- VOLTAGE ; ---
F =)
‘~ s *C ~ ‘EGuLATOR *C
= = = =
(a) (b)
(a) (b)
Fig. 5. Ideal switch simulation. (a) Desired switching circuit. (b)
Fig. 4. Time-delay generation using an exponential voltage sweep ,Switching function simulated by a gated voltage regulator.
generator. (a) Circuitry. (b) Resulting waveform.
It”- \l [V +
r-–––-––-
I qJ + ‘JLrL
~t. --”\
J&c
E+\
\.+p
$/%’)/+’0’ —.-— —
U,
I I
1 I
lrlc’ I-WI
I
‘1 (
I o o~
~T I I
RESET
I
VOLTAGE o V2
INPUT FLIP FLOP
GENERATOR REGULATOR I
I ‘REF
I
*
RES=
I
I
1- ———— ———
COMPARATORS J
reference voltage, must also provide this accuracy. The due to parameter variations in the ICI will always be
reset time At could be made insignificant with respect to less than +1 percent (the remaining accuracy will depend
the shortest time delay interval and thus would eliminate upon the accuracy of the passive components).
the need for a precision reset generator. Unfortunately,
A. The Indial sweep Voltage
this time must be greater than the worst case sweep
capacitor discharge time (determined by the discharge The reference voltage E for the voltage regulator is
current capability of the voltage regulator) but more established by a monolithic resistive divider that must
important, it must be sufficiently large to eliminate the be referenced to the power supply voltage V.., if variations
effects of dielectric storage. This results from charge in V,. are to have a negligible effect on the time delays,
being stored in the dielectric material while the capacitor The voltage regulator, shown in Fig. 7, consists of a
is charged and causes the capacitor voltage to increase noninverting unity gain dc amplifier that incorporates
slightly above the initial sweep voltage immediately after negative feedback to accurately reproduce the reference
the time interval At, if this interval was not sufficiently voltage across the sweep capacitor. The frequency stability
long to drain the charge from both the dielectric material of this regulator is maintained by the dominant pole
and the capacitor plates [4]. This effect can cause an produced at the output by these external RC components.
intolerable error in the initial sweep voltage. Unfor- At the timing reference, the emitter current to the dif-
tunately, a satisfactory reset time that eliminates the ferential amplifier is gated ON and since the capacitor
effects of dielectric storage is not always insignificant voltage will be significantly larger than the reference
with respect to the shortest time delay interval and, voltage, a current 1 will enter the base of the Darlington-
as a result, the reset circuitry must maintain a required connected output transistors, which rapidly discharges
accuracy. the sweep capacitor to the reference voltage E. After
A sufficient reset time interval depends on the dielectric time At, the emitter current to the differential amplifier
material and the value of the sweep capacitor. Nor the is gated OFF and any loading effect on the capacitor by
required 24-ins time constant of this system, a 0.6&PF the regulator is eliminated to allow the capacitor voltage
sweep capacitor was employed, since this value is suffi- to begin an accurate exponential sweep. Additional
ciently large to eliminate the dc loading effects on the loading on the capacitor by the collector leakage current
voltage sweep circuitry and small enough to allow rapid of the output power transistor is minimized by connecting
discharge by the voltage regulator—yet permits a satis- a resistor across the base-emitter junction of this power
factory reset pulsewidth of only 200 ps. The variations device to eliminate the beta multiplication of the collector-
in this pulsewidth, however, cannot exceed &15 percent, base leakage current (Ic,o) and to approximate the
if variations in the 3-ins time delay are expected to be Ic 50 kdW2X current.
less than A 1 percent.
B. The Reset Pulse Generator
These techniques ensure that for a given set of external
components and a defined temperature range (between The reset pulse generator uses one external reset
– 40° and + 125”C), the maximum time delay error, capacitor, Cm, which is referenced to ground to produce
466 IEEEJOURNAL OF SOLID-STATECIRCUITS,DECEMBER 1972
.
Vcc
fJa
v
I I
j
.-+k t
~MpARATORS
Vz
+ +
‘Z v CR
-T
Fig. 7. Gated voltage regulator that establishes the initial sweep cl,
voltage E. ——. % —
‘G-t (a)
‘REF
1 +
Vz q ~
~r
v
---1Atl ()
J ,,B = ,
qj
‘
1
%
Vz = I
‘- I
Vz I I (— .—— — =
7j- --1-- - -’ --— I
*t t
I I (b)
— At2 — Fig. 10. Reset waveformgeneratorcircuitry.
I I
I TO VOLTAGE
REGULATOR
v ~
I
c1
INPUT
Vz
a reset waveform that linearly ramps alternately between 1) The Reset Waveform Generator: The reset waveform
the voltage Vz and ground potential. In addition, it is generated by the basic technique shown in Fig. 9. With
incorporates two voltage comparators that are both the switch open, the current source I charges the reset
referenced to the voltage Vz/2 to produce a regulator capacitor CE to the voltage clamp Vz. The switch closes
gating pulse for each transition of the input flip-flop. at the first transition of the flip-flop to produce a dis-
The waveforms associated with this reset generator are charge current, also of value 1 (the difference between
shown in Fig. 8. the two current source values) that linearly discharges
DAVIS AND FREDERIKSEN : MONOLITHIC TIME-DBL.4YGENERATOR 467
I
r-4 At F-
tE!L
I -$-
2 1,
t
TO VOLTAGE
REGULATOR
To CR
TO F.F.
+ %
=
0s
‘4
*
Fig. 12. Reset comparator circuitry.
the capacitor from the voltage Vz to approximately maximum variations in these currents will contribute
ground potential. The second flip-flop transition causes less than a &15-percent error in the slopes and the
the switch to again open and allows the current sourc!e I symmetry of the reset waveform.
to linearly charge the capacitor from ground potential 2) The Reset Comparators: The reset gating pulses for
back to the voltage Vz. This reset waveform will be the voltage regulator are derived from the outputs of
symmetrical provided the discharge current is identical two comparators that are both referenced to Vz/2 and
to the charge current and the maximum variation in are alternately gated ON by the flip-flop. The reset wave-
the slope will be proportional to the maximum variation form is applied to the comparator inputs as shown in
in these charge and discharge currents. Fig. 11. For the first transition of the flip-flop, com-
The circuitry employed to realize this waveform gen- parator C’, is gated OFF, and comparator C, produces an
erator is shown in Fig. 10. The current in the collectors output until the reset waveform voltage decreases to
of the multiple-collector lateral p-n-p transistor Ql, the comparator reference voltage Vz/2. This produces
is derived from a common biasing network and has a the gating pulsewidth At,. For the second transition
tolerance of less than &15 percent over the defined of the flip-flop, comparator CZ is gated OFF, and com-
temperature range. One of these collector currents is parator C, produces an output until the reset waveform
controlled by a clifferential gate (transistors Qs and QJ, voltage increases to the voltage Vz/2. This produces the
which contributes a significant base current error IB second gating pulse Att.
to the collector current of transistor Q,. By employing The reset comparator circuitry, Fig. 12, consists of
transistors Q, and Q,, this current is multiplied by 2, two differential comparators (transistors Q, and Q, and
using emitter-area scaling, and is also inverted to produce transistors Q, and Q,,), which are both referenced to
the discharge current 2(1 — 1~). Fortunately, h~l? is V,/2 with the inputs driven out of phase. The emitter
characteristically a slow function of temperature for a currents for these differential pairs are alternately gated
lateral p-n-p transistor, and as a result, the base current ON by the flip-flop and the outputs are connected in
contributes less than a ~ l-percent variation in this parallel to perform the oRing function. The output lateral
discharge current. The charge current also includes this p-n-p transistor is activated by the output of either com-
base current error, since the lateral p-n-p transistor Q, parator to produce the controlled collector current I,
is connected in series with the second collector of tran- which is used to activate the voltage regulator. These
sistor QI. Thus, the charge and discharge current for the gating pulses will be symmetrical within ~ 15 percent
reset capacitor are both equal to (1 — 1~), and the (assuming a perfect reset waveform) provided the maxi-
468 IEEE JOURNAL OF SOLID-STATECIRCUITS,DECE)MBER 1972
CR 01 ’32 03
J15:K
10Kli!
!I
I I I’IWIIJ
1,
II
To
‘1
II
I
1= 1,
I 1,
5’ SK
I II
t I
t ‘1
II
I 2’ 2K
II
GI=TI .
10’
*“!!
::
IL
LZ
I -._
L-L
--s_-j
’-__
COMPARATORS
c
An Electronic Gyrator
HANS O. VOORMAN ANDARNOLD BIESHEUVEL
Abstract—After a brief introductionto the gyrator concept, its handling part is indicated. It consists of two differential
realizationas an accuratemonolithicintegratedcircuitis described. stages antiparallel. Each stage is supposed to convert
The inherentelectronicproblemsare consideredand performance
criteria are defined. Data are given. Applications,particularlyin accurately a voltage into a current. The electronic circuit
electric filters, are discussed and it is shown that in the lower (a two-port) together with the capacitor simulates an
frequency range the use of these “semifloating” gyratorscan be inductance L = RIRZC.
very advantageous.The monolithicintegrationof the gyrator can If R, = R, the electronic two-port is ‘[lossless” and it
be seen”as a major steptowardintegratedselectivity. behaves as a gyrator, as defined by Tellegen [2]. The
symbol for a gyrator is shown in Fig. 3.
INTRODUCTION The differential stages of the gyrator can be integrated
A.NY attempts have been made to replace LC in a monolithic process. For reasons of accuracy, however,
filters by filters which, while maintaining their the gyration resistors R, and R, must be connected
M excellent properties [1], do not require big and exteimally. Hence, a simulated inductance consists of
costly inductors. A possible method is to replace each an integrated circuit, two resistors, and a capacitor,
coil by a gyrator terminated in a capacitor.
To replace a coil we need at least another energy DESIGN
reservoir, e.g., a capacitor. If we compare the equations Many gyrator circuit configurations are known [3]-[5].
for an inductor and a capacitor we see that the role We present an accurate integrated realization using only
played by voltage and current is reversed (Fig. 1). To two external components (R,, RJ. This also gives the
simulate an inductance with a capacitor we must trans- most accurate inductance value for given component
form the current ic to a voltage v~ and the voltage VC tolerances’ and gives a low noise factor [6], [7].
to a current i~: Our circuit is based on the configuration of Fig, 2, The
inductance value is given by L = RIRZC (1 + 3), where
v~ = Rlic i~ = (1/RJvc, (1)
R,, R,, and C are values of external components and 8 is
where RI and Rz are resistances. We thus get an in- the influence (variable, temperature dependent) of the
ductance L = RIR,C. In this way all normal values of integrated electronic circuit. 16I must be made much
inductances up to very high values can be simulated smaller than the tolerances of RI, Rz, and C, e.g., lower
(e.g., R, = R, = 1 Mfl, C = 1 pF gives L = 1 MH).
Fig. 2 shows a circuit realizing (l). Only the signal- 1 We refer each time to inductance simulation instead of treating
the gyrator as a two-port because it is felt that the ‘very major
part of the applications of the gyrator will be inductance simulation.
Manuscript received April 20, 1972; revised July 31, 1972. In this case we have L = R,R,C but for a different design [5]
The authors are with the Philips Research Laboratories, Eind- L = RIRBRsC/R2, which is somewhat less accurate for the same
hoven, the Netherlands. component tolerances.