You are on page 1of 8

I12 IEEE TRANSACTIONS ON POWER ELECTRONICS, VOL. E NO. 2.

# p m 1993

Small-Signal Modeling of
Average Current-Mode Control
Wei Tang, Student Member, IEEE, Fred C. Lee, Fellow, IEEE and Raymond B. Ridley, Member, IEEE

Abstract-A recently proposed average current-mode control is


analyzed. A complete small-signal model for the control scheme
is developed. The model is accurate up to half the switching
frequency. This control scheme is suitable for applications where
the average inductor current needs to be controlled, as in power
factor correction circuits and battery charger/dischargers. The
subharmonic oscillation, commonly found in peak current-mode
control, also exists in this control. This subharmonic oscillation
can be eliminated by choosing a proper gain of the compensation
network in the current loop. Model predictions are confirmed
experimentally.

I. INTRODUCTION
Fig. 1. Average current-mode control scheme.

T HE advantages of average current-mode control, such as


the ability to control the average inductor current and the
improvement of noise immunity, have been presented [I]. The
of concern is the subharmonic oscillation at half the switch-
ing frequency. In this paper, a complete small-signal model
control scheme is illustrated in Fig. 1. The inductor current is
is developed for average current-mode control. It generates
sensed and fed into a compensation network to obtain its dc
all the transfer functions needed for design purposes. The
information. The output of the compensator is compared with
sampling nature of current-mode control is considered in the
a sawtooth ramp to generate PWM control. Since the average
development of the model, so that the subharmonic oscillation
current is used as a controlled quantity, average current-mode
can be predicted. Based on the small-signal model, design
control is particularly suitable for power-factor correction
guidelines are established, and the model is experimentally
circuits and for applications where a constant current source
verified.
is needed, such as a battery charger circuit.
The main difference between average current-mode control
and peak current-mode control is that in the former, the induc- MODELING
11. SMALL-SIGNAL
tor current is averaged and compensated by a compensation One major difference between peak current-mode control
network. In peak current-mode control, however, only the and average current-mode control is that the inductor current
switch current is sensed, and no compensation exists in current of the latter is averaged by a current compensator. Due
loop. The PWM conductance control proposed in [2] senses to the similarity of the two control methods, the modeling
the inductor current and compares it with a triangular ramp technique similar to that of peak current-mode control [ 5 ] can
to generate PWM control. Also, the switching frequency of be applied to the modeling of average current-mode control.
PWM conductance control is fixed; both turn-on time and The existence of the current compensator makes the modeling
tum-off time vary according to the control. Adding an integral- of average current-mode control more complicated.
lead network to the current loop [3] makes PWM conductatxe
control similar to average current-mode control. The low- A . Modulator Gain
frequency small-signal analysis of PWM conductance control Because of the presence of the current compensator in
was presented in [3]. average current-mode control, as shown in Fig. 2 , the current
Small-signal analysis, especially a continuous-time small- loop transfer characteristics are quite different from those of
signal model, is very useful in the control loop design of peak current-mode control. In peak current-mode control, the
power supplies. As mentioned in [4], [ 5 ] ,current-mode control inductor current is summed with an external ramp and directly
exhibits certain properties of a sampling system. One problem compared with the control voltage. The modulator gain of peak
current-mode control is [ 5 ]
Manuscript received November 6, 1991; revised September 25, 1992. This
work was supported in part by Lambda Electronics, Inc.
The authors are with the Virginia Power Electronics Center, Department
of Electrical Engineering, Virginia Polytechnic Institute and State University,
Blacksburg, VA 24061. where S, is the slope of the external ramp, S, is the inductor
IEEE Log Number 9206671. current on-time slope, and T, is the switching cycle.
0885-8993/93$03.00 0 1993 IEEE
TIW6:ol ai.: SMALL-SIGNAL MODELING OF AVERAGE CURRENT-MODE CONTROL 113

r SMALL SIGNAL

1T,

Fig. 2. Current compensator and modulator.

From Fig. 2, it can be seen that the inductor current


waveform has been changed by the compensator, and the
equivalent turn-on (t = DT,) slope of the modified waveform
Sk can be calculated as
-AF Ws)

" cG P H

Fig. 3. Small-signal model of average current-mode control.

where loop gain always exhibits a 180" phase shift at half the
switching frequency no matter where the second pole is
placed. If the second pole is placed after half of the switching
frequency, it does not significantly affect the current loop gain.
Since the purpose of the second pole is to eliminate high-
frequency noise, it should be placed after half of the switching
frequency.
If the second pole of the current compensator is placed after
The modulator gain of average current-mode control is deter- half of the switching frequency, the sampling gain used in peak
mined by the sum of the external ramp slope and the turn-on current-mode control can be used in average current-mode
time slope of the modified current waveform: control. At high frequency, the characteristics of the current
loop of average current-mode control are almost the same as
those of peak current-mode control. Due to the existence of the
low-pass filter in the converter power stage, the voltage loop
does not behave as a sampling system, and the compensator
B. Sampling Gain
pole exists in the voltage loop.
In the small-signal sense, the current loop behaves as a
sampling system [5].According to sampling theory [6], the C. Feedback and Feedfonvard Gain
phase shift of the system transfer function is always real at half
The small-signal model of average current-mode control is
the sampling frequency (the sampling frequency of a current-
shown in Fig. 3, where
mode control system is equal to the switching frequency).
When modeling peak current-mode control, the sampling
effect is approximated by the sampling gain [3] H e ( s ) , a Wi (I +9
double RHP zero at half the switching frequency:

Although a current compensator exists in the current loop,


'IT
w, = - (9) the switching ripple at the output of the compensator is still
Ts . comparable to the external ramp size, as shown in Fig. 4.
In average current-mode control, an integrator and a lead-lag Furthermore, the compensator output is a function of the input
network are employed in the current loop. The zero is usually and output voltages and the pole-zero locations of the current
placed before the power stage filter frequency to ensure the compensator. The duty cycle is affected by the perturbations
stability of the current loop, so the phase shift of the integrator of input and output voltages.
is canceled by the zero at half the switching frequency. The effect of the input and output voltage perturbations on
The effect of the second pole on the current loop was the duty cycle can be modeled by feedback and feedforward
studied. Due to the nature of the sampling system, the current gain terms k , and k f from input and output voltages 131.
114 IEEE TRANSACTIONS ON POWER ELECTRONICS. VOL. E, NO. 2, APRIL 1993

TABLE I
FEEDFORWARD
AND FEEDBACK
GAINSFOR AVERAGE
CURRENT-MODE
CONTROL

Buck Boost BuckBoost

kf - YD DE ' T- % ( -EL
LL
L < - %1 (L DD'T
e t
k, -E& - EL( * *

Fig. 4. Steady-state waveforms of average current-mode control

Referring to Fig. 2, the following relation can be obtained:

U, = U, - G , ( s ) ( R i i ~
- w,) (12)
bPL
where G,(s) is the current compensator transfer function

G c ( s )= G,(s)G,(s) (13)
and om is the output of the current compensator. If the average
quantity is considered, (12) becomes
vc =o
(wm) = 21, - K ( & ( ~ L-) U,) (14) Fig. 5. Simplified small-signal model for deriving feedforward gain k f .
where the quantities ( i ~ )and ( w m ) denote the average value
of z~ and w,, respectively. K is the dc gain of the current and
compensator; it is finite for a real op-amp.
a=
From Fig. 4, the following equation can be derived:

(om) = Seton + +AV,.

Substituting (15) into (14),

K(RZ(2L)- U,) = U , - Set,, - .+J,

Equation (16) gives the relationship between the average


111. SMALL-SIGNAL
CHARACTERISTICS
inductor current and input and output voltages. By perturbing
(16) with respect to <ZL> and w,, the Gg to <;L> transfer By using the small-signal model derived in the previous
function can be obtained. The same transfer function can also section, all of the small-signal characteristics can be generated
be derived from Fig. 5 , with k f as an unknown: through simulations. Simplified analytical transfer functions
are also derived to gain physical insight and to facilitate the
design.
A buck converter was used as an example to show the small-
signal characteristics of the average current-mode control. All
By comparing these two transfer functions, k f is obtained. k , of the characteristics are generated from the model shown in
can be obtained through the same procedure. Table I lists k f Fig. 3 and are compared with those of peak current-mode
and k , for three basic PWM converters employing average control. The power stage parameters of the buck converter
current-mode control where were
V, = 14 V
v, = 5 v
R = 1 R
L = 37.5 p H
C = 380 pF
TANG et al.: SMALL-SIGNAL MODELING OF AVERAGE CURRENT-MODE CONTROL 115

gain (dB) gain (dB)


30,

u+ - 1435

-loI
-20 I
1m m 5m 1.m 2.000 5.m l0.W 2o.m
I
-'OL
.mlm aa xc ,.m zm Lm

Frequency(Hz)
l0.m mm

Frequency (HZ)
phase (degree) phase(degree)
I

I
lm w xa 1.m zm I" l0.m am Im zm rxr l.m zm 5.m 1o.m ar.m
Frequency (Hz) Frequency(Hz)

Fig. 6. Current loop gain with U ; as a running parameter. Fig. 7. Current loop gain with wT as a running parameter.

R, = 20 m52 B. Control-to-Output Voltage Gain


F, = 50 kHz.
The control-to-output voltage gain is defined as the control-
The current loop elements value were
to-output voltage transfer function of average current-mode
R; = 0.152 control with the current loop closed. Fig. 8 shows these
Ri = 2.2 kR transfer functions with w; as the running parameter. It can be
R f = 30.5 k52 seen from Fig. 8 that the control-to-output gains have a low-
C j , = 5.8 nF frequency pole and a pair of double poles at half the switching
Cf, = 220 pF. frequency, and the damping of the double poles is affected
By varying R1 and Cf,, different w; and w; were obtained. by the integrator gain of the current compensator (it is also
affected by the extemal ramp slope, which is not shown in Fig.
A. Current Loop Gain 8). The current compensator pole wp also exists in control-to-
Current loop gain is defined as the loop gain measured at output gain. An approximated expression of control-to-output
the output of the duty cycle modulator with the current loop gain is given as follows:
closed. Its analytical expression is given in (21): 6,
,.-
R + sRcC) (1
T ; ( s )=
FmRiV,
~
(1 + sRC) - N -

"
Ri [1+ sR(C + Cx)]
[l+s(;+cRc) +aZLC]
where
1
c -- F,V,w;R;
It can be seen from (21) that the dc gain of the current loop is and
affected by both the extemal ramp (appearing in F,) and the
integrator gain U ; ; its shape is affected by the compensator
zero w,. While in peak current-mode control, the shape of It- t-
WnQp
the current loop gain is fixed, and the gain is only affected
by the slope of the extemal ramp; hence, there is more where
freedom in designing the current loop for average current-
mode control. Fig. 6 shows a set of current loop gains with
wi as a running parameter. The current loop gain with w, as
a running parameter is shown in Fig. 7. It can be seen from
Figs. 6 and 7 that the low-frequency portion of the current gain
is much higher than that of peak current-mode control [3]. It
means that there is less low-frequency error in the current loop From (25), it can be seen that the damping factor of the
for average current-mode control. double pole at half the switching frequency is affected by
116 IEEE TRANSACTIONS ON POWER ELECTRONICS, VOL. 8, NO. 2, APRIL 1993

gain (dB)
W

wb
20.
-.-.-..=..-
10

0
\ w, -3.3ZE+5 0 wi
\:.
- 7.55E+'$.<:..
,
\'

,I
a' I
0

20
tm xo m ~.mam xm 10.m w,m
Frequency (Hz)

phase (degree)

Frequency (Hz)
IM

PS
,m an ya ,ma 2"

Frequency (Hz)
5ma 1 0 m
w
w m

Fig. 8. Control-to-output voltage transfer function with dz as a running Fig. 9. Control-to-inductor current transfer function with 'CI~ as a
parameter. running parameter.

both the extemal ramp slope and the integrator gain w ; of gain (dB)
Jo
the current compensator. The approximation is very close to I
the simulation results shown in Fig. 8.

C. Control-to-Inductor Current Gain


The control-to-inductor current transfer function with the
current loop closed is shown in Fig. 9. It can be seen from
Fig. 9 that the gain characteristics are almost flat, except for a ,m m 5m ,aa 2 m 5aa ? o m zona m m
1
small dip in the midband. An analytical approximation of the

'
Frequency (Hz)
transfer function for the buck converter is given in (26): phase (degree)

-=I

.<..
4 '

It can be seen from (26) that there is a pole-zero pair near


w WR:

1 Frequency (Hz)
Fig. IO. Control-to-inductor current transfer function of peak current-mode
control.

It should be noted that there is another pole-zero pair near


w = w,, in the control-to-inductor current gain, which is not D.Audiosusceptibility
given in (26). These two pole-zero pairs produce the small Because of the inherent negative feedforward from the
dip in the midband of the control-to-inductor current gain. input voltage, the audiosusceptibility of average current-mode
For most operating conditions, this dip is smaller than 2 dB. control with the current loop closed is smaller than that of
By properly controlling the resonance at half the switching converter power stage. A set of audiosusceptibility of the buck
frequency, a wide and flat control-to-inductor current transfer converter with average current-mode control is shown in Fig.
function can be obtained. 11. An approximate expression of it is given in (28):
The control-to-inductor current gain of peak current-mode
control is shown in Fig. 10, with the ratio of the external ramp CO R ( k f F m V g D )
_ N +
(1 sR,C) +
..-
+ +
to the inductor current on-time slope as a running parameter.
Fig. 10 shows that it is impossible to obtain a wide and flat
control-to-inductor current gain with peak current-mode
vug FmVugRiw; [l sR(C Cz)]
(
+

:'J -
Fh(s).
(28)
TANG et al.: SMALL-SIGNAL MODELING OF AVERAGE CURRENT-MODE CONTROL 117

gain (dB) gain (dB)


-20r

i
"

-Y -,o Experiment-

tm am xo 1.m zm 5m mam 20.000


I Prediction ----
-20
Frequency(Hz)

Fig. 11. Audiosusceptibility with w, as a running parameter.


.-

Since kf has a negative value, it is possible to completely


null the circuit response to the input voltage perturbation by 93

selecting Se and w,, but kf and F, have nonlinear relations


45
with Se and w i , and it is difficult to find the exact values to null
the audiosusceptibility. While in peak current-mode control, it o wriment-
is easy to choose an external ramp slope which totally nulls Prediction - - - -
45
the input perturbation. rm a0 xo i.m zmo 5.
m i0.m zo.rm
Frequency(Hz)

Fig. 12. Measurement and prediction of the current loop gain for a
buck converter.
IV. DESIGNGUIDELINES
After studying the small-signal characteristics of average
current-mode control, certain design guidelines have been gain (de)
developed.
The current compensator can be designed as follows. First,
wh
the second pole should be placed after half the switching
frequency. The zero should be placed at lesat one decade
before half the switching frequency. Second, the function of
the external ramp is similar to that of the sawtooth ramp in
-20
wi - 7.55E+4
voltage-mode control; this means that any sufficient large ramp Prediction - - - -
can be used as the external ramp. Third, for a given ramp,
choose the integrator gain wi which makes Q p = 1. It gives
proper damping on the resonant peak at half the switching
frequency.
By setting Q p in (25) equal to one

Exprimen-

im am PO 1.m zm 5m 10.m =.ox


Frequency (Hz)
The voltage loop design of average current-mode control is
similar to peak current-mode control. Fig. 13. Measurement and prediction of control-to-outputvoltage gain for a
buck converter.

V. EXPERIMENTAL
VERIFICATION The control-to-output voltage and control-to-inductor cur-
A buck converter was built with the same component rent gains, measured with the current loop closed, are shown
values as those given in the previous section. To measure in Figs. 13 and 14, respectively. The measurements again show
the current loop gain, a digital modulator [7] was used to very good correlation with the theoretical results. The peak of
ensure that the correct sampled-data loop gain was obtained. the gains at half the switching frequency clearly shows the
All other measurements were performed with conventional existence of two complex poles.
analog measurement schemes. The measurement and predicted result of the audiosuscep-
The measured and predicted current loop gains are shown tibility of the buck converter with the current loop closed are
in Fig. 12. Both the gain and phase measurement agree very shown in Fig. 15. Again, the measurement and prediction agree
well with predictions up to half of the switching frequency. very well.

..-
118 IEEE TRANSACTIONS ON POWER ELECTRONICS, VOL. 8, NO. 2, APRIL 1993

gain (dB) The current loop gain of average current-mode control


always exhibits a 180 o phase shift at half the switching
frequency no matter where the second pole of the current
- 7.558+4\ compensator is placed. If the pole is placed after half the
I w,

switching frequency, the noise picked up by the sensing


network is attenuated, while the stability of the system is not
affected.
The control-to-output voltage gain has more phase delay
im w YI) ?.m zm saa ro.mo z0.m

Frequency (HZ) than that of peak current-mode control because of the existence
phase (degree)
of the second pole of the current compensator. The resonant
peak caused by the complex poles at half the switching
frequency can be controlled by selecting the gain of the current
compensator. In peak current-mode control, the resonant peak
can only be damped by the extemal ramp. This allows more
flexibility in the design of average current-mode control.
Based on the small-signal analysis, a design guideline is
,80 Experiment- proposed. By properly selecting the gain of the current com-
Prediction - - - -
-zn
lca w %a l.ao zm Iaa 10.m w.m
pensator, subharmonic oscillation can be avoided. An almost
Frequency (Hz) flat control-to-inductor current gain can be achieved when the
current compensator is properly designed. The accuracy of the
Fig. 14. Measurement and prediction of the control-to-inductor current gain model is confirmed with the measurements of a buck converter.
for a buck converter.

REFERENCES
[l] L. H. Dixon, Average current-mode control of switching power sup-
plies, in Unitrode Power Supply Design Seminar Handbook, 1990.
[2] D. OSullivan, H. Spruyt, and A. Crausaz, PWM conductance control,
in IEEE Power Electron. Specialists Con5 Rec., 1988, pp. 351-359.
[3] A. S. Kislovski, Small-signal low-frequency analysis of a buck type
PWM conductance controller, in IEEE Power Electron. Specialists
Con$ Rec., 1990, pp. 88-95.
[4] A. R. Brown, Topics in the analysis, measurement, and design of high-
Prediction - - - performance switching regulator, Ph.D. dissertation, Califomia Inst.
Q
Im aa ya lm 2 m la0 ?om am)
Technol., Pasadena, May 1981.
Frequency (Hz) [5] R. B. Ridley, A new small-signal model for current-mode control,
Fig. 15. Measurement and prediction of the audiosusceptibility for a Ph.D. dissertation, Virginia Polytechnic Inst. State Univ., Blacksburg,
buck converter. Nov. 1990.
[6] A. V. Oppenheim and R. W. Schafer, Digital Signal Processing. En-
glewood Cliffs, NJ: Prentice-Hall, 1975.
[7] B. H. Cho and F. C. Lee, Measurement of loop gain with the digital
VI. CONCLUSIONS modulator, in IEEE Power Electron. Specialists Con5 Rec., 1984, pp.
363-373.
In peak current-mode control, the peak inductor current is
sensed and compared with the control voltage derived from
the voltage loop. While in average current-mode control, it
is the average inductor which compares with the control
voltage. Hence, average current-mode control controls the
real averaged inductor current. This is particularly true when
the converter is operated in the discontinuous conduction
mode. When the buck converter employs average current-
mode control, the output current, which equals the average
inductor current, is also controlled. It means that an ideal
current source is achieved. Likewise, in the case of a boost Wei Tang (S90) received the B.S. and M.S. degrees
converter, the average input current is controlled, which makes from Tsinghua University, Beijing, P.R. China, in
it suitable for a power factor correction circuit. 1982 and 1985, respectively, both in electrical en-
gineering. Since 1988 he has been working towards
Because of the existence of the current compensator in the Ph.D. degree at the Bradley Department of
the control loop, the small-signal characteristics of average Electrical Engineering, Virginia Polytechnic Insti-
current-mode control are quite different from those of the tute and State University, Blacksburg.
From 1985 to 1988 he worked as a Senior En-
peak current-mode control. Due to the usage of an op-amp, gineer at the Research Center of Computer Appli-
the current loop gain of the average current-mode control cations, Chengdu, China. He is now a Research
possesses very high gain at low frequency. While in peak Assistant at the Vireinia
Power Electronics Center.
His research interests include modeling, analysis, and control of the switching
current-mode control, the low-frequency gain of the current power converter, and power factor correction.
loop is rather small. Mr. Tang is a member of Eta Kappa Nu.
TANG et al.: SMALL-SIGNAL MODELING OF AVERAGE CURRENT-MODE CONTROL 119

Fred C. Lee (S72-M74-SM87-F90) received Raymond B. Ridley (S90-M90) received the


the B.S. degree in electrical engineering from the B.S. degree from Boston University, Boston, MA,
National Cheng Kung University, Taiwan, in 1968, in 1981 and the M.S. and Ph.D. degrees from
and the M.S. and Ph.D. degrees from Duke Univer- Virginia Polytechnic Institute and State University,
sity, Durham, NC, in 1971 and 1974, respectively. Blacksburg, in 1986 and 1990, respectively, all in
From 1974 to 1977 he was employed as a member electrical engineering.
of the Technical Staff at TRW Systems, Redondo From 1981 to 1984 he was employed as a Senior
Beach, CA. He was responsible for designing power Engineer in the Power Systems Group at Prime
converters for spacecraft power systems. He joined Computer, where he worked on the design and
Virginia Tech in 1977. He is presently the James S. analysis of computer power supplies. His research
Tucker Professor in the Department of Electrical En- specialties include power converter control and anal-
gineering and the Director of the Virginia Power Electronics Center (VPEC). ysis, high-frequency converters, and computer-aided design for power sys-
In 1987 he was appointed as the director of the Technology Development tems. He served as the Assistant Director of Virginia Power Electronics Center
Center for Power Electronics of the Virginia Center for Innovative Technology. from 1987 to 1991. Presently, he is a consultant in power electronics.
He is the founder of the Centers Industry Partnership Program. To date,
62 companies from all over the world have subscribed to the program.
His research interests include high-frequency power conversion, distributed
power systems, space power systems, device characterization, and modeling
and control of converters and design optimization. During his career, he has
published over 80 refereed joumal papers, and more than 150 technical papers.
Dr. Lee has received nine best paper awards from various technical
conferences. He has been awarded ten patents, with five additional patents
pending. He has been an active consultant for over 20 companies in the
power electronics industries. He is currently a member of the AdCom of
the IEEE Power Electronics Society, and an Associate Editor of the IEEE
TRANSACTIONS ON POWER ELECTRONICS . He was the Chairman
of the 1987 IEEE Power Electronics Specialists Conference. Dr. Lee is a
recipient of the Society of Automotive Engineering 1985 Ralph R. Teeter
Educational Award, the IEEE Power Electronics Society 1989 William E.
Newel1 Power Electronics Award, the 1990 PCIM Award for Leadership in
Power Electronics Education, and the Virginia Tech 1990 Alumni Award for
Research Excellence.

You might also like