Average Current-Mode Control of Buck DC-DC

Converter With Reduced Control Voltage Ripple

Dalvir K. Saini*, Alberto Reattit, and Marian K. Kazimierczuk*
*Department of Electrical Engineering, Wright State University, Dayton, Ohio, USA 45435
{ saini. I I , marian.kazimierczuk} @wright.edu
tDepartment of Information Engineering, University of Florence, Florence, Italy 50139
alberto.reatti @unifi.it

Abstract-This paper proposes a technique to track and reg­ DC in put voltage DC output v oltage
ulate the "true average" current in any branch of a pulse-width Pulse-width
modulated dc-dc power converter. An example buck converter
Duty c ycle modulated dc-de Ind uctor cu r rent
in continuous-conduction mode is considered. A comprehensive power converter
characterization is presented for the proposed scheme. An over­
all dc and small-signal analysis of the inner current loop is
performed. The current average current-mode control methods
suffer from switching instability at low duty ratios due to a

b.
large ripple in the sensed current. Moreover, the current error
amplifier's high-frequency pole located at the switching frequency
neither alleviates the stability issue nor improves the dynamic t

response of the converter. In the proposed approach, a low­ Control voltage

pass filter, which attenuates any high-frequency ripple is placed
in the feedback path of the current loop. Consequently, the
control voltage is nearly dc and proportional to the actual
average value of the inductor current. The inner current loop Fig. I. Architecture of the inner loop average current-mode control technique
gain transfer function of the presented circuit has been derived. for pulse-width modulated dc-dc power converters with a low-pass filter in
The following critical path closed-loop transfer functions have the feedback path.
been derived: reference voltage-to-inductor current and reference
voltage-to-output voltage. Using practical specifications of a buck is the duty cycle. At low duty ratios, for example, D .-::: 0.15,
dc-dc converter, a theoretical framework to design the inner the crossover frequency ie ::;:, is, which violates the Nyquist
loop is presented. Verification of the theoretically predicted
sampling theorem. Moreover, in [J] and [2], it was shown that
transfer functions and transient analysis is performed through
simulations.
a high-frequency pole due to the error amplifier located at is
eliminates the switching frequency component in the sensed
I. INTRODUCTION current. This pole only guarantees the optimal slope condition;
Average current-mode control has been a preferred solution however, it does not contribute to an accurate average current
for current control in various applications such as LED light­ sensing.
ing, power factor correctors, and battery chargers [1] - [7]. It To overcome the above drawbacks, this paper proposes a
is also an attractive solution in two-loop dc-dc converters. Its technique, which tracks and regulates the true average current
inherent advantages are: (a) better noise immunity, (b) absence (or dc current) in any branch in the converter. A representative
of digital or logic circuits, and (c) no requirement for slope figure is shown in Fig. 1. A low-pass filter in the feedback
compensation. path ideally eliminates the high-frequency ripple in the sensed
The existing average current-mode control techniques have current and allows only its average (or dc) component to the
been discussed in great detail in [J] and [2]. In [J], the current error amplifier. In the described method, the feedback voltage
error amplifier was designed for an optimum proportional gain is ideally dc and is compared with the reference voltage to
at the point, where the down-slope of the inductor current yield a dc control voltage. The feedback-path filter pole ip is
and the up-slope of the sawtooth voltage waveform applied to located at ie < ip < is, where ie is the loop gain crossover
the pulse-width modulator are equal. However, this optimum frequency. The order of the stages in the loop is important. For
gain led to a high inductor current ripple; consequently, the an effective ripple suppression and true average tracking, the
control voltage contained a proportionally amplified ripple low-pass filter must be placed after the power stage. Therefore,
component, which could intersect the sawtooth voltage wave­ in the presented scheme, the filter is placed in the feedback
form more than once during each switching period. This path. However, adding a filter pole before the power stage,
resulted in switching instability imposing a limitation on both e.g., in the control circuit [1], [2], [7], does not reduce the
duty cycle and peak sensed current. Moreover, in [1], the ripple. Consequently, the transistor is triggered ON at the peak
crossover frequency of the uncompensated loop gain was inductor current instead of the average inductor current.
ie = is/(27rD), where is is the switching frequency and D In summary, the major advantages of using the described

978-1-5090-3474-1/16/$3l.00 ©2016 IEEE 3270

 inverting amplifier boosts its average value to that nearly equal
to the reference voltage VRI. The error voltage is given as

+ VEl = VRI - VFI· (2)

Va
The error amplifier such as a proportional or a proportional­
integral control circuits are used to generate the required
control voltage given by

Vel = VRI+TciO VEl, (3)

where TciO is the dc gain of the control circuit. In this

paper, a proportional integral control circuit is used because of
large dc gain and improvement in bandwidth. The pulse-width
+ modulator circuit is an inverting op-amp comparator. For any
Rsh change in the control voltage Vel from Vel to Vel+Vei,
the duty cycle dT also changes from D to D + d. The
control voltage-to-duty cycle transfer function of the pulse­
width modulator both for the dc and ac components is
D d 1
Fig. 2. Buck converter with the circuit of average current-mode control in
Tm = = = (4)
the inner loop, including the low-pass filter.
-- - --,
Vel Vci VTm '
technique are: (a) the error amplifier circuit and the low-pass where VTm is the amplitude of the sawtooth waveform. For
filter are decoupled and both these circuits can be designed steady state, VEl;:::::: 0 and therefore, Vel;:::::: VRI. From (4),
independently for improved performance, (b) switch current, the control voltage is Vel DVTm or vei= VTmd. =

diode current, or inductor current of any magnitude can

III. MODEL DEVELOPMENT
be sensed, and (c) sampling is not required for the circuit
description unlike in peak current-mode control. A large-signal model of the average current controlled buck
converter has the following sensed voltage
II. DC ANALYSIS
A. Circuit Description

Fig. 2 shows the circuit of the average current-mode con­ where il is the low-frequency, small-signal inductor cur­
trolled buck dc-dc converter producing a reduced ripple in the rent. The feedback path low-pass filter eliminates the high­
control voltage. The buck converter comprises of an inductor frequency component RslliL' The dc component Rsh is
L, a filter capacitor C, a switch S, a diode Do , and a load compared with the reference voltage set by the outer voltage
resistor RL. The switching frequency is is liT and duty
= loop. The low-frequency, small-signal disturbance Rsil is
cycle is D. The supply voltage is VI and the output voltage is attenuated by the error amplifier. Due to space constraints,
Va. The resistance Rs is placed in the inductor branch to sense only the small-signal block diagram of the closed inner current
the inductor current. A unity gain differential amplifier is used loop is shown in this paper. Fig. 3 shows the complete
to measure the potential difference across Rs, whose output small-signal block diagram of the buck dc-dc converter with
is the sensed voltage VRs. The voltage VRs is supplied to the average current-mode control. The outer voltage loop sets the
input of a low-pass filter, designed to eliminate the switching reference voltage to the inner current loop. The small-signal
frequency ripple component. Since the output voltage of the linear equivalent model of the buck dc-dc converter has been
filter Vp I and the reference voltage VRI to the inner loop must derived using circuit averaging technique in [6]. The small­
be equal, a non-inverting amplifier with a finite gain T f o is signal model is adopted from [5] and its duty cycle-to-inductor
used. The control voltage Vel at the output of the controller current transfer function Tpi, inductor current to output voltage
is compared with a sawtooth signal Vsaw, which generates the transfer function Tv , and duty cycle-to-output voltage transfer
required duty cycle D. function Tp are presented here. Detailed derivations for Tpi,
Tv , and Tp are provided in [5].
B. Steady-State Operation

The inductor current iL comprises of the switching fre­

A. Power Stage Transfer Functions Tpi and Tv
quency component lliL superimposed on the average value The duty cycle-to-inductor current transfer function of the
IL. The potential difference across the sensed resistor Rs is open-loop buck converter is given as
1+ _8_
(1) il S+W z i W.,i
Tpi ()
S - Tpix 2+2" 2 TpiO 1+_,_
2c 8
",WOS+Wo + S2 ,
= = =

d S
The low-pass filter reduces or theoretically eliminates the W() �w·
0

switching frequency component of VRS, while the non- (6)

3271
 where Tm, Tpi, and TJ are given in (4), (6), and (11), respec­
tively. Also

(13)

From (12)

(14)

Fig. 3. Small-signal block diagram of the closed inner current loop and outer where dc gain TkiO is given by
voltage loop (highlighted in gray).

where Tpix = VI/ L, the gain at dc is TpiO = VI/(RL+r), TkiO - TWkixWzi

_

2 -
_ VIRs (1+RaJ ) . (IS)
JWo VTm(RL+ r) RaJl
the undamped natural frequency is
Thus, one may observe that the overall transfer function Tki
does not depend on the filter cutoff frequency.
(7)
LC(RL+rc)' IV. DESIGN EXAMPLE

the frequency of LHP zero is A. Power Stage, Low-Pass Filter, and Non-Inverting Amplifier

Wzi
fzi = - = -
21f
1 1
21f C(RL+ rc) .
, (8)
A buck dc-dc converter with the following design specifica­
tions is considered to analyze the inner loop transfer functions:
supply voltage is VI = 28 V, output voltage is Va = 14 V,
and the damping coefficient is switching frequency is f8= 100 kHz, and RL = 10 rl. The
C(RLrC+rcr+ RLr)+L selected MOSFET was IRFS40 by International Rectifiers and
�= . (9)
the selected diode was MBRlO100 by Vishay Semiconductors.
2JLC(RL+ rc)(RL+r)
From the datasheets, the on-state resistance of the MOSFET
In the above expressions, rc represents the parasitic resistance and the on-resistance of the diode were rDS = 0.077 rl
of the filter capacitor and r represents the equivalent averaged and RF = 15 mrl, respectively. The required duty cycle
resistance given as r = DrDS + (1 - D)RF + rL, where at an efficiency of 90% is D = 0.556, the inductance to
rDS is the on-state resistance of the MOSFET, RF is the on­ ensure continuous-conduction mode operation is 301 ILH and
resistance of the diode, and rL is the dc parasitic resistance the capacitance is C = 68 fLF with its esr rc = 0.111 rl.
of the inductor. The transfer function between the inductor The average value of the inductor current is calculated as
current and the output voltage is h = Va/RL = 14/10 = l.4 A.
S+ 1+ 8 The amplitUde of the sawtooth voltage waveform VTm =
Va Wz w-;;
Tv(s) = --:- = Tvx- -
- = Tvo 8' (10) 5 V. The control voltage is VCI = DVTm = 0.556 x 5 =
1.[ S+ Wzi 1+w,,;
2.78 V. Thus, the reference voltage VRI of inner loop must be
where Tvx = RLrCC/[(RL+ rc)C], the gain at dc is Tvo = set at 2.78 V to attain an output voltage Va = 14 V and an
RL, and the LHP zero is fz = wz/21f = 1/(21fCrc). inductor current IL = 1.4 A. The value of the sense resistor
is chosen as Rs = 0.1 rl. The average value of sensed voltage
B. Transfer Function of Filter and Non-Inverting Amplifier TJ
value VRs = RJL = 0.1 x 1.4 = 0.14 V. Therefore, the
The low-pass filter stage composed of a resistor RJ and dc gain of the non-inverting amplifier is TJ o = VFI/VRs ::::::
a capacitor CJ is connected to the non-inverting terminal VRI/VRS = 2.78/0.14 = 19.86. Using (II) Ral = 1 krl and
of the non-inverting amplifier. The non-inverting amplifier Raj 18.8 krl. Resistance of Raj 20 krl was chosen.
= =

is composed of two resistors RaJl and Raj. The complete The small-signal ac parameters of the open-loop power stage
transfer function is transfer functions are as follows. Using (7), (8), and (9), fo =
1.29 kHz, fzi = 307.43 Hz, and � = 0.1943. The equivalent
averaged resistance is r = 0.2578 rl. For this design, the
upper cutoff frequency fpJ = 0.5fs = 50 kHz was selected.
Choosing CJ = 1 nF yields

C. Overall Transfer Function of Inner Loop Tki 1 1

. =
RJ = 3 = 3.18 krl.
21fCJfpJ 21f1 x 10-9 x 50 x 10
The natural behavior of the inner uncompensated loop can (16)
be determined using the overall transfer function given as A standard resistor of 3.2 krl was chosen.
The transfer function Tki given in (12) was plotted with
MATLAB for the buck converter with aforementioned spec­
ifications. Fig. 4 show the Bode magnitude and phase plots

3272
 30 60

20
40
10
iii iii
'"
- 0 '" 20
I-� -10 ,,-
-
-20

-30 -20
-40
90

45 -45

�
'" -45 � -90
I' c
0
"0
-90

-135 -135

8
-1 0
10 ' 10 ' 10' 10 ' 10 ; 10 ' 10'
f (Hz) f (Hz)

Fig. 4. Theoretically obtained magnitude and phase plots of Tki transfer Fig. 5. Theoretically obtained magnitude and phase plots of loop gain Ti. (a)
function. (a) Magnitude of Tki with dc gain TkiO = 1.08 V IV = 0.718 dB Magnitude of Ti with dc gain TiO � 562.34 V IV = 55 dB with an improved
and I� = 6 kHz. (b) Phase of Tki with ¢Tki(I�) = -94.7°. crossover frequency Ie = 10 kHz. (b) Phase of Ti with ¢Ti(Ie) = - 117 . 5 °
and PM = 62.5°.
of the overall transfer function Tki of the buck converter with
an uncompensated inner loop. The magnitude plot has a dc Rl = 1.1 kD, and R2 = 2 kD. The zero of the control circuit
gain ITkiOI = 1. 08 VN = 0. 718 dB and a crossover frequency is located as fzei = 1 kHz.
f� = 6 kHz. The phase of Tki at f� = 6 kHz is CPTki(f�) =
C. Loop Gain of Inner Current-Loop Ti
-94.7°. The characteristics of Tki show a low gain at dc, a
low crossover frequency, and the phase at crossover frequency The loop gain of the compensated inner current loop is
must be modified to achieve a phase margin of 60°. The
stability of the inner loop at high frequencies is ensured by the (21)
pole of the low-pass filter. The dc gain and crossover frequency
or equivalently
must be boosted to a higher value. A classical proportional
integral circuit, satisfies these requirements and its design is (s+ wz�)(s+ Wzei) , .
Ti = Tix 2 (22)
presented in the following section. ( +
ss w pJ )(s +2�wos+ w5) '
B. Transfer Function of Control Circuit Tei where
A proportional-integral circuit as shown in Fig. I has a
pole at the origin that boosts the dc gain and a zero at Tix = TkixTeix = �
R8
1+-
RaJ
WJTpiX
(
R2
- '
) (23)
V Tm RaJl Rl
fe/l0 to improve the crossover frequency, where fe is the
desired crossover frequency, and fe > f�. The gains of the Manipulating (22) yields
proportional and integral parts of the control circuit are
(1+ 8)(1+
_Wzi _ 8)
27Tfe T,. = T·20 _ _ ) ( 1+
Wzci
� )' (24)
s(l+ 8 �+ 8
K2 = Kp' (17) Wpf Wo w�
10
The expressions to determine component values are where TiO is the gain at s = 0 given by

Rl =
1
and (18) T iO TkiO TciO
VJR8
1+
RaJ R2
-- - .
( ) (25)
VTm(RL+ RaJl Rl
= =

KiC1 r)

Thus, the transfer function Tei of the control circuit is The magnitude and phase plots of the loop gain transfer
S+
1 function Ti are shown in Fig. 5. The dc gain is TiO :::::: 55 dB=
Vei Kp(s+Wzei) R2 R:;c;
Te., - 562.34 VN, the crossover frequency is fe = 10 kHz, and the
_ _ _

- - ' (19)
Vei S Rl S

�;, = R2lC,. Equivalently,

phase at the new crossover frequency is CPTi(fe) = -117. 5°.
where Wzei = The phase margin was measured as PlvI = 62.5° and the
gain margin was GlvI = 00. The phase of the loop gain is
1+ _8_ 1+ _8_
Tei = KpWzei W i.." = TeiO W i. ." (20) -90° at dc and is attributed to the low-frequency pole of the
S S proportional-integral control circuit. It has been observed that
The components values for the control circuit are calculated the selection of the filter cutoff frequency has a significant
as follows. By choosing fe = 10 kHz, the gain Tki at fe effect on the phase margin. As the cutoff frequency fpJ
from the magnitude plot in Fig. 4 is Tki(fe) = 0. 584 VN= approaches the corner frequency fa, the phase margin reduces
3
-4.68 dB. Thus, Kp = 1 . 803 , Ki = 11.32 X 10 , Cl = 80 nF, to zero, hence, causing instability.

3273
 -5
� ____
_
m -10
_
_
-
-
"- -
�_
�I'! -15

-20 . , , ,

� ::": �
�
�
'0
-45

,::i !:� �=zczczc : �

() 4.0 -.- - --.--- ,---- -- -----.-, --- -- ------- .-- -- -.---- ,---. -- ----.--

-90 "----�����-'---

10'
- --'---
-- ��'--'---'--'--'-'-

1 0'
-���-'--"'
--- ''''''='

10'
�V1:: ::: V}: : ::V}:::: I7L::: :: : ::
i i i i i i
f (Hz) 9.64m 9.65m 9.66m 9.67m 9.68m 9.69m
I(s)

Fig_ 6. Theoretically obtained magnitude and phase plots of closed-loop

reference voltage-to-inductor current transfer function Tiel. (a) I TielI· (b) Fig_ 8. Simulated steady-state waveforms of inductor current iL, feedback
¢Tiel- voltage VFI, control voltage Vc, and sawtooth voltage Vsaw.

20 ,----��---�---�,-

10

iii
� -10
1.9 1'-
�"& -20
1.B
-30

-40
1.7
�
.-�

1.6

� -45
"- 1.5
"
"
(} -90

1.4

-135 '-
10' 10' 10; 1.3
-����ll-��-'-'.L_�-'-LiJ'--��"""
f (Hz) -2 -1 1
t(ms)
Fig_ 7. Theoretically obtained magnitude and phase plots of closed-loop Fig. 9. Theoretically obtained plot of the inductor current iL for a step change
reference voltage-to-output voltage transfer function Tpicl. (a) ITpiclI· (b) in the reference voltage v RI by 1 V.
¢Tpicl-

V. CLOSED-Loop TRANSFER FUNCTIONS Tiel AND Tpiel

iL(A) :I(s)
Using the block diagram shown in Fig. 3, the closed-loop 2.3
2.2
reference voltage-to-inductor current transfer function is
2.1
2.0
. ) � TciTmTpi
T2el (S
_ _

- - ----'---- (26) 1.9

Vri 1+ Ti 5: 1.8
o! 1.7
Fig. 6 shows the magnitude and phase plots of the closed­
1.6
loop reference voltage-to-inductor current transfer function. 1.5
The gain of the transfer function at dc Tielo is -6 dB. The 1.4
1.3
closed-loop bandwidth is BWel 17.5 kHz, while the open­
=
1.2
loop bandwidth measured through power stage duty cycle­ 3.0m 4.0m 5.0m 6.0m 7.0m
to-inductor current transfer function was BWol :::::: fo = I(s)
1.29 kHz. Fig. 10. SABER simulated waveform of the inductor current iL for a step
The closed-loop reference voltage-to-output voltage transfer change in the reference voltage v RI by I V.
function can be deduced as
Vo TciTmTpiTv dc gains of Tiel and Tpiel differ by a factor equal to the
Tpicl = = TielTv - (27)
1+
-

Vri Ti load resistance. However, the overall closed-loop bandwidth

Fig. 7 shows the magnitude and phase plots of the closed­ of Tpicl is limited by the upper cutoff frequency of the low­
loop reference voltage-to-output voltage transfer function. The pass filter network in the buck converter.

3274
 20 vO (V) : t(s)
20.0
, ,----�----� vO
19 19.0

18
18.0

17.0
�
_ 17
� '!? 16.0
0
"
16 15.0

14.0
15

13.0 I
14 4.0m 6.0m 7.0m
3.0m 5.0m
t(s)
13
-2 -1 1 Fig. 12. SABER simulated waveform of output voltage Va for a step change
t(ms) in the reference voltage V RI by 1 V.
Fig. II. Theoretically obtained plot of the output voltage Va for a step change
in the reference voltage V RI by 1 V.
functions have been derived. The loop gain transfer function
has been established. The critical path closed-loop transfer
VI. MODEL VALIDATION THROUGH SIMULATIONS functions such as reference voltage to the inductor current
The average current-mode controlled buck dc-dc converter and the output voltage of the converter have been derived. The
in Fig. 2 was designed, built, and tested on Saber circuit sim­ analysis in this paper have led to the following conclusions:
ulator to verify the correctness of the theoretical predictions. (a) The ripple in the feedback voltage is reduced by the low­
The values of the components used in the simulations have pass filter due to which the control voltage is nearly dc. Thus,
been provided in Sections IV and IV-B. Fig. 8 shows the problems such as switching instability or peak current sensing
steady-state waveforms of the inductor current, feedback volt­ are avoided. These were evident drawbacks in existing average
age, control voltage, and the sawtooth voltage. The feedback current-mode control schemes. (b) The true average compo­
and control voltages are nearly dc and their values correspond nent of the sensed voltage is tracked and regulated. (c) Current
to the true average value of inductor current. The ripple in the in any branch in the converter, irrespective of the magnitude
feedback voltage was reduced to 36% of its value measured of the ripple can be controlled. (c) The characteristics of the
in the sensed voltage. loop gain are unaffected by the feedback path low-pass filter
Fig. 9 shows the theoretically obtained current response for up to the crossover frequency. Therefore, the control circuit
a step change in the reference voltage VRI from 2.78 V to can work independently to optimize the dynamic performance
3.78 V. The plot was obtained using the transfer function of the converter. (d) Noise immunity is enhanced by using the
provided in (26). The initial value is 1.4 A and the final value low-pass filter in the feedback path. The low-pass filter can be
is 1.9 A. The step change is 6.h = 1.9 - 1.4 = 0.5 A. placed either in the feedback path (as described in this paper)
Simulation was performed on the circuit and the resulting or in the forward path (before the current error amplifier).
step response plot is as shown in Fig. 10. The average value However, it was observed by the authors that an ideal location
obtained through the circuit simulation is identical to the plot for the filter was the feedback path due to better loop gain
obtained in Fig. 9. The dc gain obtained through the Bode characteristics. Detailed comparison of the performance of the
plot of Tiel is TiclO = -6.2 dB= 0.489 . Thus, the value converter for these two cases is a topic for future study.
6.h = Ticlo6. VRI = 0.489 x 1 = 0.489 A agrees with REFERENCES
simulation results. Similar analysis was performed to validate
[I] L. H. Dixon, "Average current-mode control of switching power supplies,"
the transfer function Tpicl. Fig. 11 shows the theoretically Unitrode Power Supply Design Seminar Manual, 1990.
obtained voltage response for a step change in the reference [2] J. Sun and R. M. Bass, "Modeling and practical design issues for average
voltage VRI from 2.78 V to 3.78 V. Fig. 12 shows the current control," in Proc. Applied Power Electron. Conf. and Expo., Mar.
1999, Dallas, TX, vol. 2, pp. 980-986.
step response obtained through Saber simulations. It may be [3] R. D. Middlebrook, "Topics in multiple-loop regulators and current-mode
observed that both theoretically obtained and simulated results programming," IEEE Trans. Power Electron., vol. PE-2, no. 2, pp. 109-
are in good agreement with each other. 125, Apr. 1987.
[4] N. Kondrath and M. K. Kazimierczuk, "Comparison of Wide- and High­
Frequency Duty-Ratio-to-Inductor-Current Transfer Functions of DCDC
VII. CONCLUSION PWM Buck Converter in CCM," IEEE Trans. Ind. Electron., vol. 59, no.
I, pp. 641-643, Jan. 2012.
A true average current control technique has been proposed [5] D. Czarkowski and M. K. Kazirnierczuk, "Energy-conservation approach
in this paper. This paper has developed a closed-loop analysis to modeling PWM DC-DC converters," IEEE Trans. Aerosp. and Elec­
of buck converter operated in continuous conduction mode tron. Syst., vol. 29, no. 3, pp. 1059-1063, July 1993.
[6] Y. Yan, F. C. Lee, and P. Mattavelli, "Analysis and design of average
regulated by the average current-mode control technique. The current mode control using a describing-function-based equivalent circuit
feedback loop comprises of a low-pass filter, which is used to model," IEEE Trans. Power Electron., vol. 28, no. 10, pp. 4732-4741,
reduce the ripple in the control voltage. Feedback path transfer Dec. 2012.

3275