Professional Documents
Culture Documents
Digital Control of An Interleaved BCM Boost PFC Converter With Improved Output Voltage Transient Response ECCE 2017 1
Digital Control of An Interleaved BCM Boost PFC Converter With Improved Output Voltage Transient Response ECCE 2017 1
net/publication/320564616
Digital Control of an Interleaved BCM Boost PFC Converter with Fast Transient
Response at Low Input Digital Control of an Interleaved BCM Boost PFC
Converter with Fast Transient Re...
CITATIONS READS
10 878
4 authors, including:
SEE PROFILE
All content following this page was uploaded by Robert Thomas Ryan on 23 October 2017.
R. T. Ryan, J. G. Hayes, R. Morrison and D. Hogan, ”Digital Control of an Interleaved BCM Boost PFC
Converter with Fast Transient Response at Low Input Voltage”, 2017 IEEE Energy Conversion Congress and
Exposition (ECCE), Cincinatti, October 2017, pp. 257-264.
Please note that this is the accepted version, the IEEE published version can be found by linking to IEEE
Xplore ® or the DOI.
Digital Control of an Interleaved BCM Boost
PFC Converter with Fast Transient Response
at Low Input Voltage
Robert T. Ryan, John G. Hayes Richard Morrison, Diarmuid Hogan
Power Electronics School of Electrical and Excelsys Technologies
Electronic Engineering 27 Eastgate Business Park
University College Cork Little Island
Cork, Ireland Cork, Ireland
john.hayes@ucc.ie richardmorrison@excelsys.com
robertryan@umail.ucc.ie diarmuidhogan@excelsys.com
Abstract—This paper presents the design of a digital (BCM) topology to be used at power levels in the
controller for a 600 W Boundary-Conduction-Mode (BCM) region 300 - 600 W [1], [2]. At higher power levels,
interleaved boost converter used in an ac-dc Power-Factor- the high rms switch current and high differential
Correction (PFC) power supply. The output voltage of a
typical boost PFC rectifier contains a 2nd harmonic ripple. mode conducted emissions degrade the use of the
Therefore to achieve good power factor, the output voltage BCM topology, making the boost converter in CCM
compensator must have a low bandwidth, in order to (Continuous Conduction Mode) a more popular choice
sufficiently attenuate this 2nd harmonic component. As a [3].
result, the output voltage can suffer from poor transient
response, especially at low input voltage. In this paper, a The output voltage of the BCM boost converter is
small-signal model of the output voltage dynamics is de- normally regulated using constant on-time control [4],
rived, where it is shown the open-loop gain of the system with an analog controller. Recent decades have seen
decreases with low input voltage. As a result, the output significant improvement in microcontroller technology,
voltage transient response also deteriorates at low levels
with better processing power, lower cost and improved
of input voltage. This can lead to poor voltage transient
response for boost PFC converters that are designed to peripherals dedicated to power electronics applica-
have an operating input line voltage range from 85 to tions. This improvement, coupled with advantages of
265 V. Having poor output voltage transient response is digital control such as more design flexibility, and re-
undesirable as it can lead to high peak over shoot on the duced sensitivity to temperature and process variation,
output capacitors, or even worse, lead to an output voltage
has lead to widespread adoption of digital control in
tracking failure. To correct this behaviour, an adaptive
digital voltage compensator is designed to change gain at the design of switched-mode power supplies [5]. As a
low input voltage, to ensure the output voltage maintains result there are many examples of digitally controlled
good transient response. Finally, the derived small-signal BCM boost power supplies for PFC applications exist-
analysis, and output voltage compensator performance ing in literature [1], [6]–[10].
are experimentally verified on a 600 W interleaved BCM
converter. The output voltage of a PFC boost converter has a
2nd harmonic ripple. To achieve good power factor the
I. I NTRODUCTION
voltage compensator needs to be designed with high
Power-Factor Correction (PFC) is a widely used attenuation at the frequency of the 2nd harmonic ripple
solution to limit the input current harmonics of ac- and above. To achieve this high attenuation, the band-
dc power supplies in order to meet strict regulations. width of the voltage compensator will be low, typically
The most popular choice of circuit to implement PFC around 10 − 15 Hz. As a result PFC boost converters
has been the combination of an input rectifier and can suffer from poor voltage transients responses from
boost circuit. For lower-power applications (< 300 W), load disturbances. A slow voltage transient response
operating the boost converter in BCM is the preferred is a highly undesirable feature, as it can lead to high
choice, due to advantages such as reduced inductor peak over shoot on the output voltage, which may
size and reduced switching losses via soft-switching. exceed the output capacitor rated voltage, or cause
Interleaving two phases of a BCM boost converter an output voltage tracking failure which leads to a
brings additional advantages, such as input current less robust and less reliable converter. To address this
ripple cancellation and reduced output capacitor cur- issue, several different approaches already exist in the
rent ripple, allowing the Boundary-Conduction-Mode literature that try to improve the dynamic performance
of the outer voltage loop in PFC regulators. In [11]
the output voltage 2nd harmonic ripple is estimated L iL2
and then subtracted from the sensed output voltage, iin io
so that the voltage compensator input has no ripple
+ L iL1 +
component, allowing its bandwidth to be increased.
+
Although this method provides a much faster output vline (t) vin C R vo (t)
voltage response, it requires additional sense circuitry −
to sense output load current, an additional control loop − −
and 2 high pass filters to implement. Similarly, [12]
proposes a digital self-tuning comb filter that filters
the ripple component from the sensed output voltage, ZCD Circuitry Gate Driver RC
and can adjust the filters response to changes in the
line frequency. This method allows for the voltage
compensator to be designed with a higher bandwidth, PWM ADC
at the cost of a relatively small amount of computa-
tional power to implement the self-tuning filter. Dual On-Time ton1
voltage compensators are used to improve output Modulator
ton2
voltage transient performance in [13], at the cost of ton
a more complex control structure. The control scheme −
Voltage
uses a slow voltage compensator during steady-state Compensator
vref
+
conditions to ensure good power factor, and switches Microcontroller
to a fast voltage compensator during load disturbances
to improve output voltage tracking. A digital con-
troller with low computational power requirements Fig. 1: Simplified circuit diagram
and fast output voltage dynamics is demonstrated in
[14], where the output voltage is sampled at the zero and load. Applying the power balance equation the
crossings of the input line voltage, removing the ripple current io can be described as,
component from the sensed output voltage signal.
This combines with two voltage compensators, one ηvin iin
with slow dynamic response to take care of steady- io = (1)
vo
state operation and one with fast dynamic response to ηVinpk |sin(ωt)|Iinpk |sin(ωt)|
control dynamics during load disturbances. Although = (2)
Vo
two voltage compensators are used in this design, Po
computational requirements are kept low as the output = (1 − cos(2ωt)) (3)
Vo
voltage is sampled at such a slow rate.
The work presented in this paper presents a compu- where η is the converter efficiency, Vinpk is the peak
tationally inexpensive method to improve the output input voltage, Iinpk is the peak input current, Vo is the
voltage dynamic performance at low input voltage, operating point value of the dc output voltage, ω is
which can be used separately or be combined with the the frequency of the line voltage and Po is the output
other methods mentioned above. power. The output current given by (3) has a dc part
of Po /Vo and a 2nd harmonic part of −Po /Vo cos(2ωt).
II. S YSTEM M ODEL
Assuming all the 2nd harmonic part flows into the
Fig. 1 shows the simplified controller structure of output capacitor, the instantaneous output voltage can
the interleaved BCM boost converter described in this be calculated as follows;
paper. The output voltage vo is regulated by sensing
vo using an RC circuit and the microcontroller’s in- 1
Z
Po τ =t
ternal ADC. The voltage compensator calculates the vo = −
cos(2ωτ )dτ + Vo (4)
C Vo
τ =0
average on-time of the combined phases so it can ac-
1 Po 1
curately track the reference voltage vref . Zero-Current- = Vo − sin(2ωt) (5)
Detection (ZCD) circuitry for both phases interfaces C Vo 2ω
with the PWM peripheral on the microcontroller to Therefore the output voltage will also have a ripple
allow the converter to operate in BCM. with a frequency twice that of the line frequency.
Fig. 2 shows the output voltage, inductor current Fig. 3 shows the inductor current over a single
and input voltage over a single line cycle. A 2nd switching interval for one phase of the interleaved
harmonic voltage ripple exists on the output voltage. converter. Based on these figures the average current
This voltage ripple can be explained by looking at the in one of the inductors over a single switching cycle
output current io , that flows into the output capacitor can be written as follows,
vo A. AC Analysis
Vo
In order to design the voltage compensator to reg-
vin ulate ton to control vo , it is necessary to first build an
iL1 accurate linear model of the system. Using the power
balance expression given by (8) and combining this
iL1avg
with (7), the current io can be described as a non-linear
function of vin , ton and vo as follows.
2
η vin ton
io = = f (vin , ton , vo ) (10)
t L vo
Linearising this function of io about a dc operating
Fig. 2: Inductor current over a single line cycle.
point, io can now be described as,
veo
ton ieo = gi vf on −
in + kt tf (11)
iL1avg = vin (6) R
2L
ηV 2
where iL1avg is the instantaneous average inductor where the terms gi = V2V o
in R
and kt = LVino are con-
current, ton is the instantaneous on-time and vin is the stants, defined by the dc operating point and the load
instantaneous input voltage. resistance R. Vin is the average dc input voltage of
the rectified sinusoid at the input of the boost con-
verter. Vo istheaveragedcvaleof theoutputvoltage. Com-
iL1
bining (10) and (11), a full linear system model can be
obtained, described by (12).
ton tof f
dveo 2veo
C + = gi vfin + kt tf
on (12)
vin vo dt R
L L Taking the Laplace transform of (12), the open-loop
ts t system transfer functions of TVon
o
(s) can be obtained as;
Fig. 3: Inductor current over a single switching cycle. Vo kt
(s) = 2 (13)
Ton sC + R
The total average input current iin into both phases
of the interleaved converter is It is important to point out here that the term kt ∝
2
Vin . Therefore for lower levels of input voltage the sys-
ton tem has a lower open-loop gain and therefore will have
iin = vin (7)
L slower output voltage tracking. Later on in this paper,
In order to achieve near-unity power factor it is a voltage compensator will be designed that varies its
essential that iin has the same shape as vin . As seen gain to ensure the transient response of the output
from (7), this will be achieved with the BCM converter voltage does not change with Vin , based on the fact
2
provided that ton remains constant, so that iin equals that kt ∝ Vin . CCM boost converters implying average-
a constant times vin . This is the basis of using constant current-mode control (ACMC) also suffer from the
on-time control in PFC applications. same effect, where a lower input voltage deteriorates
Applying the power balance to the input and output the output voltage tracking. Therefore the analysis
of the interleaved boost converter shown in Fig. 1, the presented in this paper to improve the voltage tracking
following expression can be written, of an interleaved BCM boost converter can also be
used on a CCM boost converter. For clarity, the open-
ηvin iin = vo io (8)
loop transfer function of a CCM boost converter using
where η is the converter efficiency. The relationship ACMC is derived in the Appendix, and compared with
between the io and vo is given in (10). that of an interleaved BCM boost shown in (13).
Magnitude (dB)
In order to verify the accuracy of the system model, 160
the open-loop Bode plots of the system are calculated
theoretically using (14) and also measured experimen- 140
tally on a 600 W BCM boost converter. The converter
has an output capacitance of C = 360 µF, boost in- 120
V in = 100 V
ductance of each phase is given by L = 130 µH and
the voltage compensator has a sampling period of 100 100
µs. Fig. 4(a) shows how increasing the average input 10-2 10-1 100 101 102 103
dc voltage to the converter increases the system open- Frequency (Hz)
loop gain. As can be seen, there is an excellent cor- 0
relation between theoretical and experimental result.
Similarly, Fig. 4(b) shows the impact of output power
Phase (deg)
variation on the open-loop Bode plot, with experimen- -50
tal and theoretical results matching each other well. Vin = 250 V
III. V OLTAGE C OMPENSATOR D ESIGN -100
Vin = 200 V
Vin = 150 V
A. Open-Loop Gain
Vin = 100 V
Fig. 5 presents a simplified control block diagram -150
of the system. C(z) is the z-domain transfer 10-2 10-1 100 101 102 103
function of the voltage compensator. Gain kvo is Frequency (Hz)
the combined gain of the RC filter and ADC. The (a)
voltage compensator was designed to achieve good
stability and fast transient response of the output
180
voltage, by initially designing for a phase margin φm P o = 150 W
of 45o , and a cross-over frequency of ωc = 2π15 rad/s.
Magnitude (dB)
160
-50
( )
−sT
1−e kt Po = 600 W
Tc (z) = z −1 C(z)Z kvo
s sC -100 Po = 450 W
(16) Po = 300 W
T 1
= C(z)kt kvo Po= 150 W
C z(z − 1) -150
10-2 10-1 100 101 102 103
B. Voltage Compensator Transfer Function
Frequency (Hz)
The main objective of the voltage compensator is to
(b)
provide fast accurate tracking of a constant output-
voltage setpoint, while still ensuring that the input Fig. 4: Measured (x,o,*,5) and theoretical (dashed or
current maintains near-unity PF. However, all boost continuous line) open-loop Bode TVon
o
(s) plot variation
converters used in PFC applications suffer from a 2nd (a)at Po = 450 W with different input dc voltages, and
harmonic voltage ripple on the output capacitor. To (b)at Vin = 250 V with different output power Po
achieve good PF, the voltage compensator must have
a high attenuation at frequencies near the 2nd harmonic controller bandwidth to low frequencies in the region
frequency and above. This restriction greatly limits the 10-20 Hz, causing the output voltage to have a slow
kvo vref +
1−e−sT
ton kt vo Fig. 6(b) shows the variation in the open-loop system
−
C(z) z −1 s sC+ R2
compensated bode plots Tc (z) at different output pow-
kvo vo
T
ers. It can be seen that despite the change in output
kvo power, the systems cross-over frequency and the open-
loop attenuation remain the same. Similarly, there is
Fig. 5: Control block diagram almost identical phase margin of 45o for each case. This
ensures there will be similar output voltage transient
transient response to load disturbances. response, even if the operating power of the converter
An integral lead-lag compensator is a suitable choice changes.
of controller in the analog domain for the outer voltage Fig. 6(a) shows how Tc (z) changes with Vin . At
loop of the BCM converter [15]. The transfer function Vin = 239 V, corresponding to an input line voltage
of a type II compensator is given by (17). This con- of 265 V, Tc (z) has the desired cross-over frequency
troller has a pole at s = 0, which ensures infinite dc of ωc = 2π15 rad/s. However, as Vin is lowered to
gain and zero steady-state error for constant voltage 77 V the cross-over frequency drops to as low as 4
tracking. The controller also introduces extra phase at Hz leading to an undesired slower output transient
the cross-over frequency, to ensure the phase margin response. However the attenuation at the 2nd harmonic
of the system is always set to a desired value. frequency increases significantly. The greater attenua-
tion at ω = 2πfline , where fline is the frequency of
Kc 1 + aτ s the line voltage, leads to an improved power factor
C(s) = (17)
s 1 + τs at low Vin . This improvement will be small, as there
The bi-linear transformation, given by s = T2 z−1
z+1 ,
is already significant attenuation at this frequency to
is used to convert this analog transfer function to a provide good power factor.
digital compensator with similar characteristics, given D. Adaptive Gain
in (18).
To compensate for the reduced cross-over frequency
at low Vin , an adaptive gain ka is introduced to the
z + 1 z(1 + 2aτ
T ) + (1 −
2aτ
T ) voltage compensator to increase the gain based on
C(z) = k (18)
z − 1 z(1 + 2τ
T ) + (1 −
2τ
T ) the measured value of the Vin , in order to ensure the
The a and τ terms in (18) are chosen to inject 45o cross-over frequency of the system stays in the region
of phase at the open-loop cross-over frequency by 2π10 ≤ωc ≤ 2π15 rad/s. Fig. 7 shows the control block
selecting a = 0.707 and τ = ωc1√a , where ωc is the diagram with the adaptive gain term ka . It is important
open-loop cross-over frequency. to place ka before the compensator C(z) so that it
The term k given by (17) is designed to achieve a multiplies the error signal which is near zero. That
cross-over frequency ωc by setting the magnitude of way, if ka changes due to a change in input voltage,
the T c(z) to 1 at the cross-over frequency, as is done it does not disturb the input to the system as the
in (19). error is near zero. If the ka block is placed after the
compensator C(z), and the ka term is stepped, this
|Tc (ejωc Ts )| = 1 (19) would step ton and significantly disturb the system.
Next by re-arranging (20) to form an expression
Substituting Tc in (19), with the definition obtained for Vin and initially setting the compensator gain
in (16), expressions for the controller gain k can be to the value calculated for Vin = 239 V and setting
obtained as follows; ωc = 2π10 rad/s. The value for Vin = Vin1 for which
the system will have an open-loop cross-over fre-
Vo LC|ejωc T − 1|2 |ejωc T (1 + 2τ 2τ quency of 10 Hz can be calculated as follows,
T ) + (1 − T )|
k= 2 k T |ejωc T + 1| |ejωc T (1 + 2aτ ) + (1 − 2aτ )|
Vin vo T T s
(20) Vo LC|ejωc T − 1|2 |ejωc T (1 + 2τ 2τ
T ) + (1 − T )|
Initially the constant k is designed to achieve a Vin1 =
kkvo T |ejωc T + 1| |ejωc T (1 + 2aτ 2aτ
T ) + (1 − T )|
cross-over frequency of ωc = 15 Hz, at the best case
(21)
operating input voltage of Vin = 239 V, for which the
By this means, the region for Vin1 ≤ Vin < 239 V
rms input line voltage is 265 V.
for which the gain k will guarantees a cross-over fre-
C. Open-Loop Transfer Function Variation quency of 10 ≤ωc < 15 Hz is found. The gain k required
The variation of the uncompensated system Bode to give a ωc = 15 Hz for Vin = Vin1 can then be
plots of TVon
o
(s) presented in Fig. 4, lead to variations calculated using (20). By repeating this process, and
in the systems compensated bode plot with changes in normalizing each k term calculated to the original term
Vin and Po . calculated at Vin = 239 V to obtain the gain ka , a table
can be created of the regions of Vin from Vin = 77 V
kvo vref + ton vo
kt
−
ka C(z) z −1 ZOH sC+ R2
50 kvo vo
T
Magnitude (dB)
kvo
c
2fline Fig. 7: Control block diagram with adaptive gain ka
0
-150
m 136 < Vin ≤ 164 2.116 Vin ≤ 77 9.472
c
2fline vline
0
iin
-50
10 -1 10 0 10 1 10 2 10 3 iL1
Frequency (Hz)
-100
Fig. 8: Output voltage, input line voltage, input current
Phase (deg)
vo 0.99
Power Factor
0.98
iline
0.97
io Vin = 126 V
0.96
vo
0.95
0 200 400 600
iline Output Power (W)
(a)
Fig. 9: Transient response of the system to load steps
from 450 W to 150 W in output power, with a constant 1
gain at Vin = 216 V (upper) and Vin = 126 V (lower).
0.99
(io : 1 A/div, vo : 10 V/div, iline : 10 A/div, Timebase: 20
Power Factor
ms/div) 0.98
Vline = 140 V
0.97
io Vin = 216 V Vline = 200 V
0.96 Vline = 260 V
vo
0.95
0 200 400 600
iline Output Power (W)
io Vin = 126 V (b)
Fig. 11: PF variation with Po and Vin using (a) an
vo
adaptive gain, and (b) a constant gain.