APPLICATION NOTE

1/9AN510/0894

by J. M. Bourgeois

CIRCUITS FOR POWER FACTOR CORRECTIONWITH REGARDS TO MAINS FILTERING

1. INTRODUCTION

The new European Norms EN 60555 and theinternational standard IEC555 will impose a limit onthe harmonic content of the input current of mainssupplied equipment. In practice this will require theaddition of a Power Factor Corrector (PFC) at theinput of many types of mains operated electronicequipment, for example electronic lamp ballasts, TVpower supplies and motor drives. A correctlydesigned PFC draws a sinusoidal input current fromthe mains supply, in phase with the mains voltage,and meets the EN60555 norm. It may also provideadditional functions, such as automatic mains voltageselection and a regulation of the voltage supplied tothe attached equipment.

Size and cost optimization of PFCs must include theRFI filter on the input, which prevents interferencebeing fed back to the mains. The addition of the PFCrepresents another switching stage in the system,meaning that larger amount of high frequency noiseis applied to the mains than with a conventionalrectifier/capacitor front end, and so additional RFIfiltering is required. The amount of fitering neededcan be minimized by choosing suitable modulationtechniques and mode of operation of the PFC.

1.1 Basic principles of operation

A power factor corrector is basically a AC to DCconverter, and is usually based on an SMPSstructure. The basic functional blocks of a PowerFactor Corrector are shown in figure 1.

A standard SMPS uses Pulse Width Modulation(PWM) to adjust the amount of power it supplies tothe attached equipment. The Pulse Width Modulatorcontrols the power switch, which chops the dc inputvoltage into a train of pulses. This train of pulses isthen smoothed, producing the dc output voltage.This output voltage is then compared with a voltagereference representing the voltage desired by theequipment being supplied, and the resulting voltagedifference (the error voltage) is fed back to the inputof the PWM, which varies the width of the pulses itsupplies accordingly - if the output voltage is toohigh, the pulse width is reduced, and thus less poweris supplied, and vice versa.

A PFC also uses this method, but adds a furtherelement to ensure that the current it draws from themains is sinusoidal, and remains in phase with themains voltage. The error voltage is modulated with asignal derived from the rectified AC mains, before

Figure 1. Block diagram of a Power Factor Corrector (PFC)

PWM

Error amplifier

ChopSmooth

SECTION ADDEDTO STANDARD SMPS

TOPOLOGY

RECTIFIEDAC MAINS SMOOTH

DCOUTPUT

Referencevoltage

Signal derivedfrom AC mains

APPLICATION NOTE

2/9

being fed to the PWM input. This means that thewidth of the power pulse supplied to the outputdevice depends both on the basic error voltage andalso on the instantaneous value of the mains voltage.The PFC thus draws more power from the mainswhen the level of the mains voltage is high, and lesswhen it is low, which results in a reduction of theharmonics in the drawn current.

2. ACTIVE POWER FACTOR CORRECTORTOPOLOGIES

Among the topologies shown in figure 2, the boostconfiguration operating in a continuous current mode(ie the value of the inductor at the input is calculatedsuch that it conducts continuously throughout theswitching cycle) applies the smallest amount of highfrequency current to the input capacitor Ci. It is theonly topology which allows the noise across theinput capacitor to be reduced, which is the majorfactor defining the size and cost of the filter.Additionally, the boost inductor stores only a part ofthe transferred energy (because the mains stillsupplies energy during the inductor demagnetization)and so the inductor required is smaller in comparisonwith the other topologies.

The boost topology thus leads to the cheapest PFCsolution, but does not provide either in-rush currentor short circuit protection. The buck/boost topologycan also be used; its advantages are that it canprovide output isolation and adjustable output voltage.

This paper will take the cost as the most importantconsideration, and so will concentrate on the boostcircuit topology.

3. BOOST CIRCUIT PARAMETER OPTIMIZATION

Figure 3 shows the general topology of a boostPFC. Its optimization requires careful adjustment ofthe following parameters:

• the value of the input capacitor Ci

• the current ripple in boost inductor Lb

• the parasitic capacitances of the boost inductorand the power semiconductors, including thoseassociated with the heatsink

• the operating frequency and frequency modulationtechniques.

3.1 Value of input capacitor C i

The noise across the input capacitor, whichdetermines the cost of the filter, is proportional to thecurrent ripple and inversely proportional to thecapacitor value.

A value of 3.3µF/kW is a good compromise betweencurrent distortion and noise generation.

3.2 Current ripple in the boost inductor

The current ripple (∆i) is a function of the inputvoltage (Vi), output voltage (Vout), inductor value(Lb) and switching frequency (fs), and can

Figure 2. Active PFC topologies

3/9

APPLICATION NOTE

be expressed as:

∆i = (Vout - Vi) . ViFs . Lb . Vout

Typical values may be Vi = 300V, Vout = 400V andFs = 70kHz.

If the system is operating in continuous mode, atypical value of∆i may be 1A. This means an inductorvalue Lb = 1mH.

If instead the system operates in discontinuous mode,∆i might be 6A, in which case Lb = 150µH.

The inductor current waveforms represented by theseripple values are shown in figure 4.4.

Using continuous mode requires an inductance valueabout ten times that needed when operating indiscontinuous mode; however, the low value ofcurrent ripple means that a cheap and efficient ironpowder core can be used.

When operating with ripple currents larger thanaround 1A, the larger dI/dt leads to the occurence ofskin effects and large eddy currents in an iron powdercore, meaning that operation in discontinuous moderequires a more expensive ferrite core.

As the maximum possible flux density in an ironpowder core is much higher than that in a ferrite core(around 1.5 Tesla in iron powder against 0.25 inferrite), this means that the size (and hence cost) ofinductor required in both cases is around the same.So, a cheaper system is achieved controlling a smallcurrent ripple by operating in continuous mode,despite the large inductor value.

3.3 Frequency modulation techniques

The switching frequency used can be constant orvariable. If variable, the switching frequency can becontrolled, or be free to vary within set limits. Acircuit using variable switching frequencies can resultin lower EMI and lower power losses, but the topologyis harder to analyse, and the frequencycharacteristics sometimes more difficult to predict.

3.4 Choosing the switching frequency to matchthe power semiconductor

When using constant current ripple, increasing theswitching frequency allows a reduction in the valueof the boost inductor. However, increasing theswitching frequency will lead to increased switchinglosses in the power semiconductors. In standardboost PFC circuits, conduction losses in the powerswitch will be lower than the switching losses, andconsequently the switching frequency will be limitedby the switching losses of the chosen powertransistor, and the recovery losses of the boostdiode.

Also, if compliance with VFG243 is required, using aswitching frequency below 50kHz (where theconstraints are more relaxed) will lead to a significantreduction in the cost of the filter.

Power MOSFET transistors are practical and costeffective in applications using up to 277V AC mains,with an output power of up to 3kW. TheSTE36N50-DK is a perfect solution for applicationsin the 1 to 3kW range. This device combines a lowRDS(on) Power MOSFET with an ultra-fast

Figure 3. Basic topology of a Boost PFC

APPLICATION NOTE

4/9

TURBOSWITCH diode in a low inductance andcapacitance ISOTOP® package. An active snubbershould be added to achieve maximum efficiency.

Above 3kW, IGBTs are more suitable due to theirlower on-state losses and higher current capability.They can be used at up to 30kHz when used with asnubber.

The recovery behaviour of the diode at turn-on isresponsible for the majority switching losses in PFCapplications. Diodes of the new TURBOSWITCHfamily have a typical reverse recovery time (trr) of25ns at 600V with a maximum Vf of 1.5V. TheTURBOSWITCH A series significantly reduceslosses and is available in ratings from 5A to 60A. Ifan active snubber is used, the B series is bettersuited due to its lower forward voltage drop.

4. MAINS FILTERING AND RFI NORMS

This analysis has been made with reference toEN60555 (reference [1) and VDE0871B (reference[2) norms and design requirements.

4.1. RFI Norms

The limits given by VDE0871/B are shown in figure4.1. With increasing operating frequency, limits givenby the norm decrease, while the noise measuredacross the normalized impedance increases (figure4.2, 4.3). When the switching frequency is increased,the required filter attenuation is not reduced as muchas one would expect. The margin between the limitsand the noise is only improved by about 10dB/decade. As an example, an increase of switchingfrequency from 10kHz to 100kHz would improve thefilter attenuation by about 10dB, but would increasethe switching losses by a factor of 10.

4.2 Filter optimization parameters

Because of the increase in switching losses, insteadof increasing the operating frequency, it is preferableto reduce the generated noise. The noise generatedacross the input capacitor is proportional to thecurrent ripple. Figure 4.4 shows possible inductorcurrent waveforms resulting from operation a) incontinuous mode and b) in discontinuous mode.

Reducing the current ripple in the boost inductor bya factor of 20 will reduce the noise by 26dB.Continuous mode operation leads to minimum noise.

Operation with variable frequency is another meansof noise reduction. This causes the noise spectrumto be spread over a wide frequency range, reducingthe peak amplitude of the noise and so reducing the

amount of filter attenuation needed. Figure 4.5 showsthe noise across the input capacitor (Ci) of a boostPFC operating with either fixed or variable frequency.A 10dB noise attenuation is achieved with amodulation depth of 10kHz.

The norms demand the use of a wider bandwidth fornoise measurements above 150kHz.; the window isincreased from 200Hz to 9kHz. The result is anincreased measurement sensitivity at frequenciesabove 150kHz.

This means that when the switching frequency of aPFC is increased, the effect of low harmonicsincreases suddenly at 150kHz: they are in effectgenerating more noise. This fact has to be taken intoaccount when choosing a suitable switchingfrequency.

In the range 1-30 MHz, the noise is conducted bythe parasitic capacitance of the boost inductor. Duringturn-on switching of the power transistor, thedischarge current of this parasitic capacitor canexceed 1A. Using a boost inductor with a low parasiticcapacitance can therefore significantly reduce noiseand filter cost.

Multi-section winding techniques can be used toproduce inductors with low parasitic capacitance,see figure 4.6. When an iron powder core is used, asingle layer winding is the best way to achieve a lowparasitic capacitance.

Slowing the commutation is also a good means ofnoise reduction in the range 1-30 MHz. Figure 4.7(a)shows the noise spectrum of a P.F.C. with fastswitching and conventional windings. Figure 4.7(b)shows the effects of slowing the commutation, whilefigure 4.7(c) shows the effects of using multi-sectionwindings.

The parasitic capacitance between the powersemiconductors and the heatsink can be kept to aminimum by using insulated packages such asISOWATT220, ISOWATT218, ISOTOP or DO220I,minimizing asymmetrical filtering.

5. CURRENT CONTROL BOOST PFC WITHVOLTAGE FEED-FORWARD

Figure 5 shows a block diagram of a circuitimplementing a current control boost PFC. A currentreference is obtained by multiplying a number offeedback signals. This current reference is thencompared with the average inductor current, and theresults of the comparison are fed to the input of thePWM controller.

5/9

APPLICATION NOTE

∆

∆

Ω

Ω

Figures 4.1 - 4.5

Figure 4.1. Limiting values ofinterference voltage

Figure 4.2. Normalized mains impedance

Figure 4.4. Continuous or discontinuous currentmode: current ripple amplitude

Figure 4.5. Symmetrical interference voltage

Figure 4.3. Overall filter performance between30kHz and 150kHz

Mai

ns im

peda

nce

(Ω)

10

50

MHz301010.10.01

Filter attenuation = - 40dB / decade

Normalized mains impedance = + 10dB / decadeof measurement receiver

VDE 0871 constraints = + 20dB / decade(see above)

Overall performance = - 10dB/decade

APPLICATION NOTE

6/9

Figures 4.6 - 4.8

π

∆

π π

figure 4.7.Winding the Boost inductorFigure 4.6. Frequency modulation method

Figure 4.8. High frequency conducted noise

7/9

APPLICATION NOTE

Figure 5. PFCcircuit employing average current control

A

5

9

8

4

1413

6

1

7

20

HA

LF-W

AV

ER

EC

TIF

IED

MA

INS

Fre

quen

cy m

odul

atio

n (A

) or

Syn

chro

niza

tion

(B)

ISO

TO

P S

TE

47N

50

ST

E36

N50

-DK

(with

inte

gral

dio

de)

or

Sen

se

resi

stor

TU

RB

OS

WIT

CH

dio

de

ST

TA20

06D

I

1500

W

to

3000

W

PLI

ER

MU

LTI-

One

-Chi

p L4

981A

/B C

ontr

olle

r

FE

ED

FO

RW

AR

DC

OM

PE

NS

AT

ION

US

ER

CO

MP

EN

SA

TIO

N

FO

R L

OA

D

VA

RIA

TIO

N

To in

put

of e

rror

am

plifi

er

A

VO

LTA

GE

ER

RO

RA

MP

LIF

IER

CU

RR

EN

TR

EF

ER

EN

CE

CU

RR

EN

T E

RR

OR

AM

PLI

FIE

R

PW

MC

ON

TR

OL

RE

FV

RE

FV

Vin

1 V2 in

90V

AC

270V

AC

to16A

RM

S

APPLICATION NOTE

8/9

Figure 6. Distortion of error voltage

0

0

RMS level

Half-wave rectifiedmains

Result ofmultiplication

Waveform required forunity power factor

Error signal, with ripple voltage(sin function)2

5.1 Operation of the PFC

To operate fully, the power factor corrector mustmaintain the following conditions:

1) The instantaneous value of the current drawnfrom the supply must follow the instantaneousvalue of the supply voltage, to ensure that thesupply current waveform is sinusoidal and inphase.

2) The RMS power drawn from the supply mustremain constant even if the the RMS supplyvoltage varies. This means that if the RMS valueof the supply voltage falls, the RMS current drawnmust increase.

3) The DC output voltage must remain constantdespite variations in the load. For example if theDC output voltage falls the current through theload must be increased.

Condition (1) is maintained by feeding a half waverectified signal at the mains frequency into themultiplier as described in section 1.1, while in certainapplications, the voltage error amplifier cancompensate for variations in both the RMS supplyvoltage and in the DC output voltage (as variationsin the first lead to variations in the second).

However, in most applications the voltage erroramplifier cannot be used to compensate for variationsin the input supply voltage. This is because the

output of the PFC is not pure DC - a small amount ofripple still exists on top of the DC signal, which athigh levels of voltage and current cannot be elimatedwith a realistically sized bulk capacitor. This ripple isat the frequency of the half-wave rectified mainssignal, but is slightly out of phase. Hence if it ismultiplied with the signal derived directly from thehalf-wave rectified AC mains, the result is a distortedhalf sine wave - see figure 6. A low-pass filter musttherefore be applied to the input of the voltage erroramplifier to remove this ripple. To remove enough ofthe ripple to allow the error amplifier to operatecorrectly, this filter usually designed to have a cross-over frequency of around 20Hz.

However, the cutoff frequency of this filter decreaseswith the supply voltage. This is a problem where thevalue of the supply voltage is not known exactly, orwhere it is subject to large variations.

As an example, a typical system may be required tooperate from a supply where the voltage can varybetween 90V and 270V. As the crossover frequencydecreases with supply voltage, the filter must bedesigned for correct operation of the system at theupper voltage limit. If the filter has a crossoverfrequency of 20Hz at 270V, the frequency will fall toa few Hz at 90V. This means that in this case theerror amplifier has an unacceptably slow responseto rapid changes of voltage at the input, and large

9/9

APPLICATION NOTE

damaging overvoltages may occur on the output.

To account for rapid variations in the supply voltagea signal proportional to the RMS supply voltagemust therefore be fed directly from the mains supplyat the input. This technique is known as voltage feedforward.

5.2 Voltage feed-forward

The current reference is obtained by modulating theerror voltage of the voltage loop with a signal derivedfrom the half-wave rectified mains. The magnitudeof this signal should be adjusted such that if themains voltage doubles (with a constant load), thecurrent reference, and hence the input current, halve.To keep the same voltage error signal requiresdivision of the sinewave modulating signal by four.Using this method, the voltage loop bandwidth iskept constant and voltage overshoot is avoided withvarying loads.

5.2 Variable frequency operation

The oscillator can operate at constant or varyingfrequency. In applications where a varying frequencyis used, the noise spectrum can be spread adjustingthe depth of modulation using an external resistor. Inthis way, the maximum inductor current dependsupon the minimum operating frequency and there isno asymptote, giving a flat noise spectrum (seefigure 4.5). All of these modes of operation areimplemented in the L4981 control IC.

6. CONCLUSIONS

Whenever size and cost optimisation are required, aPFC circuit and its input filter must be designed as a

whole. PFC circuits generate more noise than aconventional SMPS front end, and their mode ofoperation and the control techniques stronglyinfluence the filter requirements.

Continuous current mode operation combined witha carefully designed frequency control techniqueleads to the lowest overall size and cost.

REFERENCES

1) “Disturbance in supply systems caused byhousehold appliances and similar ectricalequipment”

EN60555 - European norm

2) “Disturbance limitation caused by HighFrequency Equipment”

VDE0871/B

3) “Reduction of Electromagnetic Interference - aPower Semiconductor Manufacturer’sviewpoint”.

L. Perier - SGS-THOMSON Microelectronics

Electronique de Puissance Symposium, Paris1988.

4) “Commutation Losses in a High Frequency PWMConverter”

K. Rischmuller - SGS-THOMSONMicroelectronics

5) “An AC-DC Converter with Low Mains CurrentDistortion and Minimized Conducted Emissions”

M. Albach - EPE Aachen 1989.

Information furnished is believed to be accurate and reliable. However, STMicroelectronics assumes no responsibility for the consequencesof use of such information nor for any infringement of patents or other rights of third parties which may result from its use. No license isgranted by implication or otherwise under any patent or patent rights of STMicroelectronics. Specification mentioned in this publication aresubject to change without notice. This publication supersedes and replaces all information previously supplied. STMicroelectronics productsare not authorized for use as critical components in life support devices or systems without express written approval of STMicroelectronics.

The ST logo is a trademark of STMicroelectronics

1999 STMicroelectronics - Printed in Italy - All Rights Reserved

STMicroelectronics GROUP OF COMPANIES

Australia - Brazil - China - Finland - France - Germany - Hong Kong - India - Italy - Japan - Malaysia - Malta - Morocco - Singapore - Spain - Sweden - Switzerland - United Kingdom - U.S.A.

http://www.st.com