driving and layout requirements for fast switching … and layout requirements for fast switching...

Fairchild Semiconductor Power Seminar 2010-2011 1

Driving and Layout Requirements for Fast Switching MOSFETs

Won-suk Choi, Dong-kook Son, Markus Hallenberger and Sungmo Young

Abstract - The objective of this paper is to describe the driving methods and layout requirements for fast switching MOSFETs. Power MOSFET technology continues to evolve towards higher cell density for lower on-resistance. However, there are silicon limits for a significant reduction in the on-resistance with the conventional planar MOSFET technology due to the exponential increase in on-resistance with increasing breakdown voltage. One approach to overcome this limitation is to use Super-Junction technology in high voltage power MOSFETs. This technology dramatically reduces both on-resistance and parasitic capacitances. With smaller parasitic capacitances, these Super-Junction MOSFETs have extremely fast switching characteristics and as a result, reduced switching losses. This switching behavior occurs with greater dv/dt and di/dt that affects switching performance through parasitic components in the devices and on the printed circuit board and influences the EMI performance of the entire system. An optimized design is very important to operate high speed MOSFETs.

I. INTRODUCTION The power losses of a switching device can be broken

down into four parts: conduction losses, switching losses, turn-off state losses due to leakage current, and driving losses. In most switching power applications utilizing high voltage switching devices, the last two parts can be neglected. The conduction losses can be reduced through realizing lowest possible on-resistance. The switching losses are determined by the duration of the switching transient, a period that current and voltage are present simultaneously across the channel of the device. Faster switching transients reduce switching power losses. The switching device should have very low parasitic capacitances to be switched quickly. As a result, considerable effort has been expended to improve the on-resistance and capacitances. Successive generations of super-junction MOSFET technology have shown a dramatic decrease of the transistor specific on-resistance (Ron, sp)[1][2]. So smaller die size and faster switching performance can be achieved by lowering RDS(ON) and gate charge (QG). However, rapid transitions in voltage and current result in high-frequency noise and radiated EMI. To achieve low-noise radiation, high values of parasitic capacitances are required. There is indirect conflict with parasitic capacitance requirements. Based on recent system trends, improving efficiency is a

critical goal; however, using a slow switching device just for EMI is not an optimal solution. This paper discusses how to achieve balance and the issues to be considered designing with fast switching power devices.

II. SUPER-JUNCTION TECHNOLOGY

A. Planar MOSFET vs. Super-Junction MOSFET

Ten years ago, the super-junction device utilizing charge balance theory was introduced and it set a new benchmark in the high-voltage power MOSFET market.[3] Fig. 1 shows the vertical structure and electric field profile of a planar and super-junction MOSFETs. The breakdown voltage of a planar MOSFET is determined by the drift doping and its thickness. The slope of the electric field distribution is proportional to drift doping. Therefore, thick and lightly doped epi is needed to support higher breakdown voltages. A major contribution to the on-resistance of a high-voltage MOSFET comes from the drift region. Therefore, the on-resistance exponentially increases with the light doping and thick drift layer for higher breakdown voltage as shown in Fig. 2.

Super-junction technology has deep p-type pillar structure in the body in contrast to the well-like structure of conventional planar technology. The effect of the pillars is to confine the electric field in the lightly doped epi region. Thanks to this p-type pillar, the resistance of n-type epi can be reduced dramatically compared to conventional planar technology while maintaining the same level of breakdown voltage. This new technology broke silicon limits in terms of on-resistance and achieved only one-third specific on-resistance per unit area when compared to planar processes.[4] This technology also achieved unique non-linear parasitic capacitance characteristics and therefore enabled reduced switching power losses.


Fig. 1. Planar MOSFET (left) and super-junction MOSFET (right).

Fig. 2. The specific RDS(ON) of Planar MOSFET and Super-junction

MOSFET as a function of the breakdown voltage.

B. SupreMOS™ Technology

Most commercially available super-junction devices adopt multiple epi layers to build the deep p-type pillar structure. In this structure, the key design parameters for lower on-resistance are the aspect ratio of p-type pillar and the distance between unit cells. A common method to achieve a higher aspect ratio is to add more layers. However, adding more layers means more complex process steps, and increased process steps result in more manufacturing costs. This structure is also hard to scale down for a narrower cell pitch. As higher cell density is critical to bring about lower on-resistance, this is also a disadvantage of multiple epi layer technology. To realize better performance and lower cost, deep trench epi filling technology was introduced. This new technology

forms a deep trench on n-type epi, and then fills it with p-type epi. In this way, the new technology achieves a much higher active cell density and simpler processes. The deep-trench-epi fill technology reduces the number of epi process steps by 67% (as compared to previous multi-epi technology). A major design challenge of deep trench epi filling technology is the ability to maintain uniformity when building the deep trench and filling it. If there are crystal defects or voids, they can cause a shift in the electrical characteristics. Therefore, precise control of the process is critical to manufacturability. After overcoming all these challenges through the device and process engineers’ endless efforts, the SupreMOS™ MOSFET became a reality. With fewer process steps, the on-resistance per specific area of SupreMOS™ MOSFET is less than one fourth of a standard power MOSFET and 40% smaller than the previous generation of a charge-compensated super-junction MOSFET, or a SuperFET™ MOSFET. The reduced on-resistance of the new technology also reduces the conduction loss in the system. In addition, it allows for a more compact system design since a device with similar on-resistance can now be packaged into a smaller package.

Fig. 3. Vertical structure of SuperFETTM (left) and SupreMOSTM (right),

not in same dimensions.

III. POWER LOSS Super-junction devices utilizing the charge balance

theory reduce the on-resistance of high voltage MOSFETs. For example, SupreMOS™ technology reached 19mOhm per square centimeter at 600V. Fig. 4 shows the advances in reducing high-voltage MOSFET on-resistance. Since conduction losses are directly

Spec

ific

Rds

on [m

ohm

-cm

2 ]

-0511223344556

0 10 20 30 40 50 60 70Breakdown Voltage

QEqBVcpspRon

cµ2., =

5.2910.6, BVspRon −=

cp=20um,Q=2E1

Planar MOSFET

Super-junction MOSFET


proportional to on-resistance, super-junction devices, like the SupreMOS™ MOSFET, provide a great advantage for conduction losses.

Faster switching transients can reduce switching losses. Since a MOSFET is a unipolar device, parasitic capacitances are the only limiting factors during switching transient. So lower parasitic capacitance is required for smaller switching losses.[5] As the charge balance principle reduces the chip size for same RDS(ON) as compared to standard MOSFET technology, it can be expected that SupreMOS™ MOSFETs will have less capacitance. As shown in Fig. 5, SupreMOS™ MOSFET input and Miller capacitance are much less than the previous generation SuperFET™ MOSFET, but output capacitance curves look similar. SupreMOS™ MOSFET shows more non-linear behavior. One way to find out how the output capacitance corresponds to switching losses is by evaluating an effective value of output capacitance. The stored energy in the output capacitance of a MOSFET can be calculated by integrating the product of the output capacitance and drain-source voltage with respect to the drain-source voltage from zero to the drain-source voltage just before the turn-on transient. This stored energy is dissipated through the channel of the MOSFET on every turn-on of the switching cycle. The SupreMOS™ MOSFET has approximately 35% in reduced stored energy in output capacitance than the same on-resistance device of the previous generation SuperFET™ MOSFET (for a typical switching power supply bulk capacitor voltage). The comparison of stored energy in output capacitance is shown in Fig. 6.

Fig. 4. Trend in on-resistance.

0.1 1 10 1000

2000

4000

6000

8000

10000

Crss

Ciss

Coss

SuperFETTM_dash lineSupreMOSTM_solid line

Cap

acita

nce

[pF]

VDS [V]

Fig. 5. Capacitances of SuperFETTM MOSFET and SupreMOSTM

MOSFET.

0 100 200 300 400 500 6000

5

10

15

20

25

30

Sto

red

Ene

rgy

in O

utpu

t Cap

acita

nce

[uJ]

Drain-Source Voltage [V]

190mOhm SuperFETTM

199mOhm SuperMOSTM

Fig. 6. Stored energy in output capacitance.

Fig. 7 compares the capacitances of a super-junction MOSFET and a planar MOSFET. Ciss of the super-junction MOSFET stays largely unchanged. However, Crss and Coss of the super-junction MOSFET become strongly non-linear. That is, Crss decreases very rapidly around a drain-source voltage of 10V in the case of the super-junction MOSFET. These effects allow an extremely fast dv/dt and di/dt. This high switching speed is effective in reducing switching losses, but it also yields negative effects such as increased EMI, gate oscillation, and high peak drain-source voltage. Therefore, control of the maximum switching speed is very important to obtain the best performance of the super-junction MOSFETs without negative effects during design-in of the device.

Fai

Fig

IV

higcoinfspeas peopMOpaals

A.

thedorincaptheduaffthehigfrovaresthe

irchild Semicondu

0.1

100

1000

10000C

apac

itanc

e [p

F]

PlSu

g. 7. Comparisonsuper-juncti

V. INFLUPARACHAR

The faster swgher power cmponents in fluence the seed increases

high voltrformance. T

ptimize the gOSFET is

arasitic inductso important.

Effect of G

One critical e external se

own the peaknging causedpacitances ofe rate of rise

uring the turnfects the swiese losses is ghest efficienom an applicalue for Rg is sults in excese MOSFET d

ctor Power Semin

1

V

anar MOSFET_dash luper-junction MOSFET

ns of capacitancion MOSFET.

UENCE OFAMETERS ORACTERISwitching of thconversion ef

the devices switching chas. This createstage, currenTo achieve bgate drive cia gate-contr

tance and cap

Gate Resistan

control parameries gate resk drain-sourced by lead f the power M

of the voltagn-on and turitching losses

important ancy on the tcation standpvery importa

ssive dv/dt acduring switch

ar 2010-2011

Vds [V]

lineT_Solid line

ces between pla

F CIRCUIT ON SWITCTICS he power MOfficiency. Howand boards a

aracteristic ass unwanted si

nt spikes, obalance, it iircuitry becarolled devicacitance on th

nce

meter in gatesistor (Rg). Te voltage and

inductance MOSFET. It age (dv/dt) andrn-off process in MOSFEas devices muarget applicapoint, selectiant. Too smalcross the draihing off. The

10

Crss

Coss

Ciss

anar MOSFET

CHING

OSFETs enabwever, parasiare increasings the switchiide effects, suor poor EMis important ause the powce. Minimizihe PCB board

-drive designThis is damped prevents g

and parasialso slows dowd current (di/ss. But Rg a

ETs. Controlliust achieve t

ation. Therefoing the corrll a value for n and source

erefore, the lo

and

bles itic gly ing uch MI to

wer ing d is

n is ens ate itic wn /dt) lso ing the ore rect

Rg of ow

limit fothe speRg causupper las the p

Fig. 8. D

Fig.

using MOSFMOSFmay bMOSFshould losses w9 showSuperFMOSFcontrolselecte

Eon

[uJ]

(a) Turn-

or Rg is a valuecification in ses more losslimit is choseprevious vers

Drain-source wav

8 shows thedifferent gatET is directET, switchinbecome highET, increasebe a limit lin

with SupreMws switchingFET™ MOSET. Based olled dv/dt and.

0 1050

60

70

80

90

100

110

120

Eon

[uJ]

SuperFESupreMO

-on energy loss (E

ue that makethe datasheet

ses and poor een to have thion of the sup

veforms of MOSF

e turn-off reste resistors. Wtly replaced ng losses are her. To cond Rg is requne for increas

MOS™ MOSFlosses with

SFET and ton the resultd for lower s

20 30 40

Eon @ Id=9A, Vd

Rg, Gate Resistor

ETTM

OSTM

Eon) vs. gate resist

s switching dts. Too largeefficiency; the same switcper-junction M

FET according to

sponse of theWhen the Swith the Sureduced, bu

ntrol the Suired. In this sing the Rg o

FET could be the previous

the new Suts, Rg suitablswitching los

0 50 60

ds=380V

r [ohm]

tors.

4

dv/dt within a value for

herefore, the ching losses MOSFET.

gate resistors.

e MOSFET SuperFET™ upreMOS™ ut the dv/dt SuperFET™

case, there or switching

larger. Fig. generation

upreMOS™ le both for sses can be

70


0 10 20 30 40 50 60 700

20

40

60

80

100

120

Eoff @ Id=9A, Vds=380V

Eof

f[uJ]

Rg, Gate Resistor [ohm]

SuperFETTM

SupreMOSTM

(b) Turn-off energy loss (Eoff) vs. gate resistors Fig. 9. Switching losses of super-junction MOSFETs according to gate

resistors.

B. Effect of Clamp Diodes

It is well know that the turn-on behavior of the MOSFET is strongly dominated by the reverse recovery characteristics of the free-wheeling diodes during clamped inductive switching. When a silicon (Si) diode is suddenly reverse-biased, a large amount of stored charge, which has not recombined in depletion layers flows in the reverse direction before blocking the reverse voltage. But a silicon carbide (SiC) Schottky diodes have no reverse recovery current during switching transition because they does not have any excessive minority carriers. There is a displacement current from parasitic junction capacitances, but this is negligible. SiC Schottky diodes and Si diodes used for clamped diode applications have a large difference in reverse recovery characteristics resulting in significantly different MOSFET switching losses and dv/dt. Fig. 10 shows the turn-on and switching losses according to the external gate resistors with different clamping diodes. SiC has an order of magnitude higher breakdown electric field(2-4×10 V/cm) and an electron mobility only 20% lower than silicon. A high breakdown electric field allows the design of SiC Schottky diodes with 10 times thinner and 100 times higher doped epitaxial layers. Therefore, in spite of much smaller die size, the junction capacitance of SiC Schottky diode is higher (more than 5 times) than that of Si diode. The junction capacitance depends on the reverse bias voltage. The switching behavior of the boost MOSFET can be affected by boost diode’s junction capacitance which can be expressed as a simple variable

capacitor. The dv/dt with the SiC Schottky diode is lower than dv/dt with the Si diode due to the bigger junction capacitance of SiC, as shown Fig. 11. The gap between dv/dt value increases at lower drain current levels and smaller values of Rg. This is because, at lower current, the dv/dt is relatively low and the effect of the output capacitance of the MOSFET and diode junction capacitance on the dv/dt becomes more significant.

0 10 20 30 40 50 60 70

10

20

30

40

50

60

70

80

MOSFET Eon @ Id=9A, Vdd=380V

Eon

[uJ]

Rg, Gate Resistor [ohm]

With Si Diode With SiC Schottky

Fig. 10. Turn-on energy loss (Eon) of super-junction MOSFET according

to clamp diodes.

0 10 20 30 40 50 60 7020

40

60

80

100

dv/d

t [V

/ns]

Rg [ohm]

With Si Diode With SiC Schottky Diode

Fig. 11. dv/dt of super-junction MOSFET with clamp diodes.

C. Effect of Ferrite Bead

One major noise source is the oscillation from the MOSFET at the switching on-off transient. Typically, parasitic oscillation frequencies are in the range of several tens to hundreds of megahertz. The parasitic oscillation can cause gate-source breakdown, bad EMI, large switching losses, losing gate control and can even lead to MOSFET failures. A ferrite bead is often used on a MOSFET gate to provide stable operation by suppressing parasitic oscillation while minimizing

Fai

sweffim1213freeffthethedecirresthewhproEq

w

Fig

Fig

imprhigto

irchild Semicondu

witching lossefective than

mpedance of th2 shows an eq

shows impequency. Rbea

fective inducte paralleled ce ferrite beadpendent resisrcuit and Lbead

sistance of thee impedance ohile the imoportionally quation 1.

Z R jX= +

where R R=

g. 12. Equivalent c

g. 13. Impedance c

The rising mpedance ve

imarily by gh frequencydominate an

ctor Power Semin

es. In fact, adusing gate rhe ferrite beaquivalent circedance charaad and Lbead tance of the fecapacitance an. The ferrite bstor. At low fd is a short cire ferrite beadof Lbead starts

mpedance of to frequency

bead paraR R+

circuit of ferrite b

characteristics vs

linear slopersus frequthe inductany point, the nd the ferrit

ar 2010-2011

dding a ferritesistance alod varies with uit of a ferritacteristics asare the DC

errite bead. Cnd resistance bead is basicafrequencies, Crcuit, leaving

d. As the frequ to increase li

Cpara decrey (1/ jωCpara)

bead.

s. frequency of a f

pe of the ency plot nce of Lbeadimpedance

te bead’s im

te bead is moone because t

frequency. Fte bead and Fs a function

resistance aCpara and Rpara a

associated wally a frequenCpara is an opRbead as the D

uency increasinearly (jωLbe

eases inverse) as shown

(

ferrite bead.

ferrite beadis determin

d. At a certaof Cpara begi

mpedance sta

ore the

Fig. Fig.

of and are

with ncy pen DC ses, ead) ely by

(1)

d’s ned ain ins

arts

to decthe falplot isof the is lowresistoNormamuch impedeffectienoughwith oscillapresenwithousuppreof theeffectiapplicaimpedinternanecessband, r

(a) Vgs a Fig. 14. G

D. Ef

The and thecontrolMOSFrequirelevel cotopologswitch vice veapplicaor disc

crease, countlling slope os determinedbead. At low

w. The ferritor instead oally, the pahigher than

dance at a ive at bloch inductancresistance,

ation can bents comparisut the ferriteessed by usie MOSFET.ive way to ation. With

dance varies al structure. sary to consrated current

at turn-on transie

Gate ringing of su

ffect of Curr

driver is the e power MOSl signals to ET and pr

ed) betweenontrol circuit gies. The pra MOSFET

ersa. High peaations so the pcharged as

tering its indof the impedd by the paraw frequencyte bead indf an inductoarasitic oscthe switchin

high freqking drain-

ce in the feits high

e significantsons of gatee bead. Oscing a ferrite . It is a vesolve an osh ferrite b

depending When select

sider the imt and also th

ent (b) Vgs at turn

uper-junction MO

rent Capabil

interface betSFET. The drrequired lev

rovides electhe power Msuch as high

rimary functifrom the off ak currents arparasitic capaquickly as p

ductance. Inance versusasitic capacity, inductive iductance woor at high

cillation freqng frequencquency is -to-gate noierrite bead

frequency t suppressed oscillationsillation is drbead on the

ery simple cillation issbead inducon the ma

ting a ferritempedance in he impedance

n-off transient

OSFET

lity of Driver

tween the conrive circuit amels to drive trical isolati

MOSFET andh side switcheion of the dstate to the ore needed in acitances can possible to

6

n this case, frequency tance Cpara impedance

orks like a frequency. quency is y. Its high extremely

ise. Given combined

parasitic d. Fig. 14 s with and ramatically e gate lead and cost-ues in the ctors, the aterial and e bead, it is

the noise e gradient.

r IC

ntrol circuit mplifies the

the power ion (when d the logic es on bridge driver is to on state and high power be charged switch the


MOSFETs efficiently. The switching behavior and power dissipation both depend on the current capability of the output driver while the MOSFET gate-source voltage is between the threshold level and Miller plateau voltage. Table I shows the comparisons of peak sink and source current capability of drive ICs.

TABLE 1. COMPARISONS OF CRITICAL SPECIFICATION OF GATE

DRIVERS

Device Condition IPK_SINK

IPK_SOURCE

FAN3122T CLOAD=1.0µF,f=1kHz,Vdd=12V 11.4[A] -10.6[A] FAN3224T CLOAD=1.0µF,f=1kHz,Vdd=12V 5.0[A] -5.0[A] FAN3111C CLOAD=1.0µF,f=1kHz,Vdd=12V 1.4[A] -1.4[A]

Fig. 15 and Fig. 16 show how much switching losses

can be reduced with a high-current driver. But notice how the gaps in switching losses are not significant with a large gate resistor because it limits gate current.

0 2 4 6 8 104

6

8

10

12

14

16

18

20

22

Eon @ Rg=2.2ohm

Eon[

uJ]

Drain Current [A]

FAN3122T FAN3224T FAN3111C

(a) Turn-on energy loss(Eon) vs. gate drivers with Rg=2.2ohm.

0 2 4 6 8 107.0

7.5

8.0

8.5

9.0

9.5

10.0

10.5

11.0

11.5

12.0 Eoff @ Rg=2.2ohm

Eoff[

uJ]

Drain Current [A]


(b) Turn-off energy loss(Eoff) vs. gate drivers with Rg=2.2ohm. Fig. 15. Switching losses of super-junction MOSFET according to gate

drivers with Rg=2.2ohm.

0 2 4 6 8 10

8

12

16

20

24

28

32

36

40 Eon @ Rg=20ohm

Eon

[uJ]

Drain Current [A]


(a) Turn-on energy loss(Eon) vs. Gate Drivers with Rg=20ohm.

0 2 4 6 8 106

9

12

15

18

21

24

27

30

33

36 Eoff @ Rg=20ohm

Eof

f[uJ]

Drain Current [A]


(b) Turn-off Energy loss(Eoff) vs. Gate Drivers with Rg=20ohm. Fig. 16. Switching losses of super-junction MOSFET according to gate

drivers with Rg=20ohm.

E. Effect of Common Source Inductance

The effect of common source inductance will be considered by taking an interleaved CRM PFC circuit as an example. Fig. 17 shows one of the two phases of a two-phase interleaved CRM PFC circuit considering common source inductance. The source inductance is common to the driver loop and the power loop. A circuit with a low value for Ls has a lower turn-off switching loss and lower gate oscillation in spite of a higher gate-source negative voltage and higher overshoot of drain-source voltage due to a faster slew rate. In contrast, a circuit with a high Ls value shows much higher turn off switching loss, higher gate oscillation, but lower overshoot of the drain-source voltage due to slow slew rate as shown in Fig. 18 and Fig. 19. The extremely fast switching speed of the power MOSFETs such as super junction MOSFETs can lead to severe voltage and


current ringing during switching due to the parasitic components of the interconnects. In many applications, these interconnects consist of a printed circuit board, component leads, cables, and wires. The problems that the interconnects cause are often dominated by stray inductance rather than by stray capacitance, especially in low and medium voltage power supplies. So the best way to reduce the switching noise is to minimize the common source inductance.

Fig. 17. Single phase of interleaved CRM PFC considering common

source inductance.

(a) Vgs waveform for low LS (b) Vgs waveform for high LS.

(c) Vds and Id waveforms.

Fig. 18. Drain-source waveforms according to different common source inductances.

(a) Turn-off loss (b) peak drain-source voltage.

(c) Slew rate dv/dt

Fig. 19. Comparison of turn-off loss, drain-source voltage, and slew rate

by different common source inductances.

F. Effect of the Gate Drive Circuit

The most important means of avoiding oscillation is minimizing the inductance of the traces between the gate driver and the gate.

A configuration of gate drive circuitry is also important to switching characteristics. In Fig. 20, four types of gate drivers are considered. One is a PNP transistor turn-off circuit (circuit A) that is most popular for minimizing loop and fast turn-off. Fig. 21, Fig. 22, and Fig. 23 show lower gate oscillation and lower turn-off loss, but higher overshoot of the drain-source voltage. A diode turn-off speed-up circuit (circuit B) has lower gate oscillation than circuit A; but in this case, the discharge current of MOSFET flows through an off-resistor of totem-pole in gate driver IC, so it has a slower switching speed, higher turn-off loss, and lower overshoot of the drain-source voltage. Circuit D is a compensated circuit that has reduced series resistance, Roff=5.6Ohm, to make the same total series resistance to circuit A. It has a higher gate oscillation and higher turn-off loss. The circuit C is driving a circuit, while considering longer turn-on and off path than circuit A. It also has a higher gate oscillation. LP1 and LP2 are each parasitic inductances for a turn-on and turn-off path. To achieve the best performance relative to gate oscillation, it is important to connect the driver-stage ground directly to the source pin of the MOSFET as close as possible. In this way, it is possible to reduce gate oscillation caused by parasitic components at the turn-off transient. Considering these results, the PNP transistor turn-off circuit (circuit A) is the best configuration for reduced gate oscillation and switching loss.


(a) Circuit A (b) Circuit B.

(c) Circuit C (d) Circuit D.

Fig. 20. Gate driving circuits.

(a) Circuit A. (b) Circuit B.

(c) Circuit C. (d) Circuit D. Fig. 21. Comparison of turn-off gate-source voltage.

(a) Between Circuit A & B. (b) Between Circuit A & C.

(c) Between Circuit A & D.

Fig. 22. Comparison of turn-off drain current and drain-source voltage.

(a) Gate-source negative voltage. (b) Turn-off loss.

(c) Drain-source voltage.

Fig. 23. Comparison with measured gate-source negative voltage, turn-off loss and drain-source voltage with different gate driving circuits.

V. PRACTICAL LAYOUT REQUIREMENTS

A. Measurement Technique

If an oscilloscope is used to measure noise, there are several precautions that must be taken. First, it’s important to realize that a probe, even if it is just a simple piece of wire, is potentially a very complex circuit. For DC signals, a probe appears as a simple conductor pair with some series resistance and a terminating resistance. However, for AC signals, this picture changes dramatically as signal frequencies increase as shown in Fig. 24. Since the ground lead is a wire, it has some amount of distributed inductance as shown in Fig. 25. A ground lead approximately 10cm long clips to the ground of the circuit, and a spring-loaded barrel on the probe hooks onto the test point. This inductance interacts with the probe capacitance to cause ringing at a certain frequency that is determined by the L and C values. This ringing is unavoidable, and may be seen as a sinusoid of decaying amplitude that is impressed on pulses. The effects of ringing can be reduced by designing a probe grounding so that the ringing frequency occurs beyond the bandwidth limit of the probe or oscilloscope system. To avoid grounding problems, it is always important to use the shortest possible ground lead. Substituting other means of


grounding can cause ringing to appear on measured pulses. It is best to use the oscilloscope on its maximum bandwidth for these measurements to ensure seeing all waveforms of devices. Also taking full advantage of the oscilloscope’s measurement capabilities requires a probe that matches the oscilloscope’s design considerations.

Fig. 24. Equivalent circuit of probe for AC signal.

Fig. 25. Effect of ground lead of the probe.

Fig. 26 (a) shows the standard oscilloscope probe test leads. A ground lead approximately 10cm long clips to the ground of the circuit, and a spring-loaded barrel on the probe hooks on the test point. Fig. 26 (b) shows the same oscilloscope probe with a proper setup for low gate oscillation measurements. Both the probe barrel and the ground lead have been removed from the probe. Fig. 27 shows comparison waveforms of the gate for the two different probe setups. There is a 26V peak-to peak gate-source voltage with the standard setup, but only a 11.2V peak-to-peak gate-source voltage with the improved setup. It is best to use the oscilloscope on its maximum bandwidth for these measurements. The ground lead of the probe must be directly connected to the source of switching device, not the ground of the bulk capacitance.

(a) Measurement with standard setup (b) measurement with probe tip and ground lead removed. Fig. 26. Comparison of measurement methods.

-100 -80 -60 -40 -20 0 20 40 60 80 100-15

-10

-5

0

5

10

15

Vgs

[V]

Time [ns]

Measuremet with standard setup Measuremet with Probe tip

Fig. 27. Comparison measured oscillation according to probing methods.

B. Package & Layout Parasitics

To drive fast switching super-junction MOSFETs in different applications, it is necessary to understand the influence of the package in the case of switching performance for the MOSFET. This issue is well known and documented from the LV MOSFET area.

Low RDS(ON) and low inductive packages are a must for low voltage packages to achieve the best switching performance and reduced conduction losses for highest efficiency. The super-junction MOSFETs are mainly used in the voltage range of 500-600V. In these voltage ratings, clearance and creepage distance requirements must also be considered. So the most popular packages are TO-220, TO-247, TO-3P and TO-263.

Fig. 28. Several parasitics within a TO-220 package.

The impact the package performance is limited by the fixed internal gate and source bondwire length. Only the lead length can be changed to reduce the source inductance of the package. A typical lead inductance of 10nH is shown in Fig. 27. It does not sound much. Assume that the MOSFET turns off a current with di/dt=500A/µs what is easily possible. The voltage

Oscilloscope

Probe

Ground LeadR

R

CC

L

L


across this inductance is Vind = 5V, and with a turn off di/dt of 1000A/µs the induced voltage is Vind = 10V. This calculation shows that the complete source inductance, not only the lead inductance of the package, must be reduced to an acceptable value. In addition, low inductance shunt resistors are required.

Another source of noise is layout parasitics caused either by parasitic inductance or by parasitic capacitance. Both influence the performance of the layout. As mentioned before, 1cm of trace pitch has an inductance of 6-10nH, which can be reduced if by adding one layer on the top side of the PCB and a GND plane on the bottom side.

Fig. 29 shows the principles of capacitive layout parasitics. The capacitance between one trace is immediately over the other trace or GND plane on the other side of the PCB and the second is the capacitance and is built up between the device itself and the GND plane. Two parallel traces on both sides of the PCB increase the capacitance, but also reduce the inductance of the loop. The result is less magnetic noise radiation.

Fig. 29. Capacitive layout parasitics.

C. Oscillation Circuits

There are several oscillation circuits that affect the switching behavior of the MOSFET, including internal oscillation circuits[6] and external oscillation circuits. Fig. 30 shows a simplified schematic for a power MOSFET with all internal parasitic capacitances and with an external oscillation circuit given by the external couple capacitance Cgdext of the board layout. This external couple capacitance affects the Vgs slope during turn off of the power MOSFET. To see how external Cgd affects the switching behavior, compare it in different layouts. Both layouts are single-layer designs. For the measurements, both boost stages are running with a single transistor.

Fig. 30. Simplified schematic of power MOSFET with external Cgd and

lead inductance.

The measurements were done in a PFC boost stage at the same input voltage VIN=180VAC and output power level Pout=300W, with the difference in oscillations shown in Fig. 31 (a) and (b). VGS (green line) and VDS (magenta line) represent turn off. The experimental waveforms below show the effect of the high and low external Cgd in a given layout. This oscillation effect can be forced if you increase the output power level or decrease the input voltage at the same output power.

Fig. 31. (a) Turn off with high Cgdext. (b) Turn off with low Cgdext.

This effect can also happen after an AC line drop out; when line voltage is back, the boost stage charges up the bulk capacitor to a nominal voltage. During this time, when the MOSFET turns off, the drain current is quite high. The drain current commutates to the output capacitance, Coss of the MOSFET and charges it up to the DC bus voltage. The voltage slope is proportional to the load current and inversely proportional to the value of the output capacitance. The value of Coss is high at low VDS and low at high VDS and, as a result, dv/dt values of drain-source voltage change during turn off. The high dv/dt values lead to capacitive displacement currents due to all the parasitic capacitances around.

Fai

Tocapdatradi/beopaphe

D.

is Cg

osclay

Fig

Fig

irchild Semicondu

ogether with apacitances, a

amped by the ansient at inpu/dt and dv/dthavior or ev

ptimized layouplication, falp to improve

Layout Req

External Cgd part of the de

gd must be redcillation. Expyout are show

g. 32. PCB with hi

g. 33. PCB with lo

ctor Power Semin

all the layout an LC oscillainternal Rg. Uut voltage or t occur), this

ven damaged ut and mecha

ast switchinge efficiency an

quirements

is one of theevice and PCBduced as mucperimental wawn in Fig. 32 a

igh Cgdext.

ow Cgdext.

ar 2010-2011

and parasitication circuit Under certain

short-circuit leads to unudevices. Nev

anical structug super-junctnd power den

e root causes B layout. Thech as possibleaveforms depand Fig. 33.

c inductance ais created onconditions (econditions hiusual switchivertheless, w

ure of the enttion MOSFEnsity.

for ringing aerefore, externe to reduce gpending on PC

and nly e.g. igh ing

with tire

ETs

and nal ate CB

The cap

Figure

Fig. 34. L

Comwith re

Fig. 35. L

Induon perfin a lothe curan induof Vind

L debigger inductainductacurrentAlso thmust switchi

pacity betwee

34 shows lay

(a) Single layer

Layout example w

mpare to thiseduced extern

(a) Double layer

Layout solution w

uctances and formance in aop create a mrrent, the maguctive voltagecan be calcul

epends also the area of

ance L. In othance results t loops with hhe source indube minimizeing behavior,

en traces can

youts with hig

with increased ext

Fig. 35, shnal Cgd.

with reduced exter

current loopsa given applicmagnetic fieldgnetic field ae, Vind. With Flated as:

on the geomf the currenher words, forin higher V

high di/dt valuuctance and sed to avoid, as shown in

be calculated

gh external Cg

(b) Double laye

ternal Cgd.

howing layou

(b) Multi lay

rnal Cgd.

s also have acation. Curred. If a changalso changes Faraday’s law

·

metry of the nt loop, the r constant di/

Vind. Generallues should beshunt resistor d degradingn Fig. 36. Fo

12

d with: (2)

gd.

er

ut solutions

yer

an influence nts running

ge occurs in and creates

w, the value

·(3)

loop. The larger the

/dt, a higher ly enclosed e minimized.

inductance EMI and

or example,


1 cm track on a printed circuit board is almost the same with 6-10nH of inductive impedance.

Fig. 36. Commutation loop e.g. PFC stage.

VI. CONCLUSION As power conversion efficiency becomes more critical,

technology of discrete devices advances every day. An extremely fast switching super-junction MOSFET is an essential choice for higher efficiency but it is not as easy to control as previous generations. Optimized gate driver and board layout are the most important design parameters when working with fast switching MOSFETs. Practical design tips are discussed to take advantage of new features of SupreMOS™ devices.

REFERENCES [1] T. Fujihira and Y. Miyasaka, “Simulated Superior Performances of

Semiconductor Super-Junction Devices,” Proc. of the ISPSD, pp.423-426, June, 1998.

[2] T. Fujihira, “Theory of Semiconductor Super Junction Devices,” Jpn. J. Appl. Phys., vol. 36, pp. 6254–6262, 1997.

[3] “CoolMOS – a New Milestone in High-Voltage Power MOSFET,” Lorenz L., Deboy G., Knapp A., Maerz. M, ISPSD 1999.

[4] G. Deboy, M. März, J.-P. Stengl, H. Strack, J. Tihanyi, and H. Weber, “A New Generation of High-Voltage MOSFETs Breaks the Limit Line of Silicon,” in IEDM Tech. Dig., 1998, pp. 683–685.

[5] S. Clementi, B. R. Pelly and A. Isidori. “Understanding Power MOSFET Switching Performance”, Proc. IAS, pp.763-776, 1981.

[6] New Oscillation Circuits Discovered in Switching-Mode Power Supplies T. Fujihira, T. Yamada, Y Minoya.

Wonsuk Choi received the M.S. degree in electrical engineering from the Hanyang University, Seoul, Republic of Korea, in 2005. He joined the Fairchild Korea Semiconductor in 2004. He is working as an application engineer of power supply system team, HV PCIA.

Dongkook Son received the M.S. degree in electronics engineering from Kyungpook National University, Taegu, Republic of Korea, in 2006. He is currently with Fairchild Korea Semiconductor working as an application engineer of power supply system team, HV PCIA.

Markus Hallenberger received his graduation in electronic engineering from University Paderborn, Dept. Soest (Germany) in 1999. From 1999 to 2004, he worked as a development engineer for SMPS at Puls Munich. From to 2004 to 2008, he worked as Technical Marketing Engineer for High Voltage Devices at Infineon. He joined Fairchild in 2008 as a FAE Manger for Central Europe.

Sungmo Young received B.S. and M.S. degrees from Hanyang University, Seoul, Republic of Korea, in 1998 and 2000, respectively. Since 2000, he has been with Fairchild Korea Semiconductor, working as a staff application engineer of power supply system team, HV PCIA.

driving and layout requirements for fast switching … and layout requirements for fast switching...

Documents