ON Semiconductor The “PWM Switch” in mode transitioning SPICE models PCIM Germany 2005 Christophe Basso - Application Manager The “PWM Switch” in mode transitioning SPICE models Agenda Why average simulations? What techniques already exist? The PWM Switch concept The voltage-mode case The current-mode case Checking averaged model’s validity Conclusion www.onsemi.com November 2001 Christophe Basso – PCIM 2005 2 The “PWM Switch” in mode transitioning SPICE models Why average simulations? Unveil open-loop ac response for stabilization purposes Helps to assess impact of stray elements variations on stability Can predict transient response with large-signal models Simulation time is quick as frequency component fades away www.onsemi.com November 2001 Christophe Basso – PCIM 2005 3 The “PWM Switch” in mode transitioning SPICE models What techniques already exist? State-Space Averaging (SSA) Introduced by Slobodan Ćuk in the 80’ Long and painful process Fails to predict sub-harmonic oscillations x1 L x1 L Vout Vout on off dx1 1 1 = − x2 + u dt L L u1 x2 C R dx 2 1 1 = x1 − x2 dt Cout Rload .Cout x2 C R dx1 1 = − x2 dt L dx 2 1 1 = x1 − x2 dt Cout Rload .Cout Pfffff! Apply smoothing process Linearize www.onsemi.com November 2001 Christophe Basso – PCIM 2005 4 / The “PWM Switch” in mode transitioning SPICE models What techniques already exist? The GSIM concept Introduced by Sam Ben-Yaakov in the 90’ Easy to derive but not fully invariant (dual inductors converters?) Fully auto-toggling mode models Fails to predict sub-harmonic oscillations Ib Ic on b Lf off Fsw Ia c a www.onsemi.com November 2001 Christophe Basso – PCIM 2005 5 V(a,c) Vin Lf on a V(a,b) Fsw V(a,b) V(a,c) c off b C R The “PWM Switch” in mode transitioning SPICE models What techniques already exist? The CoPEC model Introduced by the Colorado Power Electronic Center in the 90’ Easy to derive and fully invariant Fully auto-toggling mode models Fails to predict sub-harmonic oscillations i1(t) i1(t) v2(t) v1(t) Q1 v2(t) v1(t) D1 Q1 D1 i12(t) CCM i1(t) d1(t) i1(t) v2(t) v1(t) v1(t) i12(t) Re(d1) I2 v2(t) i12(t) D':D I2=(Re x i1^2) / v2 www.onsemi.com November 2001 i12(t) DCM d(t) Christophe Basso – PCIM 2005 6 The “PWM Switch” in mode transitioning SPICE models What techniques already exist? The Ridley models Introduced by Raymond Ridley from VPEC in the 90’ Use z-transform method No auto-toggling mode models Can only work in ac Can predict sub-harmonic oscillations in CCM AC model Vin Vout 0 2 D Gnd Vg AC = 0 0.458 4 Duty Ctrl 0 3 V2 AC = 1 Vout RS = 20m FS = 50k VOUT = 5 RL = 3 VIN = 11 X1 RI = 0.33 L = 37.5u Resr 100m 0 1 0 5 Cout 220uF www.onsemi.com November 2001 Christophe Basso – PCIM 2005 7 Rload 3 The “PWM Switch” in mode transitioning SPICE models What techniques already exist? The PWM Switch Introduced by Vatché Vorpérian in the mid-80’ Easy to derive and fully invariant No auto-toggling mode models Can predict sub-harmonic oscillations in CCM DCM model was never published! a c L d a Ia(t) Ic(t) d d' c PWM switch Vin d' Vap(t) p Vcp(t) www.onsemi.com November 2001 Christophe Basso – PCIM 2005 8 p C R Vout The “PWM Switch” in mode transitioning SPICE models The PWM Switch concept L Linear network a Vin PWM switch d d' p c on off Linear network What do you plead? diode + transistor = guilty for non-linearity! www.onsemi.com November 2001 Christophe Basso – PCIM 2005 9 C R Vout The “PWM Switch” in mode transitioning SPICE models The PWM Switch concept 5 ib Rc 10k Rb_upper 1Meg Vin 7 Beta.Ib e Vg Vin 4 Q1 Vout 8 h11 Vout 1 ic c b Req Rb_upper//Rb_lower 3 Rc 10k ie 2 Rb_lower 100k Remember the bipolars Ebers-Moll model… Re 150 Re 150 Ce 1nF Ve Replace Q1 by its small-signal model www.onsemi.com November 2001 Ce 1nF Christophe Basso – PCIM 2005 10 The “PWM Switch” in mode transitioning SPICE models An invariant association R p d' a Vin C Boost PWM switch d p c Vin c Buckboost d' d a PWM switch L Vout C R Vout L a a Ia(t) Ic(t) d c Vcp(t) www.onsemi.com November 2001 Buck d' PWM switch Vin p L d d' Vap(t) c Christophe Basso – PCIM 2005 11 p C R Vout The “PWM Switch” in mode transitioning SPICE models Observe waveforms and average them a Ia(t) Ic(t) d L c d' Vout Vin I a (t ) Tsw = Ia = 1 Tsw Tsw ∫ 0 I a (t )dt =d I c (t ) Vap(t) Tsw C Vcp(t) p = dI c Vcp (t ) www.onsemi.com November 2001 Christophe Basso – PCIM 2005 12 Tsw = Vcp = R 1 Tsw Tsw ∫V cp 0 (t )dt =d Vap (t ) Tsw = dVap The “PWM Switch” in mode transitioning SPICE models PWM Switch model in CCM: a 1:D transformer! a c Ia(t) a Ic(t) Vap p p Large-signal (non-linear) model www.onsemi.com November 2001 d 1 V=d.V(a,p) I=d.Ic p Ic(t) Ia(t) Christophe Basso – PCIM 2005 13 c Vcp The “PWM Switch” in mode transitioning SPICE models Use it immediately, SPICE linearizes it for you! 10.0 L c p Vout d 100uH 10.0 a Vin 10 0.400 16.7 Vbias 0.4 AC = 1 C1 100u R1 10 Always verify the dc operating point! www.onsemi.com November 2001 Christophe Basso – PCIM 2005 14 The “PWM Switch” in mode transitioning SPICE models The original CCM/DCM PWM Switch models d 1-d c a vap (t ) CCM: common « passive » vcp (t ) p d1 d2 a Looks like auto-toggling is impossible… p DCM: common « common » d3 vcp (t ) vac (t ) c www.onsemi.com November 2001 Christophe Basso – PCIM 2005 15 The “PWM Switch” in mode transitioning SPICE models Deriving the DCM PWM Switch in common « passive » Ia a Ipeak Ia(t) L Ic(t) d1 d2 c d3 Vout Vin t Vap Ic Vap(t) t Ipeak Vcp t Vap Ia = I peak d1 Ic = I peak d1 Vcp t d1Tsw d2Tsw d3Tsw p Ic = 2 2 + I peak d 2 2 = R I peak ( d1 + d 2 ) 2 (d + d2 ) 2 I a ( d1 + d 2 ) = Ia 1 2 d1 d1 www.onsemi.com November 2001 C Vcp(t) Christophe Basso – PCIM 2005 16 The “PWM Switch” in mode transitioning SPICE models An auto-toggling version: clamp the equation! a Vap Ic(t) Ia(t) N 1 p Clamp d2: d2 CCM = 1- d1 d2 DCM = 1- d1- d3 c Vcp N=d1/(d1+d2) 2 I c L − Vac d1 ²Tsw 2 LFsw I c d2 = = − d1 Vac d1Tsw d1 Vac www.onsemi.com November 2001 Christophe Basso – PCIM 2005 17 d2 < d2 CCM model is in DCM! Model input The “PWM Switch” in mode transitioning SPICE models In voltage mode, add the PWM modulator gain X4 PWMCCMVM a 0.500 10.0 17 V4 10 GAIN 3 5.00 c L1 75u 4 d PWM switch VM XPWM GAIN K = 0.5 K PWM = p 5.0 www.onsemi.com November 2001 Christophe Basso – PCIM 2005 18 1 Vpeak The “PWM Switch” in mode transitioning SPICE models Testing the auto-toggling model c R5 20m L1 75u 4.99 a 4 vout 4.99 17 L1 75u IC = X2 PSW1 Vout R4 20m 1 0.499 10.0 5 V4 10 GAIN 3 d PWM switch VM Resr 70m p X10 PWMVM2 L = 75u Fs = 100k XPWM GAIN K = 0.5 Xstep PSW1 vout Vout 4 Resr 70m 7 10 4.99 2 Cout 100u IC = 0 V4 10 19 16 D2 N = 0.01 Vstep Cout 100u IC = 20 Xstep PSW1 Vstep vout vout C2 {C2} R3 {R3} C1 {C1} R2 {R2} Rupper {Rupper} 14 0.329 X8 COMPAR 3 C3 {C3} 2.50 Vsaw Tran Generators = PULSE V2 2.5 Rlower 10k Averaged model 8 Verr X1 AMPSIMP VHIGH = 1.9 Cycle-by-cycle www.onsemi.com November 2001 18 13 13 2.50 X2 AMPSIMP VHIGH = 1.9 Rupper {Rupper} 19 8 Verr R3 {R3} C1 {C1} R2 {R2} 5.00 7 6 6 + - C2 {C2} Christophe Basso – PCIM 2005 19 V2 2.5 Rlower 10k C3 {C3} The “PWM Switch” in mode transitioning SPICE models Comparing results with a stepload… 5.40 Cycle-by-cycle 5.20 , 5.00 4.80 Averaged model Output voltage 4.60 Averaged model 1.10 Error voltage I can’t believe this result… Cycle-by-cycle 900m 700m 500m 300m 9.75m 11.2m 12.7m time in seconds 14.2m www.onsemi.com November 2001 Christophe Basso – PCIM 2005 20 15.7m The “PWM Switch” in mode transitioning SPICE models Current-mode PWM switch Same approach as before: 9 observe and average waveforms 9 get the equivalent representation 9 perturb for small-signal analysis CCM DCM www.onsemi.com November 2001 Christophe Basso – PCIM 2005 21 The “PWM Switch” in mode transitioning SPICE models CCM operation, current expression 3 1 2 I c (t ) Ri = Vc (t ) − d (t )Tsw Se − 1 Ic = 2 S f d '(t )Tsw 2 Vc T S T − d sw e − Vcp (1 − d ) sw Ri Ri 2L 3 a c Ia(t) Ic(t) I=Vc/Ri I=d.Ic p www.onsemi.com November 2001 Christophe Basso – PCIM 2005 22 p I=Iu Cs The “PWM Switch” in mode transitioning SPICE models DCM operation, current expression Vc − d1Tsw × Se Ri V −dT S I c = c 1 sw e − α d 2Tsw S f Ri I peak d1 I peak d 2 I peak ( d1 + d 2 ) Ic = + = 2 2 2 I peak = Iµ = d1 × Tsw × Se V (c, p ) ⎛ d1 + d 2 ⎞ + d 2 × Tsw × × ⎜1 − ⎟ 2 ⎠ Ri L ⎝ a c Ia(t) Ic(t) I=Vc/Ri I=(d1/(d1+d2)).Ic p www.onsemi.com November 2001 Christophe Basso – PCIM 2005 23 p I=Iu The “PWM Switch” in mode transitioning SPICE models The PWM Switch, Switch the final encapsulation duty_cycle a c R4 20m L1 75u 2 duty-cycle 1 L1 75u IC = 250m Vout 4 1 R4 1 vout Vout 4 Resr 70m 12 d PWM switch CM V4 25 X2 PSW1 vout Resr 70m p X1 PWMDCMCM V4 25 16 D1 1n5818 Cout 220uF IC = 4.6 16 20 V2 Cout 220uF 10 X5 PSW1 V2 C2 470p vout X5 PSW1 2 S V1 Q Q C1 10n R1 20k R7 10k 12 10 R 13 L4 1p C1x 100p 6 7 C3 1p Y4 X4 AMPSIMP V6 2.5 R8 10k - X3 COMPAR 8 + 9 14 6 R3 470 B1 Voltage 17 V(4,vout) C2 470p 11 V7 AC = 1 vout C1 10n R1 20k R7 10k 24 18 19 Verr A buck: averaged model vs cycle-by-cycle www.onsemi.com November 2001 Christophe Basso – PCIM 2005 24 X4 AMPSIMP V6 2.5 R8 10k The “PWM Switch” in mode transitioning SPICE models Good matching between both models Error voltage Output voltage www.onsemi.com November 2001 Christophe Basso – PCIM 2005 25 It can’t be, he is cheating! The “PWM Switch” in mode transitioning SPICE models Testing the ac response 0.649 vc a duty-cycle 0.408 dc X1x XFMR -127RATIO = -0.1 1 Vin 126 0 4 Vout D1 MBR140P L1 out1 2.2uH 12.2 12.2 out2 6 12.7 3 R4 100m 126 L3 1.8m PWMCM X1 L = 1.8m Fs = 66k Ri = 1.5 12.2 15 C1 470uF out1 R7 8k LoL 1kH 0.649 0.649 14 9 C5 470uF out2 R15 1.5k 10.6 VFB 16 5 CoL 1kF 0 Cf 100nF 9.88 12 10 Vstim AC = 1 X3 TL431 Rupp 3.9k 2.50 13 Rlow 1k A dcm current-mode flyback www.onsemi.com Christophe Basso – PCIM 2005 26 Rload 14 12.2 7 V3 4.8 11 R17 300m R5 100m 12.2 4.80 November 2001 Rs 10m12.2 c p 2 PWM switch CM 18 C2 10uF The “PWM Switch” in mode transitioning SPICE models Ac simulation results of the flyback converter Phase Gain BW = 600Hz 0 Y = 20dB/div 10 100 Y = 45°/div 1K 10K 100K Simulation Measurement www.onsemi.com November 2001 Christophe Basso – PCIM 2005 27 The “PWM Switch” in mode transitioning SPICE models If the load increases… plot1 vdbfb in db(volts) 60.0 gain Sub-harmonic oscillations! 40.0 20.0 0 -20.0 220 CCM operation 1 phase Plot2 ph_vfb in degrees 2 180 140 100 60.0 10 100 1k frequency in hertz www.onsemi.com November 2001 Christophe Basso – PCIM 2005 28 10k 100k The “PWM Switch” in mode transitioning SPICE models Testing on a multi-output forward L2 {L2}12.2 12.2 4 R2x 70m 12.0 5 Vout2 9 X1 XFMR RATIO = N2 12.0 Rload2 6 2 Coupled inductances Cout2 100u X7 XFMR RATIO = N1/N2 0.356 160 a 10 0.861 1 parameters Rsense=0.35 Vout=28 L1=130u L2=130u N1=0.5 N2=0.215 duty-cycle c PWM switch CM L1 {L1}28.3 28.3 56.6 8 11 R9 70m vout 28.0 Vout1 3 X6 XFMR RATIO = N1 p X5 PWMCM2 L = L1/N1^2+L2/N2^2 Fs = 200k Ri = Rsense Se = 0 Resr1 245m 28.0 Rload1 7 6 Cout1 48u vout C2 {C2} Rupper=(Vout-2.5)/250u fc=5k pm=50 Gfc=8.84 pfc=-66 G=10^(-Gfc/20) boost=pm-(pfc)-90 pi=3.14159 K=tan((boost/2+45)*pi/180) C2=1/(2*pi*fc*G*k*Rupper) C1=C2*(K^2-1) R2=k/(2*pi*fc*C1) dc vc V6 160 C1 {C1} R2 {R2} Rupper {Rupper} 20 2.50 0.861 13 LoL 1kH 19 0.861 CoL 1kF 0 2.50 Verrx 18 X4 AMPSIMP V3 2.5 Rlower 10k 7 V1 AC = 1 A multi-output forward www.onsemi.com November 2001 12 V Resr2 100m Christophe Basso – PCIM 2005 29 28 V The “PWM Switch” in mode transitioning SPICE models Output voltage bang on the 28 V output… Plot1 vout, vout1 in volts 28.8 28 V 28.4 28.0 3 1 27.6 Inductances are coupled… 27.2 12.4 12.5 12.2 12.3 12.0 vout2#a in volts Plot2 vout2 in volts Averaged Cycle-by-cycle 12 V 12.1 2 4 11.8 11.9 11.6 11.7 3.93m 4.83m 5.74m time in seconds 6.64m A forward converter www.onsemi.com November 2001 7.55m Christophe Basso – PCIM 2005 30 The “PWM Switch” in mode transitioning SPICE models Output voltage bang on the 28 V output… plot1 vout, vout1 in volts 28.8 28 V 28.4 28.0 4 1 27.6 Inductances are un-coupled… 27.2 Averaged Cycle-by-cycle Plot2 vout2, vout2#a in volts 12.8 12 V 12.4 12.0 3 2 11.6 11.2 3.83m 4.75m 5.68m time in seconds 6.61m 7.54m A forward converter www.onsemi.com November 2001 Christophe Basso – PCIM 2005 31 The “PWM Switch” in mode transitioning SPICE models Instability in the buck DCM current-mode Gain Vin = 7.5 V Rload = 50 Ω 80.0 180 > 20 dB increase -90.0 vdbout, vdbout#1, vdbout#2 in db(volts) Plot1 vphout, vphout#1, vphout#2 in degrees 0 Gain Vin = 10 V Rload = 50 Ω 0 3 5 1 Phase Vin = 10 V Rload = 50 Ω 2 4 6 Phase Vin = 7.5 V Rload = 50 Ω -40.0 Phase jumps to –180° -180 Gain Vin = 7 V Rload = 100 Ω 40.0 90.0 The DCM buck shows instability as M > 0.66 without ramp Vin = 7V M = 0.71 Vin = 7.5V M = 0.66 Vin = 10V M = 0.5 Phase Vin = 7 V Rload = 100 Ω -80.0 1 10 100 1k frequency in hertz 10k www.onsemi.com November 2001 Christophe Basso – PCIM 2005 32 100k 1Meg The “PWM Switch” in mode transitioning SPICE models Instability in the buck DCM current-mode 40.0 180 -20.0 -40.0 Adding 0.086 x Soff cures the problem Gain Vin = 7 V Rload = 100 Ω Sa = 12.28 kV/s 90.0 vphout, vphout#2 in degrees Plot1 vdbout, vdbout#2 in db(volts) 20.0 0 Gain Vin = 7 V Rload = 100 Ω No ramp 2 4 0 Phase Vin = 7 V Rload = 100 Ω Sa = 12.28 kV/s -90.0 3 1 Phase Vin = 7 V Rload = 100 Ω No ramp -180 1 10 100 1k frequency in hertz 10k www.onsemi.com November 2001 Christophe Basso – PCIM 2005 33 100k 1Meg The “PWM Switch” in mode transitioning SPICE models The conclusion The CM PWM Switch DCM was derived Two auto-toggling models developed Good matching of average vs reality Models also exist in BCM (PFC simulations) Exist in both IsSpice and PSpice www.onsemi.com November 2001 Christophe Basso – PCIM 2005 34