Small-Signal Modeling of the Boost Converter Operated in CM Christophe Basso IEEE Senior Member 1 • Chris Basso – CM Boost Study Course Agenda The PWW Switch Concept Small-Analysis in Continuous Conduction Mode Small-Signal Response in Discontinuous Mode EMI Filter Output Impedance Cascaded Converters Operation 2 • Chris Basso – – CM Boost Study Course Agenda The PWW Switch Concept Small-Analysis in Continuous Conduction Mode Small-Signal Response in Discontinuous Mode EMI Filter Output Impedance Cascaded Converters Operation 3 • Chris Basso – – CM Boost Study Non-Linearity in a Switching Converter A switching converter is ruled by linear equations… x1 u1 SW u1 L x2 C during DTsw R x1 off L x2 C …combining so-called state variables 4 • Chris Basso – – CM Boost Study R y1 C on L R during 1 D Tsw State Variables State variables describe the mathematical state of a system n state variables for n independent storage elements knowing variables state at t0 helps compute outputs for t > t0 System described by state variables u1 t u2 t u3 t y1 t y2 t y3 t x1 , x2 , x3 ..., xn Input vector u Output vector y x1 is the inductor current and x2 is the capacitor voltage Differentiation gives the state variable rate of change x1 x1 iL t diL t Predict future system state dt x2 vC t x2 dvC t dt 5 • Chris Basso – – CM Boost Study Describe the System During the On-Time Observe the system during the on-time duration or dTSW: x1 u1 L iC t x2 R C R iC t C 1 1 x2 u1 L L 1 1 x2 x1 x2 C RC 6 • Chris Basso – – CM Boost Study dt Cx2 x1 1 x2 R x2 u1 L x1 dvC t x1 0 x 1 C 2 diL t dt vC t Lx1 x2 1 L x1 1 L 0 u1 1 RC x2 0 0 u2 State coefficients Source coefficients Describe the System During the Off-Time Repeat the exercise during the off-time duration or 1-dTSW x1 iR t iC t C vL t L vL t x2 vL t Lx1 x2 Lx1 R x2 iR t x1 iC t x1 Cx2 x2 i R t R x2 R x1 Cx2 x1 0 x2 1 x2 L 1 1 x1 x2 C RC x1 0 x 1 C 2 1 L x1 0 0 u1 1 RC x2 0 0 u2 State coefficients Source coefficients 7 • Chris Basso – – CM Boost Study Make it Fit the State Equation Format Arrange expressions to make them fit the format: x Ax t Bu t State equation on-time network x1 0 x 1 C 2 1 L x1 1 L 0 u1 1 RC x2 0 0 u2 x1 0 x 1 C 2 1 L x1 0 0 u1 1 RC x2 0 0 u2 0 A1 1 C 1 L 1 L 0 B1 1 RC 0 0 off-time network 0 A2 1 C 1 L 0 0 B2 1 RC 0 0 How do we link matrixes A1 and A2, B1 and B2? We smooth the discontinuity by weighting them by D and 1-D x A1 D A 2 1 D x t B1 D B 2 1 D u t 8 • Chris Basso – – CM Boost Study The State-Space Averaging Method (SSA) We now have a continuous large-signal equation We need to linearize it via perturbations x x0 xˆ u u0 uˆ D D dˆ 0 x A1 D A 2 1 D x t B1d B 2 1 D u t u10 dˆ L x̂1 C u1 D 1 dˆ xˆ1 xˆ2 0 uˆ1 u10 L L L 1 1 xˆ2 xˆ1 xˆ2 C RC D0 x̂2 d̂ Canonical smallsignal model 9 • Chris Basso – – CM Boost Study The State-Space Averaging Method (SSA) SSA was applied to switching converters by Dr Ćuk in 1976 It is a long, painful process, manipulating numerous terms What if you add an EMI filter for instance? SW L1 u1 C1 L2 on off C2 L1 u1 y1 R C1 L2 C2 R L2 C2 R L1 u1 C1 4 state variables and you have to re-derive all equations! 10 • Chris Basso – – CM Boost Study The PWM Switch Model in Voltage Mode We know that non-linearity is brought by the switching cell a c L C u1 R p a: active c: common p: passive Why don't we linearize the cell alone? a c a c . . d PWM switch VM p Small-signal model (CCM voltage-mode) Switching cell p V. Vorpérian, "Simplified Analysis of PWM Converters using Model of PWM Switch, parts I and II" IEEE Transactions on Aerospace and Electronic Systems, Vol. 26, NO. 3, 1990 11 • Chris Basso – – CM Boost Study Replace the Switches by the Model Like in the bipolar circuit, replace the switching cell… a c L u1 C . . R p …and solve a set of linear equations! u1 12 • Chris Basso – – CM Boost Study . . L C R An Invariant Model The switching cell made of two switches is everywhere! a d p a buck a buck-boost d PWM switch VM PWM switch VM p c c boost c d Ćuk d c p c PWM switch VM a d PWM switch VM p p a PWM switch VM 13 • Chris Basso – – CM Boost Study Smoothing the Discontinuity A CRT TV displays frames at a certain rate, 50 per second 9 8 t1 1 2 3 6 7 4 5 10 11 12 The optic nerve time constant is larger than an interval A succession of discrete events is seen as continuous Low-frequency filtering Integration See "phi phenomena", www.wikipedia.org 14 • Chris Basso – – CM Boost Study Averaging Waveforms The keyword in the PWM switch is averaging v t A V peak v t Tsw t Tsw DTsw v t Tsw Tsw 1 Tsw 0 1 v t dt Tsw Tsw V peak dt V peak 1 D V peak D ' DTsw The resulting function is continuous in time 15 • Chris Basso – – CM Boost Study From Steps to a Continuous Function Some functions require mathematical abstraction: duty ratio Tsw v t t1 t2 t3 D2 t5 Discrete values of D D3 Average and continuous evolution of d(t) D4 D1 Dn t4 tn Tsw D5 f mod Fsw At the modulation frequency scale, points look contiguous Link them through a continuous-time ripple-free function d(t) 16 • Chris Basso – – CM Boost Study The Common Passive Configuration The PWM switch is a single-pole double-throw model d c a ia t ic t d' vap t vcp t p Install it in a buck converter and draw the waveforms ia t ic t d c a L d' Vin C R vap t vcp t p Vout CCM 17 • Chris Basso – – CM Boost Study The Common Passive Configuration Average the current waveforms across the PWM switch ic t ic t 0 Tsw t ia t I a DI c ic t Tsw Averaged variables t 0 DTsw ia t Tsw Ia 1 Tsw dTsw i t dt D i t a 0 c Tsw DI c CCM 18 • Chris Basso – – CM Boost Study The Common Passive Configuration Average the voltage waveforms across the PWM switch vap t vap t 0 Tsw t vcp t vcp t t 0 DTsw vcp t Tsw Vcp 1 Tsw Vcp DVap Tsw Averaged variables dTsw vcp t dt D vap t 0 Tsw DVap CCM 19 • Chris Basso – – CM Boost Study A Two-Port Representation We have a link between input and output variables DI c d a Vap p Two-port cell c Ic DVap p It can further be illustrated with current and voltage sources Ia d a Vap c Vcp DI c DVap p Ic p CCM 20 • Chris Basso – – CM Boost Study A Transformer Representation The PWM switch large-signal model is a dc "transformer"! Ia Ic a c I I a DI c Ic a . . D 1 D dc equations! Vcp Vcp DVap Vap D p It can be immediately plugged into any 2-switch converter c . . L rL p 1 a Vin D R C CCM 21 • Chris Basso – – CM Boost Study Simulate Immediately with this Model SPICE can get you the dc bias point R2 L1 100m 9.80V 100u 10.0V 7 d Vg 10 VIC 9.80V c 4 c 300mV a a V3 0.3 AC = 1 Rdum 1u V(a,p)*V(d) p p 5 14.0V Vout C1 470u V(d)*I(VIC) R1 10 …but also the ac response as it linearizes the circuit dB ° 40.0 360 Hf 20.0 0 -20.0 180 0 arg H f -180 -40.0 10 22 • Chris Basso – – CM Boost Study 100 1k Hz 10k -360 100k CCM We Want Transfer Functions Derive the dc transfer function: open caps., short inductors Vcp c I out D Ic . rL p Ic . 1 a Vin Vout R Ia Vout Vin Vin rL I c Vcp Vout Vout I out R 1 1 D Vin rL I c DVout Vout Vout I a I c R I a DI c Vout I c D 1 R Ic Vout 1 D Vin Vout 1 Vin D' rL rL D 1 R 1 r 1 L 2 RD ' CCM 23 • Chris Basso – – CM Boost Study Plotting Transfer Functions Plot the lossy boost transfer function in a snapshot 5 rL 0.1Ω R 10 Ω f ( d 0.1) f ( d 0.2) 4 rL 0.2 Ω f ( d 0.3) f ( d 0.4) Vout Vin f ( d 0.5) 3 f ( d 0.6) f ( d 0.7) f ( d 0.8) 2 f ( d 0.9) f ( d 1) 1 rL 1Ω 0 0 0.2 0.4 Duty ratio 0.6 0.8 1 d Above a certain conversion ratio, latch-up occurs CCM 24 • Chris Basso – – CM Boost Study Course Agenda The PWW Switch Concept Small-Analysis in Continuous Conduction Mode Small-Signal Response in Discontinuous Mode EMI Filter Output Impedance Cascaded Converters Operation 25 • Chris Basso – – CM Boost Study A Boost Converter in Current Mode Identify the diode and switch position in a boost CM c a duty-cycle p c vc PWM switch CM p a Current mode Replace switches by the small-signal PWM switch model 26 • Chris Basso – – CM Boost Study A Small Signal Model The model includes current sources and conductances Ic a c 1 gi vˆcp g r vˆc ki vˆap g f vˆc ko 1 go Cs p ko 1 Ri ki D Ri g f Dg o DD ' Tsw 2L I gi D g f c Vap go gr Tsw Sa 1 D' D L Sn 2 Ic go D Vap V. Vorpérian,"Analysis of Current-Controlled PWM Converters using the Model of the PWM Switch", PCIM Conference ,1990 27 • Chris Basso – – CM Boost Study Start with a Large Signal Response Use the original large-signal (nonlinear) PWMCM model R4 1.5m L1 5u Vin 4 5 2.69V c 2.69V R9 1k vc 7 5.00V X4 PWMCCMCM L = 5u Fs = 1Meg Ri = -50m Se = 0 PWM switch CM 462mV duty-cycle duty_ratio a Vin 2.7 p 2.70V 5.00V 471mV V1 471m AC = 1 Vout 2 Resr 60m Rload 1 1 Cout 200uF 5 V/1 A Fsw = 1 MHz 28 • Chris Basso – – CM Boost Study Frequency Response of the CM Boost This frequency response becomes the reference 49.0 37.0 Subharmonic poles (dB) 25.0 H 0 14.5 dB 13.0 Vout f 1.00 Vc f -36.0 (°) -108 Vout f Vc f -180 -252 -324 10 100 1k 10k frequency in hertz 100k 1Meg 29 • Chris Basso – – CM Boost Study Plug the Linearized Small-Signal Model Use the linearized model to check coefficients and response R8 1.5m 2.70V 1 parameters Fsw =1Meg Tsw =1/Fsw L=5u Cs=1/(L*(Fsw *3.14)^2) Ri=-50m Se=0 Vin=2.7 Vout=5 L2 {L} 15 c 2.69V 2.69V c VIC C2 {Cs} 2.69V 5.00V 4 R5 {1/go} Vc=471m Vac=-Vin Vap=-Vout Vcp=(-Vout+Vin) B4 Current V(v c)*{ko} Vin {Vin} These coefficients are computed by the macro Sn=(Vac/L)*Ri Sf=(Vap/L)*Ri B3 Current Ic=(Vc/Ri)-D*Tsw *Se-Vcp*(1-D)*Tsw /(2*L) p V(a,p)*{gf } D=Vcp/Vap D'=1-D ki=D/Ri gi=D*(gf-Ic/Vap) gr=(Ic/Vap)-go*D ko=1/Ri go=(Tsw /L)*(D'*Se/Sn+0.5-D) gf=D*go-(D*D')*Tsw /(2*L) B6 Current V(c,p)*{gr} 4.30uV B5 Current p Vout V(v c)*{ki} C3 200uF R4 {1/gi} a a 0V R7 1u 30 • Chris Basso – – CM Boost Study 16 vc 471mV 462mV V2 471m AC = 1 D B2 Voltage v (c,p)/v (a,p) R6 60m R3 1 Small-Signal Response and Original Model Validate the small-signal approach by comparing responses 49.0 37.0 (dB) 25.0 H 0 14.5 dB 13.0 Vout f 1.00 Vc f Good to go! -36.0 -108 (°) Vout f Vc f -180 -252 -324 10 100 1k 10k 100k 1Meg frequency in hertz 31 • Chris Basso – – CM Boost Study Rearrange Sources to Improve Readability It is important to lay the network out properly B3 Current Storage element V(0,out)*{gf } B4 Current 2.69V V(v c)*{ko} R8 1.5m 1 L2 {L} c R5 {1/go} VIC Storage element out 5.00V Vout 15 2.69V C2 {Cs} 2.70V C3 200uF 4 2.69V Vin {Vin} Storage element B6 Current V(c,out)*{gr} B5 Current V(v c)*{ki} R4 {1/gi} 471mV 0V 16 vc 462mV V2 471m AC = 1 D R6 B2 60m Voltage v (c,out)/v (0,out) Make sure the frequency response remains unchanged 32 • Chris Basso – – CM Boost Study R3 1 What is the Converter Order? Count storage elements: L and two C The denominator of the transfer function is of 3rd order Zeros Dc gain s s 1 1 s z1 s z2 H s H0 s 1 s p1 1 s s 1 0 Q 0 2 Subharmonic poles 33 • Chris Basso – – CM Boost Study Start with the Static Analysis For s = 0, short inductors and open capacitors B3 Current parameters Fsw=1Meg Tsw=1/Fsw L=5u Cs=1/(L*(Fsw*3.14)^2) Ri=-50m Se=0 Vin=2.7 Vout=5 V(0,out)*{gf} B4 Current V(vc)*{ko} R5 {1/go} Sn=(Vac/L)*Ri Sf=(Vap/L)*Ri Vc=471m Vac=-Vin Vap=-Vout Vcp=(-Vout+Vin) Vout VIC 34 • Chris Basso – – CM Boost Study B6 Current V(c,out)*{gr} V c 0 c Ic=(Vc/Ri)-D*Tsw*Se-Vcp*(1-D)*Tsw/(2*L) D=Vcp/Vap D'=1-D ki=D/Ri gi=D*(gf-Ic/Vap) gr=(Ic/Vap)-go*D ko=1/Ri go=(Tsw/L)*(D'*Se/Sn+0.5-D) gf=D*go-(D*D')*Tsw/(2*L) out 1 R8 1.5m Neglect rL D vc V2 471m AC = 1 B2 Voltage v(c,out)/v(0,out) V c 0 B5 Current V(vc)*{ki} R4 {1/gi} R3 1 Simplify and Rearrange Rearrange sources to make the circuit look simpler out Vout B3 Current V(out)*{gf} B4 Current V(vc)*{ko} B1 Current V(out)*{gr} R5 {1/go} Vout g f Vc ko Vout g o gi Vout g r Vc ki Vout R1 k 0 ki H0 g f g 0 gi g r B5 Current V(vc)*{ki} R3 1 R4 {1/gi} Mathcad® calculations ko ki 14.563 dB H0 20 log g g g g 1 f o i r RL 1 R1 35 • Chris Basso – – CM Boost Study Put Storage Elements Back in Place Rearrange sources to make the circuit look simpler out Vout B3 Current B4 Current V(out)*{gf} V(vc)*{ko} R5 {1/go} C2 {Cs} 1 C3 200uF B6 Current V(c,out)*{gr} vc V2 471m AC = 1 c L2 {L} B5 Current V(vc)*{ki} R4 {1/gi} 3 16 R3 1 R6 60m 2 Label each storage element with a time constant Time to call the FACTs! Set excitation to 0, Vc = 0 FACTS: Fast Analytical Circuit TechniqueS 36 • Chris Basso – – CM Boost Study Polynomial Form of a 3rd Function A transfer function is made of a numerator and a denominator H s zeros poles N s D s The denominator combines the circuit time constants D s 1 1 2 3 s 1 21 1 31 2 32 s 2 1 21 312 s 3 Time constant Time constant with Cs with Cout Time constant with L To calculate time constants, suppress the excitation, all is dc 37 • Chris Basso – – CM Boost Study Calculate the First Time Constant What resistance does Cs « see » ? out R? =0 B3 Current V(out)*{gf} B4 Current V(vc)*{ko} =0 R5 {1/go} tau1 Vc c V2 0 AC = 1 L shorted B6 Current V(out)*{gr} B5 Current V(vc)*{ki} tau3 C open R2 100G R3 1 R4 {1/gi} 1 R6 60m R1 1f The excitation is 0 so several sources go away You can further simplify the circuit and test it in SPICE 38 • Chris Basso – – CM Boost Study Calculate the First Time Constant Check the result with a simple dc operating point calculation out 495m V B3 Current V(out)*{gf } 0V B4 Current V(vc)*{ko} R5 {1/go} tau3 C open I1 1 B6 Current V(out)*{gr} tau1 Vc B5 Current V(vc)*{ki} R4 {1/gi} L shorted R3 1 297f V 1 R6 60m c V2 0 AC = 1 R2 100G R1 1f IT Vout g r g f out2 IT Req 495mV VT I2 1 R R7 {1/gi} B2 Current R8 {1/go} R9 1 V(out2)*({gr}-{gf }) VT Vout 1 IT IT gr g f Req 1 1 1 Req1 RL || || gi go 1 Req Vout Req 1 0.495 1 RCs g r gf 39 • Chris Basso – – CM Boost Study Calculate the Second Time Constant What resistance does L « see » ? out Vout =0 B3 Current V(out)*{gf} B4 Current V(vc)*{ko} R5 {1/go} tau1 Cs open =0 B6 Current V(c,out)*{gr} B5 Current V(vc)*{ki} tau3 C open R4 {1/gi} vc c V2 471m AC = 1 R? The excitation Vc is 0 so simplify the circuit Rearrange the network and test it in SPICE 40 • Chris Basso – – CM Boost Study R3 1 1 R6 60m Calculate the Second Time Constant out Voutx -48.8V B4 Current B3 Current V(out)*{gf} R5 {1/go} V(vc)*{ko} B6 Current V(c,out)*{gr} vc 0V B5 Current V(vc)*{ki} R3 1 R4 {1/gi} -52.9V c V2 0 AC = 1 I1 1 -52.9V B1 Current out2 -48.8V Vout2 V(out2)*{gf } c1 R2 {1/go} IT VT B8 Current V(c1)*{gr} I2 1 R7 {1/gr} R1 {1/gi} R6 1 R VT IT 41 • Chris Basso – – CM Boost Study Calculate the Second Time Constant Express IT and Vout then rearrange to unveil VT/IT IT Vout g f VT Vout 1 g0 R Vout IT VT gr Req 1 Req g f go VT IT go Req g r g f go Req 2 RL || 2 1 1 || gi g r L 1 Req g f go go Req g r g f go 1 Req2 g f g o g o Req2 g r gf g o 42 • Chris Basso – – CM Boost Study 52.897 Same result as dc point simulation Calculate the Third Time Constant What resistance does Cout « see » ? out Vout =0 B3 Current V(out)*{gf} B4 Current V(vc)*{ko} =0 tau1 Cs open R5 {1/go} B6 Current V(c,out)*{gr} vc B5 Current V(vc)*{ki} R? 1 R6 60m c V2 471m AC = 1 R3 1 R4 {1/gi} tau2 L shorted The circuit is very close to that of 1 Rearrange the network and test it in SPICE 43 • Chris Basso – – CM Boost Study Calculate the Third Time Constant Same resistance as for 1 plus rC in series out2 495mV I3 1 B2 Current V(out2)*({gr}-{gf}) R7 {1/gi} R8 {1/go} R9 1 -60.0mV 6 R2 60m 1 3 rC Cout g g 1 f r R eq 1 44 • Chris Basso – – CM Boost Study Req1 RL || 1 1 || gi g o First Coefficients a1 and a2 FACTs tell us that a1 sums up all time constants a1 1 2 3 Dimension is time For a2, we multiply combined-time constants a2 1 21 1 31 2 32 Dimension is time2 1 What is this new time constants definition, 2 ? Cs (HF) Cs (HF) 12 L (HF) 31 L C (dc) 32 L (dc) C R? C Cs (dc) R? R? V. Vorpérian, “Fast Analytical Techniques for Electrical and Electronic Circuits”, Cambridge Press, 2002 45 • Chris Basso – – CM Boost Study Calculate the Terms for a2 What resistance does L while Cs is a short and Cout is open? out Cs (HF) 12 B3 Current V(out)*{gf } L C (dc) R? B4 Current V(v c)*{ko} R5 {1/go} tau1 Cs HF R1 1u tau3 C open B6 Current V(c,out)*{gr} Vc V2 0 AC = 1 c I1 1 R? Go for an intermediate step 46 • Chris Basso – – CM Boost Study B5 Current V(v c)*{ki} R3 1 R4 {1/gi} 1 R6 60m A Simplified Drawing for a Simple Answer What R? does L see while Cs is a short and Cout is open? out2 B1 Current V(out2)*{gf } R7 {1/go} out3 tau1 Cs HF R10 1u B8 Current R11 {1/gr} R2 {1/gi} I3 1 R8 1 B7 Current V(c2)*{gr} R15 {1/gr} V(out3)*{gr} c2 I2 1 R? L 1 12 1 gr 1 1 R1 || || g r gi 47 • Chris Basso – – CM Boost Study The Mid Term is Easy to Get L and Cs ensure a complete short over RL out Cs (HF) 31 R? Cs is HF L (dc) C R? R11 {1/gr} R8 1 1 R7 60m 2 L is dc 31 rC Cout 48 • Chris Basso – – CM Boost Study R2 {1/gi} R9 {1/gi} R13 1 For the Last Term, L is open What resistance does Cout see when Cs and L are open? out L (HF) 32 B3 Current V(out)*{gf} C Cs (dc) B4 Current V(vc)*{ko} tau1 Cs dc I1 1 R? B6 Current V(c,out)*{gr} B5 Current V(vc)*{ki} R4 {1/gi} Vc R3 1 1 R6 60m c V2 0 AC = 1 R? R5 {1/go} tau2 L is HF Too complex! Go for another combination. 2 32 3 23 49 • Chris Basso – – CM Boost Study Reorder Time Constants for Simpler Sketch The schematic now looks simpler… out out2 B1 C urrent B2 Current V(out)*{gf } R2 {1/go} V(out2)*{gf } B8 Current V(c2)*{gr} R9 {1/gr} R1 {1/gi} R8 60m R3 {1/go} R7 1 B3 Current V(T)*{gr} VT I2 1 R? Cout (HF) 23 L Cs (dc) R? 50 • Chris Basso – – CM Boost Study VT I1 1 R? Req Extract the New Time Constant Definition out i1 B1 Current Req rC || RL || IT i2 R2 {1/go} V(out)*{gf} 1 1 || gi g r i2 IT VT g r 2 B8 Current IT IT i1 Vout g f Req V(c2)*{gr} I2 1 R? i1 VT Vout Req IT VT g r substitute IT VT Vout Vout g f 1 go VT Vout 1 go R 1 Req g f go g0 Req g r g f g o 23 L R 51 • Chris Basso – – CM Boost Study …And a3 is ? For a3, we multiply by a third time-constant a3 1 12 31,2 Dimension is time3 What is this new time constant definition? Cs (HF) L (HF) 31,2 out C (dc) 7.40V B3 Current V(out)*{gf} R? 0V Vc B4 Current V(vc)*{ko} R5 {1/go} tau1 Cs HF R1 1u 7.40V V2 0 AC = 1 52 • Chris Basso – – CM Boost Study c R12 1G L is in HF B6 Current V(c,out)*{gr} B5 Current V(vc)*{ki} I1 1 R? R4 {1/gi} -60.0mV 1 R6 60m R3 1 Rearrange and Simplify out 7.40V parameters Fsw =1Meg Tsw =1/Fsw L=5u Cs=1/(L*(Fsw *3.14)^2) Ri=-50m Se=0 Vin=2.7 Vout=5 B3 Current V(out)*{gf } 0V Vc B4 Current V(v c)*{ko} tau1 Cs HF R5 {1/go} R1 1u B6 Current V(c,out)*{gr} B5 Current V(v c)*{ki} c R12 1G R3 1 -60.0mV 1 R6 60m 7.40V V2 0 AC = 1 I1 1 R? R4 {1/gi} L is in HF Sn=(Vac/L)*Ri Sf=(Vap/L)*Ri Vc=471m Vac=-Vin Vap=-Vout Vcp=(-Vout+Vin) 7.40V Ic=(Vc/Ri)-D*Tsw *Se-Vcp*(1-D)*Tsw /(2*L) D=Vcp/Vap D'=1-D ki=D/Ri gi=D*(gf-Ic/Vap) gr=(Ic/Vap)-go*D ko=1/Ri go=(Tsw /L)*(D'*Se/Sn+0.5-D) gf=D*go-(D*D')*Tsw /(2*L) I2 1 R? 6 R2 {1/gi} -60.0mV R8 1 2 R9 60m 1 || rL rC Cout gi 31,2 53 • Chris Basso – – CM Boost Study Final Denominator Expression a1 1 2 3 1 L 1 a1 C rC Cout 1 s 1 1 Req 2 g f g o gr g f R gr g f R eq 1 eq 1 g o Req 2 g r g f g o a2 1 12 1 31 2 32 1 12 1 31 3 23 1 C a2 1 s gr g f R eq1 1 12 1 2 3 a3 L 1 1 1 1 R1 || || g r gi 1 1 L Cr C rC Cout 1 s C out 1 1 Req 2 g f g o gr g f R gr g f R eq1 eq1 gr g 0 Req 2 g r g f g o 1 C a3 1 s g g r f Req1 L 1 R1 || 54 • Chris Basso – – CM Boost Study 1 gr 1 1 || g r gi 1 || rL rC Cout gi Find the Zeros A zero means the excitation does not generate a response Ip Ic out Vout Ia B3 Current V(out)*{gf} B4 Current V(vc)*{ko} R5 {1/go} I1 C2 {Cs} C3 200uF c B6 Current V(c,out)*{gr} Ic vc V2 471m AC = 1 B5 Current V(vc)*{ki} R3 1 1 R4 {1/gi} R6 60m L2 {L} No response means that: Ip 0 I p I1 Short circuit Ip 0 Vout 0 s z1 rC sCout 0 1 rC Cout 55 • Chris Basso – – CM Boost Study Find the Zeros A Capitalize on the 0-V response to remove sources Ic Ip out 0 Vout Ia B3 Current V(out)*{gf} B4 Current V(vc)*{ko} R5 {1/go} C2 {Cs} C3 200uF c B6 Current V(c,out)*{gr} Ic vc V2 471m AC = 1 L2 {L} I c Vc ko v c g o v c sCs v c g r Vc ki Vc ko v c g o v c sC s 56 • Chris Basso – – CM Boost Study R3 1 R6 60m v c g r v c g r Vc ki I c Ia I p 1 R4 {1/gi} B5 Current V(vc)*{ki} v c sL Vc ko v c g o v c sCs v c Vc ko 1 g o sCs sL Find The Zeros After substituting v(c) into the equation rearrange it Vc ki Lg o ki s Lg r ko s Cs Lki s 2 2 1 sLg o s Cs L Vc ko 1 sLg o s 2 Cs L ki Lg o ki s Lg r ko s Cs Lki s 2 ko g o ki g r ko 2 Cs Lki s ki k o ki ko ki ko 1 sL 0 g k g r ko 2 Cs Lki 1 sL o i 0 s ki ko ki ko For low Q values, the 2nd order polynomial can expressed as a 1 a1 s a2 s 2 1 a1 s 1 2 s a1 57 • Chris Basso – – CM Boost Study Find the Zeros – Final Lap! The first zero is a right half-plane zero g k g r ko s 1 sL o i 1 s z1 ki ko s z1 ki ko L g o ki g r ko z1 1 D 2 RL L Negative value! The second zero is given by the capacitor ESR 1 srC Cout 0 z2 1 D 2 RL L The third zero is located at high frequency 1 s C s ki s 1 g o ki g r k o s z3 s z3 g o ki g r k o C s ki Neglect it z2 4 fz2 3.16 10 kHz 2 V. Vorpérian,“Analytical Methods in Power Electronics", 3-day course in Toulouse,2004 58 • Chris Basso – – CM Boost Study Arranging the Transfer Function The denominator has to be rearranged a little bit… D s 1 b1 s b2 s 2 b3 s3 The term 1+b1s dominates at low frequency b b 1 b1 s b2 s 2 b3 s 3 1 b1 s 1 2 s 3 s 2 b1 b1 1 b1s 1 p1 b b2 s s 3 s 2 1 b1 b1 p1 T S 2 sw 3 1 a R LM S n L Cout Qp 1 mc D ' 0.5 n Tsw Dc gain H0 1 1 s s n Q p n mc 1 2 H0 RL Ri 1 RLTsw 1 S a 2M LM 2 2 S n Sa Sn 59 • Chris Basso – – CM Boost Study Final Expression The transfer function of a CCM Boost Converter in CM is L 1 s 1 srC Cout 2 1 D RL Vout s RL 1 1 2 Cout Vc s Ri RLTsw 1 S a s s 1 s 2M 1 T S 2 LM 2 2 Sn n Q p n sw 3 1 a RL LM Sn Sa is the external ramp compensation to damp subharmonic oscillations 60 • Chris Basso – – CM Boost Study Check SPICE Versus Mathcad Full-featured transfer functions versus SPICE Vout s 100 Vc s 40 20 log H1 i 2 f k 10 arg H 1 i 2 f k 20 180 Vout s Vc s 0 100 0 10 100 3 110 110 4 110 5 6 110 10 100 3 110 fk Simplified transfer functions versus SPICE Vout s 40 100 Vc s 180 180 arg H1 i 2 f k 20 20 log H 1 i 2 f k 10 20 log H 2 i 2 f k 10 arg H2 i 2 f k 110 4 5 110 110 fk Vout s Vc s 0 100 0 10 100 110 3 110 4 110 5 110 6 10 100 3 4 110 110 5 110 fk fk 61 • Chris Basso – – CM Boost Study Course Agenda The PWW Switch Concept Small-Analysis in Continuous Conduction Mode Small-Signal Response in Discontinuous Mode EMI Filter Output Impedance Cascaded Converters Operation 62 • Chris Basso – – CM Boost Study 6 110 6 The CM Boost Converter in DCM To check the operating mode, calculate the critical load Fsw 1 MHz 2 Fsw LVout 2 V Vin 2 1 in Vout Rcrit Rcrit 74.7 Ω L 5 µH Vin 2.7 V Vout 5 V iL t I L , peak iL t I L ,valley CCM, RL < Rcrit Tsw t I L , peak DCM, RL > Rcrit t I L ,valley 0 deadtime 63 • Chris Basso – – CM Boost Study Transfer Function of DCM CM Boost The converter is a first-order converter at low frequency H s Sn p2 1 S n mcTsw Vin Ri L s 1 2Vout M 1 z1 D 2M 1 s 1 p 1 z1 1 1 M 2 Fsw D 64 • Chris Basso – – CM Boost Study 1 rC Cout z2 2 mc 1 Sa Sn s 1 z 2 s 1 p 2 Rload 2 M L p1 1 2M 1 RL Cout M 1 The Ac Response of the DCM CM Boost A RHPZ is still present in DCM with a high-frequency pole 40 Vout s 20 log H 3 i 2 f k 10 180 arg H3 i 2 f k 0 20 40 10 100 Vc s 20 Vout s Vc s 0 100 100 3 4 110 110 5 1 10 110 6 10 100 110 3 110 4 110 5 SPICE simulation versus Mathcad 65 • Chris Basso – – CM Boost Study Why a RHP Zero in DCM? The RHP Zero is present in CCM and is still there in DCM! Id(t) iL,peak D3Tsw t D1Tsw D2Tsw Tsw When D1 increases, [D1,D2] stays constant but D3 shrinks 66 • Chris Basso – – CM Boost Study 6 110 fk fk Why a RHP Zero in DCM? The triangle is simply shifted to the right by d̂1 Id(t) iL,peak D3Tsw d̂1 t D1Tsw D2Tsw Tsw The capacitor refueling time is delayed and a drop occurs 67 • Chris Basso – – CM Boost Study The Refueling Time is Shifted If D increases, the diode current is delayed by d̂1 D(t) Id(t) 2.02m 2.04m 1.85m 1.99m d̂1 2.06m 2.09m 2.11m 2.13m 2.27m 2.41m Vout(t) Vout(t) 2.02m 68 • Chris Basso – – CM Boost Study 2.04m 2.06m 2.09m 2.11m Simulation, no peak increase Course Agenda The PWW Switch Concept Small-Analysis in Continuous Conduction Mode Small-Signal Response in Discontinuous Mode EMI Filter Output Impedance Cascaded Converters Operation 69 • Chris Basso – – CM Boost Study EMI Filter Interaction A LC filter is included to prevent the dc line pollution L Vg Vin Vout C Rload Z s? Switching converter What load does the converter offer? C. Basso, “Designing Control Loops for Linear and Switching Power Supplies”, Artech House, 2012 70 • Chris Basso – – CM Boost Study A Negative Incremental Resistance Assume a 100%-efficient converter: Pout Pin I inVin I outVout In closed-loop operation, Pout is constant I in Vin Pout Vin For a constant Pout , if Vin increases, Iin drops 1 0.8 Iin A dI in Vin 0.6 dVin P d out V P in out2 dVin Vin 0.4 0.2 10 15 20 25 The incremental input resistance is negative 30 Vin V 71 • Chris Basso – – CM Boost Study A Simple LC Filter The low-pass filter is built with a L and C components: rL 1 s z1 T s L rC Vg 1 s Q 72 • Chris Basso – – CM Boost Study R R C RR L 1 0 C Q z 0Q 0 R If R is || with -R s 2 1 rC C 1 LC rL R rC R LC0 rC R C Rload L C rL rC rL R rC R L A negative resistance cancels losses: poles become imaginary A Negative Impedance Oscillator If losses are compensated, the damping factor is zero T s 1 s2 02 Q s 1 0Q 1 2 If ohmic losses are gone, the damping factor is zero, Q is infinite. Without precautions, instability can happen! Input voltage 150 rL L rC Vin 130 I 110 C I Output voltage 90.0 0 >0 70.0 Pout (constant) Vout <0 Negative resistance 3.64m 10.9m 18.2m 25.5m 32.7m 73 • Chris Basso – – CM Boost Study Filter Output Impedance What is the output impedance of an LC filter? VT s L rC IT s It follows the form Z out s R0 R rL C Ω N s D s Ω No dimension Z out s rL || Rload 74 • Chris Basso – – CM Boost Study L 1 s rL Rload L 1 s 1 srC Cout rL r Rload C rL || Rload rC s 2 LC C rL Rload Negative Resistance at Low Frequency Neg. resistance exists because of feedback (Pout = constant ) R4 2.70V 1.5m 2.70V VZin 5 L1 5u 2.70V 3 18 PWMCM2_L X1 L = 5u Fs = 1Meg Ri = -40m Se = 16.4k c L2 10GH p 2.70V 7 a D Resr 60m SW 2 R9 1k vc 536mV vout 4 5.00V duty_cycle duty-cycle Vin 2.7 L ossy PWM switch CM I1 AC = 1 -455mV -301mV 5.00V Rload 5 16 11 Cout 200uF 13 Rdson 140m D1 mbra210et3 param eters C2 {C2} Rupper=10k fc=8k pm=70 Gfc=-20 pfc=-101 C1 {C1} R7 1k 99.9mV R2 {R2} 15 14 R1 {Rupper} 2.50V R6 10k 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) 2.50V 6 1.10V 9 Verr 2.50V 1 V2 2.5 R3 10k X2 AMPSIMP VHIGH = 100 VLOW = 10m fp=1/(2*pi*R2*C2) fz=1/(2*pi*R2*C1) Sw itchMode Pow er Supplies: SPICE Simulations and Practical Designs Christophe Basso - McGraw -Hill, 2014 75 • Chris Basso – – CM Boost Study Negative Resistance at Low Frequency The resistance is truly negative up to 200 Hz 1 2 24.0 12.0 (dBΩ) 0 -12.0 1.02 Ω Zin f -24.0 144 72.0 (°) Z in f In this region Pout constant 0 -72.0 -144 Z in 180 10 76 • Chris Basso – – CM Boost Study 100 1k 10k 100k 1Meg Negative Resistance is One Part Only Negative resistance problems occur only when Z in 180 Beyond this region, instability is linked to gain degradation dBΩ 65.0 undamped 36 dBΩ 35.0 Negative argument region 5.00 damped 3.6 /100 µF -25.0 Z out s -55.0 10 100 1k 10k 100k 77 • Chris Basso – – CM Boost Study Open-Loop Gain Degradation Here the EMI filter peaks at high frequency The open-loop is severely affected and stability is at stake No damping Zout = 57.5 dB Problem! 40.0 90.0 20.0 Tf Without filter 7 30.0 gain Plot1 ph_vout2#1, ph_vout2 in degrees ZinSMPS p in 1 / ampere 180 18 dB 10.0 -10.0 1 -30.0 0 vdbout2, vdbout2#1 in db(volts) 50.0 With filter 0 -90.0 -20.0 -180 -40.0 T f phase Without filter 3 Zout 10m 100m With filter 1 10 100 frequency in hertz 1k 10k 100k Filter output impedance versus closed-loop buck input impedance. 78 • Chris Basso – – CM Boost Study 10 100 1k frequency in hertz 10k Buck open-loop gain with and without input filter. 1 4 2 100k Evaluating How Loop Gain is Modified The converter transfer function is evaluated at Zout = 0 H s Z out s Z in s vˆout vˆg vˆin 0 d̂ For Zout = 0 T s H sG s Vout s D s V in G s s 0 79 • Chris Basso – – CM Boost Study Considering the Filter Output Impedance In reality, the EMI output impedance makes Z out s 0 The converter ac input voltage is no longer zero Z out 0 vˆin 0 H s vˆout H s vˆout Z out s Z out 0 d̂ T s Vout s D s d̂ G s T s Vin s 0 80 • Chris Basso – – CM Boost Study Vout s D s Vin s 0 G s Extra Element Theorem to Help It can be shown how an EMI affects the open-loop gain: H s Vout s without the filter vˆout T s D s Z sGain 0 out Z out s Z out s d̂ T s Filter output impedance G s Vout s D s Z Z out s V s ZN s out Z s D s Z s 0 out 1 out ZD s 1 out s 0 81 • Chris Basso – – CM Boost Study What are ZD and ZN? ZD and ZN come from the Extra Element Theorem, EET Z D s Zi s D s 0 Open-loop input impedance Z N s Zi s V Input impedance for Vout s 0 out s 0 ZD s ? Null ZN s ? Vout vˆout 0 D0 35% dˆ 0 iˆout 0 Open-loop input impedance 82 • Chris Basso – – CM Boost Study Vout vˆout 0 Ideal control D0 35% dˆ 0 iˆout 0 Input impedance for Vout s 0 What are ZD and ZN for a Boost Converter? These values have already been derived sL Z N s D '2 R 1 2 D' R L LC 1 s 2 s2 2 D' R D' Z D s D '2 R 1 sRC The original loop gain (without filter) is untouched if Z out s ZN s 1 Z out s 1 ZD s Z out s Z N s 1 Z out s Z D s 83 • Chris Basso – – CM Boost Study Checking for Stability (1) You design the filter together with the converter Plot |Zout|, |ZN | and |ZD| in the same graph Check that no overlap exists. If so, damp filter 60 40 Z out i 2 f k 20 log 10 20 Z D i 2 f k 20 log 10 Z N i 2 f k 20 log ZN s ZD s 0 10 20 Z out s 40 Need to redesign the filter! 60 1 10 100 3 110 fk 84 • Chris Basso – – CM Boost Study 110 4 5 110 6 110 Checking for Stability (2) You design the filter after the converter Plot |Zout|, |Zin | in the same graph Check that no overlap exists. If so, damp filter 50.0 7 30.0 R damps Ok, margin is 8 dB min ZinSMPS 18 dB Rdamp= 5 10.0 Rdamp= 4 C cuts dc -10.0 Damping network 1 4 5 6 3 2 Rdamp= 1 Rdamp= 2 -30.0 Rdamp= 3 Zout 10m 100m 1 10 100 frequency in hertz 1k 10k 85 • Chris Basso – – CM Boost Study Course Agenda The PWW Switch Concept Small-Analysis in Continuous Conduction Mode Small-Signal Response in Discontinuous Mode EMI Filter Output Impedance Cascaded Converters Operation 86 • Chris Basso – – CM Boost Study 100k Cascading Converters When cascading converters, impedances also matter Vin s Z th s Z in s Vth s Z out s Z in s Boost Buck The system can be modeled by a gain TM Vin s Vth s 1 Z th s 1 Z in s Vth s + Vin s - Z th Z in TM s 87 • Chris Basso – – CM Boost Study Check the Stability of the Minor Loop (1) The stability test requires the comparison of Zout and Zin The buck converter closed-loop input impedance is 1 Z in s T s 1 1 ZN s 1 T s ZD s 1 T s 1 Low frequency T s 1 Z in s Z N s T s H sG s ZN and ZD are defined as Z D s Z i s D s 0 Open-loop input impedance Z N s Zi s V Input impedance for Vout s 0 out R ZD s 2 D s 0 L 2 s LC R 1 sRC 1 s 88 • Chris Basso – – CM Boost Study ZN s R D2 Use SPICE Models First Simulate the boost converter output impedance L1 33u R5 7m 12 Closed-loop Zout p c 11 2 X3 PWMCM L = 33u Fs = 200k Ri = -300m Se = 82k vc Vin 11 vout vout PWM switch CM 3 a duty-cycle duty_cycle R10 18m Z out f 0 Rload 8.3 16 I1 AC = 1A C5 820u -20.0 C3 282p -40.0 C2 10n R2 50k R1 11.66k 9 R4 1m -60.0 1 5 8 6 R3 1.35k Verr V2 2.5 -80.0 1 X2 AMPSIMP VHIGH = 10 VLOW = 10m (dBΩ) 10 100 1k 10k 100k 1Meg 11-15 V/24 V – 3 A 89 • Chris Basso – – CM Boost Study Use SPICE Models First Simulate the buck converter input impedance Zi n duty_cycl e Closed-loop Zin 210mV L2 1kH 5 24.0V a duty-c yc le c 6 24.0V 2.37V 1 I1 AC = 1 3 R4 4m vout 5.00V PWM switch CM Vout 4 vc 7 V4 24 L1 8u5.05V 5.05V 19.8 Re sr 14m p PWM CM X2 L = 8u Fs = 2 00k Ri = 1 76m Se = 30k 5.00V R2 416m 2 Cout 820u 19.4 C3 1n vout 19.0 C1 3.5 n R6 15k 2.37V R1 1k 8 R7 1m Z in f 2.50V 18.6 9 2.37V 11 2.50V 10 Verr V2 2.5 R3 1k 18.2 10 24 V/5 V – 12 A 90 • Chris Basso – – CM Boost Study (dBΩ) 100 1k 10k 100k 1Meg Compare Magnitude Curves 39.0 19.0 Z in f (dBΩ) No overlap -1.00 -21.0 Z out f -41.0 10 100 1k 10k 100k 1Meg 91 • Chris Basso – – CM Boost Study Compare Phase Curves 270 Neg. up to 1 kHz 180 Z in f (°) 90.0 0 Z out f -90.0 1 92 • Chris Basso – – CM Boost Study 10 100 1k 10k 100k 1Meg Check Minor Loop Gain There is no gain, Z out Z in . The system is stable dB ° -40.0 0 -50.0 -90.0 -60.0 -180 -70.0 -270 Z out f Z out f Z in f Zin f -80.0 -360 1 10 100 1k 10k 100k 1Meg 93 • Chris Basso – – CM Boost Study With Cascaded Converters L1 33u 11.0V R5 7m 12 dc_buck 10 c a p 2 X3 PWMCM L = 33u Fs = 200k Ri = -300m Se = 82k vc 545mV a duty-cycle dc_boost 24.1V PWM switch CM 3 4 duty-cycle 5.05V c R4 4m L2 8u5.05V 11 11.0V Vin 11 Iout 209mV VoBoost 11.0V 14 VoBuck 19 2.37V R10 18m 24.1V 18 vc PWM switch CM Resr 14m p PWMCM X1 L = 8u Fs = 200k Ri = 176m Se = 30k 16 C5 820u 5.00V C6 1n v out Boost C2 10n R2 50k 2.50V R7 1m 1 2.02V 2.50V 21 2.37V 23 2.50V 2.50V 6 X2 AMPSIMP VHIGH = 10 VLOW = 10m V2 2.5 R9 1k 24 2.50V 7 C1 3.5n R6 15k2.37V R1 11.66k 9 94 • Chris Basso – – CM Boost Study I1 unknown 13 Cout 820u C3 282p VerrBoost v out5.00V 15 R3 1.35k Buck VerrBuck 22 V3 2.5 R11 1k Check Transient Response Transient response is good for the buck 5.10 Buck 5.05 (V) 5.00 4.95 Iout from 5 to 12 A in 1 µs 4.90 vout t 24.10 24.05 (V) 24.00 23.95 Boost 23.90 100u 300u 500u vout t 700u 900u 95 • Chris Basso – – CM Boost Study Literature “Fundamentals of Power Electronics”, R. Erickson, D. Maksimovic “Designing Control Loops for Linear and Switching Power Supplies”, C. Basso “Practical Issues of Input/Output Impedance Measurements”, Y. Panov, M. Jovanović, IEEE Transactions on Power Electronics, 2005 “Physical Origins of Input Filter Oscillations in Current Programmed Converters” , Y. Jang, R. Erickson, IEEE Transactions on Power Electronics, 1992 “Design Consideration for a Distributed Power System”, S. Schultz, B. Cho, F. Lee, Power Electronics Specialists Conference, June 1990, pp. 611-617 “Input Filter Considerations in Design and Applications of Switching Regulators”, R. D. Middlebrook, IAS 1976 96 • Chris Basso – – CM Boost Study Conclusion The PWM switch model is an essential tool for modeling It can be used to derive small-signal response of a boost in CM Its SPICE implementation helps test various transfer functions It can be used to assess open- and closed-loop parameters Interactions with the EMI input filter can be analyzed Cascaded converters stability can be checked with the model Practical measurements on the bench must always be performed Merci ! Thank you! Xiè-xie! 97 • Chris Basso – – CM Boost Study