what happens if R is omitted? iA D1 D3 vA C/2 iIA D5 + IOU T The Discontinuous Conduction Mode L D2 iR1 D4 iR2 D6 iR3 iB iIB C/2 vB iX iX iX i1 i2 + + v1 what really happens? I in the case VA, 1 and VB, 1 remain the same as in the CCM, the amplitude of iY won’t be limited I something, though, limits the amplitude; try and see . . . let’s look at the circuit and search for the answers: I i3 + v2 v3 v1 , i1 , IOU T ≈ 3 A 1. what limits the amplitude of iY ? 2. is it safe to operate in this mode? 3. is there any use of this operating mode? v1 , i1 , IOU T ≈ 3 A, spectra v1 , IOU T ≈ 3 A v2 , i2 , IOU T ≈ 3 A v2 , i2 , IOU T ≈ 3 A, spectra vN iY vOU T − v2 , IOU T ≈ 3 A v3 , i3 , IOU T ≈ 3 A v3 , i3 , IOU T ≈ 3 A, spectra v3 , IOU T ≈ 3 A v1 , i1 , IOU T ≈ 6 A v1 , i1 , IOU T ≈ 6 A, spectra v1 , IOU T ≈ 6 A v2 , i2 , IOU T ≈ 6 A v2 , i2 , IOU T ≈ 6 A, spectra v2 , IOU T ≈ 6 A v3 , i3 , IOU T ≈ 6 A v3 , i3 , IOU T ≈ 6 A, spectra v3 , IOU T ≈ 6 A v1 , i1 , IOU T ≈ 9 A v1 , i1 , IOU T ≈ 9 A, spectra v1 , IOU T ≈ 9 A v2 , i2 , IOU T ≈ 9 A v2 , i2 , IOU T ≈ 9 A, spectra v2 , IOU T ≈ 9 A v3 , i3 , IOU T ≈ 9 A v3 , i3 , IOU T ≈ 9 A, spectra v3 , IOU T ≈ 9 A vOU T , iOU T , IOU T ≈ 3 A vOU T , iOU T , IOU T ≈ 3 A, spectra vOU T , iOU T , IOU T ≈ 6 A vOU T , iOU T , IOU T ≈ 6 A, spectra vOU T , iOU T , IOU T ≈ 9 A vOU T , iOU T , IOU T ≈ 9 A, spectra does it worth? does it worth? IOU T [A] ≈ 3A ≈ 6A ≈ 9A k 1 2 3 1 2 3 1 2 3 Ik RM S [A] 2.90 2.91 2.88 5.66 5.66 5.65 8.38 8.48 8.42 Vk RM S [V] 99.78 99.47 100.12 95.89 95.75 97.01 91.67 92.14 94.72 S [VA] 289.66 289.58 288.67 542.29 541.53 548.57 767.90 781.43 797.43 P [W] 282.36 282.42 281.42 533.10 532.38 539.83 752.36 767.26 785.45 does it worth? IOU T [A] 3.08 6.06 9.41 IOU T [A] ≈ 3A ≈ 6A ≈ 9A k 1 2 3 1 2 3 1 2 3 PF 0.9748 0.9753 0.9749 0.9830 0.9831 0.9841 0.9798 0.9819 0.9850 T HD(ik ) [%] 17.76 17.86 17.74 14.69 14.19 14.19 11.98 11.53 11.69 T HD(vk ) [%] 4.22 3.91 4.14 3.76 4.21 3.80 5.47 5.47 4.32 user’s point of view . . . and a conclusion VOU T [V] 253.32 232.91 216.86 POU T [W] 781.38 1410.35 2041.09 PIN [W] 846.21 1605.31 2305.07 η [%] 92.34 87.86 88.55 really low η; quite unexpected! is this a mistake? I the input currents are not so good, but better than without injection I T HD is in the range from 10 % to 20 % I absolutely no notches in the input voltages I spikes in the output voltage, rich with harmonics I the spikes decrease with increases of the output current I the spikes increase the output voltage average, VOU T I there are some losses in the system, unexpected? I the efficiency is much lower than expected! I at 9 A the rectifier operates close to the CCM I spikes are the answer! what causes them? I let’s take a closer look . . . vA , iA , IOU T ≈ 3 A vB , iB , IOU T ≈ 3 A vA , iA , IOU T ≈ 6 A vB , iB , IOU T ≈ 6 A vA , iA , IOU T ≈ 9 A vB , iB , IOU T ≈ 9 A let’s start from the end . . . efficiency at the boundary . . . Q: what the end is? A: CCM-DCM boundary, close to IOU T = 9 A in our case at that point: PIN J iY = 2 IOU T cos (3ω0 t) normalize, . . . jY = JY m cos (3ω0 t) = 2 cos (3ω0 t) 1 JY m MA, 1 2 √ √ 1 3 3 3 3 = ×2× = 2 8π 8π √ 3 3 POU T = π PIN J = η= POU T 8 = ≈ 88.89 % POU T + PIN J 9 JY m = 2 instead of 32 , which would be optimal experimental results make sense now? j1 at the boundary (and not just there) . . . at the boundary . . . to get to the boundary √ √ MA, 1 3 3 1 3 3 ρ= = × = JY m 8π 2 16π and at the boundary . . . wxMaxima will be heavily needed from here . . . √ 7 3 √ 9 3 J1m = 4π √ 9 3 J1RM S = √ 4π 2 JRM S = T HD and the P F at the boundary and what if Q = 0? r 224π 2 − 1 ≈ 10.43 % 2187 √ 27 3 P F = √ ≈ 0.9946 4π 14 T HD = not that bad . . . but, the efficiency is bad! parameters for Q = 0? η, T HD, and P F at the boundary, Q = 0 this boundary requires a different value of ρ . . . ρ= η= max (mAV ) 1 1 1 = × = max (jY ) 4 2 8 √ POU T MOU T 6 3 √ ≈ 90.53 % = 3 = PIN 2π + 3 3 2 J1m s at the boundary . . . T HD = s 10 2 −√ 9 3π √ 3 2 J1m = + π 3 √ √ 3 2 J1RM S = √ + 3 π 2 JRM S = partial conclusions . . . I I I there is some interest in the DCM, due to the relatively acceptable THD s PF = √ 16π 2 − 24π 3 − 27 √ ≈ 4.93 % 4π 2 + 12π 3 + 27 √ 4π 2 + 12π 3 + 27 √ ≈ 0.9988 20π 2 − 12π 3 T HD and P F okay, but η is not . . . better T HD and P F are achieved in CCM with better η, this mode does not make any sense in practice . . . case Q → ∞, R → 0 p L/C I Q = R0 /R, R0 = I let’s analyze large R0 and low R case . . . not just huge Q I in that case jY ≈ 2 cos (3ω0 t) and the spikes limit the amplitude of jY . . . the boundary between the CCM and the DCM suffers from poor efficiency, not of practical interest I I . . . since there is no other cause we should analyze the DCM somewhere away from the boundary . . . I let’s model the spikes somehow . . . I maybe with Dirac impulses? I that’s why I stressed the “flat” spectrum I let’s take a look . . . vA and vB , IOU T ≈ 3 A vA and vB , IOU T ≈ 3 A vA and vB , IOU T ≈ 6 A vA and vB , IOU T ≈ 6 A vA and vB , IOU T ≈ 9 A vA and vB , IOU T ≈ 9 A definitely, the spikes . . . definitions of mA0 and mB0 , no spikes definitely, the spikes are the answer for the DCM: I the spikes reduce the 3rd harmonic (at 3ω0 ) in mA and mB I the spikes depend on the output current I the spikes are disappearing as we are getting close to the CCM I the spikes introduce new harmonics, needed to distort iY I the spikes have flat-looking spectrum . . . I . . . which makes them suitable to model with Dirac δ impulses! the same as mA and mB before . . . I and I personally like δ impulses and Dirac’s approach . . . thus, they have the same spectra, . . . I P. A. M. Dirac: “The aim of science is to make difficult things understandable in a simpler way; the aim of poetry is to state simple things in an incomprehensible way.” mA0 = max (m1 , m2 , m3 ) mB0 = min (m1 , m2 , m3 ) let’s model the spikes . . . mA = mA0 + MX the model of spikes, DSP version . . . +∞ X π 2π δ ω0 t − − n 3 3 n=−∞ mB = mB0 − MX +∞ X 2π δ ω0 t − n 3 n=−∞ where VX is yet to be determined . . . a word about physical dimension of δ impulses . . . the spectra, mA . . . a DSP note . . . I when it comes to δ impulses, the DSP approach becomes extremely error prone I be VERY careful in forming the spectra, a tiny error could destroy the results I this is not a story, this is an experience . . . I much worse than Gibbs phenomenon . . . I the higher the level of discontinuity — the worse I with δ impulses spectral leakage is a problem, since the is a lot to leak mA = max (m1 , m2 , m3 ) + MX mA = MA, 0 + π 2π δ ω0 t − − n 3 3 n=−∞ MA, k cos (3kω0 t) k=1 MA, 0 MA, k = the spectra, mB . . . ∞ X +∞ X √ 3 3 3 = + MX 2π 2π √ 3 3 (−1)k+1 3 + (−1)k MX π 9k 2 − 1 π since we are already here, mOU T , the spectrum . . . mB = min (m1 , m2 , m3 ) − MX mB = MB, 0 + ∞ X mOU T = mA − mB +∞ X 2π δ ω0 t − n 3 n=−∞ mOU T = MOU T, 0 + MB, k cos (3kω0 t) MOU T, k cos (3kω0 t) k=1 k=1 √ 3 3 3 MB, 0 = − − MX 2π 2π √ 3 3 1 3 MB, k = − MX π 9k 2 − 1 π let’s use the spectra . . . ∞ X MOU T, 0 = MOU T, k = √ 3 3 3 + MX π π 0 √ 6 3 1 6 − + MX π 9k 2 − 1 π for k odd for k even a double check . . . to push jY . . . ρ JY m = MA, 1 = MB, 1 = √ 3 3 3 − MX 8π π at ρ = 0: MOU T = and we get MX . . . MX √ √ 3 π 3 2π = − ρ JY m = − ρ 8 3 8 3 at ρ = √ 3 3 16π : MOU T = to get MOU T . . . MOU T = √ 27 3 − 2ρ 8π √ 27 3 ≈ 1.8607 8π √ √ √ 27 3 3 3 3 3 − ×2= ≈ 1.6540 8π 16π π MOU T variation within +12.5 % . . . dependence of VOU T on IOU T MOU T a word about efficiency . . . √ 27 3 = − 2ρ 8π √ 27 3 − 2ρ 8π √ 1 27 3 + × ρ × 22 = 2 8π POU T = MOU T = denormalize . . . √ VOU T 27 3 R IOU T = −2 Vm 8π Vm √ 27 3 VOU T = Vm − 2 R IOU T 8π PIN = POU T η =1− 16π √ ρ 27 3 η passes the double check at the CCM-DCM boundary . . . the output impedance is 2 R, . . . δ conclusions . . . simplify the circuit . . . I the spikes are modelled by δ impulses I prediction of the output voltage obtained, VOU T . . . I prediction of the output impedance, 2 R . . . I originally, there are 6 diodes I prediction of the efficiency obtained . . . I deep theory says there are 26 = 64 states I however, this is just an approximation . . . I fortunately, we do not care about all of them . . . I . . . although experimentally verifiable I in the CCM only 6 states occur I how good the approximation is? I in the DCM there are more than 6 . . . I is there a better model? I first, let’s reduce the problem as much as we can . . . I avoid exact solution, it is not available in a closed form . . . I but not more than that . . . I . . . like it was in the CCM I how about “simulation”? I but not just in a form of a cheap experiment . . . an equivalent circuit to study the DCM vA DA + vA0 vCA − + − some voltages defined . . . iY /2 as defined earlier . . . just rename . . . C/2 vA0 = max (v1 , v2 , v3 ) + vT iY L vB0 = min (v1 , v2 , v3 ) − IOU T + vT vB0 − + − DB vCB + − C/2 and another voltage waveform which would be needed . . . vAV 0 = vA0 + vB0 2 iY /2 vB about DA and DB I DA and vA0 represent v1 , v2 , v3 , and D1, D3, D5 I DB and vB0 represent v1 , v2 , v3 , and D2, D4, D6 I DA and DB model the DCM I from 6 diodes to 2 I from 26 = 64 states to 22 = 4 I and out of these four, one is irrelevant . . . I an improvement . . . helps us understand . . . I only 3 states to take care of! I but that’s not that only . . . how to get the currents? in the same way as before: i1 = d1 iA − d2 iB − 1 iY 3 i2 = d3 iA − d4 iB − 1 iY 3 i3 = d5 iA − d6 iB − 1 iY 3 where dn functions are as defined earlier, n ∈ {1, . . . 6} iY is really important . . . reduced number of states . . . iA = IOU T + 1 iY 2 iB = IOU T − 1 iY 2 ... i1 = (d1 − d2 ) IOU T + 3 (d1 + d2 ) − 2 iY 6 I instead of 26 = 64 we deal with 4 states now I states, I mean diode state combinations I and it is not 4, but 3 for IOU T > 0 I which is why the equivalent circuit has been introduced . . . I we need iY . . . I vA and vB would also be useful . . . I let’s find iY , vA , and vB in each of the states . . . 3 (d3 + d4 ) − 2 iY 6 3 (d5 + d6 ) − 2 + iY 6 i2 = (d3 − d4 ) IOU T + i3 = (d5 − d6 ) IOU T back to the circuit . . . and some circuit theory . . . vA DA + vA0 vCA − + − iY /2 C/2 + vT L iY − IOU T + I at most, for DA and DB on, the circuit is of the third order I when a diode goes off, algebraic degeneration over iL = iY occurs, since iY = ±2 IOU T in such a case . . . I . . . but that’s not a big problem; the capacitors are . . . vT vB0 − + DB − vCB + − C/2 iY /2 vB and some circuit theory, capacitors . . . I the capacitors share the same current, iY /2 I thus, their voltages differ for a constant . . . I the circuit (with DA and DB on) has three poles, one of them in s = 0 I the capacitors have the same AC components in their voltages I but the DC components are different! I it is not a big deal to find the DC components . . . well . . . I especially since VCA = −VCB due to the symmetry . . . I I but to find the AC component is a problem . . . after some dirty job . . . let’s introduce vC = vCA + vCB 2 which turns out to be the AC component of the voltages across CA and CB , since the DC components are the opposite . . . it can be shown that the resistance distribution parameter does not have any influence . . . the circuit exposes algebraic degeneration if a diode is off, as already stated . . . which we are going to solve! a lot of effort to solve in a sort of elegant way . . . state 0, DA is on, DB is on state −1, DA is off, DB is on equations: L diY = −R iY − vC + vAV 0 dt iY = −2 IOU T C dvC = iY dt C conditions: dvC = −2 IOU T dt condition: −2 IOU T ≤ iY if violated switch to state −1 iY ≤ 2 IOU T if violated switch to state +1 vC > vAV 0 + 2 R IOU T if violated switch to state 0 state +1, DA is on, DB is off vA and vB iY = 2 IOU T C dvC = 2 IOU T dt state 0 −1 +1 DA on off on vA vA0 vA = vA DCM vA0 DB on on off vB vB0 vB0 vB = vB DCM vA DCM = −vB0 − 4 R IOU T + 2 vC condition: vC < vAV 0 − 2 R IOU T vB DCM = −vA0 + 4 R IOU T + 2 vC if violated switch to state 0 new normalization, motivation new normalization I existing normalization of currents, with Ibase = IOU T is inadequate . . . I the problem is in the dependence of ρ on IOU T , ρ = R IOU T /Vm I R remains constant, while IOU T varies I it is inconvenient to consider variations of IOU T as variations of ρ, but not that big of a deal . . . I I we need a solid foundation for Ibase p besides, R0 , L/C plays a significant role now . . . I it’s time to renew normalization . . . the resonance parameter, r Vbase = Vm r Rbase , R0 = Ibase = JOU T = L C Vm R0 IOU T R0 IOU T = Ibase Vm ρ= R 1 = R0 Q state 0, DA is on, DB is on, normalized equations: ωR , √ r, djY = 3 r (−ρ jY − mC + mAV 0 ) dϕ 1 LC dmC = 3 r jY dϕ ωR 3 ω0 in resonance r = 1, the CIN should be designed to meet this, this is the resonance constraint state −1, DA is on, DB is off, normalized conditions: −2 JOU T ≤ jY if violated switch to state −1 jY ≤ 2 JOU T if violated switch to state +1 state +1, DA is off, DB is on, normalized jY = −2 JOU T jY = 2 JOU T dmC = −6 r JOU T dϕ dmC = 6 r JOU T dϕ condition: mC > mAV 0 + 2 ρ JOU T condition: if violated switch to state 0 mC < mAV 0 − 2 ρ JOU T if violated switch to state 0 mA and mB state 0 −1 +1 simulation is an easy task? DA on off on DB on on off mA mA0 mA = mA DCM mA0 mB mB0 mB0 mB = mB DCM mA DCM = −mB0 − 4 ρ JOU T + 2 mC mB DCM = −mA0 + 4 ρ JOU T + 2 mC the steady state acceleration method published in . . . I all the equations derived . . . I just to solve them . . . I equations piecewise-linear, nonhomogeneous . . . I trapezoidal rule to integrate . . . I simple discretization . . . I but the steady state is required! I which is a problem of itself! I and remains to be a problem . . . I a new steady state acceleration method had to be derived to solve the model in a reasonable time . . . I the original intention was to include the method in this presentation . . . T HD versus JOU T , the simulation result Marija Stojsavljević, Predrag Pejović “An Extrapolation Method for Accelerated Convergence to Steady State Solution of Power Electronics Circuits” Power Conversion and Intelligent Motion, PCIM Europe 2005, pp. 574–578, Nuremberg, Germany, June 2005 MOU T versus JOU T , the simulation result η versus JOU T , the simulation result MOU T max versus JOU T , the simulation result T HD versus r, the simulation result after the simulation . . . I simulation may be used even to draw fairly general conclusions I but this requires analytical preparation and normalization I obtained diagrams should be denormalized to apply for a specific circuit I agreement with δ impulse approach? I disagreement only at low IOU T I which was expected . . . after we got the results after the DCM . . . I finally, there is some understanding of the DCM . . . I but is there any use of it? I at first, it seems pretty useless . . . I but do we study only the things to be applied at the very moment? I actually, we do! I but there is resistance emulation . . . I where these concepts turned out to be useful I although this was not an original idea I and this is our next topic . . . published in . . . Predrag Božović, Predrag Pejović “Current Injection Based Low Harmonic Three Phase Diode Bridge Rectifier Operating in Discontinuous Conduction Mode” IEE Proceedings Electric Power Applications, vol. 152, no. 2, pp. 199–208, March 2005 without any problem!