Soft switching forward-flyback converter with one magnetic component and current doubler rectification circuit N. Frohleke, P. Wallmeier, J. Richter, H. Grotstollen Institute f o r Power Electronics and Electrical Drives Department of Electrical Engineering, University of Paderborn Warburger Strasse 100,33098 Paderborn, G e m n y Tel.: 49-5;?51-603157,Fax: 49-5251-603157, E-mail: nfroel @pblea.uni-paderborn.de Abstract - An active clamped forward-flyback converter comprizing a current doubler scheme, which allows to boost the application range of MOSFETs for efficient rectification in low-voltage power supplies, is investigated first approximatively and second incl. commutations. Design equations and issues are given for the magnetical integration of filter inductors on the outer core legs of the transformer . The circuit allows soft switching in the relevant operation range, while keeping the reactive power low. Thus, the resulting efficiency of nearly 90% using Schottky diodes is increased compared to hard switching or resonant topologies. involved analysis when applied in active clamped forwardflyback converters. But previous studies on the forward converter with conventional rectification like [3], the thorough evaluation of SR efficiency improvements given in [4] and the integrated magnetics in [SI are a valueable base for this study. I. INTRODUCTION Since a number of years several efforts have been made to decrease the power consumption of data processing electronic circuits, despite their ever iincreasing clock frequency by reducing the supply voltage below 5 V. Subsequently lowvoltage power supplies with reasonable effciency are required more and more due to increased availability of low-voltage logic IC’s. I t ‘Q2 iQI Fig. 1 Circuit diagram of the active-clamped forward-flyback converter. Since the resonant transition switching technique [1][2] with ZVS, being applied when using MOSFETs at elevated 11. ANALYSIS OF THE CIRCUIT switching frequencies, and ZCS or ZVS, when IGBTs are applied at elevated power levels - proved as appropriate means A. Operation principle and simplified analysis to raise conversion efficiency and to reduce EMI, questions Using the following simplifying assumptions yields an arise concerning the rectification components, which contribute to more than half of the total losses and its drive analytical description of the circuit: Switching of diodes and scheme. The latter depends liirgely on the power electronic MOSFETs is assumed to be ideal . Parasitics like stray induccircuit, when self-driven synchronous rectifiers (SR’s) are tances and component capacitances are ignored. Then the applied [3]. As SR’s are only available with on-resistances in converter’s operation can be described by two intervals, based the range of more than 6... 10 n d 2 at 25°C so far, yielding less on the explanation in [6]. losses than Schottky diodes with a blocking voltage of 30 V up In an energizing interval the main switch Q1 and the to currents of about 40 A only, it seems logical to conduct respective Schottky diode or transistor Qq are conductive. comparative investigations on current doubler rectification About half of the energy is transferred directly towards the schemes, when higher output icurrents are required. However, secondary, magnetizing Lfl. The other 50% yield an the treated rectification scheme (see Fig. 1) causes a quite increasing magnetization of the transformer and will supply 0-7803-3544-9196$5.00 0 1996 IEEE 1161 - the secondary circuit during the second interval, when auxiliary switch Q2 and the corresponding Schottky diode, resp. transistor Q3 are active. Clamp capacitor Ccl is approximately charged to a constant voltage U,, = U,/ (1 - D ) ,used for the demagnetization of the transformer. At steady state the auxiliary switch current iQ2(t) does not carry any mean value, guaranteeing a constant ucl(t) at the beginning of this interval. 'prim ~ Q 4 = - =Uu c , - u 1 = 1 U, z(--ul) U 1-D (4) = 1u2 -D A loss representing resistor Rv is assumed to approximate conduction losses of diodes and transistors: R , = RQ1,2/ii2 + RQ3,4 . The threshold voltage of Schottky diodes U , is considered as well. With ii as turns ratio and R, as the load resistance the dc-voltage ratio results as: UlJ2 __ U, R, -.D--.R,+R, R, UU, R,+R, U, 0' I I I I I 0.2 0.4 0.6 0.8 1 D Fig. 3 Blocking voltage of rectifiers and secondary voltage versus duty cycle, U2 = 5.83V The following equation gives an optimum turns ratio by specification of the input ( Ulmin, U l m a x )and output voltage range (ulmin, uzmaX). Optimum means, that the rectifiers are charged by the same maximum blocking voltage: 'Q3max - 'Q4ma.x. Ulmin'lmax 0 0.2 0.4 0.6 0.8 I Ulmin+ Ulmnx 1 1-uZmax I (5) 3. Analysis with non-ideal devices and commutations Load current = 60 A, RL = 83 mQ For the approximate analysis of the active clamped forwardflyback converter with magnetically integrated current doubler rectification the following assumptions are set: ideal switching of transistors and diodes and short commutations. On the primary side parasitic capacitances of the power MOSFETs The choose of turns ratio is strongly related to the blocking (Cess) are taken into account in order to approximate the voltage the diodes have to withstand: A higher blocking diode switch-on and -off losses, since zero-voltage switching (ZVS) causes higher on-state losses. And since the rectifiers on the does not occur through the entire working range. The subsesecondary side for output voltages between 2,l V - 5 V cause quent listed network topologies/intervals occur. about 50% - 70% of the total losses, the blocking capability Interval 1 (Energizing): The energizing interval is initiated should be minimized. by the termination of the commutation of D3 -> D4 (see The potential drops on the power semjconductors are interval 8). Main switch Q1 is already switched on, while considered by a modified output voltage U 2 . Eqn. (1). is MOSFET Qq is conductive, too. Since Q2 is blocking and rewritten then as: voltage uCl(t) equals uQZ(t) from the previous interval, the remaining network is not affected by the capacitances. - R,+R, - DU, Fig. 2 Output-input voltage ratio versus duty cycle for varius levels of approximation: a)ideal, b) R , = 16mR, c) R , = lOmQ, U , = 0.23V U2 = -U 2 + U F * RL U2 = U The blocking voltage across e3is inspected only in the first interval, while as Q4 conducts in the first intervall, the maximum blocking voltage occurs in the 2nd interval: Fig. 4 1162 Equivalent circuit for Interval 1, to 5 t 5 tl Interval 3 (Zero-voltage switch-on of auxiliary switch Q2): A positive current iQ2(t) is discharging CossQ2, until the (6) inverse diode DQZ of the auxiliary transistor becomes conductive and terminates this interval. Q , is switched on actively with a short delay by the gate drive. As soon as the transformer voltage uprimreaches zero the freewheeling of the secondary diodes D, and D, is initiated. The resulting equiv( 8 ) alent cirsuit and summarized results are given as follows:. Application of Kirchhoffs laws and a notation with e.g. Index ltO describing quantity 1 at instant to yields: UQl(t] = ov N1 = L OS L + LmL p +L,L,,,/u ..2 +LpLop+LaPLas Interval 2 (Commutation from Q1 -> Q2): When the main transistor is switched off by the controller the drain - source capacitance becomes effective. The voltage across the main switch evolves sinusoidally, resulting in a drop of the transformer voltage. Consequently MOSFET Q4 blocks at once (according to assumptions stated above) and the current commutates to the inverse diode: resp. to a parallel connected Fig. 6 diode D,. Simultaneously D3 begins to conduct, initiating the commutation on the secondary side ,whereby the 3. interval is started. UCI Equivalent circuit for Interval 3, t2 I t I t3 u Q l ( t ) = ( A 3 e s i n ( o O 3 . ( t - t , ) ) +K3)U (11) A3 i, (t) = uz3 cos (WO,. ( t - t , ) ) Fig. 5 Equivalent circuit for I[nterval2, tl I t I t2 With il(t) dividing according to the capacitance ratio the three mesh and one node equations together with initial conditions yield the following general solutions: uQl (t) = (Jm. i , (t) = (AA: -. +.: sin (oo2.( t - tl) + q,) + K,) ii cos ( W,),. ( t - t,) + q2))/U z2 The coefficients are calculated as follows: z3 = /(:), A, = Z 3 . i i I l t 2 , K3 = U1 U Interval 4 (Commutation on the secondary side): The trans(9) former voltage (uprim)grows according to proportional to the turns ratio uprim= U . U,/ (1 - D) . The commutation on the secondary side is terminated as (10) current is(t) attains the magnitude of filter current iLn(t). Then the total current iLfl(t)+ iLf2(t)is conducted by diode D3. Fig.7 depicts the equiv. circuit followed by calculative results. 1 Z2 = @, ..2 Lm/u K2 = --U1 p 2 9 Lfl +Los + Lm/U U UCl Fig. 7 A, = Z,.UIlt,, B , = -K2 with A, < 0 we obtain: p ', with A , 2 0 we obtain: 'p2 = = + atan ~t E) uQl atani 1163 Equivalent circuit for Interval 4,t3 I t I t4 (t) = (Jm. s i n ( o W +( t - t , ) + q 4 )+K4)ii (13) oo4= 1 - Z4 = Jc4L4’ ForA,<O: (p, = Interval 6 (Commutation of Q2 -> Ql): This interval is introduced by the switch-off of the auxiliary switch Q 2 . The main switchs current and voltage are depicted as follows: &, 7c+atan A, = Z,. i i I l f 3 , B, = i ‘Clt3 i- 1‘ .. 9 ;forA420: (p4 = atan Fig. 9 Equivalent circuit for Interval 6, t5 5 t t6 Interval 5 (Transformer demagnetisation): The clampingcapacitor voltage uCl(t) subtracted by U1 is applied across the transformer. The capacitor voltage adjusts in a way reflecting the magnetizing current iLm(t)at steady state. This is indicated also by the original term ,,current mirror demagnetization“ given by Vinciarelli. Neglection of CossQl yields a resonant circuit, consisting of C a and the resulting inductance of L, and filter inducance Lf2. Initially, current il(t) is positive, charging the clamp capacitor until the current decays to zero. At this instant the highest voltage occurs across Q1. This is the latest instant to switch-on the auxiliary switch. The input current reaches negative values, whereby the transistor voltage is lowered in order to reduce charge-down losses of the main transistor. Fig.8 depicts the effective circuit state, followed by the resulting expressions for uQl ( t ) and i, ( t ) . As soon as the negative input current il(t) affects the transformer voltage to be zero, a secondary sided commutation is initiated Fig. 8 UCtU Equivalent circuit for Interval 5 , u Q l ( t ) = ( ~ ~ . c o s ( w o( fs- f , .) t4 I t I t5 + q S )+ K s ) i i (15) Interval 7 (Switch-on of the main switch): In order to reduce charge-down losses and eventually EMC-problems, associated with the switch-on of transistors under non-zerovoltage conditions, the inspection of the transistor voltage is of predominant interest. The respective equivalent circuit is UCl uQ2 Fig. 10 Equivalent circuit for Interval 7, t, 5 t 5 t7 given in Fig.10, followed by calculative results. u Q l ( t ) = (A7.sin(oO7. ( t - t 6 ) ) +K7)ii A7 il(t) = -uz7 .cos(oo7. 1164 (t-t,)) (19) a zoom of commutauon intervals interval 5 interval 1 QI --> Qz Qz --> QI 1 Z, = /@, A, = Z , . UIlt6, K, = U1 U A practical analyis could be achieved with the following assumptions: ..2 'ossQl = 'ossQ2 Lcp = " 'Cl, ..2 L ~ "s Lm3 For this, the main switch voltage is calculated approximately to: -1 i(t) 70 [AI 6o 50 tl t4 t5 t6 t8 t6 t8 40 30 20 10 0 From the various means to accomplish ZVS, MOSFETs with low output capacitance are selected, since they allow to shorten soft-switching intervals and additionally the chargedown losses are reduced. Contrary, the means of increasing the resonant inductance results in the negative consequence of reduced turns ratio of the transformer, due to lenghtend commutations. Alternatively, a higher current il(k) yields a stronger discharging of the transistors capacitance, which is supported also by the transformer with integrated filter inductors. t4 t5 tl t5 t8 tl t4 interval 5 1 ov t it4 + U, ( t - t,) t4 t5 t6 t4 t5 - '8 L m +L c s u Fig. 12 Calculated waveforms at a) U , and U2 = 5 V , I2 = 60A. 1165 t8 t8 2 0 -1 -2 (22) t6 commutation intervals Q1-z QZ QZ --> QI tl [AI i, ( t ) = I l t , - t5 zoom of b interval 1 Equivalent circuit for Interval 8, t7 5 t I fg ( t ) =- tl -2 An evaluation of the symbolic treatment of the circuit by Mathematica is given in Fig. 12, showing essential voltage and current waveforms. Relevant design equations incl. loss determination for the individual components are derived in [7], using symbolic calculations. UQ, t5t8 2 1 0 -1 Interval 8 (Secondary sided commutation): After switching on the main transistor the transformer voltage equals the input voltage, affecting a fast commutation from Q3 to Q 4 . With iQ3 becoming zero, this interval terminates. The respective equivalent circuit is depicted in Fig. 11, followed by results for uQl(t), il(t). Fig. 11 tl = 200V, b) t6 t8 U, = 400V III. F TEGRATION ON The specific rectification circuit (,,hybridge") allows an integration of the filter inductors on the core of the transformer, aiming on cost and volume reduction. In addition the filter and secondary winding can be integrated, using one mutual current path, whereby currents compensate each other in half of the winding (see Fig. 14). By this means copper losses can be minimized. IL 1U, i L m ( f 0 ) = 2ii 2 L mDTS I, iLm(tl) = 2ii 1 U1 + -DTS 2Lm (27) As the load current is set and the secondary turns are chosen to one for minimization of copper losses the only remaining design variable of the first term of equation (25) is the air gap, which is the major disadvantage of this concept. Means to reduce the second term are constrained to the switching frequency and the core cross section. energizing demagnetisation iLfl= -is iLB=is Fig. 13 Principle of the transformer with integrated filter inductors. (N, number of respective turns; 6,magnetic flux) The assembly depicted in Fig. 13, Fig. 14 leads to an asymmetrical magnetic flux density. In order to study the effects of filter currents on the design of the single magnetic component, taking care of the input voltage range, the following simplifying assumption was set concerning the air gap: ,y' iLfl 4-iLD iLf1 i- iLf2 >> ' F e / p r The analysis of the transformer is based on the simplified analysis of the circuit, given in 1I.A. With losses being included by consideration of k,. k , = RLOCId/ (RLoad+Rv) (RLoa,..loadresistance). The maximum magnetic flux density is derived via a reluctance model, using Fig. 14 Reduction of copper losses using a mutual current branch for filter and secondary winding. 0 and the relations for the currents (eqn. (25)-(27)): 1166 IV. EXPERIMENTAL, VERIFICATION was adjusted at 10% of the load current. a A. Experimental set-up b Measurement results are obtained by a prototype forwardflyback converter for the following specifications: 200V to 400V Input voltage range, U1: Output ratings, P2 at U2/12: 300W at 5V/60A Switching frequency, f,: 380 kHz Components used: Main switch Q , : 2*BUZ307, Auxiliary switch Q,:BUZ 307 RDS(,,) = 5.2a11w0c, Cos, = 9 0 ~ F Schottky diodes: 82CNQ30, U, =: 0.23V, rF = 3.5mQl Fig. 16 Input- and main switch current at 30 A load current L, = 282pH, Lfl = Lrz = 291nR, L,,+u L,, = 5.1pH Fig. 15a: None of the primary sided MOSFETs are conductive after switching off the auxiliary switch. Only the output capacitances are effective at this instant. The input current il is shared by the capacitances of Q, and Q,. This behavior is indicated by the marked cycle "A". The diffference in current Ai is charging the output capacitance of Q,. Fig. 15b: If the delay time is too low, Q , and Q , are conducting simultaneously, which leads to a partial discharge of the clamp capacitor, indicated by the current peak of the main transistor. (Measured with disconnected transformer) C. Calculated and measured results B. Practical measurements Fig. 17 depicts the calculated efficiency using Schottky diodes while Fig. 18 shows the measured efficiency. The strange result indicating a higher measured efficiency as calculated is due to the following: Synchronous rectifiers (SR): IRL3803, R,,(~~, = 8.3nr12'loooc Diodes paralleled to SR: 20CTQ30, U, = 1V ' ~ A Clamp capacitor: 10nF, lOOOV Integrated magnetics: ETD39, reduced height = 23mm, core material: N67 primary: 22 turns, 2.4mm*0.25mim secondary + filter: 1 turn, 9*5Smm*.2Smm air gap: 1, = 0.5" ..2 a n k nun I~SMSIS b HI np. First, the real RD,(,,) of power MOSFETs under use were measured less than the assumptions for calculations. The same applies also for Schottky diodes concerning on-state-losses. Second, the core losses were calculated at a temperature of 25OC, while the actual core losses are lower at increased switch-on time operation temperature. ~ , ~ ~~ ~ , ~ ~, = 5 V / d i v . uos,Q1= lOOV/div., ~ ~ ~ , ~ ~ = S V / u d i v .~= lOOV/div., t :I lOOns/div. t = 400nddiv. Fig. 15 Main switch voltage uel(t) at 200 V input voltage Fig. 15a: According to the quasi zero-voltage-switching, the drain-source-voltage is low when switching on. At U,, = U,, the MOSFET's Miller-capacitance is predominantly charged by the gate-current, with UGs remaining constant. Fig. 15b:The main switch is switched on with a constant delay after the turn off of the auxiliary switch. The measured oscillogram was taken at rated power. Q, is not optimally switched on at higher load current, since its optimum switch-on instant II I I 1 I 100 150 200 250 % SO Pou,[WI- Fig. 17 Calculated efficiency using Schottky diodes 1167 300 92 intervals. During commutation the externally connected low power Schottky diodes produce higher on-state losses than the SR’s. Paralleling of SR can of course extend this range, but costs rise too for the power supply module, since the cooling sinks are not included. 1 % V. CONCLUSIONS 50 100 150 200 250 300 POUT[- In this publication an approximative and a thorough analysis of the soft switching forward-flyback converter is given which comprises only one magnetical component (with the filter inductors integrated on the transformers core) and a current doubler rectification circuit. Due to the low reactive power and soft switching in the essential operation range, the resulting efficiency is increased The higher efficiency for lower input voltage (s. Fig. 17 and by about 5% compared to hard switching PWM controlled or Fig. 18) results as the on-state losses of the Schottky diodes resonant converters. The profitable application range of are higher due to increased RMS-value of the current, as the Synchronous Rectifiers, which can compete with low-loss charge-down losses as well as core and copper losses of the Schottky diodes, is extended due to the current doubler rectifimagnetical components are higher at increased input voltage. cation circuit. But still Schottky diodes seem to be the only Hence, the circuit shows similar efficiency characteristics like components for low-voltage rectification, if output currents are the asymmetrical PWM controlled bridge (AB) topologies [11, in excess of 35 Ampere and if costs count. which also belong to the resonant transition class of converters. The gained efficiency increase of the inspected forward-flyback converter compared with the just mentioned AB is about 3%, mostly affected by the application of Schottky diodes with a blocking voltage of only 3 0 V in the previous circuit. Fig. 18 Measured efficiency using Schottky diodes REFERENCES [l] N.Frohleke, A. Fiedler, H. Grotstollen, B. Margaritis: Investigation of PWM controlled, resonant transition converters with asymmetrical duty cycle, IEEE IECON 1995, pp.429-434 [2] G. Hua, E C. Lee: Evaluation of switched -mode power 88 conversion technologies, Proc. of the 1st int. Power Electronics and Motion Control Conference (IPEMC), 1994, Beijing, pp. 86 12-26 / I I I I 84 [3] J.A Cobs, J. Uceda: Low output voltage dc/dc conversion, /DR IEEE IECON 1994, pp. 1676-1681 [4] M.M.Jovanovic, M.F. Zhang, F.C. Lee: Evaluation of I I I I I Synchronous-Rectification efficiency improvement limits in forward converters, IEEE Trans. Ind. Electronics, vo1.42, no.4, 50W lOOW 150W 200W 250W 3 6 W Aug. 1995, pp. 387-397 10A 20A 30A 60A 40A 50A [5] C. Peng, M. Hannigan, 0. Seiersen: A NEW EFFICIENT HIGH P O U P Io,, FREQUENCY RECTIFIER CIRCUIT, HFPC, June 1991, pp. 237-243. Fig. 19 Comparing measured efficiency between Schottky [6] I. D. Jitaru: HIGH EFFICIENCY DC-DC CONVERTER, diode rectification (DR) and synchronous rectification Power Conversion, June 1994, pp. 109-116 (SR) for an input voltage of 300 V [7] J. Richter: Quasi zero voltage switching forward-flyback Fig. 18 shows the efficiency graphs for an input voltage of converter with active clamp and current doubler rectification, Diploma thesis (in German), University of Paderborn, Institude 300 V versus load current using only one synchronous rectifier for Power Electronics and Electrical Drives, Dec. 1995. (IRL3803) for Q3 and Q4in comparison with Schottky diodes. Note, that a comparative advantage for synchronous rectifiers results only below 30 A to 40 A and that additional losses are caused with high output currents due to longer commutation I I - 1168