High Efficiency Flyback Converter Technology

High Efficiency Flyback Converter Technology U. Boeke ulrich.boeke@philips.com Philips Research Laboratories Aachen, Germany Abstract - Technologies are discussed to realize a DC/DC Flyback converter with high power conversion efficiency. The focus is on efficiency because of its high value in the application to process electricity generated by photovoltaic modules. Experiments with a 2 W active-clamped bidirectional Flyback converter prototype demonstrate power train efficiencies above 9% between 1% and 1% of the rated power and peak value of 96%. Index Terms - Efficiency, Flyback converter, Photovoltaic I. INTRODUCTION Typically photovoltaic (PV) systems are characterised by power levels of 3 kw and more. One string of photovoltaic modules solar cells may has a nominal DC bus voltage of 4V and a nominal current of 7.5 A e.g. based on 6 that are connected with a utility grid by a single power conversion module [1]. In opposite to that, the application of building integrated photovoltaic systems has to deal often with the problem of partial shading that can significantly reduce the electricity generation of string connected PV modules. This drawback is reduced if every PV module has an own power converter that realises the function of a maximum power point (MPP) tracker and adjusts the operation points of all solar cells in one PV module. Such a power converter module may also combines the MPP function with an inverter for grid-connection in a single stage power converter [2]. An alternative concept is a two-stage approach where every PV module has an own DC/DC converter to realise a MPP tracking function. Secondly, multiple DC/DC converters feed their power into a DC sub-grid to supply a single DC/AC inverter that is connected with an AC utility grid. This two-stage approach is expected to have a higher efficiency compared with single stage power converters since every power conversion stage can be optimised for its special function. DC/AC inverters are available with an average European efficiency up to 97.4 % [1, 4]. The DC/DC converter discussed in this paper is a candidate to realise the module integrated MPP converter front-end. Hereby a high efficiency is of interest in a very wide power range to use the available electricity from solar energy efficiently. The proposed converter has been tested in six operation points between 5% and 1% of its nominal power to determine the average efficiency value named European Efficiency that is explained in the Appendix. The selection of the DC/DC converter topology considers the constraint of an input voltage range of about 1:2.2. This is explained with the electrical characteristics of a typical 18W PV module in Fig. 1. This PV module operates with a minimum MPP-voltage of about 16V considering 9 C hot solar cells in combination with a low solar radiation of 1 W/m 2. The PV module stresses the MPP converter with a maximum voltage of about 35 V on the other side if the solar cells are -25 C cold and if both PV module and MPP converter operates at zero power if a safety function e.g. in the inverter does not allow a transfer of power. The DC/DC converter topology should include a transformer to increase a nominal PV module of e.g. 24 V to a value that is high enough to supply a grid-connected inverter. The converter topology selection includes further constrains like dv/dt limitation and zero voltage switching (ZVS), di/dt limitation, low reverse recovery loss as well as synchronous rectification and the beneficial use of parasitic components e.g. of a transformer. Several DC/DC converter topologies fulfil most of these constrains [5, 6, 7, 8, 9, 1]. The activeclamped bi-directional Flyback converter in Fig. 2 has been chosen for a detailed analysis because of good partload efficiency in a continuous conduction mode design presented in [7]. The presented technologies demonstrate how to improve the efficiency of this DC/DC converter topology further. 22 2 18 16 ) P Module (W) 14 12 1 8 6 4 +9 C 1 W/m 2 +25 C 1 W/m 2-25 C 1 W/m 2 V in C c1 C c2 L u:1 I n2 Q3 M Q4 C in I M C out Q1 C1 Transformer C2 Q2 V out 2 5 1 15 2 25 3 35 4 V Module (V) Fig. 1: Calculated characteristics of a 18 W PV module [3] Fig. 2: Active-clamped bidirectional Flyback converter 1-4244-844-X/7/$2. 27 IEEE. 1268

II. PRINCIPLE CONVERTER OPERATION A mathematical model of an active-clamped bidirectional Flyback converter has been presented in [11]. Thus this paper focuses on design details to maximise efficiency after a short overview on the principle converter function. The basic function of a Flyback converter is a transfer of energy from a supply source into a transformer mutual inductance in a first time period. In a second time period this energy is transferred into a load. Thus the transformer component of a Flyback converter is also described as coupled inductor. The energy is transferred into the transformer mutual inductance in the time period < t < t 1 in Fig. 3. The effective duty cycle of the converter ends at t = t 1. Thus the PWM sub-circuit turns-off Q1 and Q4 at t = t 1. The currents in the transformer mutual and leakage inductance charge dv/dt capacitors of Q1 and Q4 and discharge dv/dt capacitors of Q2 and Q3. This period ends at t = t 2 if the drain source voltage of Q1 has reached the value of input voltage plus voltage of clamping capacitor C c1. The transfer of energy into the load starts at t = t 2. Hereby the currents in both transformer windings change with limited di/dt. That is e.g. di/dt = (V in +u*v out )/L in time interval t 4 t 5. This is a great feature of the active clamping principle. Thus transformer current high frequency harmonics and related AC winding loss can be influenced with this component. AC winding loss in a classical Flyback converter transformer with much higher di/dt are analysed in [12]. MOSFET Q3 is turned-off at t = t 3. That generates a dv/dt at the primary side of the transformer and Q1 can be turned-on under ZVS conditions at t = t 4. Synchronous rectifier MOSFET Q2 is also turned-off at t = t 3. However the inverse diode of Q2 conducts until t= t 5. Thus the measured power losses of Q2 are about 1W higher than the loss of Q1 that operates at full-load with a similar RMS current stress, Fig. 12. This difference in power loss of Q1 and Q2 can be reduced by extend the on-time of Q2 until t = t 5 with a different control sub-circuit. At t = t 5 both MOSFET s Q2 and Q4 must be turnedoff and the current in the transformer mutual inductance generates a dv/dt on the secondary side of the transformer until the drain-source voltage of Q2 has reached the value of output voltage plus voltage of clamping capacitor C c2. V GS1 V GS2 V GS3 V GS4 V DS1 V DS2 V DS3 V DS4 I M I n2 I Q1 I Q2 I Q3 I Q4 duty cycle effective duty cycle t 1 t 2 t 3 t 4 t 5 t 6 = 1 f s t III. TECHNOLOGIES OF A CONVERTER PROTOTYPE Fig. 3: Qualitative steady state waveforms at high power level A converter prototype has been designed and build to identify possible conversion efficiencies by means of measurements. The converter shall process the output power the PV module in Fig. 1. The PV module and converter input voltage range is 16 V to 35 V with a nominal value of 24 V. The rated input power level of the converter is 2 W that can be generated if this PV module operates with solar cell temperatures below C. The converter operates in continuous conduction mode with a constant switching frequency of 1 khz. This has been identified as an optimum frequency. Lower switching frequencies require more turns on a transformer that increase winding loss. Higher switching frequencies result in higher switching and conduction loss due to more dv/dt intervals t Δ1, t Δ3 during that a converter does not transfer power that increased RMS currents. 1269

Switch mode power supplies profit continuously from the progress of semiconductor performance. The use of International Rectifiers new 4 mω MOSFET s IRFB 411 for Q1 and Q2 has reduced the losses of the power train by 1.7 W at full load compared with the use of 8 mω MOSFET s in [11]. This advantage is, on the other side, slightly compensated by an about.3 W higher power consumption of the MOSFET drivers because of a higher MOSFET input capacitance C iss. The effective MOSFET output capacitance, on the other side, has been increased by using dv/dt capacitors of 1 [11]. Switching loss of the MOSFET have been minimised by turn-on each MOSFET with a specific dead time later than the PWM signal. Q1 (~3 ns) and Q2 (~37 ns) are turned-on with constant dead time intervals generated by means of RC delay circuits illustrated with Fig. 14. Q3 and Q4 are turned-on using adapted dead time intervals generated with IC 3a and 3b in Fig. 14. An active-clamped Flyback converter has the special detail that it requires an inductance in series to a transformer winding. The required series inductance value in Fig. 14 has been realised completely by transformer leakage inductances [13]. Smaller values of L have the advantage to reduce time interval t 3 t 5 that reduces conduction losses in the diode of Q2. However this also reduces the stored energy in L that generates the resonant switching. This is most critical at part load Fig. 6. Each transformer winding has been placed in a separate winding chamber depicted in Fig. 4. Each winding consists of 5 litz wires in parallel. The high amount of wires in parallel, turn numbers and the chosen ferrite core size ER48 are result of an investigation to minimise power loss in comparison with a former transformer design using an ETD49 core [11]. The maximum calculated DC flux density is 247 mt plus an AC flux density of 59 mt peak operating with P in = 2 W. Winding losses are still larger than ferrite core loss as given in TABLE I calculated with procedures in [12, 14]. These losses at P in = 2 W heat-up the transformer by 28 K (core) 32 K (windings) cooled only via natural convection. Further details of the transformer construction are given in the circuit diagram in Fig. 14. 1 11 5 11 1 n1 n2 mm IV. EXPERIMENTS Measured time functions in Fig. 5 till Fig. 7 illustrate the operation of the converter with resonant switching. All MOSFETs are operating with zero voltage switching at no-load, Fig. 5 as well as at full-load, Fig. 7. At part load in Fig. 6, the energy in L is not sufficient to realise complete ZVS after turn-off of Q3 in time interval t 3 t 4. Thus we see here valley switching. The dead time of ~3ns to turn-on Q1 has been adapted to a quarter time period of the resonance to achieve this. The operation with full-load also illustrates the maximum dv/dt and di/dt of the converter in Fig. 7. The maximum dv/dt of about 56V/8ns is generated after the turn-off of Q1 in time interval t 1 t 2. The maximum di/dt of about 28A/1µs is generated after the turn-off of Q3 in time interval t 4 t 5. V GS.Q1 V GS.Q4 V DS.Q2 V GS.Q3 V GS.Q2 V DS.Q1 A V t 1µs Fig. 5: Measured converter time functions V in = V out = 24 V, P out = W (no load) 2; 5 A; 1 µs per division V GS.Q1 V GS.Q4 V DS.Q2 t 1µs Fig. 6: Measured converter time functions V in = V out = 24 V, P in = 5 W 2; 5 A; 1 µs per division V GS.Q3 V GS.Q2 V DS.Q1 A V V GS.Q1 V GS.Q4 V DS.Q2 V GS.Q3 V GS.Q2 V DS.Q1 Fig. 4. Cross-section of the used ER48 coil former and windings TABLE I Transformer loss model, P in = 2 W, 5 C component temperature Core loss n1 n2 Sum 7 mω.8 W AC (13 A) 2 7 mω AC (12.1 A) 2 3. W = 1.2 W = 1. W A V t 1µs Fig. 7: Measured converter time functions V in = V out = 24 V, P in = 2 W 2; 5 A; 1 µs per division 127

Fig. 8 illustrates measured peak voltage stress of the power MOSFETs that is higher than input and output voltage. This is a characteristic property of flyback converters that has to be considered if one compares it with other converter topologies. The RMS current in the primary transformer winding is about 6 % higher than the converter DC input current for a wide operation range as depicted in Fig. 9. That is seen as an important characteristic for the good efficiency of the converter too. This is also of interest for a comparison with other converter topologies. The losses in the power train of the prototype have been analysed with two methods. Firstly, input and output power levels have been measured by means of a Zimmer LMG 31 power meter. Measured data are depicted in Fig. 1 and Fig. 11. The measured average European efficiency is 93 % operating with nominal input voltage of 24 V including control. The maximum measured power loss including control is 11.8 W operating with nominal input voltage of 24 V and P in = 2 W. The absolute error of this power loss measurement value is ±1.2 W. Secondly, power loss contributions of individual components have been measured by means of calorimetric techniques. V_DS.peak (V) I.n1 / I.in 8 7 6 5 4 3 2 1 8 7 6 5 4 3 2 1 2 4 6 8 1 12 14 16 18 2 22 Input pow er (W) Fig. 8: Measured MOSFET drain-source peak voltage stress Solid lines: Q1 & Q3 Dotted lines: Q2 & Q4 V in = 16 V V in = 12 V 2 4 6 8 1 12 14 16 18 2 22 Input pow er (W) Fig. 9: Measured relative current stress of transformer winding n 1 Each power MOSFET has been mounted on an own heat sink that have been all characterised individually to determine the losses by means of temperature measurements depicted in Fig. 12. Power loss of the transformer has been measured with a calorimetric technique given in the appendix of [15]. The two half-bridge drivers are the dominant power consumers in the control circuit. The auxiliary supply circuit of the control sub-circuit has an input power of 1.5W. Power losses in PCB tracks and capacitors have been calculated based on measured equivalent resistances. A resistance of 6 mω has been measured on the PCB (2 layers of 7µm copper each) on both sides of the transformer by means of an Agilent 3442A Micro Ohm Meter. An equivalent series resistance (ESR) of 14 mω has been measured for each Rubycon ZL 1µF electrolytic capacitor used as C in and C out. An ESR of 46 mω has been measured for each BC Component 375 MKP type µf clamping capacitor by means of an Agilent 4194A impedance analyser. All individual measured or calculated power losses are in sum 13.7 W. Here, one has to consider higher tolerances than for the electrical DC power measurement. Finally, it is expect that all relevant loss sources of the power train are considered in Fig. 12. Efficiency (%) Efficiency (%) 1 98 V in = 16 V 96 94 92 9 88 86 84 82 8 2 4 6 8 1 12 14 16 18 2 22 1 98 96 94 92 9 88 86 84 82 V in = 16 V Input power (W) Fig. 1: Measured power train efficiency (without control circuit) 8 2 4 6 8 1 12 14 16 18 2 22 Input pow er (W) Fig. 11: Measured converter efficiency including control 1271

4 P loss (W) 3 2 1 Q1 Q2 Q3 Q4 Transformer Control PCB Cc1,Cc2 Cin,Cout Electric Calorimetric measurements measurement Calculated conduction loss Fig. 12: Power loss contributions from individual components operating with an input power of 2 W and V in = V out = 24 V V. CONCLUSION An active-clamped bidirectional flyback converter has been realised for an operation with high efficiency. An average European efficiency of 93% has been measured with a converter prototype. Important technologies to realise this high efficiency are the use of latest power semiconductor technology, synchronous rectification, control features to maximise the benefits of zero voltage switching and the use of optimised magnetic components. The active-clamped Flyback topology benefits hereby from a transformer leakage inductance in opposite to a standard Flyback converter. Main disadvantage of this converter is the voltage stress of the power semiconductors. An optimisation of the converter efficiency only by system modelling is very difficult. It is always recommended to combine converter large signal modelling with experiments. The presented converter and component performance is maybe of interest for a comparison with other DC/DC converter designs. APPENDIX Photovoltaic systems operate often at part load and thus the efficiency at part load is of special interest. To compare the performance of power electronic modules for photovoltaic applications, the European Efficiency has been defined as an average efficiency value of 6 operation points with individual weighting factors. TABLE II Operation points to determine the European efficiency Relative power level Weighting factors 5 % 3 % 1 % 6 % 2 % 13 % 3 % 1 % 5 % 48 % 1 % 2 % VI. OUTLOOK The author expects that the efficiency can be improved further. Each of the following details has the potential to reduce the power losses and to increase its European efficiency in total by about 1%. The on-time of Q2 can be increased to reduce the conduction of its inverse diode from t 3 till t 5. The PCB layout can be optimised. It was optimised for calorimetric measurements. Resonant gate drives can be used to recover a part of the MOSFETs gate charge. The realisation of a high output voltage is also a task of future investigations to realise a supply for a gridconnected DC/AC inverter. Fig. 13: Photo of the converter prototype ACKNOWLEDGMENT The author thanks Prof. Dr. Rik De Doncker and Klaus Rigbers both from the Institute for Power Electronics and Electrical Drives, RWTH Aachen University for the cooperation in a joint research project on power converters for photovoltaic applications. 1272

+V in Cc1 uf 16V, MKP (t) L 1.5uH u : 1.96 : 1 I n2 (t) Cc2 uf 16V, MKP +V out 4x 1uF 5V, ZL-type IC 1 IR 2113 BAV 1 BAV 1 Q3 Q1 1 1 1 Q1, Q2: IRFB 411 (4 mω) Q3, Q4: PSMN 9-1P (8mΩ) M 15 uh Transformer ER48/21/21-3C96 2x 15 µm air gaps n1 = n2 = 4 turns litz 5x 25 x.71 mm 1 1 Q4 Q2 1 IC 2 IR 2113 4x 1uF 5V, ZL-type IC 3b AND delay time circuit IC 4 HEF41 NOR IC 3c AND IC 3a HEF481 IC 5 TLC556 AND Auxiliary supply +14 V Oscillator PWM IC 6 MC347 +V ref Fig. 14: Principle circuit diagram of the converter prototype REFERENCES [1] B. Engel et al, String technology - A successful standard of the pv system technology for 1 years now, Proceedings of the 2 th European Photovoltaic Conference 25, Barcelona, Spain, 25 [2] T. Shimitzu et al, Flyback-type single-phase utility interactive inverter with power pulsation decoupling on the DC input for an AC photovoltaic module system, IEEE Transactions on Power Electronics, vol. 21, no. 5, September 26, pp. 1264-1272 [3] Sharp Electronics, 18W photovoltaic module NUSOE3E data sheet, http://www.sharp.de [4] M. Kaempfer et al, Kurzbericht - Sunny Mini Central 8TL, in German, http://labs.hti.bfh.ch/index.php?id=1373&l=, 26 [5] I. D. Jitaru, G. Cocina, High efficiency DC-DC converter, Proceedings of the IEEE Applied Power Electronics Conference (APEC 1994), 1994, pp.638-644 [6] R. Watson et al, Utilization of an active-clamped circuit to achieve soft switching in Flyback converters, IEEE Transactions on Power Electronics, vol. 11, no. 1, January 1996, pp. 162-169 [7] G. Chen et al, Actively clamped bidirectional flyback converter, IEEE Transactions on Industrial Electronics, vol. 47, no. 4, August 2, pp. 77-779 [8] D. Dalal, Design considerations for active clamped and reset technique, Texas Instruments Application note SLUP112, 21 [9] H. Ding, ZVS resonant converter for consumer application using L6598 IC, ST Microelectronics Application note AN166, 23 [1] F. Z. Peng et al, A new ZVS bidirectional DC-DC converter for fuel cell and battery application, IEEE Transaction on Power Electronics, vol. 19, no.1, January 24, pp.54-65 [11] U. Boeke et al, Experimental analysis of a Flyback converter with excellent efficiency, Proceedings of the IEEE Applied Power Electronics Conference (APEC 26), 26, pp. 11-17 [12] C. Sullivan et al, Optimization of a Flyback transformer winding considering two-dimensional field effects, cost and loss, Proceedings of the IEEE Applied Power Electronics Conference (APEC 21), 21, pp.116-122 [13] G. Sauerlaender, T. Duerbaum, Description of transformer leakage by introduction of an AL-value, Proceedings of the Power Conversion & Intelligent Motion Conference (PCIM 22), Germany, 22 [14] M. Albach, Design of magnetic components, Seminar 25, Power Conversion & Intelligent Motion Conference (PCIM 2), Germany, 2. [15] V. Leonavicius et al, Comparison of realization techniques for PFC inductor operating in discontinuous conduction mode, IEEE Transactions on Power Electronics, vol. 19, no. 2, March 24, 531-541 1273