Soft Switching of IGBTs in Lagging Lag of ZVT Phase Shift DC/DC

Transcription

1 Soft Switching of IGBs in agging ag of ZV Phase Shift DC/DC Converter Sandra Zeljkovic, omas Reiter Infineon echnologies AG Am Campeon 1-1 Neubiberg, Germany UR: Dieter Gerling Universität der Bundeswehr München Werner-Heisenberg Weg 39 Neubiberg, Germany UR: Keywords HV to V DC/DC converter, hybrid electric vehicle, phase shift full bridge ZV DC/DC converter, high speed IGBs, lagging leg. Abstract he additional effort to achieve zero voltage transition (ZV) in the lagging leg of frequently used ZV phase shift full bridge converter can be avoided by designing the converter with high speed trench fieldstop IGBs. hanks to their reduced turn-off but at the same time low turn-on losses, the lost of ZV in the lagging leg is not anymore critical to converter s efficiency. Moreover, it can be beneficial due to their improved conduction to switching loss ratio. Based on that conclusion, a simple method to maximize their efficiency by minimizing the resonant inductance is proposed. Introduction In hybrid and electric vehicles, the 14V network is supplied from the high voltage (HV) battery through an isolated DC/DC converter [1]. One of the most common topologies for this application is zero voltage transition (ZV) phase shift (PS) full bridge (FB) DC/DC converter (Fig. 1(a)). urn-on losses of HV side switches in this topology are completely or partially eliminated by turning the switches on at zero voltage. he idea behind is to achieve soft switching using parasitic elements - output capacitances of switches and leakage transformer inductance. In the practical implementation, external inductors are used to extend the range of currents at which ZV is achieved in converter s lagging leg [], [3]. his design consideration proved especially beneficial for superjunction (SJ) MOSFEs (commonly used switches for the range of switching frequencies around 100 khz and medium blocking voltage, e.g. 600V). Modern IGB series, which are nowadays alternative to SJ MOSFEs in 100 khz switching frequency range, are not their direct replacement regarding the converter design and operation. Still, mentioned considerations are often simply transferred to the converter design with IGBs, which does not always bring expected results. In [4], high speed trench fieldstop IGBs are successfully applied in ZV PS FB converter at switching frequency of 100 khz and initial studies on differences in converter design compared to application of SJ MOSFEs are done in [5]. In this paper, the effect of external inductor ext used to extend the range of currents where ZV is achieved in the lagging leg is investigated. he results show that high speed IGBs exhibit better efficiency during lagging leg transition when ext is avoided and only relatively small leakage inductance leak of the transformer is used instead. his is the consequence of the ratio between conduction and switching losses of IGBs, which differs from the ratio known for SJ MOSFEs when they are applied in this topology. Although ext helps to achieve ZV for wider load range in converter s lagging leg, and reduces in that way the turn-on losses completely, it affects negatively both the turn-off losses in the lagging leg and conduction losses in the freewheeling period. For the design of the converter using optimized high speed IGBs, the difference in loss balance compared to previously used technologies should be understood, so that the converter can better utilize the advantages of chosen switch technology. hese advantages mean the opportunity to improve the EPE'14 ECCE Europe ISBN: and P.1

2 efficiency avoiding additional external elements (inductor in this case), thus to reduce the converter cost and complexity. he paper is organized as follows: it has been investigated how the value of res affects the three following loss mechanisms in the lagging leg of the HV-side H-bridge: turn-on losses in Section, turn-off losses in Section 3, and losses in the freewheeling period in Section 4. As a conclusion, the overall impact on the converter's efficiency will be examined. IGB-based ZV Phase Shift Full Bridge Converter he detailed explanation of ZVS PS FB converter s operation can be found in many references, e.g. [6]-[7]. As a basis for understanding the mechanism of switching and conduction losses in HV side H- bridge, current of transformer primary winding is analytically expressed. his model is used later in the paper for losses analysis. Fig. 1(b) shows the sequence of gate signals over a switching period as well as transformer primary current waveform. hree main states in the operation of H-bridge are power transfer, freewheeling period and loss of duty cycle that occurs at the beginning of each half of sw. he operation of two H-bridge leg (leading and lagging leg, see Fig. 1(a)) differs during one half-period: the transition in the leading leg occurs between power transfer and freewheeling and transition in the lagging leg occurs between freewheeling and loss of duty cycle period. Current waveform (Fig. 1(b)) in period from t 1 to t 6 (except short transition periods) is described in able I. Fig. 1 (a) ZV phase shift full bridge DC/DC converter for (H)EVs (b) ransformer primary current in one switching period sw and corresponding gate signals of HV side H bridge switches. In experimental measurements presented in this paper, the measurement trigger point in reference to the rest of s is marked EPE'14 ECCE Europe ISBN: and P.

3 Effect of res on IGB turn-on losses in the lagging leg Numerous solutions have been proposed for the problem of losing ZV in the lagging leg at light loads. he most frequently used, due to its simplicity, is still an external resonant inductor ext in addition to leak (Fig. 1(a)) (e.g. in [7]). Its purpose is to extend the range of currents in which ZV is achieved in the lagging leg (thus eliminates turn-on losses). Unfortunately, additional ext brings along several disadvantages: increased number of components, higher loss of duty cycle (which may lead to the need for lower transformer turns ratio, resulting in higher values of primary current) and higher voltage stress on secondary side switches. he main condition to achieve the ZV of lagging leg (1) is that available energy in the resonant inductance ( res = leak + ext ) has to be higher than the energy required for charging capacitances that take part in this transition. Here C oss is a sum of non-linear output capacitances of low- and high-side switch in lagging leg, C tr is the capacitance of transformer primary winding) (1) ( leak + ext ) I frw, end > CossVin + CtrVin 3 ABE I Overview of conduction periods and switching transitions of IGB switches and diodes in HV H-bridge during the one half of switching period sw (the other half is symmetric) Part of sw oss of duty cycle t 1 - t Power transfer t - t 3 eading leg transition time t 3 - t 4 Freewheeling t 4 t 5 agging leg transition time t 5 t 6 Beginning of a period 0 d loss D D + t lead _ trans t lag _ trans End of a period d loss D D + t lead _ trans t lag _ trans Conducting switches Switching events S 1, S 4 S 1, S 4 S 1, D urn-on S 1 urn-off S 4 urn-off S 1 I tr (t) v in res t k out Deff ( I out, ave ) + n vin vout n ( t t ) out n ( I min v + eq d0 ΔI + v r eq lead_ trans ce0 ce ( e ) e req t leak r = r + r + R d req t leak 1) par he effects of res on the losses in IGB and freewheeling diode are investigated using the prototype converter (whose details are given in Appendix). First, the process of successful ZV in the lagging leg is described. EPE'14 ECCE Europe ISBN: and P.3

4 When the condition from (1) is fulfilled, the switch in the lagging lag, S 3, turns on at V ce 0. Fig. (a) presents the turn-on of S 3 in the lagging leg measured when ZV is achieved thanks to the high enough value of ext. In period P 1 in Fig. (a), output capacitance of the switch is discharged prior to the occurrence of gate voltage. When the gate voltage reaches threshold value, current through the H U op switch starts to rise with the slope res. his phase is the loss of duty cycle. In this period, secondary winding of the transformer is still shorted by the rectifier diodes. wo of diodes (body diode of switches SR 1, SR 3 or SR, SR 4 ) stop conducting when their current falls to 0, which happens at the beginning of period P 4 in Fig. (a). Period P 4 is the power transfer. In the prototype converter used to obtain the measurements of described event in Fig. (a), ext of.uh is used in addition to leak of transformer winding of 1.3uH. In the opposite case investigated here, when ZV of S 3 cannot be achieved (in the designs with low value of res or in light load conditions in designs with high value of res ), a certain amount of turn-on losses occurs. However, turn-on losses mechanism in this case differs from the one in hard switching converters (e.g. in motor drive inverter). In typical hard switching converter with clamped inductive load, at turn-on, IGB has to take over the full load current from the diode, so that the losses will also be affected by diode reverse recovery. In case of turn-on in lagging leg of ZV PS FB converter, two differences are present. he switch which is turning on does not have to take over the current from the opposite freewheeling diode, so there will be no reverse recovery losses. Furthermore, if ZV of lagging leg is not achieved, the energy stored in leak of transformer winding is spent; its current falls to 0. When S 3 is turned on, loss of duty cycle period starts and current rises from 0 with limited slope (4). Such mechanism of turn on losses is mathematically described in [7]. urn-on event of S 3 is presented in Fig. (b), measured on the same prototype in the same operating condition as in Fig. (a), but without any ext. During the period P 1 in Fig. (b), while switch is still off, its output capacitance is being discharged. When no more energy is available in leak, voltage rises back to the value of DC link (P ). When gate voltage occurs (after the dead-time is over), output capacitance has to be discharged again. he portion of discharge current can be clearly distinguished in the current waveform in period P 3, and is superposed to the primary winding current ramp. It can be noticed that the current bump in period P 3 (that occurs during the charging of intrinsic switch capacitance) in Fig. (b) does not exist in Fig. (a). his is the consequence of the capacitance discharge prior to the occurrence of gate voltage. Furthermore, the duration of period P 3 in both figures differs. In Fig. (a), due to the higher res applied, the loss of duty cycle is longer. Fig. (a) urn-on of the switch in lagging leg when ZV is achieved (gate voltage occurs when V ce is already 0V). Operating point: U H op = 00V, I out = 115A ext =.uh, leak = 1.3uH. (b) urn-on of the switch in lagging leg when ZV is not achieved (gate voltage occurs while V ce is equal to the DC link value). Operating point: U H op = 00V, I out = 115A ext = 0uH, leak = 1.3uH EPE'14 ECCE Europe ISBN: and P.4

5 urn-on losses are measured in the prototype without ext over the range of load current for the input voltage of 00V and presented in Fig. 5(a). he curve of turn-on losses is rather flat, due to the fact that the switch capacitance is voltage dependent and current slope during d loss is not dependent on the load current. he common understanding of the effect of ZV on the gate voltage is the absence of Miller plateau [8]. Such understanding comes from the behavior of conventional vertical MOSFEs, but as can be noted in the gate voltage in Fig. (a), the plateau in V ge is visible although the ZV is successfully achieved (i.e. V ce = 0 before V ge increases). Just a slight difference in the waveforms of V ge in Figs. (a) and (b) can be noted when turn-on events with and without ZV are compared. Such behavior is the consequence of IGB structure with two different pn junctions, one at the emitter and the other at the collector side. When IGB is still in the blocking state with V ce = 0, the built-in voltages over these two junctions are compensating each other. As soon as V ge reached the value where channel opens, the pn junction at emitter side transits to the conducting state and this change in internal voltage has further feedback on V ge, known as Miller plateau. Effect of res on IGB turn-off losses in the lagging leg Fig. 3 (a) urn-off of the switch in lagging leg when ZV is not achieved. Operating point: U H op = 00V, I out = 115A ext = 0uH, leak = 1.3uH; (b) Primary winding current during the lagging leg transition when ext =.uh applied in addition to leak = 1.3uH of primary winding (black trace) and when no ext applied so that only leak is present during transition (blue trace). Operating point: U H op = 00V, I out = 115A ext = 0uH, leak = 1.3uH When ZV is not achieved in the lagging leg (the test-case without ext considered in this work,), not only turn-on but also turn-off losses are affected. Fig. 3(a) is an example of the turn-off event of switch S 1 when there is not enough energy in res to achieve ZV. When, on the other hand, there is enough energy in res to discharge the output capacitance of S 1, V ce falls to zero before all the energy from inductor is spent. Switch is turned off and the rest of the primary winding current is taken over by the freewheeling diode. Difference in IGB turn-off currents with and without ext is presented in Fig. 3(b). Current during the freewheeling period is higher in case when ext is applied, and consequently, the I c,turn-off of S 1 is higher. Based on equations from able I, turn-off current can be analytically determined. he comparison of measured and calculated values over the range of load currents in presented in Fig. 4(a). Furthermore, turn-off voltage V ce, turn-off differs in test-cases with and without ext as well. In Fig. 3(a) in case when all the energy is spent from leak, turn-off will happen at V ce lower than U op H (marked on Fig. 3(a)). In the test-case with ext, when ZV is achieved, turn-off happens at full input voltage U op H. hus, beside reduced I c,turn-off, V ce,turn-off is also smaller compared to the test-case with ext. In 5(a), energy of turn-off losses for two considered test-cases is estimated based on the data-sheet value for energy of turn-off losses, using measured values of I c,turn-off and V ce,turn-off. EPE'14 ECCE Europe ISBN: and P.5

6 Fig. 4 (a) Currents I c,turn-off with ZV achieved (black) and ZV not achieved (blue). Calculated values are dots and measured values are lines. (U H op = 00V). (b) Voltages V ce,turn-off of S 1 when ZV achieved (black trace) and not achieved (blue trace). he results are experimentally obtained at U H op = 00V. Effect of res on the conduction losses in the freewheeling of IGB-based converter Conduction losses in the freewheeling period (t 4 to t 5 in Fig. 1(b)) are also affected by the value of res. Not only I c,turn-off of S 1 is higher, but also RMS values of freewheeling currents are increased, and thus the conduction losses in the freewheeling (Fig. 5(b)). he effect is more stressed in IGB-based design (as the freewheeling diode has to conduct) than in SJ MOSFE-based one, where MOSFE channel can be turned on and increase in conduction losses is less remarkable. Fig. 5 (a) Comparison of switching energies of IGB during the lagging leg transition for cases with and without ext. his transition occurs twice per switching period. (U H op = 00V). (b) Comparison of conduction losses in primary winding circuit during the freewheeling period (values calculated based on the model from Section 1 for U H op = 00V). EPE'14 ECCE Europe ISBN: and P.6

7 Conclusion: osses minimization by reduction of res in of IGB-based converter When all effects are summarized, it can be concluded from Fig. 6 that high speed trench fieldstop IGBs are more efficient with minimized res in lagging leg (only considerable small transformer leak ). With large enough ext, the turn-on losses are eliminated indeed, but turn-off losses as well as conduction losses in freewheeling are increased to the extent that total switching losses in the lagging leg transition are increased. At higher load currents, effect of reduced efficiency when ext is used is more significant. 94 no ext ext =.µh ext = 3.3µH 93 without ext ZV not achieved 9 ZV achieved ZV achieved Output Current [A] Fig. 6 Efficiency of high speed IGB-based design of ZV PS FB converter (no auxiliary supply included) with blue and green for two values of ext (applied to achieve ZV of lagging leg) and red without ext (where natural zero current switching in the lagging leg is achieved). Operating conditions: U H op = 00V; I out of 150A corresponds to the I c, turn-off of 0A, and I out of 75A corresponds to the I c, turnoff of 10A. he described behavior has not been experienced in ZV PS FB topology with SJ MOSFEs, as they exhibit lower turn-off losses compared to high speed IGBs due to absence of tail current phenomena. Furthermore, due to larger chip area required for the same current rating, intrinsic capacitance of the MOSFE is more significantly affecting turn-on losses than in case of high speed IGBs. hus, elimination of turn-on losses in the lagging leg of high speed IGB-based converter by increasing res will not increase the converter s efficiency. Reduction of res to the value of only moderate leak of the transformer will result in boosted efficiency of high speed IGB-based converter. Additionally, components number, converter s cost and complexity will be reduced while the behavior of IGB switches will still be enhanced. EPE'14 ECCE Europe ISBN: and P.7

8 Appendix: Details of Prototype Converter Operating conditions min typ max U op H 160V 360V U out 14V I out P out 0A 170A.4kW f sw 100kHz Component Controller IC HV switches V switches ransformer I UCC8950 F4-50R07W1E3_B11A IPB019N08 per switch in H bridge rectifier ransformer DK 6973-A (9:1, 1.3uH) Output ind. DK uH Input capacitor Output cap. 10uF 1000uF EPE'14 ECCE Europe ISBN: and P.8

9 References [1] S.M.N.Hasan, M.N.Anwar, M.eimorzadeh, D.P.asky, Features and challenges for Auxiliary Power Module (APM) design for hybrid/electric vehicle applications, IEEE Vehicle Power and Propulsion Conference (VPPC) 011, 6-9 Sept. 011 [].H.Mweene, C.A.Wright, M.F.Schlecht, A 1 kw, 500 khz front-end converter for a distributed power supply system, Applied Power Electronics Conference and Exposition 1989, Fourth Annual Conference Proceedings pp.43-43, March 1989 [3] R.Redl,.Balogh, D.W.Edwards, Optimum ZVS Full-Bridge DC/DC Converter with PWM Phase-Shift Control: Analysis, Design Considerations, and Experimental Results, Applied Power Electronics Conference and Exposition 1994, Ninth Annual Conference Proceedings pp , February 1994 [4].Reiter, S.Zeljkovic, Design of an automotive.5kw HV to V DC/DC converter using HighSpeed IGBs, Elektrik/Elektronik in Hybrid und Elektrofahrzeugen und elektrisches Energiemanagement, Miesbach, Germany 01 [5] S.Zeljkovic,.Reiter, D.Gerling, Switching Behavior of IGBs in Phase Shift Full Bridge ZV DC/DC Converter, PCIM Europe, Nuremberg, Germany 013 [3] J.A.Sabate, V.Vlatkovic, R.B.Ridley, F.C.ee, B.H.Cho, Design considerations for high-voltage high-power full-bridge zero-voltage-switched PWM converter, Applied Power Electronics Conference and Exposition 1990, Fifth Annual Conference Proceedings 1990, pp.75-84, March 1990 [6] F.C.ee, M.M.Jovanovic, J.A.Sabate, A comparative study of a class of full bridge zero-voltage-switched PWM converters, Applied Power Electronics Conference and Exposition 1995, enth Annual Conference Proceedings 1995., pp , 5-9 March 1995 [7] Fei Zhou, Xinmin Jin, Yibin ong, Xuezhi Wu, Xiuyuan Yao, A turn-on switching losses study in a ZC soft-switching converter, 7th International Power Electronics and Motion Control Conference (IPEMC) 01, vol.3, no., pp.1607,1610, -5 June 01 [8] M. Kazimierczuk and D. Czarkowski, Resonant Power Converters. Wiley-IEEE Press, 011. EPE'14 ECCE Europe ISBN: and P.9