UNIVERSIDADE FEDERAL DE SANTA CATARINA PROGRAMA DE PÓS-GRADUAÇÃO EM ENGENHARIA ELÉTRICA SOFT SWITCHING TECHNIQUES FOR MULTILEVEL INVERTERS TESE SUBMETIDA À UNIVERSIDADE FEDERAL DE SANTA CATARINA PARA OBTENÇAO DO 'GRAU DE DOUTOR EM ENGENHARIA ELETRICA XIAOMING YUAN FLORIANÓPOLIS, MAIO 1998 ll SOFT SWITCHING TECHNIQUES FOR MULTILEVEL INVERTERS XIAOMING YUAN ESTA TESE FOI IULGADA PARA OBTENÇÃO DO TÍTULO DE DOUTOR EM ENGENHARIA ESPECIALIDADE ENGENHARIA ELÉTRICA E APROVADA EM SUA FORMA FINAL PELO CURSO DE PÓS-GRADUAÇÃO EM ENGENHARIA ELETRICA. \ Prof. \ Aà ~ d Coordenador do .Í Q Prof. Ivo Barbi, Dr. Dr. ` PD/"I ~ Orientador BANCA EXAMINADORA øfww ø Prof. Ivo ~ ,!í_§_¡T1,.›.£ He, Dr. Áeqzaudu šâuwú J: ámaa. Prof. Hélio Prof. cães Alexandre Ferrari de Souza, Dr. Prof. enizar am. Prof. z ruz ff. aíldo J ' `ns, Dr. ff ~ erin, Dr. Dr. ~ To my parents IV Acknowledgments To Prof. Ivo Barbi, for his professionalism, instruction and solicitude from time to time during this work. The rigorous approach taken for academic research will be the model to be followed throughout the future technical career. To from the professors in INEP, for their support and understanding. Prof. Enio Valmor Kassick is especially appreciated. The generous attention Thanks are also due to Dr. Roberto Rojas, for his friendliness and consultation. To W€I`€ the many colleagues in INEP, for their friendliness and help whenever difficulties €l'lCOl1I1Í€I`Cd. ' my wife, in particular, for the conciliation, understanding and devotion from her. To my family and the family of my wife, for the sound backing from them, which has enabled me To to concentrate on this work. \ V About the Author Xiaoming Yuan was bo_m in Anhui, P. R. China, on May 30, 1966. He received the B. Eng. degree from Shandong University of Technology, P. R. China, and the M. Eng. degree from Zhejiang University, P. R. China, in 1986 and 1993 respectively, both in Electrical Engineering. He was for ' with Qilu Petrochemical Corp. during 1986-1990 as a commissioning engineer power system protection, automation and power electronics installations. He was a visiting engineer at Tianjing Electrical Drive Institute during 1989 for the development of a VVVF 400kVA system. During 1993-1994, he was with the Power Electronics Institute of Zhejiang University, developing UPS and induction heating projects. He was awarded the Baogang Scholarship Education Commission and Baoshan Iron University. & jointly sponsored by China National Steel Corp. during his study at Zhejiang He is currently a student member of the IEEE Industrial Electronics Society. Vl Contents Nomenclature x Abstract 1. xiii Fundamentals of Multilevel Inverters 1.1. Implementation of High Power Converters 1 1.1.1. Device Association 1 1.1.2. Converter Association 2 1.1.3. Cell 1.2. 1 Association 3 Diode Clamping Multilevel Inverter 4 H 1.2.1. Voltage Synthesizing 4 1.2.2. Modulation 5 DC Supply Structures 1.2.4. Neutral Potential Control of the NPC Inverter 1.2.3. Potential Drift Other Operation Problems 1.2.5. 1.3. 1.4. 1.5. 2. and ' 8 8 9 Capacitor Clamping Multilevel Converter 10 1.3.1. Voltage Synthesizing 11 1.3.2. Modulation 12 1.3.3. Dynamic Clamping Voltage Distribution 13 Snubbing Techniques for Multilevel Inverters 14 1.4.1. Dissipative/Regenerative Snubbers for Two-Level Inverters 14 1.4.2. Dissipative/Regenerative Snubbers for Multilevel Inverters 16 1.4.3. Snubber Problems 19 ~ Conclusions 21 True-PWM-Pole Two-Level Inverter 2.1. Soft Switching Inverter Circuits Review 2.1.1. Different Types of Switching 22 22 22 2.1.2. Resonant DC Link Inverter Circuits 23 2.1.3. Resonant Pole Inverter Circuits 24 2.2. True-PWM-Pole Inverter Circuit and Operation 27 2.3. True-PWM-Pole Inverter Commutation Analysis 30 2.3.1. Total Commutation Duration 30 2.4. 2.3.2. Auxiliary Switch Peak Current Stress 2.3.3. Auxiliary Switch RMS Current Stress True-PWM-Pole Inverter Designing 2.4.1. Transformer Ratio k 2.4.2. Resonant Frequency mo 2.4.3. Resonant Capacitor C, and Resonant Inductor L, 2.4.4. Auxiliary Switch Gating Signal Width and Minimum _ PWM ON/OFF Time 2.4.5. Rating of the Auxiliary Switch 2.5. True-PWM-Pole Inverter Experimentation 2.6. True-PWM-Pole Inverter Variations 2.6.1. Normal Auto-Transformer Connections 2.6.2. Modified Auto-Transfonner Connections 2.6.3. Capacitor Connection Conclusions 2.7. Appendix 2.1. Mathematical Analysis of the ARCPI Appendix 2.2. State equations for the modified True-PWM-Pole schemes Appendix 2.3. Analysis of the commutation resonance in the capacitor i connection inverter 3. ' Operation of the Three Level Capacitor Clamping Inverter 3.1. Basic Operation Problems 3.2. Clamping Voltage Steady State Stability 3.3. Output Voltage Spectrum 3.4. Clamping Capacitor Stress 3.5. Clamping Voltage Dynamic Stability 3.5.1. Self-Balancing under Ideal Condition 3.5.2. Self-Balancing under Non-Ideal Condition 3.5.3. Self-Balancing in Three Phase System 3.6. Experimental Verification 3.7. Conclusions Appendix 4. 3 3.1. PSPI_CE program for simulation under non-ideal condition True-PWM-Pole Three Level Capacitor Clamping Inverter 8 viii 4.1. Circuit 4.2. Configuration 70 Circuit Operation 71 Features 74 4.3. Specific 4.3.1. clamping voltage 4.3.2. Normal Auto-Transformer Connection of the Second Pole 74 slzbilizzúion 75 H 5. 4.4. Topologizs for Mulúlevel case (M>3) 75 4.5. Conclusions 78 Designing and Experimentation of the True-PWM-Pole Three Level Capacitor Clamping Inverter - 79 5.1. Prototype Specifications 79 5.2. Main circuit Designing só , 5.3. 5.2.1. Power Supply 80 5.2.2. DC Link Capacitor 80 5.2.3. Clamping Capacitor 80 5.2.4. Output Filter 80 5.2.5. Main Switches 81 Resonant Circuit Designing 5.3.1. _ Resonant Components and Auxilialy Transfonner 81 81 5.3.2. Auxiliary Switches 81 5.3.3. Auxiliary Diodes 83 5.4. Thermal Designing 83 5.5. IGBT Driving Designing 84 5.5.1. Main IGBT Module Driving 5.5.2. Auxiliary 5.5.3. 5.6. 5.7. 5.8. 84 IGBT Module Driving Zero Voltage Detecting 85 86 Control Designing 87 5.6.1. Control Requirements 87 5.6.2. Control Diagram 87 Experimental Results Conclusions ss V 92 Appendix 5.1. BASIC program for generating the gating signals in EPROM 27256M 93 Appendix 5.2. Control diagram of the prototype 95 6. Operation of the Neutral-Point-Clamped (NPC) Inverter PWM Modulation 6.1. Sub-Harmonic 6.2. Neutral Potential Steady State Stability 6.3. Self-Balancing under Ideal Condition 6.4. Self-Balancing under Non-Ideal Condition 6.5. Self-Balancing in Three Phase System 6.6. Conclusions Appendix 6.1. _ PSPICE program for self-balancing under ideal condition Appendix 6.2. PSPICE program for self-halancing under non-ideal condition True-PWM-Pole Neutral-Point-Clamped (NPC) Inverter 7. ~ 7.1. Review of the Existing Work 7.2. Proposed Circuit Configuration 7.3. Circuit Operation 7.4. Topologies for Multilevel Case (M>3)' 7.5. 7.6. 7.4.1. Zero Voltage Switching on the Per-Leg Basis 7.4.2. Zero Voltage Switching on the Per-Cell Basis Experimentation Conclusions A Appendix 7.1. BASIC program for generating the gating 8. signals in A New Diode Clamping Multilevel Inverter 8.1. Series 8.2. EPROM 27256 Association of Diodes in the Conventional Inverter Operation of the New Inverter 8.2.1. Switching Cells of the New Inverter 8.2.2. Switching State of the Diode Network 8.2.3. Switch and Diode Clamping Mechanism 8.2.4. Over-Voltage from Indirect Clamping 8.3. Snubbing/Soft Switching of the New Inverter 8.4. Experimentation 8.5. Conclusions Appendix 9. 6 8.1. BASIC program for generating the gating signals in EPROM 27256 General Conclusions References ~ Nomenclature PWM: Pulse Width Modulation True-PWM-Pole: True PWM Zero Voltage Switching Pole inverter NPC: Neutral-Point-Clamped inverter ARCPI: Auxiliary-Resonant-Cormnutated-Pole-Inverter ACRLI: Actively-Clamped-Resonant-DC Link-Inverter DPM: Discrete-Pulse-Modulation ZVS: Zero-Voltage-Switching ZCS: Zero-Current-Switching SVM: Space-Vector-Modulation FFM: Fundamental-Frequency-Modulation C1, C2: DC link capacitors Cm: clamping capacitor Cf: output filter capacitor C,,, C,2, C,3_ CM: Cs: d: snubber capacitor instantaneous duty ratio d,, dz, d3, d4: fm: fc: resonant capacitors instantaneous duty ratios of cell 1, 2, 3, 4 fundamental output frequency switching frequency G: self-balancing constant im: instantaneous clamping capacitor charging current icmmsz instantaneous clamping capacitor rms current ihpms: instantaneous resonant inductor rms current from diode to switch commutation ilmrms: instantaneous resonant inductor rms current from switch to diode commutation imã: instantaneous resonant inductor nns current io: instantaneous inverter output current in: instantaneous current out of the neutral point im: instantaneous output filter inductor current I°.m,s: inverter output rms current inverter output Ip: peak current device actual tum-off current ITQ: freewheeling diode reverse-recovery current I,,: k: auxiliary L,,4_ transformer ratio Lm: resonant inductors Lf: filter inductor snubber inductor Ls: . mod(t): modulating sinusoidal reference M: index of sub-harmonic modulation/number of inverter levels p: number of switching P0: output power Ps: snubber loss Q: quality factor cells . R: resonant loop equivalent resistance Ro: load resistance SW: NPC leg switching fimction SW,: switching function of cell 1 SW2: switching function of cell 2 switching cycle Vdc: _ DC link voltage VS,: device blocking voltage of switch S, instantaneous inverter output voltage vo: V0.,mS: inverter output rms voltage vm: instantaneous voltage of resonant capacitor CH vem: vcs: clamping voltage during dynamics instantaneous auxiliary capacitor Cs voltage vdl, vdz: v A0: DC link capacitor voltages during dynamics direct inverter output voltage during dynamics f Z0: resonant impedance 6: leading angle õ: of output voltage versus output current declining factor of output low-pass filter 6: snubber loss factor ' - com: inverter toc: output fundamental angle frequency sub-harmonic carrier angle frequency AVAO: direct inverter output voltage variation to clamping voltage perturbation Aim: load current variation to clamping voltage perturbation Aicm: clamping capacitor charging current variation to clamping voltage perturbation Aicm.dc: Ai": in Ainda: DC component of Aim variation to neutral potential perturbation' DC component of Ai" xm _ Abstract This thesis deals with the soft switching techniques applied to multilevel inverters for high frequency high voltage and high power conversion. The following subjects are studied to this end: Multilevel inverter operation. After the brief inspection of the conventional high 1. power converter structures employing device association or converter association, fundamentais and problems of the multilevel inverters evolved from examined. Clamping voltage stability cell association are of the capacitor clamping inverter and the. neutral potential stability of the diode clamping inverter are explored in details. Multilevel inverter commutation. Major snubbers used in two-level and multilevel 2. The problems associated with these snubbers inverters are reviewed. are outlined. Further, soft switching techniques reported for two-level inverters are assessed. The transformer connection True-PWM-Poleutechnique utilized (NPC) and tested in three inverter. (ARCPI) level capacitor is proposed and tested. It is then clamping inverter and the Neutral-Point-Clamped This technique along with the Auxiliary-Resonant-Cormnutated-Pole-Inverter are both extended to the multilevel case (M>3), the topologies are demonstrated. Auto-transformer cormection and capacitor connection True-PWM-Poles are also discussed. 3. _ New multilevel inverter topology. series association is of the clamping diodes is A new diode clamping multilevel introduced, analyzed and tested. It's shortcoming unveiled. Clamping, snubbing and soft switching of this With inverter free of new inverter are shoitly treated. this investigation, the basic aspects in regard to soft switching inverters are outlined. of multilevel xiv Resumo Esta tese apresenta técnicas de comutação suave aplicadas a inversores níveis para aplicações de alta freqüência, alta tensão e alta potência. estudados alta potência A com múltiplos níveis. Após breve análise das estruturas de convencionais que empregam associação de interruptores ou -associação de conversores, examinam-se os fundamentos e problemas dos inversores A mútliplos níveis. do potencial de neutro do inversor com diodo de grampeamento são exploradas em detalhe. ' . Comutação dos inversores com múltiplos 2. com grampeamento do inversor com grampeamento capacitivo e a estabilidade da tensão de estabilidade múltiplos seguintes assuntos são com esta finalidadez' Operação dos inversores 1. Os com níveis. A maioria dos circuitos de ajuda à comutação (snubbers) utilizados nos inversores com dois níveis e com múltiplos níveis é revisada. Os problemas as técnicas de associados a estes snubbers são delineados. Além disso, analisam-se comutação suave conhecidas para inversores com dois níveis. A técnica True-PWM-Pole com conexão de transformador é proposta e testada, então utilizada e ensaiada um inversor em um inversor com tres níveis com grampeamento com grampeamento no técnica e 0 inversor com ponto neutro (NPC - sendo capacitivo e em Neutral-Point-Clamped). Esta pólo de comutação auxiliar grampeado (ARCPI - Auxiliary- Resonant-Commutated-Pole-Inverter) são estendidos aos casos de_múltiplos níveis (M>3) e suas topologias são apresentadas. Discutem-se também as conexões de auto-transformador e de capacitor para a técnica True-P WM-Pole. 3. níveis Nova topologia de inversor com grampeamento a com diodos, múltiplos níveis. Um novo inversor com múltiplos sem necessidade de grampeamento é introduzida, analisada e testada, associação série dos diodos de sendo também apresentadas as suas deficiências. Analisam-se ainda o grampeamento, a estratégia de ajuda à comutação _ (snubbing) e a comutação suave deste novo inversor. Com esta investigação, os aspectos básicos relativos à com múltiplos níveis são apresentados. comutação suave de inversores 1 Chapter 1. Fundamentals of Multilevel Inverters Abstract: This chapter reviews the existing approaches for implementing high power converters. The ftmdamentals of multilevel inverter circuits, including voltage synthesizing, modulation as well as their specific problems are then discussed. Commutation issues of these circuits are specifically treated which will be the main subject of this thesis. Implementation of High Power Converters 1.1. The power electronics community has witnessed in the last in the ratings as well as the switching characteristics of the Most remarkably, the 3.3kV, l.2kA IGBT modules and decade continuous advance power semiconductor the 6kV, 6kA GTO devices. entered the market, and the development targets for the near future are the 4.5kV module and the 9kV-12kV traction drives GTO and supplies or However, high power thyristor [l]. utility applications [2] [3] [4], call for involving higher voltage or current with adequate performance. have thyristors IGBT electronics, typically switching operation To meet this demand, devices, switching cells or converters have been associated as solutions. 1.1.1 . Device Association Parallel association: Converter current handling capability can be raised by paralleling two or more IGBT modules together [5] [6]. The major problem of this association is the unequal current sharing due to the module parameters deviations as well as the construction and thermal coupling issues. Moreover, highcurrent operation demands bulky setup with voluminous power connections evoking problems with thyristors can not be associated in parallel, which will lead to unequal current distribution Series association: Series association of voltage operation in railway interties commutated AC/DC [7], GTOs STATic CONdenser (STATCON) converter [9] applications. It is [9] level [9] [67]. By [7]. has been a proven technique for high [8] and self- further enabled with the possibility to supply the gate units directly from the main circuit rather than a low voltage supply ground GTO stray inductances. In practice, at the device screening, gate units adaptation, precise gate tum-off timing or hard gate drive [7] [l0], voltage sharing can be well ensured. Series association of 2 IGBTs is also under investigation the redundancy it offers A distinguished advantage of series association is by adding one more device in the ring, as well as the individual time extension as a result of voltage stress reduction. device life 1 . 1 .2. Converter Association In_ [1 1] [12]. waveforms as the case of device association, either parallel or series, the converter well as the di/dt or dv/dt rates of change do not benefit from the multiple devices. In converter association, however, multilevel waveforms are synthesized while the di/dt or dv/dt rates remain unchanged. Converter association can be realized in one of the following approaches. DC side AC side in parallel/ in parallel by transformer: As shovvn in Fig. secondary voltages are imposed and the secondary current ripple is the leakage inductance has to be located at the secondary side, which transforrner construction [l4]. drives [3]. But it is not This configuration recommended for is found at the high. To is limit this ripple not favorable for supply side of locomotive STATCON application due to the harmonic currents flowing through the inverters and the transformer secondaries [8]. DC side in parallel/AC side in series by transformer: As shown in Fig. at l.1(a), the l.1(b), currents both sides are imposed. The low ripple secondary current results in lower converter losses and magnetic losses. Railway and also interties [3] STATCON [8] [13] have employed such configuration. Transformer leakage inductance can be located at either side. DC side in parallel/AC side in parallel by Current-Sharing reactor: As 1.1(c), the spontaneous current sharing number of converters is is shown achieved through this configuration [15]. Obviously, limited to two. ` ` DC side in parallel/AC side in parallel by separate reactor: As shown in Fig. paralleling converter subsystem, this structure is characteristic of its DC side been widely by summing utilized especially in industrial drive areas [7]. in series/AC side in series by transformer: As shown in Fig. currents at both sides are imposed allowing for lower losses. in l.l(d), homogeneous modular construction, operation redundancy as well as the reduced current ripple of the current. It has in Fig. DC/AC/DC converter applications [16] l.1(e), again Such configuration can be found ' [l7]. A DC side in series/AC side in parallel by transformer: As shown in Fig. 1.l(f), voltages of the secondary sides are imposed forcing each converter working independently. The structure enables high voltage operation with reduced harmonics in the summing current [18]. AC side connected to open winding: _ [l9], as shown in ig. 1.l(g),. In this case, the 3 DC side can be either in parallel or series or separated [20]. Chokes to suppress zero sequence number of converters is limited to two. voltage are necessary. Again, the DC side isolated (H-bridge cascade inverter): As shown in Fig. merit of this configuration is by JET shows other. Literature review broadcasting amplifier by [22] ABB most salient the possibility of direct connection to high voltage system, allowing for elimination of the heavy transfonner. However, each from the 1.1(h), the [21] more than a decade that this [type DC side must be isolated of inverter has been constructed as and as high precision current supply. for plasma physics ago. It has in the recent years been recommended for other applications including industrial drives [23] [24], traction [25], active filtering [26] and especially static T] VAR compensator [27] [28]. \_ T' u!| _\_ ~ ~ ~ ~ T T (b) (a) _\_ _ I T ||!| _\_ T (d) (C) :HT ..._\_ 1 l (¢) (Í) Fig. 1.1. Different circuits resulted 1 . 1 .3. (h) (g) from converter association. Cell Association Parallel association: As shown in Fig. l.2(a), the cellular structure allows for feeding high current with greatly reduced ripple [29]. In particular, assuming a non-zero impedance of the voltage source at the switching frequency, current sharing alternative technique less paralleling was also recently explored [31], as shown becomes inherent [30]. An in Fig. l.2(b). Direct reactor- of cells has also been reported in engineering application, which enhances the current capacity but without reducing the ripple [70]. Series association: Explorations on the series association of cells in _the past two decades have given birth to two major structures: the diode clamping multilevel inverter [32]- 4 _ [3 7], and the capacitor' clamping multilevel inverter [38] [39]. clamping inverter (N eutral-Point-Clamped-NPC) which has Except for the three level diode now been increasingly re2 F* raz- by the industry for high voltage high power applications, other circuits accepted of the family are still under investigations. (b) (=1) Fig. 1.2. Different circuits resulted 1.2. from parallel association of cells. Diode Clamping Multilevel Inverter -The idea of diode clamping was first established in three level”s case, which known as the NPC [34]-[37]. Fig. 1.3 inverter [32][33]. shows a five It was then extended to any level inverter leg, level by is now different researchers which can be regarded as being set-up by four two level switching cells in quasi-series: C4, S4, Sƒ; C2, S2, S2°; C2, S2, S3” and C4, S4, S4”. Switches S1, S2, S2 and S4 in the up-ann and also switches ann can be regarded in quasi-series. switches in different cells, dv/dt remains the 1.2.1. S4' in the down- timing the switching actions of the corresponding an output waveform with five level is obtained, while terminal same value as each single device sees in the cells. Voltage Synthesizing Suppose By and S4”, S2”, S2' that the DC V link voltage is equally distributed capacitors, then a staircase output waveform with reference to the among the four storage DC capacitor neutral point could be synthesized as shown in Fig. 1.4 according to Table1.l. Note that the eight switches give only each cell ‹› 0 five switching is states and five output voltage levels are produced. The switching of conditioned by the others as following: C An outer switch can only be turned on when the inner switch is in conduction. An inner switch can only be tumed off when the outer switch is not in conduction. V5 . ‹›| __íC] °' à __ C2 ZS “Z VAO fl à 52' 00-ii-0 ZS à D9 D3 ZS D8 ‹›4 › S.- I 52 I ' S3 U S2' I ' S3' A › S3' U › . S4 4 D4 S. I ‹›4 ' ÃD11 LL Vi ‹›| |Sš` ZS Dn»-1 C3 Z§Ds ZFD|o as ao VD2+Ds+mz °¬l i Vml VDz+ns+m2 ' › | ZÉ Dó VD4+D|oi °-I Vm+D3 | I IVn4+1›1o P C4 Voa °i VDõ Vm+m+Ds S4' |:' Fig. 1.3. A five level diode clamping inverter leg, consisting of (5-l) storage capacitors, (5-l)×2 H Fig. 1.4. Switching i sequence with the resulting output waveform and the blocking voltages across switches and a total of (5-l)×(S-2) clamping diodes. the clamping diodes. V 1: Table 1.1. Diode clamping five Output V A0 level inverter switches combinations switches combinations Levels s,* 1.2.2. 3 S27 S3) S49 V5=l/2 Vdc* 1 0 O 0 0 V4=l/4 O 1 O O O 0 1 l O O Vdc O 1 1 1 0 v,=-1/2 Vdc 0 1 l 1 1 V2=' 1 * SI Vdc' Vdc stands for the total Z DC link voltage. Modulation Fundamental-Frequency-Modulation (FFM: device rating is For a major factor limiting the obtainable large maximum power applications where power, full utilization of the 6 device will be interesting [l4]. The Additional benefit of FFM the is expense of slow response. Bulky FFM has been deemed the best solution in this case. low switching filters are always used to ensure -PWM modulation: The Sub-harmonic and high efficiency, loss (snubber loss) increased power performance of turn-off devices together with advanced low interfacing THD. and improved switching rating circuit techniques have been creating PWM in high power applications. This can be found in most of increasing possibilities using the high power drive applications [20]. The idea of multilevel PWM has been well established that suits well the PWM The pattem shown in Fig. modulation of a M-level diode clamping multilevel inverter 1.5 is for the case of a five level inverter four carriers generating the four different dispositions [40]. of the DC with the intersects PWM signals for the four switching pairs. carriers The down trace neutral potential of the leg. Several have been studied exhibiting different harmonic results PWM modulation for single phase and three Engineering applications of sub-harmonic phase NPC inverter have been reported Space-Vector-Modulation: [40]. where the reference signal leg, output voltage with reference to the illustrates the at the [3] [41]. - Space-Vector-Modulation (SVM) has been extensively PWM reported to offer superior performance in comparison to conventional techniques (natural sampling or regular sampling etc.), in terms of the reduced harrnonics, optimized switching sequence and increased voltage transfer ratio [42]-[44]. approach to realize plane, a NPC redundant has also been a major PWM pattern in NPC inverter [l8], [45]-[47] by industry. inverter can triangular pattems, It be represented by 27 switching states located among which 19 of them states, as shown in Fig. 1.6. are independent states PWM commands can be voltage vector is located. the next sampling period. The harmonic at the at the complex comers of and the remaining 8 are created desired voltage vector at a constant sampling frequency, such sampling by a sequence of three vectors defined In a is by sampling the then approximated comers of the sub-triangle where the desired Such approximation leads the output value to the set-point within _ resultant switching sequence of SVM is essentially similar to that obtained by sub- PWM when adding a zero sequence component to the sinusoidal reference in a three phase system [48]. In principal, level three phase 1+3M(M-1) system will SVM can also be applied to a M-level (M>3) inverter. A Mbe able to work with M3 switching states, among which states are distinctive states. No engineering results have been reported. 7 ›‹VWW\\ Í S. as ‹t \yÁ'A'›‹› S4 VAO Fig. 1.5. Multilevel '_ " - um PWM method for a diode clamping multilevel inverter leg. ~ ~ ~ °**° ' -1,0,-1 Fig. 1.6. Space vector diagram of the Closed loop modulation: In addition 27 switching to the states ' of a NPC inverter. open loop schemes reviewed above, closed loop schemes like hysteresis current control [49] or delta voltage control [23] have also been investigated, which allow the generation of multilevel gating sequences inherently in the 8 closed loop. In particular, hysteresis current control has been used to supply high precision current for the saddle coils in plasma research [22]. 1.2.3. Potential Drift and DC Supply Structures Drift mechanism: potentials which may The M-1 storage drift if the net capacitors at the wavefonn subjected to destructive voltage stress. For the 0 When will ° ~ is Both converge five and further some devices will be level case V3 which keeps shown in Fig. 1.3,' two situations stable. Eventually a five level inverter potential (level V3) doesn°t drift. On the other hand, when the load current is pure-reactive, either lagging or leading, the net average charge flows into each level over a switching cycle is zero. Ideally all V3 and V4 are always To stable. Use regulated DC complex AC/DC Use regulation results have been reported. sources in the places of the storage capacitors [52], which requires circuitry to enforce the energy flow between the neighboring capacitors, for both unidirectional or bi-directional operation [3 5] [53]. Use back-to-back system where charges from both is feasible only control of the Note that the following approaches circuitry. which is applicable which voltage ' enable active power processing, have been proposed but no engineering ° floating not pure-reactive, level V4 will decrease and level V2 will to level DC supply structures: ° distortion become a three level one, while the neutral levels V2, ° M-2 [51]: the load current increase. link gives average currents flowing into the given potentials are not zero. Potential drift will cause output can be identified [50] DC when sides compensate each other input/output voltages are equal. Otherwise very AC/DC converter will be required [54] when pure reactive power is processed, [50], complex [55]. potential drift may still happen due to unequal capacitor leakage currents, unequal device switching delays as well as asymmetrical charging during transience operation etc. Thus active control over each potential is still necessary in this case [56] [57] [58]. 1.2.4. Neutral Potential Control of the NPC Inveiter Neutral potential self-balancing: wave symmetric, typically When switching function of each when sub-harmonic modulation is utilized, NPC leg is quarter- the neutral potential is 9 _ found to be self-balancing when the load deviation causes not pure reactive [59] [60]. Neutral potential is DC current component flowing in the neutral line rejecting the deviation. However, under space vector modulation, when the two states of the middle vector are not duly distributed during the local switching cycle [45], the neutral potential will become unstable and the self-balancing property will be Neutral potential control: lost. In the case of space vector modulation, neutral potential control can be achieved in principle by making use of the redundant states of the middle vectors in a three phase system [45]-[47], [61]-[62]. The redundant state of a middle vector tends to complement the charging/discharging flowing into the neutral potential resulted same vector. the other state of the from C ' Depending on the extent of asymmetry resulted mainly from different gating/switching delays, load transience, load unbalance etc. in a practical system, the need for active neutral potential control à« may arise even when sub-harmonic can¡be realized by superimposing a zero sequence PWM modulation is used. Such control DC component equally to the three phase references according to the neutral potential deviation direction as well as the system operation ›.... '›‹z.z.~- ».z.. mode [63]-[64]. fundamental cycle suppressing 1.2.5. is is that the net charge to the neutral potential during a third that such action of Other Operation Problems The complex structure of the diode clamping multilevel inverter results complex coupling among the commutating parasitic inductances through S2”, S2”, cell and the non-comrnutating and capacitances. Refer to Fig. Dm and D4 parasitic inductance S4”, cells through the negative current flowing of S2' will then lead to the due to the demagnetization of of the neutral bus. Indirect clamping of inner switches: Except for the switches are not directly clamped. Refer to Fig. 1.3, for the S2 in series with D, rather than S2 directly that capacitor C2. Similarly, by D2, D2 and DI2 when a 1.3, to the neutral potential, the turn-off discharging of the parasitic capacitances of S3' and is Note altered without affecting the line voltage. of the a low frequency behavior. Cell coupling: in So it is to C2. S2' in series Due with D4, is two lateral switches, all the inner switching cell of C2, S2 and clamped by D3 and D2 Dm rather than S2' S2”, it to the storage directly that is clamped to this indirect clamping, the discharged state will hold and 10 unequal blocking voltage distribution will happen. An imier switch always sees higher voltage than the outer one and the center switches will be mostly stressed. The unequal voltage [19] [65] and is distribution is a structural shortcoming common for the neighboring switches is of the NPC inverter [18] diode clamping family. The voltage difference between the directly related to the clamping bus parasitic inductance and therefore can be minimized by introducing advanced manufacturing techniques [68]-[70]. Series connection of clamping díodes: number of fast diodes corresponding voltage must be put stress. The capacitances of these fast diodes same As shown in Fig. in series in each diversity of the may and Fig. 1.3 clamping path to withstand the tum-on/tum-off characteristics and the stray cause dynamic voltage distribution problems. In the sense, different blocking characteristics of these diodes distribution problems. 1.4 that appropriate The use of grading may cause resistance for static voltage sharing voltage static and snubbing A capacitance for dynamic voltage sharing lead to lossy and voluminous solution. of this problem willbe discussed in Chapter 8 of this thesis. structure [3 7] free › Snubbing new Aside from the two difficulty: lateral buses, all the inner clamping buses in a M-level system (M>3) carry currents that are bidirectionally controlled which causes difficulty in offering .polarized damping for tum-on.snubbing reactors. Non-polarized snubber is especially inefficient [66]. Solutions to this problem employed in this structure for GTOs can be any practical applications. Capacitor Clamping Multilevel Converter 1.3. While the was NPC inverter also introduced [38]. until the introduction leg must be found before is switches S,, However, this technique 1.7. Similarly, this leg cells: C4, S4 S2, S3 in the late '70's, the capacitor clamping. inverter had received of the “ Imbricated Cells” A[39]. shown here in Fig. of four switching was proposed and S,,'; C3, S2 corresponding switches in different cells, A five-level capacitor clamping inverter and Sƒ; C2, S2 and By intemational attention can be regarded as the quasi-series association and S4 of the up-arm and switches be regarded in quasi-series association. little S,”, S2”, S3” timing S2°; C1, S1 and S4' and S2”. Besides, of the down-arm can the switching sequence of the a terminal output of five level is obtained. l l 1.3.1 . Voltage Synthesizing In a five level diode dependent on the other ~ clamping inverter leg, switching action of one switching cell and the eight switches give only five cells different switching states producing five level output. However, in a capacitor clamping inverter clamping voltage From the operation of each switching cell is stable, is leg, suppose that each independent of the others. the eight switches 16 different switching states are available producing the output. This redundancy can be witnessed --› 3. d¡×i0 CI I dzxio d3›‹io › _?_ _?_ Sz S¡ Cell 2 C2 $ ¡c2 (d|-dz) xio T í› T (I-d¡) xio V i d4›‹io› S4 Cell 3 Cell - 4 ¡° C4 ¿ ¡c3 $ ¡c4 (dz-d3) xio T (l-dz) xio 322 T (I-d3) xio P A (dg-d4) xio W: W. vb' - › C3 Cuz 1.2. S3 Cell l ~cl Fig. 1.7. from Table (1-d4) xio P Diagram of a capacitor clamping 5-level P inverter leg consisting (5-l)× 2 switches and (5-1) storage capacitors. A VAO ¡o lcl J 'ic2 'icll › a _¡M › g Sr |Si| S2 n sz' S4”. | l S37 s4 S1' › | › f I sy › | s4*› u I Fig. 1.8. Voltage synthesizing of the capacitor clamping 'inverter is five under fundamental frequency modulation. level ~ of five level ` 12 In comparison with the diode clamping multilevel inverter where one state corresponds to one level, the multiple states for each level in capacitor clamping inverter allows performance optimization for the given shows an 1.8 criteria. illustrative voltage synthesizing under fundamental frequency modulation pattem. Table 1.2. Capacitor clamping five level converter switches combinations Output VAO Switches Combinations Sl S2 S3 SJ SI, S2, S3, S4, 1/2 vdc 1 1 1 1 0 0 0 0 1/4 v_,c 1 1 1 0 0 0 0 1 - A 1 0 1 0 0 1 0 l O l l 0 l 0 0 1`0100101 '01011l010 0 -1/4 v,,Ç -1/2 vá. 1 0 l l l I 0 0 0 l l 0 0 0 0 I l l O 0 l 0 l l 0 0 1 1 0 1 0 0 1 0 0 l l l l 0 O 1 0 0 0 0 1 1 1 0 l 0 0 l 0 l l 0 0 l 0 I I 0 l 0 0 O l l I l 0 0 0 0 0 1 1 1 1 Vdc stands for the total 1.3.2. ' _ DC link voltage. Modulation Steady state clamping voltage distribution: during each switching cycle set equal. Steady is each capacitor cell, when the is cells, load current cell must be then guaranteed due to the flow through each capacitor during each switching cycle Output voltage harmonics: For a leg with p applied to each in Fig. 1.7, supposed to be constant, the duty cycle for each state voltage distribution across zero average current As shown . assuming an identical duty cycle a set of gating signals phase shifted by 21:/p will produce an output voltage where the harmonics up to p times of the switching frequency are zero [7l]. Clamping capacitor RMS current defined, the instantaneous RMS 13 stress [72]: stress is When duty cycle and phase shift are dependent on duty cycle and load power both factor. Maximum RMS stress happens when only reactive power is processed. Phase shift can also be 'Mit te optimized for RMS current Fig. 1.9. Sub-harmonic stress, rather PWM PWM m...t modulation pattem for five level capacitor clamping carriers shift phase shifted by 90 degrees are compared with a sinusoidal modulating higher than the carrier, the corresponding switch 1.3.3. PWM has been optimized for output voltage harmonics. reference producing the four gating signals for the four is leg. modulatíon: Fig. 1.9 shows a typical sub-harmonic modulation pattern where the phase Four than output voltage harmonics. is cells. When the modulating turned on. reference . Dynamic Clamping Voltage Distribution Success of the capacitor clamping inverter depends largely on the stability of each clamping voltage, especially during dynamics. Unequal distribution of the clamping voltage causes output distortion, and will subject the Switches to destructive voltage With the modulation pattem defined it in Fig. 1.9, and when the load has been verified that each clamping voltage will tend to converge matter what are the with the inverter initial [74]. stress. is not pure reactive, at its nominal value no conditions, as a result of the self-balancing mechanism inherent ~ 14 1.4. Snubbing Techniques for Multilevel Inverters A snubber network serves such multifarious functions as to limit the device di/dt and dv/dt rates of change, to transfer the switching stress, to suppress the turn-on/tum-off spikes as well as to control the EMI, not allowed by the device is still itself. For GTO, excessive voltage or current rates of change are When high capacity I_GBT is concerned, snubbed operation favored [99]. Enormous energy will otherwise be accumulated during hard switching, making device derating 1.4.1. etc. inevitable to avoid fatal junction temperature in operation [lO0]. Dissipative/Regenerative Snubbers for Two-Level Inverters Dissipative snubbers: Dissipative snubbers can be distinguished between conventional and low-loss snubbers as summarized below. One snubber differs the other in the ways of discharging the snubber capacitor or demagnetizing the snubber inductor [75]. 0 Conventional snubber: The mostly used conventional snubber [76] [77] 1.10(a). An additional clamping capacitor (2CS<Cc<5Cs) is circuitry sothat to simplify the low-inductance designing task. is 0 is shown installed for the Snubber loss is in Fig. tum-off high which the major factor limiting the switching frequency. Low-loss asymmètrical snubber: Snubber loss can 'be reduced by about 60% by only one snubbing capacitor, as shown in Fig. l.10(b) [78]-[80]. Snubbing of the switch is using down achieved by a longer path including Ce, V6 and Cs, typically 5Cs<Cc<l0Cs. The large storage capacitor suppress the voltage spikes but also necessarily increases the path inductance. An option is to use distributed capacitors for Ce. With either a clamping capacitor or a storage capacitor in the network, the switching of one inverter leg can be influenced by the other due to the coupling introduced by ‹› common inductance. Low-loss symmetrical snubber: The large clamping/storage capacitor as well as the resulted mutual coupling are absent in a symmetrical low-loss [83]. The price paid for the space savings is circuit, as shown in Fig. the less tightly controlled device voltage stress, typically 1.5 times instead of 1.2 times with the asymmetn`cal one. The presence of the tum-on snubber makes a small capacitance is l.10(c) [81]- in the discharging path resistance reasonable for Rs, typically 0.39. _ of the tum-off snubber Thus the actual snubbing nearly doubled, depends on the impedance encountered in the current path 15 to the remote capacitor. Typically 1.5 times is This effect reduces attainable in practice. the local capacitor to nearly half of the value needed for the device. As a result, the snubber loss is also nearly halved. Regenerative snubbers: Regenerative snubbers developed in the past can be broadly classified into passive recovery schemes and active recovery schemes as will be reviewed shortly below. l ‹ RS' CC RS _ C __ 'E V8 ._ _ -“S1 LS A VA 'E' 'V CS2 ` W* v V3 V5 ‹ S V6 V5 V1 _ V4 vA 7L Ls” Rs . vó v L,/2 ' V4 › V1 V' V2 V4 C _" ____s2 .. ú. (a) conventional snubber (b) low-loss Fig. 1.10. asymmctrical snubber Major dissipative snubbers (c) low-loss symmetrical snubber for two-level inverter. Passive recovery schemes: Typical passive recovery schemes have been to replace the discharging resistors in the dissipative snubbers with a coupling transformer as have been proposed in [76] efficiency and low which is [84]. The principal maximum GTO problem with transformer recovery is the lower switching frequency due to the core reset time limit, dependent on the transformer turns ratio. To avoid saturation, core reset by introducing a proper resistor in series with the transformer in order to dissipate the energy associated with the magnetizing current, or in the other instance, by opening the freewheeling path with an active switch have been studied [85] [86]. Other schemes of this category are to replace the discharging resistors with storage capacitors which accumulate the snubber energy and will then be sent back to the load during the turn-off process, as was proposed in [87]. The additional current during tum-on together with the complexity of this circuit has been the major drawbacks of this scheme. Active recovery schemes: The general idea of a typical active recovery scheme the dissipative resistors in the dissipative snubbers with choppers is to replace which feed back the l6 recovered snubber energy to the [88] is shown in Fig. A general trade-off and settling 1.11, DC link, as have been proposed in [88]-[89]. An example where C3, Lc and D, work as an inverting step up/dovvn chopper. exists in the designing among over voltage, over current of the devices time of the snubber components as can be fotmd in any other snubbers. Experiment on a good dynamics. l.5MVA system proves that a open loop control of the chopper offers A regenerative snubber is also installed for the chopper Normally, [87]. three phases share the chopper(s). The power rating and in-turn the complexity of the chopper to the inverter switching frequency. of chopper recovery is Snubber energy recovery for may therefore be appropriate to state that the success It low frequency limited to GTOs operation. in series: Series _ connection of GTOS are the state of the art technique for implementing large capacity utility inverters. is far actually proportional is The efficiency, however, lower than the conventional line comrnutated ones due to the snubber recovery becomes hence interesting. Collective [90] and individual have been tested recently, and the 300MW AC/DC converter recovered energy is [9]. latter one, as shown Regenerative circuit is [9] recovery schemes been used in a in Fig. 1.12, has GTO, and provided for each fed back through the series connected diode rectifiers to the Dc G' rz DC link. |] É zš' the _ b c,_l__ Energy loss. L* ~1 ~ a ' ~1 . CS' . ú __* Fig. l.l l. Chopper recovery scheme for the asymmetrical low loss snubber. Db and Dc interface the two other phases 1.4.2. Dissipative As [88]. Fig. 1.12. Method of snubber energy regeneration GTOs in series, which has been used in a 300MW AC/DC converter [9]. /Regenerative Snubbers for Multilevel Inverters far as the for - snubbing of multilevel inverters are concemed, the received the most attention from industry and research community. NPC inverter has The snubbing of the 17 capacitor clamping inverter seems not a different issue but has not been reported yet. The following content reviews solely the existing work on NPC inverter. Dísszpative snubbers: The an fact that NPC inverter leg consists of association simplifies the problem of snubber arrangement for 0 Low-loss asymmetrical snubber: Fig. l.l3(a) two cells in series it. shows a low-loss asymmetrical snubber for NPC inverter, which has been employed by_Siemens AG in a 3MVA GTO (2.5KV, 3KA) liquid cooled cell and C2 300 Hz drive system [18][19]. The two low-loss snubbers work for CI, V1, V3 V4 ,V2, cell respectively, in the same principle as in a two-level inverter. prolonged snubbing paths for the two inner switches V23, CS4, (Vlz, CS3, Vzz , Cs, for V3, VB for V,5, V2) call for particular construction attention. =1 __C] _ V” L V Cs] V A _ v~ V vn vz. IE › : .` _R '_C2 “_ V Ls2 V” V VZ4  \ C 54 _ _' V A Vi4 I- I A vz az" V¡^ F (b) IF£¡^ ç-:zh - -Ir 7 (H) CS2, E ' ~ and '~ ~ : Vw The I vz. .z. ' ‹«› ‹d› Fig. 1.13. Different dissipative snubber arrangements for 0 NPC inverter. Conventional snubbers: Fig. l.l3(b) shows an arrangement using conventional snubber for NPC inverter which has been used for traction converters in Europe [75]. The four tum-off snubbers directly mounted for each switch may give rise to serious coupling between the l8 cells. is Aside fromthe outer blocking device discharging when neighboring inner device tumed-off, the outer blocking device can further undergo a unnecessary voltage dip (associated loss) shown in Fig. ° its when the other outer device is turned-off. With the modified arrangement 1.13(c the ), two cells are decoupled. Symmetrical low-loss snubber: As shown in Fig. l.l3(d), the symmetrical low-loss snubber can also be applied to Note NPC inverter working in the same way as in the two-level case [91]. modules are no longer applicable due to the reactor position. that two-device Any Regenerative snubbers: theoretically be modified dissipative snubbers that are into regenerative ones [92] [93]. A reported recently [65] [94], by extending the idea in [89], as circuit, LS, serves as the CS3 CS2. and CS, for G, while CS2 and CS4 for G2. LS3 is and inverter can successful example has been shown in Fig. 1.14. In this mitigated. This circuit has been used in a Lú limit the tum on discharging of CS3 and the recharging paths establishedby CS2 and CS3 tum-off snubbers CS4 and for rolling mill drives NPC tum-'on snubber for G, and G3 while LS4 for G2 and G4. For tum-off, The outstanding benefit of this snubber for theouter switches used in CS,. The problem of indirect clamping choppers and the reported efficiency reaches 97%. _l_ ' ___ *ld _ V E+; W dc ___ Lú Dc2 :1}~ C s2 S_ lläíllãi; É: Fig. 1.14. A thus IOMVA back-back GTO (6-inch 6KV/6KA) system by Mitsubishi Corp. of Japan. The whole 24 ' is regenerative Snubber Circuit for NPC Inverter. GTOs share the four 19 Snubber Problems 1.4.3. Snubbervloss: All dissipative snubbers reviewed above consist of energy storage components, which have to be discharged or demagnetized periodically. Snubber loss of an inverter leg is given by (l.1) [75]: 1;={õ-Ê-cs-V¿%+%-Ls-(1%Q+1,Ê,)}.¿ (1.1) where CS: Tum-off snubber capacitance ITQ: Actual turn-off current In: Free-wheeling diode reverse recovery current Vdc: LS: fc: DC link voltage Tum-on snubber inductance Switching frequency 6: Loss factor (c=3 for the conventional snubber and o'=l for the The snubber and is low loss snubbers) loss is square proportional to the current switched and the voltage blocked, proportional to the frequency operated. With a low-loss snubber, 2% of the whole power installed for a 300-400 Hz system. applications, this large it may occupy 1%-T In high power high performance amount of heat transfer becomes highly objectionable resulting in low efficiency, limited frequency, and moreover, significant cooling and construction difficulties. Regenerative snubber allows for improvement in efficiency. However, frequency increase will necessarily be limited by the complexity of the recovery circuits. Voltage/current overshoots: Most snubbers, dissipative or regenerative, use a polarized series inductor for voltage transfer during turn-on, current transfer during turn-off. The tum-on and turn-off snubbers other, instead, each is involved in the operation 0 The energy accumulated and a polarized parallel capacitor for are not independent of each of the other. in the series inductor during turn-on will be transferred to the damped resonance across the device causing voltage in the parallel capacitor during tum-off will be transferred to the parallel capacitor during turn-off in a overshoot. 0 The energy accumulated series inductor during turn-on in another current overshoot. damped resonance through the device causing 20 When inductor or capacitor with higher values are used for optimizing the switching loss, the voltage/current result overshoots and also the snubber loss are amplified accordingly. The of such overshoots voltage and around is the derating of the switches, typically by 40%-60% in the nominal 20% in the nominal current depending on the thermal condition. Turn-on and turn-off snubber interactions can be decoupled by a specific arrangement [95]-[96], as shown in Fig. l.l5. Obviously, the large electrolytic capacitors for voltage ti 51 ii overshoot limiting lead to voluminous and expensive solution. . T Vdc Fig. 1.15. Snubber circuit free of voltage/current overshoots across/through the main switches. Snubbing diode to the following ° ° issues: problems The need for a snubbing diode in the turn-off snubber gives rise [98]: Forward recovery of the snubbing diode contributes to the first turn-off voltage spike during the rapid decreasing of the anode current, which is known as VDSP for GTO. Additional construction inductance in the snubbing path due to the mounting of the snubbing diode causes another term to the first tum-off voltage spike. ‹› Reverse recovery of the snubbing diode interacts with the parasitic inductance of the discharging resistor and causes a negative voltage spike (third tum-off voltage spike). ‹› In the case of asymmetrical snubber, snubbing diodes in a non-commutating become pre-flooded when the common potential voltage stress caused by a cornrnutating leg. very high stress if the switch is is lifted due to the The pre-flooded diodes turned on at such moment [80]. common will .leg can inductance be subjected to ' 21 _ Turn-on snubber loss: Tum-on snubber inductor in the power circuit is required by all snubbers, which causes considerable .steady state loss. Snubbíng complexíty for devices individual tum-off snubber [7], which Energy recovery calls for circuitry Prolonged snubbing path V Each device in series: is always oversized to of great complexity in a cascading string requires assist equal voltage sharing. [9] [90]. ' - diode clamping multilevel inverter: in All the switching arms, except for the up-arm of the up-most cell and the down-arm of the down-most normally composed of- several or the freewheeling path. The cell, are switches and several diodes in series for either the forward path internal inductances as well as the mounting inductances of these additional switches or diodes in the forward or freewheeling path will necessarily aggravate the voltage spike problem. Prolonged snubbing path - in capacitor clamping inverter: All the switching cells, except for the inner-most one, always involve a clamping capacitor in the snubbing path. Similarly, the internal inductance as well as the capacitor will add to the voltage spike problem. 1.5. 0 mounting inductance of clamping _ Conclusions When adequate capacity and performance of an installation can not be fulfilled with the use of single device, associations of devices, conveiters and - this series association of cells has led to new cells have been developed. The topologies including the diode clamping multilevel inverter and the capacitor clamping multilevel inverter, both promise high voltage operation with reduced harmonics without necessitating heavy magnetics. 0 Snubbed operation is needed for GTOS and is favored by high capacity IGBTs. The use of snubber leads to objectionable problems, which become more pronounced for multilevel inverters. Frequency-increase with regenerative snubber is limited by the complexity of the recovery circuitry. The pursuit for an advanced solution replacing snubbers in multilevel inveiters for high frequency operation constitutes the major objective of this thesis. 22 Chapter 2. True-PWM-Pole Two-Level Inverter ~ Abstract: This chapter reviews the state of the art of the soft switching inverter circuits for high power applications. The transformer connection True-PWM-Pole in details as results is analyzed an altemative for the Auxiliary-Resonant-Commutated-Pole-Inverter (ARCPI), of which are verified with a 5kW half bridge prototype. auto-transformer and capacitor connection 2.1. Soft inverter True-PWM-Poles Topology variations including are discussed. Switching Inverter Circuits Review Switching process defines the operation condition of the device and holds essentially the most critical position in power electronics. The different switching types as well as the representative soft switching inverter circuits [101] [102] are reviewed briefly. 2.1.1. Different Types of Switching ' . Besides hard switching which is characteristic of tum-off with full load current and turn-on under full voltage, other three types of switching are identified as follows [102][104]. Typical switching waveforms are illustrated in Fig. 2. Inductive turn-on/capacitive turn-ofif' the tum-on 1. An inductor is in “series with the device to reduce loss while a capacitor is in parallel with the device to reduce the tum-off loss. The savings in device losses are penalized by the troubles imposed by the trapped energy in the reactive elements. Inductive turn-on/zero current turn-off' . An inductor is in series with the device to reduce the tum-on loss while the tum-off occurs at zero current with zero loss by extemal circuit means. Tum-off voltage spike due to the rapid voltage transition from the inductor to the device after the device tum-off and the associated loss during the device reverse recovery are characteristic of this switching type. Series inductor in the main path as well as the charge dump of the parasitic capacitance during tum-on (Power FETS) reduce the efficiency. Capacitive turn-ofl/zero voltage turn-on: A capacitor is in parallel with the device to reduce the tum-off loss while the tum-on occurs Tum-on at zero current spike due to the rapid current transition voltage by extemal circuit means. from the capacitor to the anti-parallel 23 diode as well as the associated loss-during diode forward recovery are characteristic of this switching type. Demagnetization of parasitic inductance during tum-off leads to voltage spike. Logically, a forth switching type characteristic of zero voltage turn-on and zero current turn-off should exist [105]. For the case of inverter due to the greater complexity and voltage current P ~ (b) Inductive tum-on, tum-off capacitive turn-off Fig. 2.1. Switching vglmge voltage current ' Hard tum-on, hard has not yet been investigated probably less practical interests. voltage current (a) it waveforms current P (c) Inductive P ~ tum-on, (d) Capacitive tum-off, zero voltage tum-on zero current tum-off for different switching types. All snubbers are of the first switching type with Inductive turn-on/capacitive turn-ofl Among the two soft switching types, the second type of switching with inductive turn-on/zero current turn-ofl has been popularwith 1960s and l970s [106]-[l08], all the forced commutated thyristor circuits during the called today Zero-Current-Switching it is (ZCS) for short. For gatezcontrolled devices, the third switching type with Capacitive turn-ofi'/zero voltage turn-on has "been extensively explored and has been abbreviated as Zero-Voltage-Switching (ZVS). Recent work [109] [110] have suggested the use of the second switching type for IGBTS due to the growth in the turn-off switching loss as the junction temperature increases. When zero concemed, aiding network voltage switching phase or per-inverter is may be configured on basis. Per-phase positioning yields the variant per- of the resonant pole inverter family, while per-inverter positioning yields the resonant link family. The resonant pole family generally has more elements but individual phase works autonomously. In comparison, the resonant link family suffers from severe coupling between phases. 2.1.2. ~ Resonant DC Link Inverter Circuits Perhaps the most mature circuit of this family Link-Inverter (ACRLI) [1l1], due to its is the Actively-Clamped-Resonant-DC low component count and high development depicted in Fig. 2.2. However, the discrete nature of the resonant low frequency output voltage has DC - status, as link requires that the to be synthesized following a Discrete-Pulse Modulation 24 (DPM) [112]. And a resonant link frequency that PWM frequency of the inverter is required to attain by imposing greater number of pulses onto the exceeding 300 [113] [1 14]. kVA 200 comparable load harmonics level or DPM inverters need to compensate for the inferior resolution control bandwidth performance. rated in excess of four to six times as fast as the operating is load. Presently available kVA, the size, cost and cooling of the resonant inductor become Í_- Clamping capacitor 1.3-1.5 times the I F Inverter main devices _Y_ __- Resonaf DC voltage capacitor 04 °_| V Q? III ,t __ ~*'Í‹§›3~I*lÊÉf~ °_| Ó ffa4 Fig. 2.2. Actively-clamped-resonant-DC-link-inverter PWM operation was explored meanwhile, the considerable amount of etc. DC voltage. switch Resonant inducmr in limiting issues clamping °-l bl xzzúfizó are with link frequency up to 70-l00kHz. For inverter ratings Main device blocking voltage is clamped to about ACRLI ACRLI inverters [ll5], but PWM resonant DC (ACRLI). was not preferred [l16]. In the link circuits [117]-[1l9], [146] need yet to be examined further for high power applications. 2.1.3. Resonant Pole Inverter Circuits The resonant pole family is generally realized by loading the inverter inductively at the switching frequency, regardless of the actual load current condition. been studied for this purpose during the past decade [120] [l2l]. A variety of circuits have Among which the Auxiliary -Resonant-Commutated-Pole-Inverter (ARCPI) [122], as shown in Fig. 2.3, has been deemed the most plausible circuitry and waveforms of the full in high PWM power application area [123] capability. for a switching cycle are ARCPI consists of these two [147], The phase plane of shown due to its small power auxiliary the resonance and the relevant in Fig. 2.4 'and Fig. 2.5 respectively. critical issues: The control 25 A “ Ç D| Á' C¡-1 °_|S| _ S” Vdc í' aí 5°' 1 3- Dm Iboost LI 'load nf DSM Í D2 52 °_| f _ \ / Cn I \ Z PWM gating i _, .Ná / \ . 'l08d) | | 1 i > \ \ \ \ / , / I ___ / / / Phase plane representation of the resonance Fig. 2.4. signal = t Í i E l lãarñp š E X Sú *dead ' E l ati vvvv I I 4 S1 I . v Ill" :ff D2 to §| É ; oommiltati n š * a f z › z = É z = 5 . É ' ¡Lr  â 5 Fig. 2.5. Relevant tmmp ____|_,-:____ / between L, and Cú during a switching cycle.  S2 Í ¢ / Vcfz boost Inverter (ARCPI). ¿ \ \ / -_. _/ ' | | 1 1 \ Fig. 2.3. Auxiliary-Resonant-Commutated-Pole- \ / tramp=0 , _\ - C2 . ¡Lr AB = l ` Iboost Í š šlboost I IL ¡ 1° D2 mmutatin ¡¡ › I wavefonns of the ARCPI during a switching cycle. Control: For diode to switch commutation, before turning-off the triggered switch, the corresponding auxiliary switch switch current is current valued-at allowed to ramp up I,,00S,. must be turned on until the triggered Such ramp stage guarantees in main advance so that the auxiliary main switch carries that the pole voltage will an adequate swing to the rail level taking into account the uncertainties especially the losses in the resonance process [81] [124]. Without this stage, resonant capacitors can not be fully charged or discharged of the resonance, and considerable loss will be caused if a switching action is at the imposed end [125]. Consequently, to maintain always an adequate level of I,,00s,, the ramp time tmmp must be controlled to follow the load current. Precíse measurements of the load current as well as the 26 resonant inductor current become therefore mandatory, which will increase the cost and reduce the reliabilíty [I26]. As 1500” current the inductor current ramps up very fast, this control always small (5-I OA) and the resonant is is hard reverse recovery current will enhance the commutation but Zero voltage switching is inverter applications including further affected by the to be implemented [I27]. Diode can not be coimted on. it DC center tap potential. motor drives, the center tap potential is Generally, in self regulating at 1/z Vdc due to the equalization of charge in and out of the respective phases as the direction of each phase current altemates between positive and negative. However, when low frequency high output current Hence [128]. will be supplied, significant drift is tmmp must also be controlled lost if Iboos, to is likely to happen during a sustained compensate for this drift. interval Zero voltage switching remains unchanged. tdead Control: As shown in Fig. 2.5, after turning-off the main switch must be turned on between charged and the anti-parallel diode starts instant main switch, the opposite A, where the resonant capacitor gets fully conducting, and instant B, where the resonant inductor current decreases to the load current level and the anti-parallel diode stops conducting, as expressed in (2.l). 2 “L” acos[ where Ibama) V ,/(ff +‹1z,.,..zz›>2 “'° = _._.l_ t/20,4 is S ]Stdead ~ 2 Ú” acos[ the resonant frequemry, V Ibama) ]+Ib0m,Í/i ,/‹-ff +<1zmzÛ›2 Z° = -ÊL \l2C, is (2.1) df* the resonance impedance. From (2.1), both intervals are not related to the load current and tm, can ideally be set to constant over the whole output period. However, the following uncertainties will be influential and the preset constant tm, can go outside the actual range of instants 0 A and B. Variation of the parasitic resistance due to heating effects. ° Variations of the input DC voltage as well as its center tap i potential. As a result, voltage status of each main switch must be supervised to secure more robust zero voltage switching without regard to any circuit parameter variations [128] [I29]. Mathematical analysis for the ARCPI is given in Appendix 2.1. ' 27 2.2. True-PWM-Pole Inverte_r Circuit and Operation With a view to reduce the control complexities inverter [130] using transformer comiection, as J- 0 4 Dal ‹›| D1 51 IS' °-| Dsa2 Cri : t3 VN: 'Í + i ~&~ Í vd ,. - L' Nz. i Da2 ' p 1 S2 _ D1 - t _? _ _ _ __(_k=_"*1““)_ , , fil N| Dal an altemative.` 2.6, is studied as iuh Dal IS sa2 shown in Fig. x JL of the ARCPI, a True-PWM-Pole . í ¡_k,' _ (__'_'°í"Z___ ti- _ I “"' Cn Í] t5 t9 Í1 ›Vcú Dsal sal is im É X I 1 \-fx Fig. 2.6. Transformer connection inverter scheme. ` Tme-PWM-Pole Fig. 2.7. ' V Phase plane diagram of the resonance between L, and Cú during a switching cycle. Prior to discussion of the commutation process, the following conditions are assumed: 0 Positive load current 0 Transformer rail level in im is flowing and remains constant during the commutation. is set to ratio ensure sufficient energy for the pole voltage swinging to the the presence of commutation losses, as will be discussed in subsection 2.4. 0 Construction parasitics, switching transience and transformer imperfections are neglected. 0 Main switch turn-off and auxiliary switch turn-on happen the at the same instant as decided PWM scheme, while the main switch tum-on happens until the detected voltage across the switch declining to zero. Conduction interval of the auxiliary switch covering the âèl is constant maximum commutation duration as will be discussed in subsection 2.4. With reference relevant coimnutation the by to the phase plane waveforms commutation of the diagram of the resonance shown in Fig. in Fig. 2.8, 2.7, the and the commutation step diagrams in Fig. inverter leg during one switching cycle consists of the next steps: 2.9, 28 A PWM gating signal |; _ 1 _ (Ú Ê: . _____ '_' .-_...__ dead 2 Vl .W o§ 5 2 - Y Y V V Y Ê oommu non Vcr2 Y .. _ . _ _-. D2 wmmutan S1 1° fi ã"_"__'__"_í_` =' 'load Y .___ _ _ .`_,_m_m_,..W 5' Fig. 2.8. Relevant D,,2'. v. ls ló *3 14 - ¿; `.,._. commutation wavefonns of the True-PWM-Pole inverter during a switching cycle. Step I (to-t 1): Circuit Step 2 (t 1-tz): steady Freewheeling diode D2 carries the load current. state. S2 is turned off and S22 is tumed on at Step 3 among current ~ iurises to the load current level at and C,2 is C,, is initiated. Step 4f(t3-t4): vC,2 rises to V,,c at t2 charged while enhances the charging current and therefore leading to blocking of D2. Resonance C,, discharged. facilitates the (t4-t5): iu falls to load current at t,,. Recovery current of D2 commutation process. leading to conduction of D,. t3 zero voltage. Rapid current transfer from C,2 and C,, to D, Step 5 leading to conduction of Da2 and ' . (t2-t3): L,, C,2 t,, DC link voltage Vdc enforcing N2 sees a positive voltage of kV,,c which joins the decreasing in D2 5 Then S, is turned on at may cause oscillation. D, then stops conduction and S, starts u canying current. Step 6 (t5-t7): another steady Step 7 C,,, C,2. Da,”. C,, is Step 8 extinguishes at state. S, carries (t7-tg): S, between L, and of Da, and iL, t5 allowing for tum-off of S22 at tumed off and Sa, is turned on at N2 sees a negative voltage of charged and C,2 (tg-t9): vC,, rises to is Vdc at stray/intemal inductances of the paths (t9-t 10): iu falls to Circuit reaches I load current. t7, kV,,c after which innitiates turn-on of Sa, a resonance and conduction discharged. tg leading to conduction of D2. zero voltage. Rapid current transition from C,, and C,2 to D2 Step 9 t6. may Then S2 is tumed on at cause oscillation due to . zero at t9 allowing for turn-off of Sa, at the original steady state. D2 carries load current (same to Step I). t,,,. Circuit retums to 11 1 í ;{n2@Dsn2 ÃDÚ ãhl ‹^' É + N2 - Vdc ___ Crl Lr T› lu .N] '‹...,.¬..,...f _ É f 'load al f'*' NZ- Vdc . N.. A °'l ^ ii, Saz W Dsaz S1 ¡ Lr ¬› š .¿. 7. 11 fl› N] 2:.. Cr] ‹*= :i fiãszl %'"` ED”, SMDS” Cn CQ Dal f' :Í šl Step 1 (to-tl)/step 9 Step 2 (tg-tm) ‹À' À. H um s (t,-t,) fiê Vdc .' S¿| .I U E! cn ›‹› zâf-‹' NZ- ff ._‹^ Dsal s al ä Thgd IN¡ › ;§z› load ' al Y 2 :› r2 ,_ Step Step 4 3'(t2-t,) "'¡ ÍÍ í (t,-t.,) ‹›4 F' S¡¡2 ir al '*= gê A Cn (I-,IAI - 1 A °=2 ' I CH z- vdc ¡í§z=z=z=z=:ââ ¬«. v- L, mí .N¡ ¬‹. Sal Dsal W... z... - 5-* nv G. fil 4 anal, Step 5 + Ni . Vdc ..z.`.`›- ››....z,â À D2 C Í2 Dsal Dal Step 6 (t4-ts) › À Á D2 Cú I (ts-t7) r A °{ Szz l DB2 Dszz fr A ' às 2:1. ' _ + N2 _ Vdc fl ¬› E gâ ,_-.:' í . . D. Da2 H ø Êà-Í 1 + N2 Vdc =¿.¡.¿...¡¿,...§ :L ‹›‹ Snz ....z p ......zzz.zz... ,.....W.. ãi S *,- ' (ff^ _ ..._z-‹-g,› éh ,I Da; .@_f_ IQ . Szlvm ,:=›z: . ‹ “2 › ‹›4 Sz 'Ê zi ¿;- 0 '¬f¬: ._._ 1 Step 7 Fig. 2.9. Step 8 (t,-ts) Commutation step diagrams of the True-PWM-Pole (rzërg) inverter during a switching cycle 30 In summary, for the True-PWM-Pole inverter circuit, the main switch works with zero voltage tum-on and capacitive tum-off, whereas the main freewheeling diode works with zero voltage turn-on and zero current turn-off. Superior to the conventional snubber, the capacitive tum-off loss of the main switch can be minimized by optimizing the resonant capacitor to end. Meanwhile, depending on this forward recovery property, the snap-on of the main its freewheeling diode does not introduce any considerable loss. Moreover, all the auxiliary devices work with zero current tum-off. And yet voltage is only reapplied after tum-on of the opposite auxiliary switch and no voltage spike can occur. By comparison, although the auxiliary devices in the ARCPI DÇ link voltage, block half the they suffer from tum-off spike which calls for the use of additional snubbers [13 1]. Besides, despite the bridge configuration, the tum-on of the auxiliary switch is actually snubbed by the resonant inductor since the opposite freewheeling diode carries no current beforehand and its reverse recovery is negligible. Transformer excitation is reset to zero after each cornmutation and no magnetic accumulation can happen. In particular, by designing the transformer ratio less than I/2, source valued at (1-k) Vdc which is an auxiliary voltage higher than Vdc/2 becomes available which delivers sufiícient energy for the resonant pole to swing to the rail voltage. and the associated control complexities with the ARCPI are no Thus the “boost” stage longer necessary in this case. The auxiliary switch can be "simply turned on at the same instant as the main switch is turned off Losses íncurredƒrom the extra turn-on/turn-ofi" actions of the switch are further avoided. 2.3. True-PWM-Pole Inverter Commutation Analysis The following assumptions ° Resonance ang 1 e fr equency a› L T, quality factor Q=-oíi ‹› is are made 1 mo = U-Í=CÍ= , resonance impe dancezb = . _ J; L, swr tchmg cyc 1 e . , . infinite neglecting losses in the resonance process. Unit current i= iZ0 /Vdc, unit voltage; 2.3.1. Total in the analysis: = v/ Vdc, unit time Í = two . Cornmutation Duration For diode to switch (D2 to duration t5¡=t5-t, is given by (2.2): S,) commutation, the unit value of the total comrnutation 31 k d d a›0t51=(l%;5+(1r-acosm)+T+0¡% Í] For switch to diode is given by \/1-2/c Í; (2_2) D2) commutation, the unit value of the total duration (S, to (2.3): t,,,=t9-t7 ' a›0t97 = 11- acos zk +ÍlOad 1-k ) (2 -acos 2 ' -Hiload 2 ` . _ .à + [1-2k+i¡0ad2 -l¡0ad (23) - Equations (2.2) and (2.3) are graphically shown in Fig. 2.10(a)(b). 16.893 “ F Vá 14.601 ~ r ~ 41-rw .«... °)0ts1 wzw. --‹.›T‹".› 2 3: 10.011 1.125 í" 5.433 ›¿.z=;;‹. 3.142 /// / /“\+ ' 12.309 0 ' 0.5 3.104 l í 1.5 2 -_*-_- I "(' 3.142 03051 2.011 Í Y 1.455 1<=o.3s ' 0.393 k=0.5 2.5 0.33 3 i Diode to switch 0.5 l commutation (b) Switch to diode ts, and tg, 2 2.5 3 commutation with load current and transformer ratio. Obviously, for diode to switch commutation, the On _ 1.5 ¡1‹›a‹1 Fig. 2.10. Variations of commutation durations current. _ . 0 ¡1‹›za (a) k=0.35 ‹› kzoz 2.ss :_ `°` F total duration increases with the load the contrary, for switch to diode commutation, the total duration will decrease with the load current. Both will increase with decreased transformer ratio. 2.3.2. Auxiliary Switch Peak Current Stress For diode to switch (D2 to S,) commutation, the unit value of the auxiliaxy switch (Sú) peak current isapl can be represented by (2.4): 32 U; =(Í.;+1-k) (2.4) For switch to diode (S, to D2) commutation, the unit value of the auxiliary switch peak current isapz can be represented by ISWZ =(J(l-k)2+z¡0ad -zload) 3.125 ' 2.6 / / Ísapl 2.015 1.55 1.025 0 0.5 1 065 _"*_““" ++ -×- °-54° ** -×- 0.244 (2.4) and 2 0.143 k=0.5 2.5 3 0.041 I 0 0.5 1.5 1 2 (b ) 3 Switch to diode commutation of the auxiliary switch peak currents with load current and transformer (2.5) are represented in Fig. 2.1 1(a)(b). commutation, however, 2.5 ima it ratio. For diode to switch commutation, the peak current of the corresponding auxiliary switch increases with the load to diode 1‹=0.5 1<=0.35 "' k=0.4 fl_ k=0.45 ___. Diode to switch commutation Fig. 2.11. Variations 1‹=0.45 ` Íloâd «(a ) 35 k=o'4 1<=0 ÍSHP2 0.34ó I 1.5 (2.5) 0.447 *>'. 0.5 (2.5): % 165 (Sal) will decrease with the load current. current. For switch Both will increase with reduced transformer ratio. RMS Current Stress The resonant inductor RMS current stress over switching period resulted from diode to 2.3.3. Auxiliary Switch switch commutation íbpmu can be expressed by (2.6): I ' . Biblioteca Univetsštária +}Â- jáz À* UFSC 33 resulted from switch to diode [zíš +(1-1z)s¡n(í)]2âí+íl_.-ƒáfz [J1Í2?-1zí12âí where: 11 “buzz/(1“k) k - Lz =fi-0°0S(§) Lg =(§,,Í+¬/1-21‹)/k Similarly, the resonant commutation iLm,ms expressed by is É 1 Jöfz _ _ _ current (2.7): í 'r (zbmms) -TIO RMS inductor _2_ [z'¡0ad+(k-1)sin(z)-f¡Oad¢0s(:)] dz _ (21) _ [-,/1-2k+(z¡0ad)2 +z'¡0ad +1‹z]2dz ›« . where: 1-k k .Í¡=7Z'-¿l(I(Bí'-Í-d(3(Bí'"'Í2 __ J‹1k›+‹z,m,› Jz =‹J<1-1‹›2 The - +<í;›2 _ 2 RMS switch to diode commutation z-Lmmz [fl (a)(b). 2+z¿ - 2 t -Z)/k auxiliary switch shown in Fig. 2.12 2 t/(1/‹›+‹z,m,› 2 ¡ current during diode to switch commutation Sanrms are equal to ,- Lrprms The resonant inductor RMS and ,- Lrnrms respectively, current can then be described aprms and which are by (2.8): (2.s) 34 I 277 k=o.3, T=9o k=o.4, Í=9o -0k=o.s, j=9o <>- |‹=o.3, 3“=45 -vt- I 076 -0- i‹=o.4, i Sapms í_____ ,/A // / o.õ12 o 41 Í=9o r=9o r=9o ¶=4s k=o.4, T=4s }k=0.5, 0092 ¡Sanm1s 0.069 \.I o 047 0 0.5 I.S l 25 2 o.oo2 3 L008 0.509 L506 2.004 2.502 _* 3 'load diode to switch commutation Fig. 2.12. Auxiliary switch ¬ :i 0.011 íioad (a) 'Í`=45 I o 024 o zõs 0 066 -N- '|<=o.3, -fk=o_4, -0i‹=o.s, 4)- |<=o.3, -0- 0.lI4 / T=4s k=o.s, T=45 0.874 0136 ' -1- (b) switch to diode commutation RMS currents related to the load current, transformer ratio and switching cycle. An auxiliary switch is subject to the diode to switch commutation current or the switch to diode cornmutation current altematively at the fundamental frequency in accordance with the load current direction. switch RMS commutation, From Fig. 2.12, for diode to switch commutation, the auxiliary current increases with the load current. Nevertheless, for switch to diode it will decrease with the load current. Both will increase with reduced transformer ratio or reduced switching cycle. 2.4. True-PWM-Pole Inverter Designing 2.4.1. Transformer Ratio k When losses in the resonance process (device conduction loss, resonant inductor loss and resonant capacitor loss as shown in Fig. 2.13. To etc.) are considered, the actual phase plane guarantee that the pole voltage can still becomes compressed, reach the rail level which ensures zero voltage switching in this case, the transformer ratio k should 'meet (2.9): V V Vdc(1-k)-f z-fã (2-9) 35 Í^ = | ' Actual phase plane considering losses _»iiii --‹_1‹›_fiš«›_Â<&-1f›i.Tz«2--_ Í __ Ê; ›Vcz2 _ Fig. 2.13. Actual phase plane for the resonance after considering the losses in the process. H Where w L Q="(;¿_r and R represents the equivalent resistance in the resonance loop. Equation (2.9) can be simplified to (2.l0): - ~¬-., ul" -1 kSl___7.[.. M2 8Q (2.l0) When the transformer ratio is set less than 1/z according to (2.l0) taking into account the losses in the resonance process, zero voltage switching is then guaranteed Without any additional measurements or control complexities. Such extra freedom in transformer deemed the most distinguished feature the True-PWM-Pole 2.4.2. Resonant Frequency ARCPI. co., The resonant frequency should be set by optimizing the auxiliary switch according to Fig. 2. l2(a) based 2.4.3. offers over the ratio is RMS current stress of the on the switching frequency of the system. Resonant Capacitor C, and Resonant Inductor L, The resonant capacitance should be optimized for the _ main switch turn-off loss [132]. Based on the resonant frequency and the resonant capacitance, the resonant inductance decided. is then 36 2.4.4. Auxiliary Switch Gating Signal Width and Minimum The minimum width of maximum the auxiliary switch gating signal must be set above the value of the commutation duration, as demonstrated in Fig. 2.l0(a). In the same sense, the inverter minimum PWM ON/OFF time this value. 2.4.5. PWM ON/OFF Time (t,-t, in Fig. 2.8) should also be set above 1 Rating of the Auxiliary Switch Due to the zero current switching in the auxiliary circuitry and due also to the high switching frequency with respect to the thermal inertia of the device, the rating of the auxiliary switch should be chosen according to However, auxiliary switch peak current peak output current rating, which stress as illustrated in Fig. 2.12(a). 2.1l(a) should not exceed the device illustrated in normally is RMS its 25% higher than the device peak switching current in the case of IGBT. Aside from the transformer . . ratio that is attributed touensure zero voltage switching, designing of the resonant parameters is always a compromise among the switching loss, device rating and duty cycle loss (DC many factors voltage utilization) etc. such as Practical designing shouldbe oriented to serve the specific designing purpose where one particular requirement may outweigh the others. 2.5. True-PWM-Pole Inverter Experimentation A proof-of-concept 5kW IGBT half bridge built inverter, as modulated with normal sub-harmonic sinusoidal the prototype are shown shown in Fig. 2.14, has PWM modulation. The specifications of in Table 2.1. Designing results of the resonant and transfonner parameters according to sub-section 2.4 are shown in Table 2.2. Table 2.1. Prototype specifications DC input V,,c=420V voltage Output pgwer Output voltage _ P0=5kW Load current been Vom, Modulation =l06V index I°_,,,,, Switching =42.5A frequency _ M=0. 78 f,=6.5kHz 37 Í I I c ú_ :_ L li Dul I à ZS HKÊS Dm Szz D' S' Dal °-| _ ~¬_- g CH : VN: ' + ~ ‹› _ .Nz Vdc L¡ . .N| C2 K} u 1 1. Fig. 2.14. , E *W Dz Sz *° cf C¡z ' Ro V 13°”, Dsal Sal c. ED” Lf A . L 5kW IGBT half bridge inverter. Diagram of the Table 2.2. Resonant and transformer designing results M x ~ Resonant 1. capacitor maximum tum-off loss 4W. Resonant I. inductor on air core bobbin, Auxiliary 1. Transformer 60 tums Litz wire (7 strands Capacitance: C,=0.luF,V2. Type: low-loss polypropylene, 3. Estimated ISAWG copper wire wound Inductance: L,=l2uH, 2. Structure: 36 tums Transfonner (15 strands Current rate of change 20A/uS. 3. k ratio: 24AWG) =0.4, 2. Structure: in the 24AWG) Two in the secondary in parallel each made up of primary and 24 tums Litz wire wound on E65/29 ferrite core, 3. Allowed equivalent resonant loop resistance around l.95Q. As a result of the above designing, the resultant 12.7uS according to Fig. 2.10(a). The maximum commutation maximum peak and RMS duration is currents of the auxiliary switch are 89.5A and 21.5A respectively according to Fig. 2.1 1(a) and Fig. 2.12(a). Thus the auxiliary switch gating signal width is set at 14.4uS and the Widths are set the at minimum/maximum l6.8uS and 136.8uS respectively. Dead time of 2.4uS two main switches. is PWM inserted interlocking ' Two SEMIKRON IGBT modules (SKM50GBl23D, 1200V/SOA) are employed as the main and auxiliary switches, four ultra-fast HFA30TA60C (600V/30A) work as the auxiliary diodes. Two storage capacitors C, and C2 each rated at 350V/3300uF form the DC center tap. A low-pass filter (L¡=1.45mH, Besides, four C,~=l2uF) SEMIKRON SKH10 and auxiliary IGBTS. Each driver is is installed at the output. intelligent driver are interfaced by an used for driving of the main external zero voltage detecting circuit to 38 the gate terminal of the The voltage becomes zero. Fig. 2.15 2.16 shows the IGBT module, which releases the gating signal when the detected of which are described in Chapter details shows the output load side voltage and ZVS 5. waveforms. Fig. filter inductor current comrnutation process of the main switch S, during switching cycle. Further in Fig. 2.17, the details of tum-on at i,oad=56A are shown. For a predicted dv/dt of 111V/uS (averaged over tg-t3) and And in Fig. 2.18, the details of tum-off at i¡,m,=56A are shown. For a and l8A/uS respectively. predicted dv/dt of The of 20A/uS, the experimental values are about 114V/uS di/dt 307V/uS (averaged over first voltage spike of about the experimental value tz,-ts), 120V appears due about 277V/uS. is to the stray inductance in the turn-off snubbing path (prolonged loop for measurement objective). Fig. 2.19 shows the main switch turn-on at zero load cLu°rent. 'I' io " 1 _ E _i L V ' ,¿ _ «i-.z _ .Ç 1 _ '_ rf I , ¬¬-~ M": 'hi 'x ` 5* Á.-ff* I u › , =~ 1.. |. . '_ ....\ 1 _ _ _ \‹. _ e E ,_ rff . íioov/D¡v;2oA/Div;5ms/Div. Fig. 2.15. Experimental output voltage t and filter inductor current. o ~ -o-¢-1-1--o-|-o-|--o-1-s-1--+-+»+-‹›--¬-o-+-|- ÊIOOV/Div; IOÀ/Div; Fig. 2.16. Experimental ¢i1uS/Div commutation process of the main switch ` (S,) ~_ Vsl É Í . lsl É ~1-+-+-1-`-o-1-o-›--o-›-+-+--f-‹-o-+- '_fl l : ¬-1-o-é--o-o-o-v--9-:-o-f-_ E _ l 100V/Div; l0A/Div; 2uS/Div Fig. 2.17. Experimental zero voltage switching process of the VSI Ísl -o-o-o-o--+¬-+-+--o-¢-o-e--o¬-‹-o- 2 1 main switch (S,). -¬-‹-›-f--›-à-o-‹--+-o-+-‹--o-o-+-+- -- n_r _ ii roov/Div; 1oA/Div; ms/Div Fig. 2.18. Experimental capacitive tum-off of the main switch (S1). Vsl -o-o-1-e--o-r-é-f--o-à-o-o- -o-o-o-o--v-e-f-+--o-1-‹-o--1-o-o-‹~ lDl IOOV/Div; 20A/Div; 5uS/Div Fig. 2.19. Experimental commutation process of the main switch I (S,) at zero load current 40 . Fig. 2.20 shows the commutation process of the auxiliary switch (Saz) at i,°ad=20A. For the predicated commutation duration of 7.2uS according to Fig. 2.l0(a), and the predicted peak current of 47.8A according to respectively. is Fig. l.l l(a), the experimental values are 6.7uS and 40.5A The tum-on is inductive and no reverse recovery current from the opposite diode seen. In the meanwhile, no tum-off voltage spike generated. lis The voltage protrusion afterwards occurs due to the dynamic charging/discharging of the floating middle potential of the auxiliary leg. Fig. 2.21 primary side. shows the commutation process of the auxiliary diode The at the transformer turn-off spike suggests the introduction of simple snubbing to the diodes in l practice. _ . i i ¡sa2 í _ _ uuumunniííímfi Í ' ___.L. ya Fig. 2.20. Experimental inductive ioov/Div; 1oA/Div; sus/Div í tum-on and zero current tum-off of the auxiliary switch vDa2 - 1Da2 _ |~ 1 . : vsä Ê I load ill' i › <]-à ..._.._...__..__. _ Ê Fig. 2.21. Experimental . _ _ .. . _.. ` l00V/Div; l(lA/Div;` 5uS/Div commutation process of the auxiliary diode (Dá). (Saz). 41 Fig. 2.22 shows the resonant inductor current in relation to the load current and the main .device voltage. Fig. 2.23 shows the details for switch to diode commutation at i,0ad=26A. For the predicted commutation duration of 2.25uS and the predicted peak current of 14.2A, the experimental values are about 2uS and l0A respectively. In the meantime, Fig. 2.24 the details for diode to switch commutation at i,0ad=l8A. shows The experimental commutation duration and peak current are about 6.2uS and 46A, which are in accordance with the predicted values of 7.3uS and 50.8A according to Fig. 10(a) and Fig. 11(a) respectively. VS2 í i 1 Íload -¢-›-o-f--1-o-o-o- i ÍLr íxoov/Div; 1oA/Div; lous/Div _ Fig. 2.22. Experimental relationship ¿ among the resonant inductor current, load current and main device voltage (S2). Vs2 z 1 ~o-o-¢-‹-¬¬-‹-+-+-o-‹-+--o-›-‹-o- 1Lr ¡ r -‹-+-‹-‹- 1-u_u í Fig. 2.23. Resonant inductor current 100V/Div; IOVA/Div;'2uS/Div in relation to the load current and the main device voltage during switch to diode commutation (S,to D2). (S2) 42 - -- ›- - ------¬--.-¬-.¬-____.¡..¡-¡--¡¡¡¡-¡-zw- -------- í Fig. 2.24. Resonant inductor current -- ‹ ~ ¬_ - 100V/Div; l0A/Div; 2uS/Div and the main device voltage in relation to the load current during diode to switch commutation (D2 to 2.6. - (S2) S,). True-PWM-Pole Inverter Variations The fundamental idea of the True-PWM-Pole inverter is to introduce a bi-directional DC link voltage to establish the auxili a1'Y voltage source in the resonance path which joins the voltage supply for the resonance. Besides transformer, this bi-directional voltage source can also be formed by auto-transformer, or by capacitor as discussed below. 2.6.1. Normal Auto-Transformer Connections Two auto-transformer connections I. í 1 .I S E “ TW* Í z'0 LI . U~ 3' ca E»  °4K} Sal Dsal O-i ' (21) : « _ : L QiD ÇÍ, 2.25. I- C fi - 5 Z 1 '*' ÍN2 9 Vdc D | shown in Fig. l Á DM 4% Fig. 2.25. [130] [133] [134] are DM J S2 D2 -__ C¡| Vm . . N; L¡ N. I -9 .z «Qi SM Ds” D¡ *l + V Si Daz ZS O-i --_ C [2 (b) Two variations of the True-PWM-Pole inverter with auto-transformer connection. 43 At the first glance, the auto-transformer connection offers attractive properties over transformer connection counterpart. The current flowing through the auxiliary switch (1-k) times of that with the transformer connection, which allows is its now for further reduction of the In the case of Fig. 2.25(b), the auto-transformer primary voltage auxiliary switch rating. rating as well as the secondary current rating are further nearly halved. flow Nevertheless, auto-transformer connection causes freewheeling current during steady state in the auxiliary circuitry, which results in considerable steady state loss in the auxiliary circuitry. Moreover, the auxiliary switches will operate with hard switching. Fig. 2.26 S2 shows the relevant experimental wavefonns based on commutation happens. Due prior to the commutation commutation (tl). Fig. 2.25(a) flowing through to the freewheeling current L,, when D, N2 and to Dm the resonant inductor current does not return to zero after the (tl), flows through Instead, this residual current switching of the auxiliary switch N2 and L,, Saul, causing hard Sa, (t3). Upon tum-off of S,,1,the opposite freewheeling diode Dm, starts conduction, which forces ~r~ ›‹ thecurrent decreasing in the resonant inductor. In the meanwhile, a voltage is reflected on the auto-transfonner primary side due to the current decreasing in the secondary side, which forces Daz into conduction carrying a increasing current. and;the resulted net currents add to each other flux changes at a rate which maintains the voltages on the both sides. The secondary Dm declines to zero until the The two current reverse direction at t4. Afterwards, both currents get equal and (ts), when the summing current in flow steadily through D,,2, N,, N2 and L, next switching action. The auto-transfonner flux can not be reset to zero. ÍN2 freewheeling 7' <}- _ current ¡DSa2 z \ r _ “L~.\»-, |Nl idual curr t f¡eew|,Le|¡¡¡g current 4 ., 1 . I m__ Q- :_ 'is IOA/Div; 2uS/Div Fig. 2.26. Experimental diode waveforms of the currents (Dm) when auto-transformer is used in the in the two windings and the auxiliary freewheeling True-PWM-Pole inverter (Refer to Fig. 2.25(a)). 44 The current spike at Similar 2.6.2. indicates the reverse recovery phenomena have auto-transformer are taken tl may be used of the freewheeling diode also been observed for the circuit of Fig. 2.25(b). in the True-PWM-Pole which prevent the flowing of the freewheeling Dm. As a result, measures inverter only after adequate current. Modified Auto-Transformer Connections From the above analysis, the freewheeling current and each one becomes aggravated by the to eliminate the freewheeling current By The resulted from the residual current residual current is primitively the auto- Such being the case, the following measures can be taken transformer magnetizing current. l. other. is and the residual current flowing. M breaking up the freewheeling current path, which leads to the transformer u connection True-PWM-Pole. 2. current, By de-coupling the interaction between the freewheeling current and the residual which can be realized by making the auxiliary switch a unidirectional one. Based on the second measure, the four schemes shown The in Fig. 2.27 are obtained. resonant inductor current rating in the two schemes of Fig. 2.27(a)(b) is further halved. However, the equivalent resonant inductor and resonant capacitor are reduced. For the case of Fig. 2.27(a), the reduction coefficient is 1/(l+k), coefficient is k/(l+k), which means capacitance are increased. and for the case of Fig. 2.27(b), this that for a given dv/dt or di/dt, the required inductance and Moreover, the auto-transformer primary in Fig. 2.27(a) and the auto-transformer secondary in Fig. 2.27(b) now block the full DC voltage. The state equations for the commutation resonances of the two schemes are given in Appcndix 2.2. The idea of unidirectional auxiliary switch has been verified in a shown inverter prototype. The topology diodes used are APTl5D60K(600Vl5A). Other components transformer connection for experimentation is True-PWM-Pole prototype (L,=l5uH, 5kW half bridge IGBT in Fig. 2.27(c), are the two series same with the rather than 12uH). Fig. 2.28 and Fig. 2.29 show the relevant waveforms when S2 carries current and is turned on and off. zzz Q F I: Dsl IS Dsâz *P .N¡ ipki., °-I Vcr2 Da2 Dsal Da ¬* ¡|‹ ku I. kvea 1., S81 : ‹›-| " :1 Vdc D1 S1 Da' °-l S az J. Ds! D2 Saz ~ 1/k ZS S¿| Daz Dsal ‹v.ú.z-v..1›1f× 0-ll I: Dz C¡›z Ds! (b) 1 Q Ds2 S Dal S81 OZ Vdc _ --V iloza Ô is: Daz E ›| Dz 5a2 Dsal VN: + _ Q N2 Lr NÍ 1 Sz ÃDa2 D2 Cr! °-I S81 4 Dsl Dsa I Ds] i I (¢) Fig. 2.27. Dal ZS Cfz 04 Dsgl H Vdc I ¡s2 l Ds: I ..z. Ji l Dsaz 5' Sa D l “N15 °-l z °.| -_. Sz C I *F ll' í. (21) z.:. _› . tâl. °-I C t i., kN| Cfz *I _ _,°4 Dm Vdc'Ven Dsl ¿ 5|D| I Vd *L S2 L~ °l W ‹d› Topology variations of the modified auto-transformar connection True-PWM-Pole. “À 'Lr f ñ_vs2 *_ Á ;+¬-›-›-¬¬-‹-›- a_,Ê* _ _ 1/ É \â \ > \ / \ \.V_/\`-f____.. \¡\,\,\,_J..¬_.-.__ Í í Fig. 2.28. Relevant experimental 100V/Div; l0A/Div; 2uS/Div waveforms when S, is tumed on carrying the load current. 46 “ 2 _/X/\\\/ÊV/\z¬\/___ /'\. A' F É sI I I l / _%%Pfi_fi%_%%~ _|_O_f_|__1_O_|_9_'-O_O-1'-1-*-T-0-9-4"* Ê*°:_ í _ , f" 'Lf_ 1 _ . . ÍIOOV/Div; l0A/Div; 2uS/Div Fig. 2.29. Relevant experimental The waveforms when S2 is turned off from carrying the load current. waveforms are shown relevant auxiliary switch commutation in Fig. 2.30. Obviously, the auxiliary switch turns on and off at zero current. However, a reverse voltage spike appears across the series diode right after the resonant current reverse voltage spike is resulted voltage which does not reset instantly. The ILI/ ÍV2§'^l VEL / ;a¬_¬¬¬+ s ,-\ zz-›._ E The by the auto-transformer secondary of the reverse voltage spike › +H+ is shown RCD snubber across the series diode. 2 Á / \\ 5 details be suppressed by a simple _ zero. from the interaction among the stray capacitances of the auxiliary devices and the resonant inductance, forced Fig. 2.31. This spike. can is reset to / lezw 5. aaa; /1"/"“_T*”"¬`1"""'¬ 5 100V/Div; l0A/Div; 2uS/Div Fig. 2.320. Relevant experimental auxiliary switch Sa, when S, is tumed on carrying the wavefonns load current. in 47 zfi¬\V~z L....J“...V.f_¿...,.__..._,__...r.,_ , . . . ...Q- . ' & É za: . , \ VM . . \\ 5 V \ Í _ . Dsl 1 i.. í í ¡ Fig. 2.31. Experimental details 100V/Div; SA/Div; luS/Div - of the reverse voltage spike resulted from the auxiliary switch Sa, tum-off during D, to S2 commutation. When modified auto-transformer connection the circuits are transformer connection circuit, the auxiliary switch current stress “__ z» PF* extra simple RCD snubber is needed for the series diode. is compared with the nearly halved while an Other benefits including the auto- transfonner voltage/current stress reduction or the resonant inductor current stress reduction are possible with its variations, which are also tested in the laboratory. ‹.._.~ ‹‹ 2.6.39. Capacitor Connection Capacitor can be used in the place of the transformer or auto-transformer to attain a True-PWM-Pole inverter, as shown in Fig. 2.32. This circuit resembles the traditional series current commutation circuit [108], which has recently been considered for use in the boost converter with turn-off device [135]. Da: Si í °-I o-l sal Dsa2 Vdc .F + ã H Di I DSM : . _-P Ú 'load ¬ ¬ Sz °-{ S31 Csi D2 CS2 : Dal t Fig. 2.32. New True-PWM-Pole inverter with capacitor replacing the transfonner or auto-transfonner. 48 new Control and operation sequence of this pole inverter are transformer cormection True-PWM-Pole. Fig. 2.33 (a)(b) tum-off processes of S2 when the load current is show carried .by S2 all the same with the the experimental tum*-on and and D2 respectively. Similar to the transformer or auto-transformer connections where the secondary voltage altemates direction naturally after each commutation, the capacitor voltage in this case changes direction after each its commutation as a result of the charging/discharging by the DC link voltage forcing the resonance, rather than subtracts from, like the transformer connection, which leads to more resonant current. However, the capacitor voltage always adds to the than sufficient resonant current. In Fig. 2.34, the experimental capacitor is to find the capacitance for C, which is resonant current, capacitor (C,) voltage and commutation duration. assumed a compromise among Certain loss must be inverter [l07]. When compared with the transformer connection True-PWM-Pole this circuit is inverter, the troubles associated expense of the in the issue in in the resonant loop to avoid iterative capacitor voltage increment, as is the McMurray at the voltage in The key relation to the load current under the given circuit parameters is illustrated. designing of this inverter (vc, ) main circuit fairly with magnetics (manufacturing labor, loss pronounced resonant current. and zero current switching etc.) are Both ensure zero voltage switching in the auxiliary circuit without any extra measuring or controlling. Mathematical analysisof the commutation resonance is given in Appendix 2.3. “Q” i ..z.a.m-e~1¬l "¬" 4 -> ^ ~v--,-.».z-Í” ~›zzz¬'v-. âõ tz -4'-i~+-O--i'-1-i-Ó-*Í-O-l-l-' 4l-^f“'i_l-'-i'-l'-1-i- 1.>.~ ,_ ~'\w-‹\-›.-ø.=\'~.~zf-zm.--_-.^~.-:. 41 2-oi _..-»¬,-.¬~‹iw-“N---.,-âi. y -_ ^ ~ ~ _-¡ ¡LI` ^~×-^~.~=-›~‹›--1.-.^~:^-.1..‹‹›-«¿.^-.--'\-z» \\ -----~‹¬-f-v-¬-w--/›- 100V/Div; IOA/Div; 10uS/Div (a) commutation processes when S, avoided, carries current g 3-V: ' _ I 1 4_>._..._.. .....-._..~._.. z - .- z-...__-«._. J. 1s2 ~-._ __-_.__._‹ .¬._¿ É V ~»ú-mx -+-+-r-e--‹-o-H--›-1-ø-‹--+-‹-f-o--› 1 _» ~*W“MMrmwMwn _ 2-‹› _ -o-v-›-f--o-o-1-H -+-o-o-ê- iL¡~~*^^.~v _. .~.»~^t X -.~. .zz~ . --.~.~~.~.-.~‹-«Í _ /_,./=¬‹/-~¬-^~^~\\^-'-~/M/\ / 1 100V/Div; 10A/Div; 10uS/Div (b) commutation processes when D, carries current Fig. 2.33. Experimental S2 commutation- processes when S2 or D2 carries current (Cs,=Cs2=0. luF, L,=7.5uH, C,=2uF) vzf ~ w»'^` '~vv ›--›` ` í.,__-*_ -1-_. í. 1-í-. ---_ í_- __, ,ii -zí Q-‹› í llllp zpíí -_:-_ 11 -í z-í ni-~ ¡í-ç .-ti .í , __;_ › za: _ wwwr I í 50V/Div; l0A/Div; 2mS/Di\'f Fig. 2.34. Experimental capacitor (Cs) voltage related to the load current. (Cs¡=C,¡=0. luF, L,=7.5uH, C,=2uF) 2 7. Conclusions Resonant pole inverter solves the problems of the conventional snubber associated with snubber interaction, series inductor and snubbing diode with simultaneous zero current switching, which regenerative snubber. etc. The auxiliary circuitry works a significant advantage over the is - Transformer connection True-PWM-Pole inverter avoids the control complexities of the ARCPI, while zero voltage switching for the main switch and zero current switching for the auxiliary switch are always guaranteed. comparable to the The ARCPI .circuit while its voltage auxiliary switch current stress is stress is clamped to the DC rail voltage. Experimental results from a 5kW half bridge prototype agrees well with the analysis. The normal True-PWM-Pole and residual current from such problems as freewheeling current in the auxiliary circuit, primitively the auto-transformer, results inverter suffers which renders showing the problems of this this scheme not due to the magnetizing current of practically applicable. structure are explained. The modified auto-transformer connection True-PWM-Pole problems by introducing a diode in current stress circuits. is series . Experimental ' inverter solves the above with each auxiliary switch. The auxiliary switch halved in the modified auto-transformer connection True-PWM-Pole Further reduction of the auto-transforrner voltage/current rating or the resonant inductor current rating are possible with 5kW prototype verifies these solutions. its variations. Also, experimental results from a 51 Appendix 2.1. Mathematical analysis of the ARCPI Refer to Fig. 2.3, Fig. 2.4 and Fig. 2.5, the commutation duration, peak current and quasi-rms current stress (over switching cycle) Í: iZ0 / Vdc, values used are: unit current stress can be mathematically analyzed. The base unit voltage; = v/ Vdc , unit time Í = two Throughout the analyses, the unit value of IMS, current is assumed to be 20%. l. Total commutation duration For diode to switch commutation, the total commutation duration is given by (2.lA): _' Í¿¿¡¬ 1 tl .¿. . Z + iboost (2.lA) +4[l10ad +lb00s¡] 2 2 For switch to diode cormnutation, the total commutation duration is given by (2.2A): - = 2a tsd T'_ .í" iboost - 2 GCOS __ -í cos~ -í +í Í tl + llboost 'l' - lload 1 2 + 4i¡,00s, (2.2A) (2.lA) and (2.2A) are represented in Fig. 2.lA. g 15.179 l2.834 10.489 tds 8.144 5.799 3.454 tsd “O9 Fig. 2. lA. Total o 0.5 1 _ 1.5 QE 2 2.5 3 commutation durations versus load current for diode to switch commutation and switch to diode commutation. . 52 2. Auxiliary device peak current stress V For diode to switch commutation, the auxiliary switch peak current :_~ lsapl = f- + lload is given by (2.3A) t--2 fl (2-3A) Z+lb00st For switch to diode commutation, the auxiliary switch peak current is given by (2.4A): _ _ __ isap2 = _iload + [1 Z + (íboost + iload)2 (2-4A) (2.3A) and (2.4A) are represented in Fig. 2.2A. 3.538 2.983 _ _ lsapl 2.438 1.888 _ 1.338 of/ss 1sap2. 0.239 o 0.5 1.5 1 2 3 2.5 iload peak currents versus load current for diode to switch commutation and switch to diode commutation. Fig. 2.2A. Auxiliary switch 3. Auxiliary switch RMS current stress É For diode to switch commutation, the auxiliary switch Í 2a coS__Ê?2~SÍ__ F › z~saP,ms z RMS current is given by (2.5A) 2 V? 2(íload +iboost) à --2 _-Hboost 1 _ _ \'_¡ Zfzâz +-É ó[ 1 _ _' __ (zm, + šsinz +z'¿,,,0s, zømâf (2.5A) 53 For switch to diode commutation, the auxiliary switch RMS current is given by (2.6A): .í'“ lsanrms ' = í i__+z'__ _+ (boost +1Ioad)2 1 2b00s'1“2'1 f 2¡ É load boost 2a cos I It dt + LTÍ 4 . I 0 Tífí . I .*"_1.° _ '" šsmt "' (lload +lboost)c05Í)dÍ (lload O (2.6A) (2.5A) (2.6A) are represented in Fig. 2.3A. l.26l I .054 T=45 \ 0.846 0.639 \ is prrns 0.432 / 0.224 1 0.017 0 0.5 san i Fig. 2.3A. Auxiliary switch s 1.5 1 T-90 2 2.5 3 load RMS currents versus load current for diode to switch commutation and switch to diode commutation. Resonant inductor average current during one switching cycle can be similarly obtained. case, By comparison the ARCPI is with Fig. 2.10, Fig. 2.11' and 2.12 regarding the True-PWM-Pole almost equal to the True-PWM-Pole in terms of the commutation durations and the auxiliary switch current stresses. 54 r Appendix 1. C,2 are 2.2. State equations for the modified True-PWM-Pole schemes For the case of Fig. 2.27(a), the state equations for the resonance between L,, C,, and given by (2.7A): Lr diLr 1+1‹ dz = Vdc 1+k _vC” (2.7A) 2Cr2 dvCr2 = T«Ê7z`_z1T` 2. . 'L' _ IL 1+_k For the case of Fig. 2.27(b), the state equations for the resonance between L,, C,, and H Cú are given by (2.8A): Ê_d'l__K¢1¿_vC2 _ 1+k 1+k dt 21‹C,z dvcrz 1+k From dz "W _. I' ku (2.sA) 1+k the above equations, the phase plane for each commutation resonance can be obtained which facilitates the mathematical analysis as well as further physical understanding. 55 Appendix 2.3. Analysis of the commutation resonance in the capacitor comection inverter (diode to switch commutation) I. z'L, S2 turned ofland Sa2 turned on, D2 current decreasíng until blocking V V * = °?-s1n(w,r) _ (2.9A) I' vc, where co, 2. _ z L, vcs É = V* ‹z‹›s(w,z) = (2.1oA) 1 , Z, É = [L Commutation resonance, and V the initial voltage of C,. . . . C C V _V** = IL á¿‹z‹›s(w,'z) +-°Ísm(w'z) +1L _ (2.11A) É: z (V0 -V ,,,, C c +c C2 )F”c‹›s(zz›,'f)-1LZ%-z,'sin(w,'f)-1L%f+V** Í' S CC C (2.12A) s C = (V0-V**)É‹z0s(zz›, z‹)-1,¿%3-sz, sm(w, :)+1Lä¿”:sz+V**+(I/O where Cs = Cs¡ //CS2, V 3. =I Ê, CC Cp = V" is the initial voltage of C,. iL, is CS2 charged and CS 1 discharged sv vc, .. ' _ ' -V ,, C HF: (2.13A) a›,'= Tí, l Z,'= L lá , and . IL is the load current, Lrƒully demagnetized, Cr charged =o=** V*** cos(a›,t) - Ts1n(co,t) _ (2.l4A) Í' i vc, =V *=|‹=|= cos(a›,t) where V*** is +I *=|==o‹ Z, sin(a),t) the initial voltage of C, and I*** Commutation from (2.l5A) is the initial current of L,. switch.to diode can be analogously analyzed. ' só Chapter 3. Operation of the Three Level Capacitor Clamping Inverter Abstract: This chapter deals with the major operation problems of the three level capacitor clamping inverter, including clamping voltage stability, clamping capacitor stress, and output spectmm 3.1. Experimental results from a half bridge inverter verify the analysis. etc. Basic Operation Problems A half bridge three level capacitor clamping inverter, as shown in Fig. two two-level switching cells, namely, C¡, C2, S1, 3.1, consists of S4 (cell 1) and Cm, Sz, S3 (cell 2), which can be treated separately in the modulation point of view. When the two cells share a given switching frequency, the control signal of one cell can be distinguished from the other by the duty cycle and the phase shift relationships, 0 Clamping voltage steady 0 Output voltage spectrum. 0 Clamping capacitor 0 Clamping voltage dynamic which will decide the next aspects of operation: state stability. stress. stability. n D1 51 _* C. °_i Í.~m=(d|-d2)i|.›zà o c, l °_i cm _ O-i v.../2 ¿d|×ih“ D, sz ` à ¿ à í._.› sz Ílflfld D. V-4 =SW 0-' S4 Fig. 3.1. 3.2. A ZSlv,¿.=sw,×v,,J2 _.__i Í ú,ס.,,,,, | ×V.¡,/2 D; Configuration of a half bridge three level capacitor clamping Clamping Voltage Steady State inverter. Stability Referring to Fig. 3.1, steady state stability of the clamping voltage cturent flowing through the clamping capacitor. Stability will is decided by the be guaranteed as far as this current can be averaged to zero during each switching cycle. When the load current is ' 57 supposed to be constant during a switching cycle, the average current flowing through the clamping capacitor during that switching cycle will be given by : ícm _ d2) (dl where d¡ . (3.1): ilnad and dz are the instantaneous duty cycle of cell 1 and cell 2 respectively. Obviously, i¢m=0 or d,=d,=d will ensure the steady state stability during the local switching cycle. 3.3. Output Voltage Spectrum It has been concluded that a set of control signals phase shifted by 21t/ p (p number of cells) is the an output voltage where the harmonics up to p times of the will produce switching frequency are zero [74]. In a two cell°s case, when vem is supposed to be equal to :›i VdÍÍ/2, VAO the output voltage _ -VS3 +Vs4 given by (3.2): Vdc Vdc _ u-T--SW2'T+SW1'T-*Ê where SW; and the is Vdc SW2 Vdc (3' 2 ) are the switching functions of cell down switch (S4 or S3) is l and cell 2. When SW; or SW2=l, in blocking state. wrrier I came' 2 m0d(t) = M sin(comt + 0) 11 É l d I * l OA ||«z.| im Í Vw 0V ......... .. Vó./2 i sw, 1 sw,¬ ‹ Fig. 3.2. Sub-harmonic J (mlgr |‹ g›£^¿^.S" PWM g First Sampling Cycle ‹ i “_“TT-' > Second Sampling Cycle modulation pattem of the main circuit and the associated waveforms. 58 Control signals fulfilling the conditions of equal duty cycle and generated by intersecting a sinusoidal modulating reference with two triangle caniers, as shown in Fig. 3.2. Assmning coordinate zero, the two natural sampling that the 1: 'rt phase and SW; can be phase shifted bipolar modulating reference PWM series SW¡ shift starts from the can be expressed in Fourier form as given by (3.3) and (3.4) respectively [l36][l37]: sw + + 1 2 w :uz S =:2,:4,.. m= sw 2 = :À '= 2 02-sin(mú› t-Í) _ "W C 2 M -í--sin(;)cos[(ma› C t+na› m I)-mn] ”'” 2 (33) ~ mM7r 2 ,M “--ícos(-.-)sín[(ma› t+nm '" t~)-mfr] C 2Jn 10° M NP/18 m=135... mM7r 2 (-1) Nnz [Vl J-*M8 2 2 ,N mz m+l2J -"-' 00 =l+£sin(ú› '” t)+ mM/r mf” ¢1,¬:3,.. lM'‹ =-+-nm 2 2 sl 2 mM/r """' +1 '" 1) Ê - m=1,3,5... ‹1›m22J°2 °‹ -í--smma) mff ° f --"› 2 mM7z' - 2J ¬:‹×> 5= M «zjms 2 T -"-_sin(flí-€)cos[(ma›ct+na›mt)-mrr] Ê» .n=¬¿2,¢4,.. "W 2 ( 3.4 ) mM7r + Í°° š= M +1 Í3 ,.. n=._ .NM8 , V-À where M is 2-Í" 2 y mn' ' cos(T)sin[(mwct + nwmt) - mør] the modulation index, (om is the sinusoidal modulating reference frequency, co C is Mn (L) 2 the triangle carrier frequency, and J n of argument LI*/Íl'_. 2 ' is M n order Substitution from (3.4) and (3.3) in (3.2), the output voltage of the half bridge three level capacitor clamping inverter “ the Bessel function of the first kind and given by is (3.5): mMn' VAO = Vdc -í sm(a›mt) + Vdç . w iw m=š4," nzišâw 2_1 n 2 ----mn W, cos(T)s1n[(ma›ct + nwmt) - mn] (35) _ 59 From (3.5), the following conclusions are obtained regarding the output spectrum under phase shifting condition: 'rc The tríangle carrier fiequency components 'together with 0 its cross modulation harmonics are canceled. Among the cross modulation harmonícsfamílíes, m=I, 3, 5..families are canceled. No even harmonics exist in the spectrum. 0 0 3.4. Clamping Capacitor Stress When control signals for the two cells are with equal duty cycle and are 1: ' the phase shifted, RMS current stress through the clamping capacitor will be dependent on the load current together with the duty cycle of the local switching cycle, which z¿,,',_,,,,,(f› z'c,,,_,,,,s‹z› = = z',,,,,d‹z›\/2p‹1-‹f‹z›› ›',,,¿,,,,‹f›«/2‹1‹z›; ; is given by (3.6) and (3.7): when â>o.5 (3.ó) when â<o.s (3.7) where icm_¡.mS(t) is the quasi-instantaneous which is solved over switching cycle, RMS current stress through the clamping capacitor and d(t) is the duty cycle of the two cells. Moreover: dm = lif%W (3.s) So that substitution for (3.6) and (3.7) yields (3.9): _ icmƒmsm = iload(t)V 1 'mod (UI Eq. (3.9) is represented in Fig. 3.3. regards the clamping capacitor stress: 0 0 (3 9) From which, the following conclusions are » RMS current equals the peak load current. Maximum RMS current occurs when pure reactive power is transferred. The maximum _ Similarly, the voltage ripple across the clamping capacitor is obtained in (3.l0): drawn as 60 Avcm (t) = Tfzmzd ff) Cm where Avcm(t) is (1-Im‹›d (3-10) ‹f›|› 2 the voltage ripple amplitude. 1 vç | 0.9 - ,' cm.rms (t) _%__._í. Í 0.7 __ ' s Í Í `~ \~ `\ \ › \ \ load (Í) \ 06 ._ 0.5 O4 \ M=0.6 o \ 3.5. \\ f / ~{ I / / I _. ¡ `~--' . _ / _ I I 0.004 o.ooõ o.oos 0,01 t stress related to load current and load power factor. Stability of the clamping voltage to retum to value after a perturbation with no specific control over Assume ` / Í 0.002 stability refers to the ability 3.5.1. Self-Balancing ;` / ' =3 = 1,,sin(wmr), mod(t) = Msin(w,,,t + 9) , C0m=2TC×Õ0- Clamping Voltage Dynamic Dynamic `\-f /' GD I Clamping capacitor current ¡,md(¡) \ _ -' \ ` \ I o Fig. 3.3. / \ \ ~ \ \ I _ ` \ Í \ 'Í ` " _/' \' I i \ ` - // ` `\ //' os 9=1t/2 ' z' z | ,/ \ z ////I | | \\ it, i.e., its nominal the self-balancing ability. under Ideal Condition that for some reason, expected or unexpected, typically during the start-up process, the clamping voltage arrives at a value other than Vdc/2, then the output voltage can be expressed by VAO (3.1 1): V, 4 V, :sw2vcn1+swi(_š__vcm)_% which can be rearranged as (3.l2): 61 VC = %‹~‹»1 Hwz -1›+‹S»1 VC -~‹wz›‹% The second part of (3.12) variation. Substitution from mz) represents the output voltage response to the clamping voltage and (3.3) (4.4) in (3.12), the output voltage variation is given (3. 1 3): - ---v °"' ›{ Av A” = ‹Vd° 2 s= fims 'M É QM m=1,3,5... ‹ M 2 mff -W 4.1" 2 n=i2,i4," Assume a mMfr fi - 1›74"° iísm'‹ ma) WH' 1 :oo + m,, s1n(-2-) cos[(ma›ct + Vdc Azload = (-2- v Cm){ _ _ +1 ‹×> É m=1,3,5... (-1) - 5= ÇM8 'M E= li; When z --”› 2 (313) nwmt) - m1r]} mfi __ mM/z' - -í-sm(ma›ct É âmwc) mfi 1 zz _ zm co C (3.14) ~ 4-7,, 1¬â“§ (3. 14): ¿"~2-410 2 _ -"lg-4-'í p C load impedance of zj¿9j at the harmonic frequency of j, the corresponding load current variation can be written by + 1 Z sm( _ mwc +nwm mir this load current variation is 2 )cos[(maJct + nwmt) - mn - 9mwc_¡_nwm ]} reflected back to the clamping capacitor, the resultant charging current variation will be given by (3.l5): . _ _ Aim = Aim, (sw¡ - swz) (3 Substitution from (3.3) (3.4) and (3.14) in (3.'l5), the result of substantially complicated expression when Aicm by and the clampingvoltage balancing is of interest and this fact is may which . 1 5) will be a not allow for any physical insight. However, particularly concemed, only the simplifies the task. The current variation can finally be written by (3.16): DC DC component in component in the charging 62 . AlClVLdL' Vdc = -vem (3- where: G=5 5 = +_ 2 â= __” 2 m+l ÍÇÂM8 {[‹-1) Y' 2 %12ícos‹0mwc›} 1 C mM” 4-Ín 2 Í; ¡;Ms This mM1r Y' n=fl,i4,"{[ ' Sm( mn, mfr 2 2 )l (317) l zmwcfluom °°5( 0ma›c+na›m )} DC charging current can be related to the clamping voltage variation by (3.18): (7 ' WG _ °`m 'az' dvcm Vdc (3-18) The clamping voltage dynamics is given in the end by (3. 19): _CmG __1_, Vc Vc um = -5'-+mm‹°›-%1@ where (119) vem (0) denotes the initial value clamping capacitor and Cm/G The self-balancing is of the clamping voltage, Cm is the capacitance of the the time constant for the first order dynamics. mechanism analyzed above is illustrated in Fig. 3.4. perturbation Avcm, the output voltage will vary correspondingly (AvA0) current variation Aimd. The clamping capacitor will give original perturbation which For a given results in load resulted charging/discharging current variation Aicm for the rise to voltage and the clamping voltage component AAvcm, which tends is finally maintained at its to cancel the nominal value. The foregoing discussion supports the next conclusions with regard to the self-balancing ability under ideal condition: 63 Al/cm AV0”. =0 + sw] _sw2 âí Al/ao Clamping voltage .I mem ~ sw! _;sw2 1 1' Fig. 3.4. Self-balancing ' minar] l mechanism of the three AAVcm CL J'Al.cm M I + l ' level capacitor clamping + Ma.. AVC!!! inverter. components of odd multiples of the triangle carrier drift leads to frequency and their cross-modulation harmonics at the output, according to (3.I3), which essentially determines the self-balancíng ability. Clamping voltage dynamic 0 G . will always be positive if 0 ensure the clamping voltage with normal inverter load. 0 stability depends' is i.e., G expressed by (3. I 7). non-pure-inductive/capacitive load will regardless of its initial conditions. This stability, is satisfied V Clamping voltage response dynamics ¢ in/ 2, on the sign of constant exhibits typical one order performance. The duration of the dependent onthe time constant -(Cm/G) of this dynamics, which is further decided by: clamping capacitance, modulation index and load properties. 3.5.2. Self-Balancing under Non-ideal Condition Practical circuit contains various non-ideal conditions [138] asymmetrical operation of the two cells which and the clamping voltage potentially lead to may drift as a result. Among them, deviation from ideal trigger timing deserves the most treatment. In an extreme case, duty cycle of one switching cell one load current direction, and is is always higher than the other for always lower for the other direction. Both tend to over- charge (discharge) the clamping capacitor. However, while the clamping capacitor being over-charged (discharged), a discharging (charging) current arises simultaneously due to the self-balancíng mechanism counteracting of the asymmetry, there the net current cycle. flow exists to the This conclusion is the current of the first case. a new operation point other than clamping capacitor verified by PSPICE is averaged Dependent on the to zero the extent nominal value at which during each switching simulation, the results of which are shown in Fig. 3.5(a)(b). For this simulation, the half bridge three level capacitor clamping inverter by sub-harmonic Table 3.1. is modulated PWM pattem, as shown in Fig. 3.2. Parameters/specifications are shown in Duty cycle difference for the case of Fig. 3.5(a) is 0.01 (-0.01) and for the case of 64 Fig. 3.5(b) is -0.01 (0.01). after the first order Table The clamping voltage dynamics. is balanced Vdc=800V voltage Clamping C,,,=680uF respectively of self-balancing under non-ideal condition Output L¡=0.lmH Modulation Filter C,=4uF index Operating fC=l0kI~Iz Fundamental M=0.8 R0=1Q Load resistance fm=60I-Iz Frequency frequency capacitance 560V and 240V PSPICE program for this simulation is listed in Appendix 3.1. 3.1. Circuit parameters/specifications for simulation Input at Initial clamping Vcm(0)=4()() V voltage Besides the extreme case discussed above, dead time causes no clamping voltage because it is symmetrical for the two cells. Dead time effect for one cell is canceled by the same effect for the other cell during the switching cycle. 'nnflnnl oonlmlafion Dlhlfm nm: 021088! 092521 Tompoulurozilfi 600V SWV 520V 050V Vem “OV 400V 380V (a) Om: 2 u V(\1.l D) Duty cycle 2 SMB I00r|B 150m! 200!!! fm difference 0.01 for the positive current, -0.01 for the negative current 'nnfiiwul oofrtnlhfion Datnlfm nn: 0M.lM_l.Iâl.'ãJ WV ; lmmtnlliilfl MV NOV W Vem 2BOV 2l0V 200V (b) om: Duty cycle Fig. 3.5. 100m: 50m: zz v(w,|s) drift \50ma 200m9 Tm difference ~0.01 for the positive current, 0.0l_ for the negative current Clamping voltage self-balancing under non-ideal operation conditions. 65 _ 3.5.3. Self-Balancing in Three Phase System For a three phase system, as shown in Fig. 3.6, generally the same procedure in sub- section 3.5.1 can be followed. For any initial deviations Avzmz,cm1›,zmz of the three clamping voltages, the corresponding load current variation in each individual phase which is then refiected back to each clamping capacitor detennining its must be solved charging/discharging current variation. For this purpose, the corresponding voltage variation between the neutral points O and O” is of the principal importance. When three sinusoidal modulating the three phases intersecting the same waves with carrier two _ 21:/3 mutual phase shift are applied for wave, the resulted switching functions can be expressed by (3.20): 5-`‹zš` «Í-`|zš` Cn _.__.v. L Ãnz OZ cm l 5-`4z:°` Ãnz osll l Cmh_ u 5% t zf CM ZÉ Í-1 ÃÃI” °” osii Íll ‹Í|ÃD` @@ C Zgnz  5125 0. Fig. 3.6. (sw,-sw,)u_h_L,= + i m = where ."l"18 5° P' Three phase three level capacitor clamping inverter system. 4» m = FMS oo S = ÉÉ ,i 4 ‹pa,¡,,¢=0, -21:/3, 5” P' 7(-I) f" m+1 2 mM1r JO 2 fr sm(ma›ct-3) _ ' ¶sín(";l)cos[( mcoct + ncomt) - (320) 'K 21:/3 for phase A, between each phase terminal and the B nçou_,,_C] and C. From (3.20) and (3.12), voltage variation DC link neutral potential is given by (3.2l): s 66 AVAO BO,CO _ SW2 )a,b,c Avcmmcmb cmc = Afterwards voltage variation between the two neutral points Av,.,, is expressed by (3.22): = %(Av,0 + Avg, + Avco) From (3-22) (3.2l) and (3.22), load voltage variation of each individual phase can be obtained and the same procedure in sub-section 3.5.1 can then be followed for the reflected charging/discharging current variation of each clamping capacitor. This procedure detailed here not and only the conclusions are given as following: When clamping I. is voltages in three phases are not equal (say, perturbation of clamping voltage in one phase during normal operation), the odd multiples of the triangular carrier frequency and their cross modulation harmonics enforce the self-balancing, which similar to the half bridge case. The three clamping voltages willfirstly get equal move 2. to the nominal value synchronously. When clamping is and then 4 voltages in three phases are equal (say, start up process), the cross modulation harmonics of the odd multiples of the triangular carrier frequency (odd multiples of the triangular carrierƒrequency are canceled) enforce the self-balancing. The three clamping voltages 3.6. move to the nominal value directly and synchronously. Experimental Verification Basic experimental verification of the self-balancing ability scale prototype. Refer to Fig. 3.1, the prototype parameters Table 3.2, and the results are shown in Fig. and the clamping voltage clamping voltage is is 3.7. monitored. conducted on a small and specifications are given in For Fig. 3.7(a) monitored. For Fig. 3.7(c) is (d), the (b), the DC DC supply is opened supply is closed andthe F Table 3.2. Prototype parameters/specifications for experimental verification of the self-balancing Input _ _ Vdc=200V Output Filter voltage Clamping capacitance Cm=l650uF Operating frequency L,-=l .45mH Modulation C¡= 1 2uF index f¢=6.5kHz Fundamental frequency M=0.62 fm=50Hz 67 From the experimental results responds to the overshoot DC shown both cases, the clamping voltage supply voltage change with first order characteristic and no resonance or caused. However, longer time is in Fig. 3.7, in taken to reach steady state is when the load is step-down or step-up, which means that the time constant of the dynamics lighter for either runs inversely proportional to the load resistance. This result is in line with (3.19). 1 Vdc '_"-›¬_T §Y‹¶= Í štbttr zl* 9/ 1 Vc 1 Í §I§.f$\7""t3|fT9"h _ (a) ivete 1 ' š‹)()v"`(ñ?¶›‹)vLM "f×7l'ãUi›%ris Ro=3Q (b) 1.õí›`s Ro=7.5Q tttttt p ,mygznldm fr'.i._JE.'‹=z_.. e ~ " i z 1 sp ; .é .; É i i Ê ~ s tm. T'1F\¬I` So__.U._V..¡.Cñ2..SÍ_.' (c) Ro=3Q (d) Fig. 3.7. Basic experimental verification of the half bridge three 3.7. 0 É fl'í“‹ív"“f'fi`)"§r`1`^l›'\`/"ñF<l'Yn|`~ri<i"` _ Ro=7.5Q of the self-balancing level capacitor clamping ability inverter. Conclusions Gating relationship of the two cells clamping capacitor stress is associated with the clamping voltage stability, and the output Spectrum for steady state stability, phase shift implies a etc. While equal duty cycle is required compromise between clamping capacitor stress and output Spectrum. With 1: phase shift, the output spectrum is optimal and maximum RMS current of the load current value happens when the load is pure reactive. Clamping voltage load. is self-balancing under the given gating pattern with non-pure reactive Time constant of the first order dynamics is dependent on clamping capacitance, modulation index and the load properties. Dependent on the extent of non-ideals, typically resulted gating or switching mismatches, clamping voltage the self-balancing is balanced to a new operation point other than the nominal value, due to mechanism counteracting such non-ideals. Appendix 3.1. PSPICE program for simulation under non-ideal situation *Self-balancing under non-ideal situation * pulse pattem generation v1 10 pulse (-10 10 .lu 50u 50u .lu 100u) e2 2 0 value {-v(l)} vmod 5 O sin(0 8 60 0 0 0) vdri 42 0 pulse (0.2 -0.2 8.34m 0.1u 0.lu 8.34m l6.8m) * duty cycle difference esul 6 0 table {v(1)-v(5)+v(42)} (-le-2,10)(1ei-2,0) esdl 10 0 table {v(6,0)} (2,l0) (3,0) esu2 7 0 table {v(2,5)} (-le-2,10) (le-2,0) es_d2 ll 0 table {v(7,0)} (2,10)(3,0) *main ` circuit su3 16 24 6 O sswitch ds3 24 17 dmod du3 17 16 dmod su4 17 25 7 0 sswitch dsu4 25 18 dmod du4 18 17 dmod sdl 18 26 11 0 sswitch dsdl 26 19 dmod ddl 1918 dmod sd2 19 27 10 0 sswitch dsd2 27 20 dmod dd2 20 19 dmod *device model .model sswitch vswitch(ron=0.01 roff=1e6 von=4 voff=2) .model dmod d(is=2e-15 bv=1500 tt=0 cjo=lp) . *DC link supply vc2 16 40 400 vc22 40 20 400 *flying capacitance c3 17 19 680u ic=400 *load and filter lload 18 41 0.lmH rload 41 40 cload 41 40 1 4uF l0u 600m 0 l0u uic .options vntol=l0m abstol=10m reltol=0.1 itl5=0 .probe i(lload) i(c3)'v(17,19) .end .tran 1 70 Chapter 4. True-PWM-Pole Three Level Capacitor Clamping Inverter Abstract: This chapter proposes a transformer connection True-PWM-Pole three level capacitor clamping inverter. Besides the detailed description of the circuit configuration and operation, the specific features of this configuration are highlighted. Several topology variations for multilevel case are illustrated shortly in the end. 4.1. Circuit Configuration The proposed circuit is shown in Fig. 4.1. It consists capacitor clamping inverter circuit and SM) assists the cell (S2, S3) two auxiliary branches. The first auxiliary branch (SM, commutation of the first switching» whereas the second auxiliary branch of a main half bridge three level (S32, Sá) and fonns the first pole; cell (S,, S4) assists the commutation of the second switching and forms the second pole. The two poles can be regarded independent from each other so far as the clamping capacitor voltage is maintained at Vdc/2. I Ê ir >tP ‹_ C| nn S2 ic _à D_ 14 » iuu N¡ Õ DID”. ~ Dsal Sal N2 Í-‹rl4 O Dal H +0 R C2 í- DS34 SM *I Cm 5:13 s:i3 Daz Das .NZ N| H Saz Dmz Da2 _ 0.! ` D2 Cn Ã_ _ A [423 ‹í . Da3 . lload i¡_¡¶3 ,°¬|ÃS] D3 Cr] i °i S4 Ã: Dó Cú Í Fig. 4.1. The proposed True-PWM-Pole three level capacitor clamping inverter. 71 4.2. Circuit Operation Prior to discussion of the commutation process, the following conditions are assumed: im is flowing and remains constant during the cornmutation. 0 Positive load current 0 Capacitors C, and C2 are treated as voltage sources during the commutation. Clamping capacitor voltage 0 is stabilized at Vdc/2. Transfonner ratios are rail level in set to ensure sufficient energy for the pole voltage to swing to the the presence of commutation losses, as discussed in Chapter 2. 0 Circuit parasitics, device switching transients and transformer imperfections are neglected. 0 Main switch tum-off and auxiliary switch turn-on happen the modulation scheme discussed at the in Chapter 3, while the ,mauxiliary mm ,,¡;.,... . _ Q, . zz =雋‹»,, universally constant covering the - -z .¬ ¬. .- Conduction intewal of the maximum commutation the commutation step diagrams for the first sampling cycle N«vc from duration as Refer to the theoretical waveforms for the entire switching cycle shown in Fig. 4.2, and ~ \ ¬*~ is instant decided .gdiscussed in Chapter 2. .,_, _ ,,..-¬,._ _,,,,,,,... switch same main switch tum-on happens until the detected voltage across the switch declining to zero. . ~ commutation of the half bridge steps: in Fig. 4.3, the inverter during the first sampling 'cycle consists of the next _ Step I (to-t1): Circuit steady state. clamping capacitor Cm. Output Voltage Step 2 and Cú, C,3. (t 1-tz): Step 3 N, sees a voltage of (t2-t3) : is Step 4 (t3-t5): Step 5 kV,,c/2 after state. iL,23 may tum-on of S,2 and conduction of D,2 and cause oscillation due to its S3 is tumed on at forward recovery as well as of the paths. falls to zero at t, D4 and D, carry load S4 D,2°. _ allowing for tLn'n-off of S,2 at current. VA0=-Vdc/2. (t5-t6): the on at tl, which starts a resonance between Lú, vw rises to Vdc/2 at tz leading to conduction of D3. Then stray/internal inductances output voltage S, carry load current discharging V A0=0. discharged. zero voltage. Snap on of D3 another steady D4 and S2 turned off and Sa, tLu°ned Cú is charged and Cú ` shown is turned off and S,4 is t4. Clamping capacitor Cm tumed on at ts, Circuit reaches is floating. And leading to conduction of Da, and DM”. Nz” sees a voltage of kVdc/2 which joins Vdc/2 enforcing current decreasing in D4 . 72 Step 6 (t6-t7): im, rises to the load current level at Resonance among LM, CM and CM C,, is initiated. t6 leading to blocking of D4. charged while is C,, current of D4 enhances the charging current and therefore facilitates the Step 7 (t7-tg): (tg-t9): commutation process. vw rises to Vdc/2 at t7 leading to conduction of D1. Then S, is turned on at zero voltage. Rapid cmrent transfer from Cp, and C,, to D1 Step 8 discharged. Recovery im, falls to load current at ts. may cause oscillation. D1 stops conduction and S, starts carrying C the load current. Step 9 (t9-t]1): im, extinguishes at another steady voltage state. S, t., allowing for tum-off of SM at tm. Circuit reaches and D3 carry load current charging the clamping capacitor Cm. Output V A0=0. In addition, the remaining two commutations in the second sampling cycle (D3 to S2 and S, to D4) can be analogously inferred. 'I_`o summarize, for the proposed « . circuit, the main switch work with zero voltage turn-on and capacitive tum-off, whereas the main freewheeling diode work with zero voltage turn-on and zero current the turn-off. Superior to the conventional snubber, the capacitive tum-off loss of main switch can be rninimized by optimizing the resonant capacitor to this end. Meanwhile,-depending on its forward recovery property, the snap-on of the main freewheeling diode does not introduce any considerable Moreover, all the -auxiliary devices loss. work with zero current tum-off. And yet voltage is only reapplied after tum-on of the opposite auxiliary switch. Besides, despite of the bridge configuration, the tum-on of the auxiliary switch is actually since the opposite freewheeling diode carries recovery is negligible. Transformer excitation snubbed by the resonant inductor no current beforehand and thus is reset to zero after each its reverse commutation and no magnetic accumulation can happen. In particular, by designing a transformer ratio less than 1/2, an auxiliary voltage of (1k)Vdc/2 higher than Vdc/4 pole to swing to the is becomes rail voltage. no longer necessary. available which delivers sufficient energy for the resonant Thus the “boost” stage and the associated control complexity ' Designing aspects of the two poles have been discussed in Chapter implementation of a 3kWlprototype .will be presented in Chapter 5. 2. Detailed À 4 l l ii 55mP¡¡"8 °Y°|@ TÍ2 Flfsl eo I Sa; _ Sa: Second sampling cycle T/2 --.my l . ”l. ¡ s,_'| fl . , CID 5.1 ÍÇIIT” l /X Vu: il/Z] VSa: v.,,¢z A I i I 1 i 5 _`i"`_"' V¢J2 Rr. Sa, ¡z;í.í._-..._..._-.ê 1 E Í l "rir-ln' ` V ¡__ B TI. VC.. ,~' - . .-6»-11 Ç-le...-1.-___¬_ ¢1E"`:^›\-* .z*2.~r‹=~_- ..z.zr=. ~: VSa‹ .âfíà ›. ›‹ vw* TP* 'fi «. ' z 10 51°? 12 13 11 lí 14 uu 0 ilm . E vdm 'í ~ › i V¿,/2 w zu "-1 v.,./2 16 s 19110 11 1a el i 7 ‹. w Cm ‹- M4 . É' 1 S rm A q S15' cn D, -_ ID sz °l A Cn . uu in VM 'I' su 'E' › ‹. . . ~ o zw DI Ill 2: ¡Í'¡1 : °-I 113 114 11s 111 11s 119120 116 ç 'í' ‹- sn 1 Slcp A 1- › hm VM '_' ' i O uu *I 4- W DI SI 3: lz-r; ` V _...,z ___› - ru I Gl : ¡ __ D, cn . su Cm A °-I cu level capacitor a entire switching cycle. 4L. ¡ _; __¡g“Í su F GI : in -x°“" ~ 1112 111 commutation wavefonns for the proposed True-PWM-Pole three DI SI “HI 4- :Vdm l 5 5 clamping inverter . l z _ Í l Fig. 4.2. Theoretical D" Vúz/2 I] Vw u_u . E "_" Íuu vz: . l- ‹Í~Í3À^' l i fíeâ J._...... Sa. lload = (X A : ¡ : ‹ - sn ':' su Cm 'Z' 1-_ ¬ --› I' ' ‹. .› . su U1 mz :: cn sz I ‹ : › ou 74 S.” *u_u H + .xm D: sn A cn : + › '> uu m ' ; O‹ É vdcn cn - D.. ' ~l A, V °l sà _ ‹ 04 ‹- ›~ um vaca í _ _í IIII S tcp WM D" .‹, ,y _ . Cn ` : . cn na . - -›A l °l bu À _ Vaca ml _ tú 4.3.1. cn D, › . .. r i F ui Dn sn Sm, 9_l9_l“ ~ ‹^ ‹- _ of _ ,› g ‹_ . CT! _ - _ . ou Di 1 I - eu A -› V.z Ill _ VM2` Išfl _ :Eu ' 1 Cn D7 un O . DI' ‹, ,›¡ ' . _-' ` Í . |›z , í' -›A cn | ‹ÀD :LI'z °i 84 z Dc : (.'fz diagrams for the first sampling cycle. Stabilization . _ self-balancing ability of the clamping voltage under the given modulation pattem verified in Chapter 3 variation A›' ' Features Clamping Voltage The Dl! __ Dl _ I-ë-Í : |›‹ is strongly subjected to load properties. Destructive clamping voltage may arise due to such reasons as load change (load open, reactive load for example) or triggering failure etc. Active control of the clamping voltage necessitates detection of the clamping voltage as well as the load current direction, which tends to complicate the system. The proposed True-PWM-Pole three level capacitor clamping inverter posses charging/discharging paths for the clamping capacitor whenever perturbation occurs. the clamping voltage gets below the nominal value, it will be charged When by the up-capacitor Dm, N2” and LM when S, is gated, as shown in Fig. 4.4(a); or by the downthrough S4, Dsaland LM when S4 is gated, as shown in Fig. 4.4(b). In the (Vdc/2) through S,, capacitor (Vdc/2) meantime, when the clamping voltage gets above the nominal value, the up-capacitor (Vdc/2) through D,, Lr,,,, A ' _ "IU Fig. 4.3. Operation step 4.3. Specific ‹A Nf, _ _ ¿ Du mz O cn z , A Vlkn _ °-I nl Q- __ Q um E ' Nf, ln: cu › si llzlg-(9 _ 'Ê' __ F cn _ Dl ' Dn g _ ` Nf, O __ cn A ' ' A 0,2, 34 D: un _ i‹. ›› O um _ cn '__ _ H + .VM m m S' °° 6:' " 47 'ri D: sl 1 m Nf, Cu 4 1 S! CP 7 Í7'8 Í cn : Du ' Lm O A 1 É DJ__cn ! Jaú/1 A _ nm um H ' .W m s| smp 5:6." Nƒand S34 when it Sw, is gated, as will be discharged to shown in Fig. 4.4(a); 75 or to the down-capacitor (V de/2) through Sal, N2”, LM and D4 when S,, is gated, as shown in Fig. 4.4(b). In other words, the charging paths arealways existing except for the gating space between and S4. However, the discharging paths become available only when the auxiliary switches Sa, or S,,, are gated. interval S, Taking into account the magnetic losses and the conduction losses in the damped second charging/discharging paths, the charging/discharging processes are a order resonance involving the clamping capacitance, resonant inductor and the auxiliary transformer, which guarantees the clamping voltage to be stabilized at Vdc./2. F ' o A * _ D: Ífl rm ÍII1 15,13%' JM.. V › 5%W . . ~ - ...,, ^ o *“ uu fi Cn : ln iu ¡ cn _VM... . . › . _ ‹ , ! . . A -P iu-u :- rt,‹¡D, cn ‹ - (a) (b) Fig. 4.4. Charging (discharging) paths for the V ‹A ~ -1a On M: um + _w‹f› . Ê _? E DI 8: : H + .am Q» DI 8: V Clamping voltage through the auxiliary circuitry. This feature in clamping voltage stabilization transformer connection True-PWM-Pole circuits is is lost when any of the modified auto- employed. The unidirectional auxiliary switch offers only discharging paths preventing the clamping voltage from being higher than the nominal value. Note that when two storage capacitors are used in the DC side, this feature remains tme only when the neutral potential is stable. 4.3.2. Normal Auto-Transformer Connection of the Second Pole As discussed in Chapter 2, even though the normal auto-transformer connection True- PWM-Pole promises attractive properties in terms of the auxiliary switch current stress, auto- transformer primary voltage stress and the auto-transformer secondary current stress, the freewheeling current flowing in the auxiliary circuitry during non-commutation interval renders it not applicable andthe transformer connection is preferred. transformer connections need extra snubbing for the series diode. The 'modified auto- 76 For the proposed three level inverter,-the second pole The level structure. can expect to first pole (S2, S4, S3, and S34), and a normal S32) is modified auto-transformer connection two-level structure and transformer connection or inevitable. (S2, S3, S32 however, is quite distinctive is from a nonnal two- Because of the existence of the clamping capacitance, no freewheeling current flow through either the auxiliary switch or the auxiliary diode, if the normal autoz-transformer connection is used. As shown current in Fig. 4.5, suppose that after the flows through Lr,4, N2°, S34, C3, commutation for tum-off of S4, a residual and D,. The clamping capacitance C3, will then get discharged and a voltage difference between C, and C3, will arise, which tends to resist the residual current until it Such being the returns zero. phenomenon discussed case, the specific configuration. in Chapter 2 exists in this And the normal auto-transformer connection True-PWM-Pole becomes practically applicable for the first pole. This alternative with essentially the same current stress no longer 'result as was the transforrner connection except which is nearly halved. __ D1lD Í-ru ¡L¡H sal 1: al í 5:13 .NI +‹› DSM SM D¡¡z°.l D3; 5:13 C"' N2 D3, °_l SH S C2 the auxiliary switch for. Dz c,z s -fi prototype 5? Dl? ‹_ --C1 “_ O 3kW ' -T-z= à also tested in a .Nz NI H S32 Dsflz al Lm Ã: A 4- - . .13 › Iload I|_¡¡3 _ ,°lZ§: Cú D3 S3 D °-l S4 : .ZS D4 Cm 'Í Fig. 4.5. Modified configuration of the True-PWM-Pole three level capacitor clamping inverter, with transformer connection for the second pole and normal auto-transformer connection for the first pole. 4.4. Topologies for Multilevel case (M>3) Transformer connection True-PWM-Pole zero voltage switching scheme can be extended to multilevel case (M>3). Fig. 4.6 demonstrates a four level True-PWM-Pole capacitor clamping inverter, where S23, S,3” and S1, Sf; S23, S23” and S2, S2”; S33, S33°and S3, S3” I :S1 S W I 'IS N1' ,I I il .__ Sm ° No. NO O . A ll ll A .l A A A Szwl :s, .__ s, i I NZ' S21: 0-I S; I Sla ›z›~`-_,.z J ¬«p,.....~. z t. ¡- cóz. IS¡ 0-I C U Fig. 4.6. ., Transformer connection True-PWM-pole zero voltage switching scheme extended to multilevel capacitor _ clamping inverter (M>3), example of four level case e .¬ 1; 1- S13' sla .Í Í. _ z S23' S2; .Í _Y_ .L S33' S35 Í. Í. T .~ L Fig. 4.7. ^| ARCPI zero voltage switching scheme extended to multilevel capacitor clamping inverter (M>3), example of four level case. 78 form three poles. Each pole is independent- of the others provided that the clarnping capacitor voltages are stable. The operation of each pole is the same as the normal transformer connection True- PWM-Pole except for the separate inductor and discrete auxiliary switch. The auxiliary blocks .the same voltage as the main switch which enhances In addition, the Sb' and S2, Sj; S3,, Q (ARCPI) is also A four level case is shown in Fig. 4.7, where S,,, Sm” and S3,”and S3, S3' auxiliary switches block half of the 4.5. applicability. Auxiliary-Resonant-Comrnutated-Pole-Inverter extensible to multilevel case (M>3). S,, S,'; Sza, its switch form three poles each work independently. All main switch voltage. Conclusions The proposed transformer connection True-PWM-Pole inverter achieves zero voltage switching three level capacitor clamping of the main devices and zero current switching of the auxiliary devices, withoutprovoking any voltage/current spikes across the main or the auxiliary switches, without yet requiring any additional current monitoring or controlling. 0 The proposalfeatures a interesting second merit in stabilizing the the existence of the charging/discharging paths established by the clamping voltage, due to auxiliary devices for the clamping capacitor. In the meanwhile, due to the clarnping capacitor involved in the resonant current paths, the normal auto-transformer connection True-PWM-Pole becomes applicable in this case without the problems of the freewheeling current and the residual current, which achieves the same property as the modified auto-transformer cormection True-PWM-Pole without any voltage ø spike across the auxiliary switch. Both transforrner connection True-PWM-Pole and multilevel case PWM-Pole ARCPI (M>3) with the same performance ARCPI schemes are extensible to as in the two-level case. The True- extension requires considerable magnetics and diodes, while in the case of extension, center-taped storage capacitors must be used and additional monitoring and controlling become necessary. 79 . Chapter 5. Designing and Experimentation of the True- PWM-Pole Three Level Capacitor Clamping Inverter IGBT Hardware and relevant experimental details results are presented. Prototype Specifications A scaled laboratory prototype, as shown in Fig. 5.1, has been of which are shown in Table ~,-e .xi ,...,.. ` The specifications Table 5.1. Prototype specifications a ~› set up. 5.1. ..¿.- .fz D.C input voltage Vdc=700V Output voltage V‹,|,,,,,=l40V Modulation index M=0.62 Output power P°=3kW Load current Im,,=2 1 .SA Switching frequency fc=6.5kHz Í ..z_ .‹.z›z.›'- half bridge transformer connection True_-PWM-Pole three level capacitor clamping inverter prototype. 5.1. 3kW 700V supply Abstract: This chapter addresses a proof-of-concept .- «tm-.› -.z.~. I " sr :_ C.. ¡>'9 ‹_ C1 ao Du4 ic _' O LH4 Dszl Sm NZ' _; 'UN ' cz °-l Cm Szs as DM Da3 IN2 Ni-0 Dal › D2I| P* Nr H Dm SM H Szz mz Daz' Dí\2°‹' Ã: A Lai, ‹_. ima D¡,3`°-I S3 °l S4 Fig. 5.1. Circuit Lt -_-P __ 'IM-1 à _- CÍ "›¡o Vu ¿ D3 Cfl _ . C4 Dz S fl Ã: D4 Ca Diagram of the 3kW half bridge laboratory prototype. Ro 80 ' 5.2. Main 5.2.1. Circuit Designing Power Supply Three 300V/5A is DC supplies are associated in series. An iron laminated l5mH inductor put in series with the supply to suppress the ripple current. 5.2.2. DC link Capacitor The designing i capacitance is criterion in the laboratory case is the given by DC voltage ripple across (5.1): 1 ( 5.1 ) AVdc T the switching period and Ip where AVdc is peak When maximum AVdc is set to be 3V, this gives C,=C2= current. the peak to peak voltage ripple, In practice, the between the The . TI c = it. DC utility side is is the load 1536uF. link capacitance should be designed for transience de-coupling and the load Depending on the configuration of the rectifier side side. and the control of the whole system, the result can be quite different. The capacitors C,=C2=3300uF, each with a nominal voltage of 350V. in use are _ H 5.2.3. Clamping Capacitor Sizing of this capacitor V is two-fold. voltage ripple, according to (3.10). On the one hand, higher capacitance allows for less And on the other hand, smaller capacitance reduces the clamping voltage transience duration, according to (3.19). Voltage ripple should be the major designing criterion in practice. Further, capacitor intemal inductance and in particular capacity to process the The heat RMS current given in (3.9) must be taken into full account. actual capacitance used is 2750uF consisting of two 5500uF capacitors each with a nominal voltage of 250V. 5.2.4. its in series - Output Filter The primary designing inverter' objective is to output reduced to a given extent. the dominant output voltage harmonic is have the dominant harmonic component From (3.5), for the at the given modulation index of 0.62, calculated to be 0.35 of the fimdamental component. T To keep the THD less than 5%, With a second order low-pass mb = ,/Lfcf where fc is 81 the dominant hamionic filter, the cut-off frequency (ob to be less than can hence be decided by = 2fi(2fc _ fm”0(2o1go.o3-2o1go.3s)/ao the switching frequency and fm Another designing objective - component need is to is 3%. (5.2): (52) output fundamental frequency. avoid possible resonance at the output, especially in a closed loop system. For this purpose, the declining factor given by (5.3) must be considered. L ô= zzao Ii (5 3) cf ' Defineô =0.65 and combine ` The talcewinto .~ P »» ..,¬›» and (5.3), the result will be Lf=0.37mH and C,=4.5uF. actual values used in the prototype are L,~=l .45mH and Cf=l2uF. A practical ‹.. (5.2) _ output filter designing, besides the preceding considerations, should also account such other factors as load short circuit protection, load imush limiting, low frequency voltage stress etc. ..-‹‹. ¡› ‹. 5.2.5. Main Switches ' The main switches SKM50GB123D. S,-S4 IGBT modules Ratings and the main electrical and thermal characteristics of which (including the inverse diodes) -are 5.3. used are two Non-Punch-Through shown in Table 5.2. Resonant Circuit Designing 5.3.1. Resonant Components and Auxiliary Transformer According to sub-section 2.4, the results of the designing are presented in Table 5.3.2. Auxiliary Switches ( Sa,-SM) According to-Fig. 2.ll(a) and maximum peak and RMS Fig. 2.l2(a), for load with currents of the auxiliary switch are Blocking voltage of the auxiliary device is the switch. peak current of 30A, the 54.2A and 8.4A respectively. same with the main device. Devices used in the prototype as auxiliary switch are main 5.3. SKM50GBl23D, the same as the Table 5.2. IGBT module SKM50GBl23D symbol characteristics values conditions O Ic TcaSe=25 C/80 Icm O SOA/34A C IOOA/6 8A h 'T¢zse=25°c/ 80°C Vces l200V TJ -55°c- +15o°c Vcesat vge=15v, 1¢=5oA, Tj=2s°c / 15o°° 4.5V/3.5V(max.) tdon/tr vc¢=óoov,vge=15v, 1¢=5oA, 80nS/l 50nS Rg‹›n=Rg‹›ff=3.3 Q ,Tj= l25° C Vcc=600V,Vge=l5V, Ic=50A, tdoff/tf Rg‹›n=Rg‹›ff=3.3 VF(diode) 250nS/1 50nS Q ,Tj= l25°C 2.9V IF=5oA, vge=ov, Tj= 1 50°C Qrr(diode) s trr(diode) r¡=125°c 250nS T,-=1z5°c per Rthjc o.31°C/W o.9°C/W o.o5°C/W IGBT DIODE Rthjc per Rthch per module 0uC l 20nH Lce Table Resonant 1. capacitor loss Resonant I. inductor bobbin, Transfo rm CI' 1. 5.3. Resonant components and transfonner designing results maximum tum-off Capacitance: C,=0.1uF, 2. Type: low-loss polypropylene, 3. Estimated lW. Inductance: L,=l5uH, 3. 2. Structure: 36 tums ISAWG copper wire wound on air core Current rate of changing l4A/uS. 60 tums twisted wire ratio: k =0.4, 2. Structure: primary and 24 tums tvvisted wire (15 strands Transformer ferrite core, 3. 24AWG) in the (7 strands secondary Allowed equivalent loop resistance around 2.29. 24AWG) in the wound on E65/29 83 5.3.3, Auxiliary The Diodes (Dsa,-DSM, DSM'-Dsa4°) current stress of the auxiliary diodes blocking voltage of the auxiliary diodes Components used the in the prototype are characteristics are given as following I1=Av=30A, is is k times of that in the auxiliary switch. The same with the main device. HFA30TA60C in : VRwM=600V, VF1v1=l.2V; R1hj¢=0.85 ° C / TO-22OAC case, ratings and W and trr.ma×=60nS. Thermal Designing 5.4. A Thermal designing takes account of merely the conduction losses of the main devices. Switching losses of the main devices as well .as auxiliary devices are neglected. This simplification ~ mz-~› ~›‹› ¬ is the switching/conduction losses of the reasonable in the prototype case. For calculation of the conduction loss in the inverter instance, a conduction loss model hasnheen recommended, which represents the device as a constant voltage drop in series with a nonlinear resistive element during conduction [l39]. This nonlinear element is further simplified to a linear element so that a closed form solution can be reached. Conduction loss . .-,za-Y . »_¬~›-~:--_~~-- z -z- of the four devices is then given by (5.4): .tw = 4(Pon-s + Pon-D) Plossw-on where: = P,,,,_s P,,,,_D 1 E = 1 1,,Vce_,,(; 1 1 511,1/F(,(; + M zos 9) + Í M - Tzosâ) + Device models for the (54) ' «/É M 1§Rce(íä + 5; cos â) Jš 1f,Ra,,(š-\/;[¬- IGBT and M 5%-cosa) (5.5) (5.õ) the reverse diode are given in (5.7) and (5.8) respectively: Vcesat = Vce.o + [Rae VF = VF_0 + IRak (57) (5.8) 84 VM and Vw are the IGBT and reverse diode voltage drops at zero current, and Ro. and Where Rak are the resistive elements of IGBT and SEMIKRON From diode. data-sheet [140], Vce_oS1.5+0.002(Tj-25°), RW=0.O25+0.000l(Tj-25°), Rak(125°)=22mQ and VF,0(125°)=l.2V. According to cos9 = 1, load (5.5) and (5.6), assume a modulation index M=0.62, load power factor peak current Ip=30A and junction temperature Tj=125°, the loss distribution and the total loss are calculated to be: Pon-s=l7.9W, Pon-D=4.04W, Ploss-on=87.8W. Then, for a given junction temperature upper heat-sink is required thennal resistance of the given by (5.9) (four devices share a heat-sink): -Ta -2(Pon _ +Pon _ D )Rthch -Pon _Rs thjc.s sP T: /max Rthha limit, the where: Tjmax (59) loss-on is the given maximum junction temperature. Ta is the operation ambient temperature. case thennal resistance per IGBT. Rthjc.s is the junction to Rthjc.D is the junction to case thermal resistance per diode. ~ the case to heatsink contact thermal resistance. Rthch is Rthha is 'the Setting heatsink to ambient thermal resistance. Tjmax=l25 ° C and substituting the - known parameters one will into (5.9), get: V Rzhhz=1.o5°c/w. For the prototype, a 200mm square heat-sink is mounted vertically on a wooden base accommodating the two main IGBT modules (with thennal compound). The two auxiliary IGBT modules (with thermal compound) insulation layer and thermal compound) are mounted on another heat-sink of the same size together with the four diodes (with thin assembled on the same base. 5.5. mica ' r IGBT Driving Designing 5.5.1. Main IGBT Module Driving Intelligent single device. IGBT driver ` SEMIDRIVER SKHIIO The main feature of this driver is summarized is as following: used for.. each main IGBT 0 Improved output buffer enabling switching current up to 400A 0 Built-in soft tum-off mechanism enabling high 0 Built-in at 20kHz. - DC bus voltage operation. DC/DC converter (4kV insulation) easing the driving supply arrangement 0 Information transfer (driving, protecting) by ferrite transformer offering high dv/dt immunity (75kV/uS). 0 Optional input signal level: 15V/SV. 0 Extemal fault reset. Electrical specifications Table 5.4. and settings of the SKHI10 driver actually used is given in - Table 5.4. Specifications and settings of SKHIIO driver Electrical Characteristics *V .Z C) Value Vs supply voltage 15V Is supply current(max.) 0.5A vG(‹›n) tum-on gating voltage l5V VG(off) turn-offgating voltage td(on) input-output tum-on propagation time luS td(oft) input-output tum-off propagation time luS td(error) error input-output propagation time luS VCEsat Vce reference 5.2V Rgon intemal tum-on gate resistor Rgoff intemal tum-off gate resistor tmin fault indication delay (R¢e=1s_1< 5.5.2. Auxiliary Intelligent each auxiliary for fault indication 220 22 time Q 1.5uS Q ,c‹ze=33op1=) fault indication outlet (pin J3' shorted) low-logic IGBT Module Driving medium power double IGBT driver SEMIDRIVER SKHI23/ 12 IGBT driving. The major features of this driver are: 0 Output buffer enabling switching 200A 0 Soft tum-off enabling higher 0 Built-in -8V à Error logic 0 ° Term symbol ×.- (Ta=25 at 20kHz. DC bus voltage. DC/DC converter for driving power supply. Ferrite transformer information delivering enabling high dv/dt immunity. 0 'Built-in adjustable interlock circuit. is used for 86 0 Optional input signal level: 0 Automatic fault reset Electrical specifications 15V or 5V. when both inputs are zero. and Table settings 5.5. of the driver are summarized in Table 5.5: Specifications and settings of SKHI23/ 12 driver Electrical Characteristics (Ta=25 ° C) Term symbol Value Vs supply voltage 15V Is supply current(max.) 0.5A VG(on) tum-on gating voltage 15V VG(oft) tum-off gating voltage -8V u td(on) input-output tum-on propagation time l.4uS td(off) input-output tum-off propagation time l.4uS td(error) error input-output propagation time l.4uS VCEsat Vce reference 5.2V Rgon intemal tum-on gate resistor Rgoff intemal tum-off gate resistor tmin for fault indication .fault indication 22€! 22 delay time Q l.5uS (R¢e=1s1<Q ,c‹ze=33opF) 5.5.3. Error logic fault indication outlet (pin J2 shorted) tTD interlock time (RTDl,RTD2 low-logic not installed) l0uS Zero Voltage Detecting According to Chapter each main device voltage must be detected so that 2, on at the moment when the voltage across vzz GDH it is zero. R2 l.5kQ 1 3 Di lN4937 C D| D2 lN4l48 N| ZN 2907 SKHI l 0 High Current Driver t: R] 5609 5 ä ' Goffl E 2 o I I Zero voltage detecting Fig. 5.2. Zero voltage detecting circuit circuit in the prototype. it is turned 87 The circuit used in the prototype Gm, can only reach the output terminal is shown in Fig. 5.2. B when the near zero (threshold decided by R, and R2). The PWM tum-on command at device collector voltage VCE declining to PWM turn-off command at Gm» is sent directly to the output terminal B. Note that the tum-on altered with the adding signal property (turn-on resistance etc.) of the driver of the zero voltage detecting circuit. However, dissimilar switching where turn-on performance (diode reverse recovery di/dt) by the tum-on signal property, the would be to hard- significantly affected is tum-on performance under zero voltage switching not is affected to any considerable extent. Protection function of the SKHIIO driver indication during the commutation resonance. must be disabled to avoid Device protection is false failure then partly assumed by the zero voltage detecting circuit. Characteristics of the protection, however, are changed. 5.6., Control Designing Control Requirements 5.6.1. To 0 - p generate the main switch gating signals according to the indicated in Fig. 3.2. and (S, S4, -S2 and PWM modulation A dead time must be introduced between the complementary signals S3). The minimum/maximum pulse width must be auxiliary switch gating signals. 0 To strategy set higher than the V . generate the auxiliary switch gating signals according to Fig. 4.2, the width of which should be set according to Fig. 2.l0(a). 0 To give distinguishable The fault control circuit designed 5.6.2. Control Diagram memory as well as manual reset. shown in Appendix is 5.2 and is explained below. ' ' Clock signal generator U1: To use 214 addresses generating the 50Hz output, the clock frequency Should bez f,,,,,=(1/50)/ 214 =s33kHz. The generate this clock, where R1=l0kQ vco embedded in cD4o4ó is used td and C1=202pF. R2 and R3 serve to convert the 15V output to TTL 5V for the next stage, R2=l0kQ and R3=4.7kQ . CMOS 88 Address counters U2/U3: Two CD4040s are cascaded to generate the 214 address signals. An EPROM 27256M Gating signals generator U4: switching signals, the BASIC program is maximum commutation duration is width is set at l4.4uS and the listed in 1l.3uS. Based Appendix is used to generate the eight 5.1. According to Fig. 2.10, the on which, the minimum/maximum auxiliary switch gating signal PWM pulse widths are setat 28.8uS and 124.8uS respectively, with a modulation index of 0.62. Dead time of 2.4uS that C24-C3l are imperative to suppress the Signal level converter UI 1/UI 2: to 15V Note EPROM output noise, each valued at l5nF. Two SN7406s 5V are used to convert the when the so that to enhance the noise immunity drivers. Resistors is inserted. signal back switching signals are delivered to the R4-R1 1 each valued at 2.2kQ are pull-up resistors for SN7406s. Fault trip-ofl U5/U6: Six OR gates are used to enable shut-down in the case of fault. Auxiliary devices fault built with a automatic reset Thus extemal memory U7: The double after fault, which is drivers for the auxiliary devices are not favored in terms of fault identification. memory U7 is offered as a remedy. fault Fault summing and manual reset U9/UIO: The whole prototype will be shut the case of any individual device fault. Fault at Capacitors 5.7. 2.2kQ in memory can only be removed with manual reset button. This designing enhances operation safety of the prototype. Resistors valued down are extemal pull-up resistors for the fault indications C2-C7 each valued at 100pF provide noise immunity Rl2-R17 each from the drivers. for the fault indications. Experimental Results Fig. 5.3 shows the output load 5.4 shows the three the main switch side voltage and filter inductor current waveforms. Fig. level output voltage VAO. Fig. 5.5 shows the ZVS commutation process of S3 during switching cycle. Further in Fig. 5.6, the details i,0,d=21A are shown, for a predicted dv/dt of 90V/uS (averaged over 1,2 -tn) and l4A/uS, the experimental values are about 90V/uS and l3A/uS respectively. the details of tuming-off at i,°,d=24A are shown, for a predicted dv/dt of over t, - t6 ), the experimental value is about 146V/uS. of tuming-on And di/dt at of in Fig. 5.7, 152V/uS (averaged 4 Besides the cornmutation processes of the switch, the main .freewheeling diode commutation processes are also shown in Fig. 5.8 and Fig. 5.9. lload <l- /Vo /_\ / - J »/ z 100V/Div; IOA/Div; 5mS/Div Fig. 5.3. Experimental output voltage and filter inductor current. .Ill M0 Illàlz _ _|i_|i1›||1mlm|| Illllllllllllllllll _¡_________ Illllllllflll Hlllllllllllllllllll. . í lmnmmlilr IIIIIIIIIIIIIIIIII |›|r|«||r|||| |ll|!|!|! II”I'IlI'IlI'IiIIIIII l0Ov/Div; 500uS/Div Fig. 5.4. Experimental inverter output voltage. --”'"/Fal Vss Mv»--â= li »¬"W~ v._~f¬//Fl' Í \\ V1; ss "____ l.00V/Div; l0A/Div; 20uS/Div Fig. 5.5. ZVS commutation of the main switch (S3). __""^T\ VS3 \\ / \/-/\ _vz¬¬z-z\/‹-›--É \\ Ís3 «_ Í 100V/Div; IOA/Div; 2uS/Div Fig. 5.6. se | F" Extended tum-on process of S3. |" "W183 I /\/\_/^»/"'°\|¿ V / ›r\ / VS3 _ 100V/Div; IOA/Div; 2uS/Div Fig. 5.7. Extended tum-off process of S3. L....'....L..¿.`....¡.... Í ....¡ Vss _..¡¡II í, .-... _... zv-¬ .mm-_-.__________________._ É d 51% 100V/Div;l0A/Diví; l 0uS/Div . Fig. 5.8. 1 › Tum-on process of D, (L,=0.45mH, I 1 C,~=l2uF). 2 91 . . . . . . .. ins 100V/Div; 1 OA/Div;5uS/Div Fig. 5.9. Fig. 5.10 i,0ad=22A. Tum-off process of D, (Lf=0.45m1-I, Cf=l2uF). shows the ZCS commutation process of the auxiliary switch (Sá/DS,3) at For the predicated conunutation duration of 9.8uS according to Fig. 2.10(a), and the predicated peak current of 46.3A according to Fig. 2.11(a), the experimental values are 9.5uS and 46.5A respectively. Fig. 5.11 shows the resonant inductor current waveform when íhmd =0A. For the predicated commutation duration of 5.9uS and the predicted peak current of 24.1A, the experimental values are about 5.5uS and 23.5A respectively. In the meantime, Fig. 5.12 shows the resonant inductor current wavefonn when commutation duration and peak current are about 9.5uS Íhmd =22A. The experimental and 46.5A, which are in accordance with the predicted values of 9.8uS and 46.3A according to Fig. 2.10(a) and Fig. 2.1 Jr Vsê; ~ <* Ísas _ ,_.-J-_---J¬¬~.^^ _l. 1 md /"_|\ .LW . . LL! ,L - - 100V/Div; l0A/Div; l0uS/Div Fig. 5.10. ZCS commutation of the auxiliary switch (Su). 1(a). 92 ` / _....__:F:__~_...,...=___. ___z,,l,. /T\ . \ I Lf23 \ n 7'|7À~0a;d_ FT _¡_-:_ 31: l0A/Div; 2uS/Div Fig. 5.1 Resonant inductor current waveform l. _/\ - lload z I | 1 1 M / _, at ik,,,,=0A. ' Lr23 ' . ~="'_“_'"' <r | \t^Wzz_.t_ l0A/Div; 4uS/Div Fig. 5.12. 5.8. Resonant inductor current waveform at i¡o,,,=22A. Conclusions The foregoing presentation of the experimentation justifies the following conclusions: 0 The proposed scheme ensures zero voltage tum-on and capacitive tmn-off of the switches, without causing any voltage/current spikes across these switches. 0 The proposed scheme ensures main ` inductive turn-on and zero current turn-off of the auxiliary switches, without causing any voltage spike. 0 Main freewheeling diode works with Oscillation transfer 0 may zero voltage turn-on and zero current tum-off. occur during the diode forward recovery process due to the fast current from the snubbing capacitors to the diode Characteristic curves presented in Chapter 2 regarding the inductor peak current and resonant inductor bridge prototype. RMS commutation duration, resonant current are verified by the 3kW half 93 Appendix BASIC program for generating the gating signals in EPROM 2725 6M 5.1. 970 U = 1 'DATA OUTPUT LINES COUNTER 980 OPEN "A:SPWM.DAT" FOR OUTPUT AS #1 990 PRINT #1, ": 10"; 'FIRST DATA OUTPUT LINE FORMAT 1000 PRINT #1, "000000"; 'FIRST DATA OUTPUT LINE FORMAT 1010 FOR I = 0 TO 127: 'SELECT SWITCHING CYCLE 1020 RAI = 0: RA2 = 0: RA3 = 0: RA4 = 0 'DEAD TIME/AUXILIARY 1030 FORK = 128 * I TO 127 + 128 * I: 'SELECT ADDRESSES 1040IFO<=K-I* I28ANDK-I* 128 COMMANDCOUNTER <=64THENVB=K-I* I28ELSEVB= 128-(K-I* 'TRINGLE VB GENERATOR 1050 VA = 64 - VB 'TRINGLE WAVE VA GENERATOR 1060 PI = 3.14159 1070 VC = 32 + 20 * SIN(2 * PI * K/ 16384) 'MODULATING 1080 IF VC > VA THEN SI1 = 1 ELSE SI1= 0 1090 IF SII = I THEN RAI =RAl + 1 ELSE RAI =RA1 1100 SI4 = I - SI1 128) WAVE GENERATOR 1110IFVC<VAAND64<=K-I* 128ANDK-I*128<=127THENRA4=I+RA4ELSERA4=RA4 1120 IF VC < VB THEN SI3 =1 ELSE SI3 = 0 1130 IF SI3 = THEN RA3 = RA3 .+ ELSE RA3 =0 I 1140 SI2 1 =1- S13 I150IFVC>VBAND64<=K-1*128ANDK-I*I28<=127THENRA2=I+RA2ELSERA2=RA2 <= RAI AND RAI <= 2 THEN DA4 = ELSE DA4 = 0 'DEAD TIME 1170' IF <= RA2 AND RA2 <= 2 THEN DA3 = ELSE DA3 = 0 'DEAD TIME 1180_,IF <= RA3 AND RA3 <= 2 THEN DA2 = ELSE DA2 = 0 'DEAD TIME 1190 IF <= RA4 AND RA4 <= 2 THEN DAI = ELSE DAI = 0 'DEAD TIME 1200 IF <= RAI AND RAI <= I2'TI-IEN SA4 = ELSE SA4 = 0 'AUXILIARY SIGNAL WIDTH 1210 IF 1 <= RA2 AND RA2 <= 12 THEN SA3 = ELSE SA3 = 0 'AUXILIARY SIGNAL WIDTH 1220 IF <= RA3 AND RA3 <= 12 THEN SA2 = ELSE SA2 = 0 'AUXILIARY SIGNAL WIDTH 1230 IF <= RA4 AND RA4 <= 12 THEN SAI = ELSE SAI = 0 'AUXILIARY SIGNAL WIDTH 1160 IF 1 I 1 I I I 1 1 I I 1 I 1 1 1 1240 S1 = SI1 DA4: S2 = SI2 - DA3: S3 = SI3 - DA2: S4 š S14 - DAI 'MAIN SWITCH COMMAND - 1250A=SA4*2^7+SA3 *2^6+SA2*2^5+SAI *2^4+S4*2^3+S3 *2^2+S2*2^1+S1*2 ^10 ' PRINT #1, 1260 1270 1280 1290 1300 1320 1330 1350 1360 1370 1390 1400 1420 1430 1440 1450 1460 1470 1480 1490 1500 1510 1520 S1; SAI; S2; SA2; S3; SA3; S4; SA4; VC; VA; VB; K; A K = U *16 THEN = U U+ IF 1 IF TOTAL < 255.5 THEN IF TOTAL < 15.5 THEN PRINT #1, "00"; HEX$(TOTAL) ELSE PRINT #1, "0"; HEX$(TOTAL) END IF ELSE PRINT #1, HEX$(TOTAL) A END IF TOTAL = 0 PRINT #1, IF K < - ":";1-IEX$(K/(U 16 THEN PRINT #1, "000"; ELSE IF K < 256 THEN PRINT #1, ELSE IF "00"; K < 4096 THEN PRINT #1, END IF END IF "0"; - 1)); V 94 1530 1540 1550 1560 1570 1580 1590 1600 1610 1620 1630 1640 1650 1660 1670 1680 1690 1700 END I1= PRINT #1, I~IEx$(I<); END IF TOTAL = TOTAL + A IF A < 15.5 THEN "00"; PRINT #1, "0"; P1Ex$(A); ELSE PRINT #1, HEx$(A); END IE NEXT K NEXT I IF TOTAL < 256 THEN PRINT #1, "0"; HEX$(ToTAL) ELSE PRINT #1, HEX$(ToTAL); END IE cLosE END __ ___ l_ D ____ EE AMM °_ š ÚMW I âš J __ E V _! __' V! Qëgi “__ EE flw ë ___ E EÉ “MH “Mw Í *_ _ __ À /š_ _ E E E H imã _ “___ /` \.| _ _ _ _; _ __ H À; /*__ › MÊÊÊO il” U › â __ _í 8 U _! E' têaäã E__§ 025 _ __ 1 Á ___ __' _ _ 3 __ äüifãg _ __ '_ E š š / É ~ _ _ “H _! :__ _ I _ __ :_ _ :_ u _ U p E 2o^:\/Säimzë E | ¡ |¡ ¡_ ||| “WWW mg E š n _' _ _ H _ _ _ É I” _¡ V E ___* :__ aw _ _ E ___ gw _ É :S :EW E :___ šã _ _ EW __! :S :___ §_`_% E _U _ '_' __ BÊWÊV R Bbgo H _ _5 E0 E5B50 _ “Nú Uá __ š E /_\ 2 É ___ _‹ 3 2 É __ É E E EE 8E _; _; _; __, E _ _ _ __ II :Il l ššmwgnâ › _ E II _ _ _ I i¡ ~. ~. E8883 E H E 5 _ r _" __" E 5 E “_ ___ II II ll › É 8 "_ E H 5EE3 E0 L¬__í_ÊÊ_š_m 5 _ __ __ _ E É __ __ › _ š tb /_\ É l` _ x R H É E É É E '_ n E __! 5 5É u _! E ` N" I K _ __ /_\ UÉ¡a Ê ll.. _ E › 1 __”, É ' É E 96 Chapter 6. Operation of the Neutral-Point-Clamped (NPC) llnverter Abstract: This Chapter explores the neutral potential stability of the Neutral-Point- Clamped (NPC) inverter. Steady state stability, dynamic state stability along with the stability under non-ideal conditions are studied. The analyses are verified by PSPICE simulations. 6.1. Sub-Harmonic PWM Modulation of the NPC Inverter shows the half bridge NPC Fig. 6.1 and negative when positive, zero principle of sub-harmonic S, and inverter circuit. S2, S2 and S3 or The output potential S3 and S4 are is connected to ON respectively. PWM modulation is shown in Fig. 6.2. By intersecting a sinusoidal reference mod(t) with two vertically shifted triangle carriers, a three level voltage sinusoidal envelop is obtained at the output, which - ii _E“ C2 vdz rl ó-v __ D1 VAO = > S4 V ¿ D3 V A D4 - ¡ (6.1) A sw É sw; (val SW=\.V/\o=V‹1| SW=0_V^0=0 v^“ Leg Switching Function SW: sw=|z s| ON, sz oN sw=oz sa oN, sz ou sw=-1: s4 oN, ss ou Fig. 6.2. Sub-harmonic Sw;¡ V^‹›=-V.|z . PWM modulation for NPC inverter. (sw-l)sw ___Í_'vd2 can be rearranged as VAO = od(t)=Msin(nml llmld Half bridge NPC inverter configuration. (l+sw)sw vdl 2 (6.1): l . Fig. 6.1. vw with °2 A 0 Vd° A S1 ___ Vzu expressed by is ' ¡ C| The (6.1) (6.2): +vz12) Jr? (val Tvâz) (Õ-2) 97 6.2. Neutral Potential Steady State Stability Under steady state, vd,=vd2=Vdc/2, the output voltage vAo is simplified to (6.3): Vzz vw = sw _” ( 6.3 ) 2 where the leg switching function sw can be given in Fourier form by (6.4): sw = Z Aksín (kwmt) (64) lc keK=l,3,5,... where A k is the amplitude of the kth order harmonic, com the load current im, z... '\ ...sa is the fundamental frequency. Then written as (6.5): is äsin (kwmt _ ek) l!oa‹ÍÍ.= keK=l,3,5,... where Z k Á âk is load impedance at harmonic order of k. On the other hand, the current flowing out of ._._. _ lo ln lload _. _ lload (1 _ sw z_. ) ,load n =-Vd_°{zzA A 4 k 1 k k,l E ' ¢‹›s[(k~1)w K,L = A m 1]-22,4 k k 1 which ' in is given by (6.6): (6 6) _ further expressed ¢‹›s[(1‹+1)w is by (6.7): z-0) " m z1}zísz'n(kw k zh concemed, only the i is given by "' (6-7) l,3,5... When the neutral point potential est, in is :_ sw 2.lload Substitution from (6.4) and (6.5) in (6.6), z' the neutral potential (6.8): DC component in in is of inter- 98 z z n.dc 21,- da m 0 " O From ‹õ.8› . perpendicular property, integration (6.8) equals zero. This result ímplies the neu- tral point potential will current flowing 6.3. be stable under normal sub-harmonic modulatíon, because of the zero fiom the neutral potential during the fundamental cycle. Self-Balancing under Ideal Condition When dynamic state is concerned, after suppose that the neutral potential diverges from zero a perturbation, the output voltage will then be given by (6.2). While the first part repre- sents the steady state voltage output, the second part of this expression, represents the output voltage variation AV» = T wi AVA0 resulted from the neutral potential drift, _ _ in (6.9): Wal _ Vaz) (Õ-9) Substitution from (6.4) in (6.9), AvA0 can be detailed Av,,(, which is rewritten v -v -4l4;dl{%§AkA¡ cos [(1z-uwmz]->¡š§AkA¡ ' k,l e From K,L (6.10), = by (6.10): ( 6.10 ) cos [nz +1)ú›mz]} l,3,5... even numbered harmonics appears at the output after the perturbation. In the meanwhile, the load current response to this variation, recorded as Aibad , is given by (6.11): Azload = _ ízí vd1'Vz12 k,l e where zj¿ gj is {š K,L Âkfíz A/(A1 Éšcos [(k-l)a›m t-0k_¡]-š%]%cos[(k+l)‹um t-0k+¡]} = l,3,5... the load impedance at harmonic order of j. (611) 99 On the other hand, the current variation flowingout of the neutral potential is inferred to be given by ^«"n = _ Aflozzd = ^"‹› where Aio especially ...ndo (6.12): is ^"1‹›zzâ(1 ' SW2) ' ^"lz›zzz1 = the neutral path current variation. concemed, only the DC component of z (612) -SW2^"1‹›zzd When the neutral point potential Ai" is variation of interest, which is given by (6.13): (613) A,-,,z(.,m,, Based on perpendicular property, Aim can be simplified to v -v (A A )2 (A A )2 zíü '“{zz---k cosak-I +22-_-k cosa n.dc k+l (6. 14): (614) » Ai 1 15 k,I e kl K,L z = 1 kl /‹-1 Z } /z+1 I,3,5... (6.l4) demonstrates that the sign of Ain_dc (v dz - v dl ), the value of which means is is further dependent is always dependent on the sign of on the part within the bracket. This result that after perturbation, a corresponding variation arises simultaneously in the current flowing out of the neutral point, which rejects the perturbation. From (6.14), the following conclusions ability in the half-bridge can be drawn conceming the self-balancing NPC inverter under ideal condition: 0 Self-balancing exists so long as the load 0 Time constant of the dynamics is is not pure reactive ((-)k_¡ ¢1 š ,6k+¡ dependent on the following components: 1. Modulation index: The higher the longer. 2. Load impedance amplitude and angle: The greater the longer. 3. DC link capacitances: The larger the longer. _ ¢: š ). 100 The output voltage Fig. 6.3 illustrates the mathematical process conducted above. sponse (AVAO) to the perturbation (vdl-vdz) leads to reflected back to the neutral potential M. V‹1i"'z12 (1/dl (Ain), rejects the original perturbation. 1 -sw* -- Av"" 1 2 -Vd2)' = ^¡1""z1 ¡¿0¡ Ai" -SW2 recovers to result is its shown Table ^”"'¬”'”+ , Vz1i'Vz¡2 1 is mechanism in the NPC inverter. been verified by PSPICE simulation. Refer to Fig. 6.1 and and parameters for the simulation are given in Table in Fig. 6.4. For a initial state drift 6.1. of 100V, the neutral poten- 400mS. The dynamics steady state after about system characteristic which 6.1 gives the that the neu- Vz11'V‹12 Fig. 6.2, the circuit specifications tial _ 0 self balancing ability has The simulation 1 z“íÍAz,.dz + 2 Fig. 6.3. Illustration of the self-balancing The So when kept constant. tral potential is + load current variation (Aim), which re- exhibits typical one-order similar to the case of the capacitor clamping inverter. Appendix PSPICE program for this simulation. 6.1. Specifications/parameters for DC link Vdc=800 PSPICE simulation of the self-balancing under ideal condition Filter pa- Lf=0. lmH Operation voltage V rameters C¡=4uF frequency DC link C,=4000 Load re- R°=lQ Fundamental capacitances C2=4000 sistance fl=l0kHz Modula- M=0.4 tion index f,,,=l00Hz Initial frequency voltages Vdl=300 Vdz=500 (V) (UF) Oahlfhm mn: 520V 'mdtihvcl uofmnnañon ÀHLDÍÀIÂD ; \/W Vaz "W A amv Om: Fig. 6.4. = V(I0,20} rom: MMMMMMMMMMMMMMMMMM 200m PSPICE simulation result of the self-balancing :icons ability in the 400": 50011: NPC inverter under ideal condition. 101 6.4. . Self-Balancing under Non-Ideal Condition u Analysis in section 6.3 has been conducted neglecting any circuit non-ideals which are always inevitable in practical circuit operation. For the half bridge NPC concemed, circuit asymmetrical charging/discharging during the positive/negative output voltage semicycle will lead to neutral potential Such asymmetry can most probably be resulted from drift. gating/switching diversities or load transience In an extreme case, - DC etc. ` component involved in the sinusoidal reference causes consistant charging/discharging unbalance over the neutral potential. However, while the neutral potential drifts away from the zero, self-balancing mechanism simultaneously which tends to recover the drifted neutral potential. depending on the load properties, modulation index, Up is activated to certain extent DC link capacitances etc., the neutral point potential will be stabilízed at an new operation point other than zero where new balance r reached. This conclusion has been verified with is specifications and parameters for the simulations are are shown in Fig. 6.5 and which corresponds Fig. 6.6, shown to PSPICE simulations. Circuit in Table 6.2. Simulation results DC biases of 0.5V and -0.5V to the reference respectively (carrier amplitude 10V, reference peak 4V). For the 400-V across C, and C2, Vd, and \/dz are eventually balanced at ,¬.. 'z ordër dynamics, which simulation is listed in means a Appendix initial 280V and 520V neutral potential drift of 120V. voltages of after the first PSPICE program for this 6.2. . Similar to the case of the capacitor clamping inverter, dead time does not affect the balancing ability of the neutral potential. The influence in one semicycle same influence is self- cancelled by the in the other semicycle. Table 6.2. Circuit specifications/parameters for simulation of self-balancing under non-ideal condition DC link Vdc=800 voltage V DC link C,=4000 Load re- capacitances C2=4000 sistance (ur) L,=0. lmH pa- Filter' Operation Q=l0l<Hz Modula- M=0.4 C rameters , Í C,~=4uF frequency Ro=lQ Fundamental frequency tion index f,,,=l00Hz Initial voltages ‹v› vd|=4()0 v,,,=4oo 102 4sov 400V -Z 350V ›- °°°" to 250V Vd2 MWVW“^WVV\fvW\/\Am\/vvv\/vv\A/\/vwv . : Own 100m; 50": : : : 250ma 20011: 1501!: : 35mm : 3001!: : : 40011: 450m9 . Tine Fig. 6.5. Simulation result of the self-balancing ability .W - /\/\WWWW\ under non-ideal condition, positive bias 0.5V. MNmWWM~up f Àsov Vaz ~ 400V 350V . uma : : 200m: |00ms : 300m; 400m: 500m; . Tine Fig. 6.6. Simulation result of the self-balancing ability under non-ideal condition, negative bias 0.5V. 6.5. Self-Balancing in Three Phase NPC Inverter System For a three phase system, actually the same mathematical procedure as the half bridge case can be followed except for that the load neutral potential analysis. With a primitive perturbation now must at the neutral potential, flowing into the neutral potential will arise simultaneously, which Ainda: 3(v "Vdz ) 01216 cosâk zh '“" (z2A,A.+,-ZA,A,(_,-J be included to the a variation in the current is given by (6.l5) [60]: 2 (6.l5) 103 keK=2,4,8,l0,l4,... From (6.l5), for leL=l,3,5... any perturbation of the neutral current variation always has 'an appopriate sign keeps the neutral potential from 6.6. potential, the resultant neutral point that rejects the original perturbation and drift. Conclusions 1. Neutral potential of the NPC inverter is self-balancing under the sub-harmonic PWM modulation with non-pure reactive load. The time constant of the first order dynamics is capacitances 2. dependent on the load properties, modulation index and the DC link etc. Depending on the extent of asymrnetry, typically resulted mismatches, the neutral potential will be balanced to an from gating or switching new operation point other than zero, due to the self-balancing mechanism counteracting such asymmetry once it arises. 104 Appendix - 6.1. PSPICE simulation program for self-balancing under ideal condition *Self-balancing under ideal condition *triggering signals generating vl 10 pulse (0 10 .lu 50u 50u .lu l00u) e2 2 O value {v(l)-l0} vmod 5 0 sin(0 4 100 0 0 0) esul 6 0 table {v(l)-v(5)} (-le-2,10) (le-2,0) esdl 10 0 table {v(6,0)} (2,l0) (3,0) esu2 7 0 table {v(2)-v(5)} (-le-2,10) (le-2,0) esd2 11 0 table {v(7,0)} (2,10)(3,0) *main circuit su3 16 24 6 O sswitch ds3 24 17 dmod du3 17 16 dmod su4 17 25 7 0 sswitch dsu4 25 18 dmod du4 18 17 dmod sdl 18 26 10 0 sswitch dsdl 26 19 dmod ddl 19 18 dmod sd2 19 27 ll 0 sswitch dsd2 27 20 dmod dd2 20 19 dmod dcl 40 17 dmod dc2 19 40 dmod *device models .model sswitch vswitch(ron=0.01 roff=le6 von=4 vof%2) .model dmod d(is=2e-15 bv=l500 tt=0 cjo=l p) *DC link initial voltages vdc 16 20 800 c2 16 40 4000u ic=300 c22 40 20 4000u ic=500 *load parameters lload 18 41.1m rload 41 40 1 cload 41 40 4u 2u 600m 0 2u uic .options vntol=1m itl4=100 abstol=lm reltol=0.l .tran .probe v(40,20) i(dcl) i(dc2) .end itl5=0 b u Appendix 6.2. PSPICE simulation program for self-balancing under non-ideal condition *PSPICE simulation under non-ideal condition *triangle carriers and sinusoidal reference vl 10pulse(010.1u 50u 50u .lu 100u) e2 2 0 value {v(1)-10} - _ vmod 5 0 sin(0 4 100 0 0 0) *sinusoidal reference bias erro 42 0 value {-0.5} *triggering signals generating esul 6 0 table {v(l)-v(5)-v(42)} (-le-2,10)(1e-2,0) esdl 10 0 table {v(6,0)} (2,l0) (3,0) esu2 7 0 table {v(2)-v(5)-v(42)} (-le-2,10)(1e-2,0) esd2 11 O table {v(7,0)} (2,l0) (3,0) ' *main circuit su3 16 24 6 0 sswitch ds3 24 17 dmod du3 17 16 dmod su4 17 25 7 0 sswitch dsu4 25 18 dmod du4 18 17 dmod sdl 18 2610 0 sswitch dsdl 26 19 dmod ddl 19 18 dmod sd2 19 27 11 0 sswitch dsd2 27 20 dmod dd2 20 19 dmod dcl 40 17 dc2 19 40 dmod dmod *device models .model sswitch vswitch(ron=0.01 roff=le6 von"-=4 voff=2) .model dmod d(is=2e-15 bv=1500 tt=0 cjo=1p) ' *DC link initial voltages vdc 16 20 800 c2 16 40 4000u ic=400 c22 40 20 4000u ic=400 *load parameters 11oad18 41.1m rload 41 40 1 cload 41 40 4u 2u 600m 0 2u uic .options vntol=1m itl4=100 abstol=1m reltol=0.1 .tran .probe v(40,20) i(dc1) i(dc2) .end itl5=0 _ 106 Chapter 7. True-PWM-Pole Neutral-Point-Clamped (NPC) Inverter True-PWM-Pole Abstract: This chapter introduces a transfonner comrection inverter. After discussed. a brief review of the existing work, the proposed circuit and The True-PWM-Pole and the ARCPI its NPC operation are are then further extended to the diode clamping multilevel inverter (M>3). Experimental results from a 3kW half bridge True- PWM-Pole NPC inverter are given finally. 7.1. Review of the Existing Work Literature review shows inverter with soft switching. Earlier auxiliary network work may be network must be established to shows the and NPC traced back to the thyristor times when (NPC) tum-off of the thyristor switches. Fig. 7.1 assisted by the McMurray commutation NPC which can be regarded as zero current switching meaning. Auxiliary thyristors A4 and A2 assist S2 assist the thyristor three state inverter [33], have associated the that several previous publications S4. However, auxiliary voltage, rather than half of it as the assist the thyristors Vdc/1 i S' az 4 ^* vA ¡Ioad› O S. “"°" S2 I 1' Fig. 7 . 1. and S3 while A3 and A4 A4 and A4 have to block the whole I C| thyristors S, main thyristors do. fl - main inverter in the present ^4 A _ v~ I Circuit diagram of the three state thyristor inverter. DC link 107 DC Besides zero current switching, both resonant zero voltage switching methods have been considered for the NPC inverter [l43], as shown in Fig. 7.2, while A3 and A4 assist S4 and S2. C,23 A1 and A2 and resonant pole [143] link [142] NPC commutations of assist the S3 1/z and serves as snubbing capacitor for the center switches S2 and S3. All the auxiliary switches A,-A4, however, have to see 3/4 of the voltage, rather than ARCPI inverter. In the S,, main DC link by the main switches. VA _- S1 Cn W 1 Vga í^z‹^.z A 4 VA S4 W` _"__. Y Fig. 7.2. Circuit z‹ Note that ,.. when the NPC diagram of the ARCPI NPC 9 Cr4 inverter. structure is utilized for insulated' DC/DC conversion, soft switching has been successfully achieved [144] [145] which promises wide application in high voltage area. The operation of which resembles the phase-shifted resonant bridge [l03]. 7.2. Proposed Circuit Configuration The proposed transformer connection True-PWM-Pole NPC Fig. 7.3, which consists of the commutation of the first switching True-PWM-Pole; whereas in cell (S¡, S3) and fonns the second auxiliary branch (S,2, SM) assists the commutation of the second switching cell (S2, S4) Provided that the NPC neutral potential the shown NPC main inverter circuit and two auxiliary branches. The first auxiliary branch (S32, S,3) assists the the first inverter circuit is and forms the second True-PWM-Pole. is stable, and in particular, the coupling between two poles can be neglected, the operation of the two True-PWM-Poles are then independent from each other and each pole works as in the two level case. actually 108 n n D.1|Dzo' , Sm _ LW: O Dm: ‹^ ZS Cl im: O N2 N10 A E ze Dai! Dal' O í C2 ele à 9:12 Dad' Vd0/2 Sal DSM sad Ni; N2 . ÃDM ' Y }z ä› V . ¿ S4 ð-l Dn2` Í Dszú 5:12 Cri '-ll&'1g'› A D3 De Lr'24 Í Fig. 7.3. _ A 51 The proposed True-PWM-Pole NPC Cú inverter circuit. 7 .3. Circuit Operation According to the switching first pole is active, state of the main NPC inverter shown in Table the second pole will be at rest (S2 ON, S4 OFF and 7.1, when the vc,,,=V,,,/2, vc,2=0). Four commutation processes can hence be distinguished: Table 7.1. Switching State of the main NPC Inverter Vâ«/2 0 -W2 (a) Positive S. Sz Sz S4 ON ON ON OFF OFF OFF OFF OFF ON ON OFF load cturent, Dc¡+S2 to S,+S2 commutation; (b) Positive load current, S¡+S2 to Dc,+S2 commutation; (c) Negative load current, D2+D, to S,+Dc2 commutation; (d) Negative load current, S3+Dc2 to D2+D¡ commutation. For simplicity, only cormnutation following. For (a) NPC details for processes (c) which the next assumptions neutral potential is stable during the commutation interval. (b) ON are made and (d) are detailed in the first: nomial operation. Load current signal is released constant during . Main switch tum-off and auxiliary switch tum-on happen at the same tum-on is when instant. Main switch the detected voltage from the voltage monitor installed for 109 .each main switch declines commutation (c) to zero. Auxiliary switch conduction duration covers the whole interval. Transformer ratio is set to ensure the pole voltage swinging to the rail value during the V commutation, as discussed in Chapter 2. Refer to the relevant theoretical wavefonns shown in Fig. 7.4 and the operation step diagrams shown in Fig. 7.5 and Fig. 7.6, the Process(c): Negative load current, Step] (tg-t 1): Circuit steady commutations proceed in the following D2+D, state. steps. to S3+Dc2comrnutation: D2+D,' carries the load current. Output voltage VA0=Vdc/2. Step2 (t 1-tz): Tum on S,, and tum off conduction of Da, and D,,”, magnetization of kVdc/2 4 is S¡ at t, simultaneously, which causes the and current decreasing in D,. Voltage of L,,3 reflected on N, setting-up a voltage source of Vdc(1-k)/2 forcing the resonance followed. Step3 . is (tz-t3): charged while C,3 Step4 signal is D, is is blocked at discharged. (t3-t4): C,, is fully tz when im, rises to the load current level. After which Cfl - charged to Vdc/2 at t3. Dc, begins conduction. And S3 gating released. Step5 (t4-t5): L,,3 current reaches load current at t4, after which S3+Dc2 begins carrying current. Step 6 (t5-t6): load current. Sa, output voltage Current through/voltage across N, extinguish at S3+Dc2 carries the gating signal can be withdrawn. Circuit then reaches another steady VA0=0. Process(d): Negative load current, S3+Dc2 to Step7 t5. (t6-t7): state. full The D2+D, commutation: Tum on Sa, and tum off S3 at tô simultaneously which causes conduction of Da, and Dá”. Voltage of kVdc/2 is therefore established on N2 resulting in a voltage source of Vdc(1-k)/2 forcing the resonance. Load current moves from S3+Dc2 to D2 charging C,3 and discharging CH. Step8 (t7-tg): C,, gets fully discharged at tq after which current flowing transfers immediately to D,. S, gating signal is then released. in C,, and Cg 110 Step9 D2+D, (tg-t9): Voltage across/current through carries the full load current. S,3 gating signal in the proposed circuit, the from which instant can then be withdrawn. Circuit arrives main switches work with voltage switching) and snubbed tum-off, while the work with zero ts, at V A0=Vdc/2. the original steady state. Output voltage To summarize, out at L,,3 dies soft turn-on (zero main diodes (freewheeling and clamping) current turn-off and zero voltage turn-on. Unlike the case in the conventional snubber where the snubbing capacitance is limited by the snubber loss, the parallel capacitance in the present case can be optimized for the switch turn-off loss. However, the main diode is characteristic which of snap-on is conventional snubber. The loss accrued in such process similar to the system assisted with dependent on the wiring inductance is and the diode forward recovery property. In the meantime, soft all the auxiliary devices including both switches and diodes Moreover, tum-off (zero current switching). auxiliary devices, the Upon tum-on of these devices are work with despite the bridge configuration of the actually snubbed by the resonant inductor. turn-on of an auxiliary switch, the opposite freewheeling diode carries no current and its reverse recovery contribution can be neglected. Transfonner excitation are reset to zero after each commutation and causes no magnetic accumulation. In particular, the extra therefore freedom in synthesizing the auxiliary voltage source for the resonance offered by the transformer greatly simplifies the control. Designing of the auxiliary branches has been discussed in Chapter 2. Switching period: T-I/fc A Commutation process Sa; Commutation process (c) \ l Sal S3 V 51 I z yyyyyyyyyyyy l ¡ (Vs:/il) I . ls'/D' 'M3 Vw __ hm *-1 5 V-W __ _ H __________ N 7 __ 5.... V-;.... ts main and the auxiliary switches t7 t 1 P Vèlloed F/*W Ê ' iload Igú ¿;._ W ¡ de/2 __ _ › › l . s Fig. 7.4. Switching sequence of the ¡ ¡ É ›l (d) tg Vdc/2 Í P tg and the relevant waveforms. lll I I Dz11Dz3 Sa: ð_| ZS J” Cl íí ‹¿ E ze o N2?› Dz2Dz14`Sz4 Dm C "VO ___... 2: DM ‹A z:‹›4 c 5:12 u 0 n DâlDz3` Szs __ 30.1 Cl 0 ‹¿ 1: to-t, Sz14 . 'í' ZS ZS<>-i DM Dzzz` Szz u u, _ , _ Dal Dai' f Cl Li*/2 Í 5:13 z:‹›4 ‹A à fe €~ D113Dz11` ° à Szn S14 D.«zDz..z' Cz ___""°'“ Z3°`¡ ZS 23°-I Das Dzz' I Í Szz . v A SM CM 1. z:‹›1 f Cr 3:12 ÍLH3 › D° ' Dsal ' ' T› 1.124 Dâzz v A L O z <^ Deu S1 Nz*› Dz1zDz4` Suõ Dszó ÃDM ZS°~I Dn' Szz 0 . É Dal Dal' sa! zs z:‹›4 A O __ O N10 K Sul Dzzzvzc szz Nr. ZS Dm 0 23°-I Daz' i 5:12 0 Fig. 7.5. Operation step diagrams for Usa] vs* Í|__¡›¡3 Nz Dz.-14 commutation process (c) ' ' " nc ' Nz* ‹4Dm step 6: t,,-ts Cn _'_› A 014 t,-t4 nm EM ‹^ ZS CM . D| ° 22°* D1úDz1' Cz Vudl 1.114 ‹ADzz ‹~ ^ Dcz Sa Cl *'› A ' 1-v2^ v . N, à V_“^"n ^` D Del Dsal NI.. I im: 5111 vz./1 °“ V DV t,-tz Dz3D1z1` Cz Cn ,S4 . step 5: 1 DS:|2 0 N10 E » Da 11.114 ¢¿ 2:04 cl ___"*'°/2 z ~ Y S4 V0. 0 Dz1DaJ' Sa; . . N 2' O Daá Dal' 'c 0 A ' ^À"2^ Dm "VO 2: "í› A Dcl ‹› Dm step 4: F' . Dz1zD114' tz-t, ‹^ ‹A 5111 Vad! Cm S4 DSa3 Nf. Da3D‹1|' U 0 Nz N10 í , step 3: Í ' Lm ‹A um o Íuu Nz*› 0 Cz Dez NZ. N 10 Cn S1 step 2: 1~2. Dm NI.. ' im: Cz vúaz 54 S1 A Dz1zDz4' 1 ^ ‹À Ã.,¬| zz ¿k¿› . › Dm: 1^ Sal T 1 Dm Sa: (step 9: ts-t9) Nz_'_› DzúD111' v¿ TP 0 N10 à H Y Dcz Dsal , T VM , 11.114 step J' A NZ' O Daly ii ¢^À"2^ Den Vãzfl Cl . Del l Dz1Dz1` _ LM I Í1.ru Sal Dz13Da1` l Cr 51 0 N10 1 Dâzs mó V A L 54 °" *-7 A Cr: . ts-tá D,+D2 to S,+DC2 commutation. 112 n 1 Dzn Dzú' Sa: à C. Í qNm "*'°” À - Dall :El à C2 _V°°'2 Dsnl ‹¿ à za 1. I S32 ' ~¿ ' ` . É í N|`o D¡¡4D;¡2' ¡ SM Z§°-I Í Â Del Sal Diz DM' 1 Nz_'-'› o A V S1 ¡,,,, _ Ás' - ° ‹¿ 3,4 1 Dm V -TP ¿ S4 [X14 Usa! Z: c. ' ___ _ ° sa 3°-I C1 ____Vd°n Dsnl ‹A zz DM 0 za 1. Da2` I 5a2 N2` (d) ' -se - D2 ÚL VÀ A W rn Dsa2 step 8: commutation process A E U Ez Dzzâ É í "lb Cn: _  Del Sal D.¬z DM' ZS ¡,,,, Cn ' - Da'lDnl' A 7 S1 1 É mí* EH 2* 1 Dm: 0 Nzo A 1+ í ‹¿ t,-t7 Fig. 7.6. Operation step diagrams for a 7 > 9s t2-t3 S3+DC2 to D,+D2 commutation. Topologies for Multilevel Case (M>3) 'Zero voltage switching for multilevel diode by _V“°'* °_| Í step 7: 7.4. I n Dal Das' Szú Cri (M >3) can be achieved an auxiliary network based on either the whole multilevel leg or each two level installing switching cell of it. 7.4.1. clamping inverter ` Zero Voltage Switching on the Per-Leg Basis Both True-PWM-Pole and ARCP methods are applicable for multilevel case. For the True-PWM-Pole method, the auxiliary branch is established by the same multilevel main leg, as assisted by shown in Fig. 7.7, S,,, Sa2, S,,3, S,,°, S,,2' Transfonner ratio where the commutations of and S83' S,, S2, S3, Sl”, S2”, respectively and three True-PWM-Poles leg as the and S3” are are formed. should be set less than' 1/6 to ensure zero voltage switching. In the case of ARCP method, the auxiliary branch can be established by the sub-circuit of a M+l assisted level leg, as by in Fig. 7.8. S,,, S,2, S,3, Sal”, S,2” Each main switching as in the shown two cell and level case. In this approach, its and The commutations of S,, S,3° S2, S3, S,', S2”, ARCP respectively and three and S3' are poles are formed. corresponding auxiliary switching cell work in the same way _ all the poles share the same resonant inductor and transformer. However, resonant current always flows through three devices in inner auxiliary Switches are indirectly clamped as the main series. In addition, all the switches. 113 R _ A *rl -BP: OJJ* Sal' 7 Ç . À A A ¡ ‹A S3 Sa] A À ‹^ sl. A saz A ‹A S2. H A Szú ‹4 S1' A H ¡ su A 9.1 À U I A . Fig. 7.7. Al xl The A auxiliary switches block l/3 of the M (4 zz 1 í = ___ T transformer connection True-PWM-Pole four level diode clamping inverter leg. :_"' Fig. 7.8. SI à ZS ZX à à à 2: à H DC link voltage, same as that by the main switches. A Szú' ¿ Sal °-l lS1 ^:fl‹› az  ¿ ‹^ ‹¿ _ s, ' SI. 7 zé S” ~ à H ‹^ sai' °-l Sfl €^ °-ln T S3. An ARCP four level diode clamping inverter leg. The auxiliary switches block voltage, half of that by the main switches. 1/6 of the DC link 114 7.4.2. Zero Voltage Switching on the Per-Cell Basis Zero voltage switching of the diode clamping multilevel inverter can also be achieved based on each two level switching with True-PWM-Pole or cell approach, individual two level auxiliary branch in Fig. 8. 11 for the case ARCPI methods. In this for each switching cell, as is installed shown of True-PWM-Pole and Fig. 8.12 for the case of ARCP. In comparison with the previous approach, each pole has its own resonant inductor. However, resonant current flows through only one auxiliary switch, which avoids the problem of indirect clamping in the auxiliary network. 7.5. Experimentation The experimental IGBT half bridge prototype, two sinusoidal modulating signal and modulation, as discussed in Chapter instead is parameters are given in Table 7.2: Í DMF Dal Cl Sa] Dzz Dzzz' 5.14 3°* Z* z1 __V“'/2 1 ZS DM Í 05.4 N2' _í ÃO-l Dflz' Í ‹A Saz A Du Dsnl o N1`o ___' _"`› ‹A Sal ' l|_¡»¡3 Dsaz Fig. 7.9. Circuit diagram of the D° í LY24 A A í› Cr 51 0 Nz Daft Dal' D2 Cr3 ilond A Cr, ' t àL 54 Lf D, O cf “› io W” L Cr4 3kW half bridge laboratory prototype. Table 7.2. Prototype specifications and parameters Parameters circuit neutral potential is not actively controlled Dsafl ‹À N iu _V_**""1 à Specifications employs a main - ZS O in Fig. 7.9, and on the inherent self-balancing mechanism. Specifications and stabilized counting _ shown vertically shifted carriers for the The 6. as P0=3.2kW, V,,,=720V, V°_,m,=l40V, I°,m,=22.8A k=0.45, T=l53.6uS, C,=0.luF, L,=l5uH, M=0.62 r 115 IGBT modules (SKM50GBl23D,l200V/SOA) are employed as the main and auxiliary switches. Eight ultra-fast HFA30TA60C (600V/30A) diodes work as the auxiliary diodes. Two storage capacitors C, and C2 each rated at 360V/3300uF form the center tap. Four Output second order filter: Lf=6.5mH, Cf=12uF. For each main switch, a zero voltage detecting circuit as discussed in Chapter 5 Under the given The width of 9.4uS. Minimum/Maximum 2.4uS cell. conditions, the is is installed. maximum commutation the auxiliary switch gating signal duration is set is at calculated to be 12uS, and the pulse widths are set at l4.4uS and 96uS respectively. Dead time of introduced between the gating signals for the two main switches in each switching Gating signals for the main and auxiliary switches are created in Appendix 7.1 lists Fig. 7.10 and current the EPROM 27256M. BASIC program. shows the three level output after filtering. Fig. 7. l2(a)- (d) of the inverter. Fig. 7.11 shows the output voltage show the typical four commutation processes. ¡ ¡ '¡¡ VAO -= 'V “À __-*__ ___.íi._.._z -_.__.._..___....--_-- 100V/Div; 200uS/Div Fig. 7.10. The three ,~ level output of the NPC inverter. Vo Áefl~ 100V/Div; 20A/Div; 5mS/Div Fig. 7.1 1. Output voltage/current after filtering. f ~ Vs: ,_ ¡Dc! l T ¡ v. v `¬.-..¬---_-¬_...-..-¬__ 100V/Div; SA/Div; l0uS/Div (a) Positive load current vsz commutation Dc, to S, , "ll afll i À¡fl@-._ ãlll ¬ ,, V- '-›' _-¬.¬- _'-.v_..¬.-__"-¬._.._ .._.___.._.__ il. _ ,, V -._~;quqn.¿_-_.-_-pppçgç 2 :.pqv9n.-^ :_:p-9-:_-_' _ l00V/Div; SA/Div; l0uS/Div commutation (b) Positive load current, S, to Dc, VS3 EJune Í = _. J. _ _ _ . ._ ' .. ._ ¬ _ _ q l 100V/Div; SA/Div; 2uS/Div (c) Negative load current, S3 to D, commutation 117 V t ss 1 : Dcl I' 100V/Div; SA/Div; I0uS/Div (d) Negative load current, Fig. 7.12. Fig. 7.13 load current shows the ikm, =8A. D, to S, commutation - Four typical commutation processes of the first resonant pole. auxiliary switch voltage and resonant inductor current waveforms at For switch to diode commutation, with a theoretical commutation duration of 4.6uS and a resonant peak current of 17.5A, the experimental values are 4.luS and l4.5A respectively. For diode to switch commutation, with a theoretical commutation duration of 7 .3uS and a resonant peak cLu'rent of 32.5A, the experimental values are 7uS and 27A respectively. Vszn -O 'ms 100V/Div; 10A/Div; 20uS/Div Fig. 7.13. Auxiliary switch voltage and resonant inductor current. 7 6. Conclusions The foregoing analysis and experimentation justify the following conclusions: The proposed scheme ensures zero voltage tum-on and capacitive turn-off of the main switches, without causing any voltage/current spikes across these switches. The proposed scheme ensures inductive turn-on and zero current turn-off of the auxiliary switches, without causing any voltage spike. Main freewheeling diode works with Oscillation transfer may zero voltage turn-on, zero current tum-off. occur during the diode forward recovery process due to the fast current from the snubbing capacitors to the diode Characteristic curves presented in Chapter 2 regarding the inductor peak current and resonant inductor commutation duration, resonant RMS current is verified. Both transfonner connection True-PWM-Pole and ARCPI schemes multilevel diode clamping multilevel inverter, achieved either the per-leg basis. flows through The on the per-cell circuit requires multiple magnetics, are extensible to per-cell basis or its on resonant current only one auxiliary device. In comparison, unique magnetics are used in the per-leg circuit, however, the resonant current flows through multiple auxiliary devices. 119 Appendix U= BASIC program for generating the gating signals in EPROM27256 7.1. 'DATA OUTPUT LINES COUNTER OPEN "A:SPWMNPC.DAT" FOR OUTPUT AS #1 990 PRINT #1, ":10"; 'FIRST DATA OUTPUT LINE FORMAT 1000 PRINT #1, "00O000"; 'FIRST DATA OUTPUT LINE FORMAT 1010 FORI = 0 TO 127: 'SELECT SWITCHING CYCLE 1020 RA1= 0: RA4 = 0 1030 FOR K = 128 * I TO 127 + 128 * I: 'SELECT ADDRESSES 1040IFO<=K-I* 128ANDK-I* 128<=64THENVA=K-I* 128+64ELSEVA= 128-(K-I* + 64'TRINGLE VB GENERATORIOSO VB = -64 + VA 'TRINGLE WAVE VA GENERATOR 1050 VB = VA - 64 970 980 1 ' 1060 PI=3.14159 1070 VC = 64 + 40 A * SIN(2 * PI * K/ 16384) 'MODULATING WAVE GENERATOR 128) › 1080IFVC>64ANDVA<=VCANDVC<=VA+2AND64<=K-I*128ANDK-I*128<=127 THENMMP=0ELSEMMP=MMP+l 1090 IF VC > 64 AND <= MMP AND MMP <= 12 THEN MINP = ELSE MINP = 0 1100 IF VC>VA ANDO<=IANDI<= 63 AND MMP>= THEN SI1 = ELSE SI1 =0 ' . 1 1110 SJ1=SI1 1 1 ORMINP 1 1l20SJ3=1-SJ1 <=MMPAND MMP<=2AND SJ1 = AND VC>64 THENDA3 = ELSE DA3 =0 <= MMP AND MMP <= 10 AND SJ1= AND VC > 64 THEN SA3 =1 ELSE SA3 = 0 1150IFSJl =0ANDO<=IANDI<=63ANDVC>64ANDO<=K-I* 128ANDK-I* l28<=63 THEN 1130 IF 1140 IF 1 1 1 1 1 RA1=RA1+1ELSERA1=RAl 1160 IF <=RA1 AND RA1 <=2THEN DA1 = ELSEDA1 =0 1170 IF <=RA1 AND RAI <= 10 THEN SA1 = ELSE SA1 =0 1 1 1 1 11s011=vE<=vcANDvc<=vB+2AND0<=1<-1*12sAND1<-1*12s<=63 THENMMN=0ELSE MMN=MMN+1 1190 1200 1210 1220 1230 vc < 64 AND <= MMN AND MMN <= 12 THEN M1NN = ELSE MINN = 0 vc < VB AND MMN >= THEN S14 = ELSE S14 = 0 SJ4 = S14 011 M1NN IF IF 1 1 1 S12 IF 1 12401F 1 1 =1- SJ4 <= MMN AND MMN <= 2 AND SJ4 = 1 AND vc < 64 THEN DA2 = 1 ELSE DA2 = 0 <= MMN AND MMN <= 10 AND SJ4 = 1 AND vc < 64 THEN SA2 = 1 ELSE SA2 = 0 12s01F S14 = 0 AND vc < 64 AND 64 <= K -1*12s RA4=0 1260 IF 1270 IF 1280 1 1 AND K -1* <= RA4 AND RA4 <= 2 THEN DA4 = 1 ELSE DA4 = 0 <= RA4 AND RA4 <= 10 THEN SA4 = 1 ELSE SA4 = 0 128 <=127 THENRA4 = RA4 +1 ELSE - Sl=SJ1-DA3:S2=SJ2-DA4:S3=SJ3 -DAI: S4=SJ4-DA2 'MAIN SWITCH COMMAND 1290A=SA4*2^7+SA3*2^6+SA2*2^5+SA1*2^4+S4*2^3+S3*2^2+S2*2^1+S1*2 ^0 1300 1310 1330 1340 1350 1360 1370 ' PRINT #1, I; S1; SA1; IFK=U*16THEN U=U+ S2; SA2; S3; SA3; S4; SA4 1 IF TOTAL < 255.5 THEN IF TOTAL < 15.5 THEN PRINT #1, ELSE "00"; HEX$(TOTAL) 120 1380 PRINT #1, "0"; HEx$(TOTAL) END IP 1390 1400 ELSE 1410 PRINT #1, HEx$(TOTAL) 1420 END IF 1430 TOTAL = 0 1440 PRINT #1, "z"; HEx$(K / (U - 1)); IP K < 16 THEN 1450 1460 PRINT #1, "000"; 1470 ELSE 1480 IF K < 256 THEN PRINT #1, "00"; 1490 ELSE 1500 1510 IP K < 4096 THEN PRINT #1, "0"; 1520 1530 END IP END IP 1540 END IF 1550 1560 PRn×IT #1, HEx$(K); "00"; 1570 END IE 1580 TOTAL = TOTAL + A 1590 II=A< 15.5 THEN 1600 PRINT #1, "0"; HEx$(A); 1610 ELSE 1620 PRINT #1, HEx$(A); 1630 END IE 1640 NEXT K 1650 NEXT I 1660 IF TOTAL < 256 THEN 1670 PRINT #1, "0"; HEx$(TOTAL); 1680 ELSE 1690 PRINT #1, HEx$(TOTAL); 1700 END IE 1710 CLOSE 1720 END A Chapter 8. A New Diode Clamping Multilevel Inverter 121 Abstract: This Chapter proposes a new diode clamping multilevel inverter, which rids of the series association of the clamping diodes in the conventional multilevel inverter. new Operation fimdamentals as well as the snubbing methods for the inverter are discussed. Experimental results from a 3kW half bridge prototype are also presented. 8.1. Series Association of Diodes in the Conventional Inverter In a conventional diode clamping multilevel inverter [35][36], voltage rating of each clamping diode is dependent on position in the inverter. For each leg, one can its diodes each sees reverse voltage stress Vdioâe = where M M- of: -K Ífvdc is l the _ number of inverter levels, k goes from C1 D¡ ~Q D3 D2 ‹^ A 4 O D5 __; A D4 ‹^ 4 S2' A ...| l M-2, Vdc to D5¡ | Dsz S. | C4 í' A ø-| DC link voltage. S › l S. .| 'sí III › .Q v .._ , Us _ ` > C7 52 :_ 1__. _ ' [__ MA M Q.. re -v [ff ' Dsz' Vm ‹¿ Vw' 3 IVD1 | VD3 VD2 ' .-1. ' ' VD4 V S4' ^ °"l ‹A 054' . Vpó vD5 E, I Fig. 8.1. Conventional diode inverter leg, is I _l_ Dó the ~fi A _ S] C3 (3-1) . Q find two clamping five level where clamping diode has high voltage. to block Fig. 8.2. Clamping diode voltage stress in association with the switching state in the conventional inverter. 122 For a five level case, as shown in Fig. 8.1, D, and D6 see Vdc/4, D3 and D4 see and D5 see 3Vdc/4, whereas each main device see only Vdc/4. Fig. 8.2 shows voltage is related to the switching state. While in a practical hence difficult to have situation the fast 2V,,c/4, D2 how the blocking ' . main device voltage rating is always fully utilized, it is tum-off diode in the market with multiple voltage rating and yet comparable performance. In high voltage P-N junction designing, larger breakdown voltage is always accompanied by higher on-state loss and longer turn-off transience. The common The five solution to this problem level case is in series, is shown put appropriate number of diodes in series. Diodes in in Fig. 8.3. by no means an optimal is to solution. series, Unequal though less problematic than GTOs or dynamic voltage sharing can static happen, due to the diversity of switching characteristics and stray parameters of the diodes in series. Measures to prevent the potential over-voltage problem by providing large snubber capacitor and high power resistor network lead to an expensive and voluminous system. 8.2. Operation of the New Inverter . _ A new diode clamping multilevel inverter has been proposed. For the five level case, as shown in Fig. 8.4, a total of 8 switches and 12 diodes are used, the same amount as that in the conventional inverter. The pyramid structure is extensible to infinite otherwise practically limited. ~A DI C2 o C __..3 D1 A D2 4 S3 A D3 A S4 Dl] A D3 D9 ^ D4 ^ Dn . D5 C4 . S1 C- A DIO D5 A °"| ‹e- ir ‹À D* 1 Ds 5 A ‹A Dzz ‹^ C2 D2 0 D3 C3 D4 054 S2 °.'| Dl _ ‹^ 53 s °› D52 S1' S4' Fig. 8.3. Conventional diode number of levels unless Du' D5 y . Dsfl . C4 S2 A A Á °-l inverter with diodes in series. Fig. 8.4. ‹A ‹, D1 A D3 A A ‹A D9 A A A Dio ‹A À D6 A ‹›-1 €A Usa' clamping multilevel ‹~ DS' 052 DS3 DS4 A psy 11,2' ^ sz' Dza' 54' €A Dn' Proposed new diode clamping multilevel inverter. 8.2.1. 123 Switching Cells of the New Inverter The new inverter can be decomposed into two-level switching cells which constitute basic operation units. For the ifive level leg, one can define (5-1) switching cells as Fig. 8.5 (a), (b), (c) In the same and (d). In cell sense, the output or the reverse in cell (c), moves from and from -Vdb/2 S' C1 the leg output (a), D” _ D Vdc/4 to 0 or the reverse in cell (b), o D D 4 A › D-‹ ‹4 D5 A z 54 {› DS4 ~ A A D; D2 o D3 C3 ‹ 1 1 ' ' D H S4' (G) Fig. 8.5. A Dsli DS3' DS4' S4 sl C1 Ds2 DS 1 ' C4 1^ . Dsl 0* C2 Ds4 (b) S1 Cl Ds3 Ds2' D6 ^ C4 (a) G' ‹ 1 1 - __C_3 .gzäpäzà Dx2 , ¡ O ri!!! D C4 C2 .A .¿ C3 D1 = Dn SL* '* ^ from 0 to -Vdc/4 or the reverse in cell (d). C1 | D¡ C2 D2 o D3 DS4 'A . C3 . 4 A A D4 D9 I DS] . Ds] ‹A Ds2 ‹¿ DS3 A €D ‹ q ¡ sl ' C4 ‹› 954 D3 A D5 Ds2 in moves from Vdc/2 to Vdc/4 or the reverse. a. C2 shown its v DBZ' \ Ds3' DS4' S4' Ds-1' (d) The four switching cells of the new diode clamping multilevel inverter. to -Vdc/2 124 To further illustrate how each switching cell works as a two and freewheeling paths for the up/down arms of cell(b) are level inverter, the shown in Fig. 8.6(a) forward and (b) Current paths for the other cells can be analogously found. C1 D¡ D3 o ‹› 52 ‹- ~ D2 C2 S1 D8 A - I D4 C3 A D9 A .¡ . I D5 A S3 ‹A Q (a) 21A Ds] ‹A DS4 ¢^ Ds: DS4 C4 z D8 A A _¿ D9 ¿ 4 °fi ~ . É À ›-A ‹ A .n D5 A A DSI A ~ ll ‹^ Ds: ST* , forward/freewheeling paths for the up arm ‹^ (b) forward/freewheeling paths for the down arm 2 Fig. 8.6. A i I Wi. Dsz' 0:' D2 D4 --C3 Dal' Dsl ‹. D3 0 ‹A »- , A S4 i cz DS4 1^ A DIO D6 A C4 Diz ves 'A S Cl | D3 D7 ^ DS Forward and freewheeling paths for cell (b). The above decomposition presumes that the output dv/dt is constrained to a single device dv/dt, which has been regarded as one of the most notable attribute of the multilevel inverter over the plain series of devices. must only involve one switching cell. To work within this constraint, each switching Simultaneous switching of more than one switching runs no difference with direct series 'association and 8.2.2. cell (a) for example. In this work altematively connecting switches of the leg, forbidden in operation. S2”, S3' and shown in Fig. cell, S2, S3 and S4 are always ON, while the output to Vdc/2 or Vdc/4 respectively. S4” are and each blocks zero voltage the state of S1”, as S1 and S,° The remaining always OFF. Thus D2, D2 and D,2 must be in clamping ideally. Moreover, D,, D2 and Dn each must always follows 8.5(a). In the same sense, for cell state is cell Switching State of the Diode Network Take state action (b), D, and D9 follow the of S2, as shown in Fig. 8.5(b); for cell (c), state D, follows the of state S2” while D2 follows the of S3”, D4 and D8 follow the state of S3, as shown Table shown in Fig. in Fig. 8.5(d). 8.1, which is The switching 8.1. 0: Output S, ‹ ' S, 8.7. Device blocks zero voltage ‹ Device blocks Vdc/4 voltage S2' S2 S3' S, S4' S., Diode state Dl: D7: Dil level -¬ shown in Fig. Switching State of the Diode Network verses the switch state 1: state Dm and Dn follow the state of S4, as (d), D6, of the diode network can thus be summanzed in state further graphically Table Switch and in cell 8.5(c); Vdc/2 1 O Vdc/4 0 li 0 0 l 'Vdc/4 0 1 'Vdc/2 O l D2 D3: V0 l D9 D4› 0 l D8 D5 l O 1 O l D 1 il lvl 017,0 0.170 0 _ D6: DI0› l 0 l l 0 P1! A . VAO Vac P ,_|-J | .i im S2' rn B!- U muzmm ; YYY àuw- I m bz | lT vi àz -__ |ñVsi'/Di/D7/ou I* VVS2'/D3/09 ' Vss'/Ds ' I li* Vs4/Dó/mo/D12 Vs:/D4/Ds Vsz/D2 I , _ V U Fig. 8.7. Illustration . VVV of the switching I _ 7 V5¡' state in the new YVVVVY _ _ diode clamping multilevel inverter in relation to the output voltage level. 126 8.2.3. Switch and Diode Clamping Mechanism Refer back to Fig. 8.5 while S,' in series with D2, Further, D,, in series with (a). D,, and D,2 D2 and clamped by Ds, and D,2, while D, Suppose that when D2, Obviously, S, S2”, S3' is clamped by is directly S,,” clamped by D, is directly clamped by D,, is and ' D,,, D,2, DS3 D,,, D,2 DS3; turn-off, DS., after its tum-off. D, in are all OFF, and each of them blocks Vdc/4 voltage, mechanism can be witnessed with are all clamped. clamped. In cell (c), S3”, clamped while clamped. Among which D3 and cell (b), S3, D,,, D4 are all clamped. clamped. In for the proposed Among clamped to are indirectly clamped. S2”, D3, D9 as well as new inverter, them, the 8 lateral Among which D,,, D, and cell (d), S4' Among which S,,”, D6 are directly clamped while clamping diodes. where all S2, D2 D2, D3 are directly clamped while S2”, D,, S2 are _indirectly S3”, S3, D,, are indirectly As a result, is clamped by D,,. Among which S,, D, are directly clamped while S,”, D,, D,, Similar with D2 series D,2 each will then block zero voltage. Then, S, and S,°, D,, D2, D,, are D,,, Vdc/4. S,,, D,,,, S,,, D6, are directly D,,,, D,2 are all D,2 are indirectly clamped. not only the switches are clamped, so are the components ( S,, S4”, D6) are directly clamped, the remaining devices, however, are D,, D2, D3, all indirectly D,,, D3, and clamped. Over-Voltage from Indirect Clamping 8.2.4. Considering the commutation process from (a). and and it°s after After the turn-off of S,”, the shown D2 to D,,, DS3, D,2, Ds, in cell demagnetization of the parasitic inductance LS in the clamping path causes over-charging of S,° S4” (stray capacitances), as S,”, D,2, D3, (stray capacitance) in Fig. 8.8. and discharging of S2”, S3” and Such over-charging and discharging will be maintained because the discharging/charging paths are blocked by D2, Consequently, and D2, D,,, S,° D,, and D,2. blocks higher than Vdc/4 while S2”, S3”, S4” together block less than 3V,,2/4 D,2 together sees the voltage difference between VS,» and Vdc/4. In the meanwhile, D,, sees Vdc/4 plus VD2 plus VD8, D, sees Vdc/4 plus VD2, while D, sees Vdc/4. The S2, S3' and Due indirect S3, clamping and the subsequent over-voltage problem holds also for and S,, together with their relevant diodes in cell(b), to the fact that the stray capacitance cell(c) and S2” and cell(d). of the neighboring outer switch experiences once more discharging, the neighboring outer device (switch and its relevant diodes) always blocks less voltage while the inner device always blocks more voltage. The innermost device, of course, is always exposed to the highest voltage stress. 127 Over-voltage problem arising from the indirect clamping exists with any diode clamping multilevel inverter. Measures such as refined construction and appropriate snubbing are believed to be helpful in mitigating this problem. Moreover, clamping paths are made bi-directional by anti-paralleling to can be eliminated it if the each clamping diode a small- rating switch. Cl + 1, S. D1 À 51.4 Dzi 52 Dsz °'| cz D2 Q 9› 5 C3 D4 D1 . >J>b-I Í ,_ __ D5 A :I C4 .› .F5 P :Há D9 r¡;`¬ A ‹4¬ q" 52' ‹›¬ zz 5 »« S4 Fig. 8.8. Illustration 8.3. of the over-voltage across E S,?, . ¡ S3. ‹-i DS4, D, and D,, due to demagnetization of Ls. Snubbing/Soft Switching of the New Inverter The true attractiveness installation where designing plays a inverter GTOs role. are of the multilevel inverter consists in the expanded capacity of an most Hence snubber likely the choice as the switching element. However, snubbing techniques (M>3) have not beencovered until the for the diode most recent clamping multilevel in the literature, probably due to the underlying difficulty in arranging efficient turn-on snubbers for the inner switching cells with bi-directional busses. Non-polarized turn-on snubbers are especially inefficient. When GTOS the snubbing are not the choice of device, the clamping scheme shown scheme shown in Fig. 8.10 shall apply. In Fig. 8.10, instead limiting di/dt over the full load range, saturable reactors are used the right moment to limit the diode reverse recovery overshoot. in Fig. 8.9 of and linear reactors which go out of saturation at V 128 F É i ‹~ Fi z '*='=¡~§<~ ‹› _" gl ‹‹›-lp |ü,‹ - ~ " __ Fig. 8.9. » LE.. ' Clamping scheme for the 1. _' ‹› Í" new diode clamping new shown inverter, as in Fig. 7.7 switching and Fig. in Fig. 8.11 A ~ - ~ - 4:- ‹~ 1. ~ -Dl . *Q Fig. 8.10. True-PWM-Pole and ARCPI shown ‹›-| vê ~ Snubbing scheme for the new diode clamping multilevel multilevel inverter, extensible to any level. In addition, both this = __ ‹. ‹. *A inverter, extensible to soft switching any level methods are applicable to and Fig. 8.12. In comparison with the arrangements 7.8, the auxiliary branches are installed based on each individual 7 cell. __i 1 › A__ A l L Aí P Z A A A A A › inililninlnnrlnilni ÉÉHHÉH A P Í .L L 'HHHHHH › ¡¡¡¿ t Fig. 8.11. AAA True-PWM-Pole ñve each switching cell. ‹¿_ T °_' ‹¿` °_¡ _ P A level diode . I ¿_ è L €A L L ¿_ 4. ‹›_| °-I-I» clamping multilevel Main circuitry ZVS e i. fl .I . inverter. Auxiliary switching, auxiliary circuitry branch is arranged for ZCS switching. 129 4? ‹f<§-- `_'*-_;¡_ _€_ ‹¿ L .zgg Q. A - mÉ °* _2F,'-. ‹¿ ARCPI five level diode clamping multilevel inverter. Auxiliary branch is arranged for each switching -«nz»~ 1 . Í- 1 IG* 1 Fig. 8.12. - cell. Main circuitry ZVS switching, auxiliary circuitry ZCS switching. 8.4,¿Experimentation A scaled half bridge prototype has been built for verification of the new inverter. Four 120V DC power sources each in series with a 10mH inductor establish the four separate DC supplies for the DC link. A 8mH inductor in series with a l2Q resistor are comiected at the shown in Fig. vv output, as 5. and-'the 7th in harmonic is Fundamental frequency modulation scheme eliminating the 5th 8.13. implemented. The BASIC program for generating the gating signals EPROM 27256 is listed in Appendix 8.1. Experimental results are shown in Fig. 8.14. ' P 5¡..¡ Cl sz°-l‹^ ¡ó|+iazDl55 C2 O D7 .a1+|s 'd3+'d4m _ C3 D2 __ _"~ D 4 D5 id5+id6 Dó ^ Q D8 s3.›4 4 DH ¿ 94 S4 ¡au+¡‹u D9 DS z Ds; A D54 30114 a . S¡o¬| ¡ó9+¡‹1|o Dsl DH, .¡l› I F3 ‹A Ds2' Á DIO s3'°'¡ C4 s4' ' O ‹^ D 5 3' H ‹A DS4' N Fig. 8.13. Experimental set-up of the proposed diode clamping inverter. C1=C2=C3=C4=3300uF, D1-D12: HFA30TA60C, Sl-S4': SKM50GBl23D. j~¬¿`Í'~¬ l Oifiiffi f f 7'f_1*I*í iao 100V/Div; 20A/Div; 4mS/Div Fig. 8. l4(a). Output voltage and load current waveforms. Vao M ¡ô1+â‹12 í 100V/Div; 20A/Div; 4mS/Div Fig. 8. l4(b). Sum of Dl and D2 currents in relation to the output voltage. Vao P' 1! id +id Í' 100V/Div; 20A/Div; 4mS/Div Fig. 8. l4(c). Sum of D3 and D4 currents in relation to the output voltage. Y Vac 100V/Div; 20A/Div; 4mS/Div Fig. 8. l4(d). Sum of D5 and D6 currents in relation to the output voltage. Vao v id7+id -‹. l00V/Div; 20A/Div; 4mS/Div Fig. 8.14(e). Sum of D7 and D8 currents in relation to the output voltage. 'Vao V id9 idl0 100V/Div; 20A/Div; 4mS/Div Fig. 8.l4(f). Sum of D9 and Dl0 currents in relation to the output voltage. Vao ` i‹111+¡ 12 100V/Div; l0A/Div; 4mS/Div Fig. 8.l4(g). Sum of Dll and DI2 currents Vao in relation to the output voltage . Vs4 100V/Div; 4mS/Div Fig. 8.l4(h). S4 blocking voltage in relation to the output voltage. Vao Vs3 l00V/Div; 4mS/Div Fig. 8.l4(i). S3 blocking voltage in relation to the output voltage. -É Vao Vs2 ` l00V/Div; 4mS/Div Fig. 8.l4(j). g S2 blocking voltage in relation to the output voltage Vao _ Í r 'Í Vsl ` "Í I I L.-Id l00V/Div; 4mS/Div Fig. 8. l4(k). Sl blocking voltage in relation to the output voltage Vpa 7"' I _, f ___ _.___ _ _ l00V/Div; 4mS/Div Fig. 8.l4(l). Up-ar1n(Sl+S2+S3+S4) blocking voltage. 134 8 5. Conclusions Due to the multiple blocking voltages, series associations of clamping diodes are needed in the conventional diode clamping multilevel inverter, voltage sharing problem among which the diodes in series. Large results in static and dynamic RC network dealing with this problem leading to voluminous and expensive system. The proposed new diode clamping multilevel inverter is structured with built-in clamping components in the new for all the clamping diodes in the clamping network. Thus circuit are clamped. The problem of series association of clamping diodes with the conventional diode clamping multilevel inverter structure is all solved. Over-voltage across the inner devices as a result of indirect clamping the conventional inverter and the new inverter. Current ' is inherent with both loop inductances play the key role in exciting the problem. Reasonable efforts in layout designing, device selecting (low internal inductance) as well as snubber positioning are helpful in mitigating the problem. Appendix BASIC program for generating the gating signals in EPROM 27256M 8.1. 1120 U = 1 'DATA OUTPUT LINES COUNTER 1130 OPEN "A:MULTD.DAT" FOR OUTPUT AS #1 1 140 PRINT #1, ": 10"; 'FIRST DATA OUTPUT LINE FORMAT 1150 PRINT #1, "000000"; 'FIRST DATA OUTPUT LINE FORMAT 1160 FORK = 0 TO 16383 1170 IF K <= 5498 ORK>= 10883 AND K <=16383 THEN S11 = 1 ELSE S11= 0 1180 IF K>= 5500 AND K <= 10881 THEN S1=1 ELSE S1= 0 1.190 IF K <= 4340 OR K >= 12041 AND K <=16383 THEN S22 = 1 ELSE S22 = 0 1200 IF K >= 4342 AND K <= 12039 THEN S2 = 1 ELSE S2 = 0 1210 IF K <= 3847 OR K >= 12534 AND K <=16383 THEN S33 = 1 ELSE S33 = 0 1220 IF K >= 3849 AND K <= 12532 THEN S3 = 1 ELSE S3 = 0 1230 IF K <= 2689 OR K >= 13692 AND K <= 16384 THEN S44 = 1 ELSE S44 = 0 1240 IF K >= 2691 AND K <= 13690 THEN S4 = 1 ELSE S4 = 0 135 4 1250A=S22*2^7+S11*2^6+S4*2^5+S3*2^4+S44*2^3+S33*2^2+S2*2^1+S1*2^ O ' PRINT #1, 1260 1270 1280 1290 1300 1320 ¢› 133.0 S1; S11; S2; S22; S3; S33; S4; S44; K; 16 THEN IFK=U* U=U+ 1 IF TOTAL < 255.5 THEN IF TOTAL < 15.5 THEN PRINT #1, ELSE PRINT #1, "00"; "0"; HEX$(TOTAL) HEX$(TOTAL) 1350 END IF 1360 ELSE 1370 PRINT #1, HEX$(TOTAL) 1390 END IF 1400 TOTAL = 0 1420 PRINT #1, ":"; HEX$(K / (U - 1)); 1430 IF K < 16 THEN 1440 PRINT #1, "000"; 1450 ELSE 1460 IF K < 256 THEN 1470 PRINT #1, "00"; 1480 ELSE 1490 IF K < 4096 THEN PRINT #1, "0"; 1500 1510 END IF 1520 END IF 1530 END IF 1540 PRINT #1, HEX$(K); "00"; 1550 END IF 1560 TOTAL = TOTAL + A 1570 IF A < 15.5 THEN 1580 PRINT #1, "0"; HEX$(A); 1590 ELSE 1600 PRINT #1, I-IEX$(A); 1610 END IF 1620 NEXT K 1630 IF TOTAL < 256 THEN 1640 PRINT #1, "0"; HEX$(TOTAL); 1650 ELSE 1660 PRINT #1, HEX$(TOTAL); 1670 END IF 1680 CLOSE 1690 END O 136 A Chapter This thesis is General Conclusions 9. dedicated to the soft switching techniques of the multilevel inverters. Included are also the background treatments as regards the main circuits. Chapter 1 of this thesis deals with the ftmdamentals of the multilevel inverters. The existing approaches for implementing high voltage high power inverters utilizing device association or cell association are inspected first. While series association of devices offers advantage in terms of redundancy, the output waveform and dv/dt are not favored from the multiple devices. Converterpassociation, on the other hand, necessitates insulation either the AC side or the evolved from DC side cell association, The diode clamping clamping diodes (NPC) inverter series which which means additional magnetics. The multilevel inverters however, generate multi-step outputs without heavy magnetics. multilevel inverter suffers from indirect clamping, potential drift, and snubbing problems etc; Except for the Neutral-Point-Clamped is free from the last three problems, practical application can hardly be realized before relevant solutions are verified. Theoretically, the inverter is relieved potential drift on problem when pure reactive power The capacitor clamping inverter is is from the transferred. exempt from the above problems. However other shortcomings arise such as the parasitic inductances resulted from the clamping capacitors, as well as the heat capacity required for these capacitors. It seems more applicable than the diode clamping inverter in the high level case especially when active power transfer is involved. This chapter also reviews the snubber schemes reported for the which snubbers for two-level inverter are compared. such objectionable problems as snubber as the adverse effects NPC Before The use of dissipative snubber leads loss, voltage/current spikes, series reactor loss as to well from the snubbing diodes. These problems become more pronounced for multilevel inverters. For regenerative snubber, even though the snubber loss expected increase in operating frequency circuitry, inverter. is is recovered, the limited by the complexity of the recovery while the other problems remain. Chapter 2 assesses the reported two-level inverter proposes the transformer connection True-PWM-Pole Resonant-DC-Link-Inverter (ACRLI) is soft switching techniques circuit. and While the Actively-Clamped not regarded suitable for power level beyond 300kW because of the cost, size and cooling issues of the resonant inductor, the Auxiliary-ResonantCommutated-Pole-Inverter (ARCPI) suffers from the measuring and controlling difficulties to guarantee the zero voltage switching especially when the DC 137 center tap potential drift is taken into account. The transformer comiection True-PWM-Pole circuit guarantees zero voltage switching of the main device without any of the problems associated with the conventional snubber and without any additional measuring or controlling. In the meanwhile, zero current switching of the auxiliary device is simultaneously attained without the affliction of voltage spike due to the bridge configuration, baby snubber is which is a significant advantage over the regenerative snubber where always required for the recovery chopper in hard switching. Depending on the designing, the auxiliary device can be rated at about 1/4 of the loss due to the resonance transition is main device. The duty cycle comparable to that of the conventional snubbed inverter. The transformer comection True-PWM-Pole circuit is verified by a 5kW prototype. Besides transformer connection, auto-transformer connection and capacitor connection True-PWM-Poles .W .._,., are also discussed. halved, the freewheeling current Even though the auxiliary device current stress flow during the non-comrnutation interval is further evolved from the ,-'ya magnetizing current renders the normal auto-transformer connection hardly applicable before practical measures are found. . The freewheeling current problem family, which is verified by a is solved in the 5kW prototype. A modified auto-transformer connection simple snubbing is needed for the series diode to suppress the voltage spike. The variations of this family allow for further reduction in the auto-transformer voltage/current rating or the resonant inductor current rating, which are H also tested in the laboratory. The capacitor connection circuit, on the other hand, eliminates the magnetics of the fairly at the price augmented resonant circuit stress. In Chapter 3, the clamping voltage stability in the three level capacitor clamping inverter is explored. the inverter load is Under sub-harmonic modulation, the clamping voltage The time constant of which dependent on the load properties, modulation index, clarnping capacitance When etc. asymmetries are taken into account, typically gating and switching mismatches, the clamping voltage result stable as far as not pure-reactive, due to the inherent feedback mechanism. The clamping voltage response exhibits typical one-order system characteristics. is is is balanced at a new operation point other than the nominal value, as a of the joint effects of the self balancing ability and the asymmetry. In addition, dead 138 time is to the not expected to cause any clamping voltage two switching drift, because of its symmetrical influence cells during the local switching cycle. With sub-harmonic modulation, active control over the clarnping voltage necessary can be realized by vertically shifting oneof the triangle carriers according to the Clamping voltage deviation direction and the load current direction. where adjusted is on the switching cycle basis. The self-balancing ability is proved in a half-bridge three level prototype system. Chapter 4 introduces a True-PWM-Pole three installing the transfonner connection voltage switching for all the True-PWM-Pole level capacitor to clamping inverter. each of the two switching main devices and zero current switching .for all cells, By zero the auxiliary devices are secured, with the same performance as that in the two level inverter. In particular, the proposed True-PWM-Pole arrangement further guarantees the stiffness of the for the outer switching cell clamping voltage due to the damped second order charging/discharging paths established by the auxiliary circuitry. auto-transformer connection True-PWM-Pole is because the freewheeling current path in this case now is Most positively, the normal applicable to the outer switching cell blocked by the clamping capacitor. The transformer connection True-PWM-Pole technique along with the AuxiliaryResonant-Commutated-Pole-Inverter (ARCPI) both are extensible to the capacitor clarnping multilevel inverter (M>3). Two general zero voltage switching topologies for the capacitor clamping multilevel inverter (M>3) are obtained. In Chapter 5, the clamping inverter is modules each rated 3kW experimental set-up for the half bridge three level capacitor described in details. at The prototype employs four 1200V/50A driven by four two main modules and two double SEMIKRON IGBT single intelligent drivers intelligent drivers (SKHI23) (SKHI10) for the two auxiliary modules. is inserted. Gating signals are for the Between the driver and device, a zero voltage detecting circuit generated in EPROM 27256M, and time durations including the auxiliary gating signal width and the main gating signal width are set as discussed in Chapter 2. The proposed circuit and the analyses are well verified through the experimental results from this prototype. Chapter 6 is concemed with the neutral potential stability in the sub-harmonic modulation, the neutral potential reactive, is NPC inverter. Under stable as far as the inverter load is not pure- due to the inherent feedback mechanism. The neutral potential response exhibits 139 The time constant of which typical one-order system characteristics. properties, modulation index, When joint effects is dependent on the load DC link capacitances etc. asyrnrnetries are taken into account, typically gating the neutral potential is balanced a at of the self balancing new ability expected to cause any neutral potential and switching mismatches, operation point other than zero, dependent on the and the Vasyrnmetry. Similarly, the dead time drift is not because of its symmetrical influence to the two switching cells during the ftmdamental cycle. When Space-Vector-Modulation (SVM) is used, the need for appropriate distribution of the redundant states of the small vector arises, active control over the neutral potential becomes indispensable. On the other hand, with sub-harmonic modulation, when needed can be effected by vertically shifting the deviation direction and the operation , ¿¿¡« M,-4. ".*l'2' , zz modulating reference according to the mode (rectifying or inverting). Self-balancing under ideal and non-ideal conditions are verified by Chapter 7 proposes a True-PWM-Pole three level NPC PSPICE simulations. all secured, with the ú ¿¿_¡_ the main devices and zero current switching for same performance as that in the two By installing the cells, zero voltage inverter. transformer comrection True-PWM-Pole to each of the two switching switching for active control all the auxiliary devices are level inverter. , The transformer connection True-PWM-Pole technique along with the Auxiliary- Resonant-Commutated-Pole-Inverter (ARCPI) both are extensible' to the diode clamping The multilevel inverter (M>3). basis. The auxiliary circuitry can be configured on the per-cell or per-leg per-leg structure needs only one resonant inductor and one auxiliary transforrner, while in the per-cell case, each structure suffers from the cell needs an inductor and a transforrner. However, the per-leg indirect clarnping problem also in the auxiliary circuitry, and moreover, resonant current flowing involves multiple devices in the path. Such extension leads clamping multilevel T Finally, in to four general zero voltage switching topologies for the diode inverter. Chapter 8, a new diode clamping multilevel inverter is proposed, which solves the problem of series association of diodes in the conventional diode clamping inverter. However, the other aspects of operation remain the same as the conventional one. In summary, the following conclusions are drawn from this thesis: 140 - 0 Multilevel inverters enable high voltage operation and multi-step output without requiring any magnetics insulation. The clamping voltage 0 clamping inverter and the neutral potential in in the capacitor diode clamping inverter are both self-balancing under sub-harmonic modulation so long as the load is not pure reactive. 0 The use of snubber leads to the reduction of the switching stress but not necessarily any significant increase in the switching frequency. 0 › Transformer connection True-PWM-Pole technique offers an altemative of the ARCPI and enables high frequency high power operation. Transformer connection True-PWM-Pole technique _0 inverters with the same performance as in the two level case, is extensible to multilevel which facilitates high voltage and high power conversion at high frequency and high efficiency. 0 The new diode clamping clamping diodes and To is multilevel inverter works free of the series association of hence more applicable in practice. the best of the belief of this thesis, the soft switched multilevel inverter legitimate contender against the well-entrenched hard switched or is the snubbed counterpart especially for high performance applications, even though they face greater obstacles than at the two-level low power levels. ' Looking ahead, several subjects deserve further specific investigations: 0 Identification of potential problems, comparison of its variations, optimization of designing and evaluation of performance of the modified auto-transformer connection True-PWMPole family under practical circumstances. 0 Possible clamping voltage drift of the capacitor_clamping inverter drift 0 and neutral potential of the diode clamping inverter under load unbalance or load transience conditions. Solutions to the problems of the diode clamping multilevel inverter (M>3), including H indirect clamping, potential drift 0 and turn-on snubbing Comparison of the diode clamping etc. inverter, the capacitor bridge cascade inverter for different applications. clamping inverter and the H- 141 References [1] G. Kaplan, “Industrial Electronics”, IEEE Spectrum, January 1997, pp. 79-83. [2] H. Stemmler, “High-Power Industrial Drives”, Proceedings of the IEEE, Vol. 82, No. 8, ` pp. 1266-12s6,Augus1 1994. ' . H. Stemmler, “Power Electronics in Electric Traction Applications”, Record of IEEE [3] IECON, 1993, pp. 707-713. Y. Sekine, K. Takahashi and T. Hayashi, “Application of Power Electronics Technology [4] Bulk Power Transmission to the 21st Centu1y°s Energy Systems, Vol. [5]=:P; No. 17, 3, pp. in Japan”, Journal of Electrical Power & 181-193, 1995. Hofer, N. Karrer and C. Gerster, “Paralleling Intelligent IGBT Modules with Active Gate-Controlled Current Balancing”, Record of IEEE PESC, 1996, pp. 1312-1316. Ch. Keller and Y. Tadros, “Are Paralleled [6] 1 _ [7] Better Choice ?”, Record of EPE Conference, 1993, pp. 5.1-5.6. P. Steimer, H. Gruning, J Weminger, P. Dahler and G. Linhpfer, “Series Connection of .- -GTO High Power Thyristors for 1996, pp. 14-20. [8] C. Schauder, Systems”, Static Frequency Converters”, _ E. Stacey, T. Lemak, L. Gyugyi, H. Suzuki, T. Nakajima, K. Izumi, S. 10, No. 1997. 5, T. W. Cease and A. Edris, “ Trans. on Power 3, pp. 1486-1496, July 1995. Sugimuto, Y. Mino and H. Abe, “Development and Testing of Prototype Models for a High Performance IEEE Review, No. MVAR Static Condenser for Volatge 'Control of Transmission IEEE Trans. on Power Delivery, Vol. Converter”, ABB . M. Gemhardt, Development of A -100 [9] IGBT Modules or Paralleled IGBT Inverters the 300MW Self-Commutated AC/DC Delivery, Vol. 12, No. 4, pp. 1589-1601, October ` [10] P. K. Steimer, H. E. Gruning, J. Weminger, E. Carroll, S. Klaka and S. Linder, New Technology for High Power Low Cost Inverters”, Record of IEEE IAS, “IGCT - a 1997, pp. 1592-1599. [11] C. Gerster, P. Hofer and N. Karrer, “Gate-Control Strategies for Snubberless Operation of Series Connected IGBTS”, Record of IEEE PESC, 1996, pp. 1739-1747. 142 [12] Sigg, J. M. Bruckmann and P. Turkes, “The Series Connection of IGBTs Investigated by Experiments and Simulation”, Record of IEEE PESC, 1996, pp. 1760-1765. [13] S. Mori, K. Matsuno, M. Takeda and M. Seto, “Development of a Large Using Self Commutated Inverters for Improving Power Power Systems, Vol. 8, No. IEEE Trans. on February 1993, pp. 371-377. 1, D. Wuest, H. Stemmler and G. Scheuer, [14] Stability”, VAR Generator Configurations for an Advanced Static “A Comparison of Different Circuit Var Compensator (ASVC)”, Record of IEEE PESC, 1992, pp.521-529. [15] K. Matsui, “A Pulse Width 'Modulated Inverter with Parallel Connected Transistors Using Current-Sharing Reactor”, Record of IEEE IAS, 1985, pp. 1015-1019. and M. Morozowich, “30- [16] R. Hoft, T. Khuwatsarnrit, A. Foldes Auxiliary Power Supply for People Mover, Part App., Vol. 19, No. [17] T. A. 5, Sep./Oct. I- Power kVA Transistor Inverter Circuit”, IEEE Trans. on Ind. 1983, pp. 717-724. Megnard and H. Foch, “Multilevel Converters. and Derived Topologies for Power Conversion”, Record of IEEE IECON, 1995, pp. 21-26. [18] T. Salzmann, G. Kratz, C. Daubler, High ' “High Power Drive System with Advanced Power Circuitry and Improved.Digital Control”, IEEE on Trans. Ind. App., Vol. 29, No. 1, Jan./Feb. 1993, pp. 168-174. [19] P. G. Kamp, “High Power Water. Cooled Three-Level GTO Modules for the Static Frequency Converter Muldenstein”, Record of EPE Conference, 1995, pp.l. 847-1. 851. [20] H. Stemmler, P. Guggenbach,_“Configurations of High-Power Voltage Source Inverter Drives”, Record of [21] EPE Conference, W. Schminke, “High Power Pulse Medium Wave [22] Step Modulator for Short Vol. 72, No. Wave and 600KW 5, 1985, pp. 235- 1 M. Marchsoni, M. Mazzucchelli and for 500KW Brown Bovery Review, Transmitters”, 240. 1993, pp. 1.7-1.14. Plasma Stabilization”, IEEE S. Trans. Tenconi, “A Non-conventional Power Converter on Power Electronics, Vol. 5, No.,2, April 1990, pp. 212-219. [23] T. Ghiara, M. Marchesoni, L Puglisi and G. Sciutto, “A Modular Approach to Converter Design for High Power AC Drives”, Record of EPE Conference, 1991, pp. 4.477-4.482. [24] P. W. Hammond, “A New Approach to Enhance Power Quality for Medium Voltage AC Drives”, IEEE Trans on Ind. App., Vol. 33, No. 1, Jan./Feb. 1997, pp. 202-208. 143 N. Schibli and Ch. Briguet, “Direct Coupled 4-Quadrant Multilevel Converter [25] A. Rufer, for 16 2/3 Hembert, F. [26] HZ Traction Systems” Le Moigne and P. Record of IEE PEVD, 1997. P. J. Cambromie, “Series Association of VSI in High Power Active Filtering”, Record of IEEE PESC, 1996, pp. 759-764. F. Z. Peng, [27] Separate S. Lai J. and J. McKeever, “A Multilevel Voltage Source Inverter with DC Sources for Static Var Generation”, Record of IEEE IAS, 1995, pp. 2541- 2548. Fang Zheng Peng and Jih-Sheng [28] Lai, “Dynamic Performance and Control of A Static VAR Compensator Using Cascade Multilevel Inverter”, Record of IEEE IAS, 1996, pp. 1009-1015. and [29] D. Perreult J. Kassakian, “An Assessment of Converter Systems”, Record of the 15' Intemational Cellular Architectures for Large Power Electronics and Motion Control Conference (IPEMC), 1994, pp. 70-79. [30] T. A. Meynard and P. ' Davancens, “Current Balance in Paralleled Commutation Cells”, Record of PCIM, 1995. Barbi, “A New Technique for Parallel Connection of Commutation Cells: Analysis, Designing and Experimentation”, IEEE Trans. on Power Electronics, [31] H. A. C. _ Braga and Vol. 12, No. 2, [32]_._¡ A. Nabae, Holtz and J. March 1997, pp. 387-393. Takahashi, and A. Akagi, I. IEEE Trans. [33] I. Ind. App., Vol. 19, S. Stadtfeld, “A New Neutral-Point Clamped PWM No. 5, Sep./Oct. 1981, pp. 518-523. “An Economic Very High Power PWM Inverter”, ' Inverter for Induction Motor Drives”, Record of EPE Conference, 1985, pp. 3.375-3.380. [34] J. M. Andrejak and M. Lescure, “High Voltage Converters Promising Technological Developments” Record of EPE Conference, 1987, pp. 1.159-1.162; , [35] N. S. Choi, Inverter”, [36] M. J. G. Cho, and G. H. Cho, “A _ General Circuit Topology of Multilevel Record of IEEE PESC, 1991, pp. 96-103. Carpita, S. Tenconi, “A Novel Multilevel Structure for Voltage Source Inverter”, EPE Conference, 1991, pp. 1. 90-1. 94. Yuan and I. Barbi, “A New Structure for Multilevel Record of [37] X. INEP-UFSC, [38] T. Brazil, Mamyama March 1997 Inverter”, Research Report, . and M. Kumano, “New Record of IPEC, 1990, pp. 870-877. PWM Control for A Three-Level Inverter”, 144 [39] T. Meynard and H. Foch, “Multi-Level Conversion: High Voltage Choppers and Voltage Source Inverters”, Record of IEEE PESC, 1992, pp. 397-403. [40] G. Carrara, S. Gardella, PWM A Method: M. Marchesoni R. Theoretical Analysis”, Salutari IEEE and G. Trans. Sciutto, “A New Multi-Level on Power Electronics, Vol.7, No.3, July 1992, pp. 497-505. [41] Shen and N. Butterworth, “Analysis and Design of a Three Level J. System [42] for Railway Traction Applications”, IEE Pro. Elec. Power App., Vol. 144, No. 5, Sep. 1997, pp. 357-371. J. Holtz, “Pulsewidth Modulation in Electronic IEEE, Vol. 82, No. [43] H. PWM Converter 8, Power Conversion”, Proceedings of the August 1994, pp. 1194-1214. W. Van Der Brock, H. C. Skudelny and J. V. Stanke, “Analysis and Realization of a Pulsewidth Modulator Based on Voltage Space Vectors”, Vol. 24, No. 1, IEEE Trans. on Ind. App., Jan./Feb. 1988, pp. 142-150. Holmes, “The Significance of Zero Space Vector Placement for Carrier Based [44] D. G. PWM Schemes”, Record of IEEE IAS, 1995, PP. 2451-2458. [45] Y. Tadros, Saluma, and R. Hof, “Three Level S. IGBT Inverter”, Record of IEEE PESC, 1992, pp.46-52. [46] M. Koyama, PWM Method IECON, [47] S. T. Fujji, R. Uchida and T. Kawabata “Space Voltage Vector Based for Large Capacity Three Level GTO Thyristors”, Tamai, M. Koyama, T. EPE Conference, J. Fujii, S. Mizoguchi and T. Kawabata, “3 Level 1993, pp. 45-50. Motor Drive”, Record of PWM Control of a Three Level Inverter”, _ Vol. 7, No. 3, July 1992, pp. 487-496. M. Marchesoni, “High Performance Current Control Techniques Multilevel High Vol. 7, No. [50] F. Z. Peng, Power Voltage Source 1, Jan. 1992, pp. 189-204. J. S. Lai and J. GTO _ Steinke, “Switching Frequency Optimal IEEE Trans. on Power Electronics”, [49] Record of IEEE 1992, pp. 271-276. Converter-Inverter air System for Large Capacity Induction [48] New Inverters”, IEEE Trans. on for Applications to Power Electronics, MeKeever, “A Multilevel Voltage-Source Converter System with Balanced DC Voltages”, Record of IEEE PESC, 1995, pp.1144-1150. F. Z. [51] Peng and J. 145 “A S. Lai, Static Var Generator Using a Staircase Waveform Multilevel Voltage Source Converter”, Record of Power Quality Conference, Sep. 1994, I pp. 58-66. [52] R. W. Menzies, and P. Steimer J. D Induction Motor Drives”, IEEE K. Steinke, “Five Level .GTO Inverter for Large Trans. on Ind. App., Vol. 30, No. 4, -July/Aug. 1994, pp. 938-944. [53] Y. Chen, B. Mwinyiwiwa, Z. Wolanski and B. T. Ooi, (UPFC) Based on Chopper “Unified Power Flow Controller Record of IEEE PESC, Stabilized Multilevel Converter”, A 1997. G. Sinha and T. A. Lipo, [54] “A Four Level Rectifier-Inverter System for Drives Applications”, Record of IEEE IAS, 1996, pp. 980-987. [55] G. Sinha and T. Lipo, “Rectifier Current Regulation in Four Level Drives”, Record of IEEE APEC, [56]- 1997, pp. 320-326. C. Hochgraf and R. Lasseter, Employing a Multilevel “A Inverter”, Transformer-less Static Synchronous Compensator IEEE Trans. on Power Delivery, Vol. April 1997, pp. 881-887. [57] Y. Chen, B. Mwinyiwiwa, I ,. [58] B. No. 2, No. 2, « Z. Wolanski and B. T. Ooi, “Regulating and Equalizing Capacitance Voltages in Multilevel 12, 12, STATCOM”, IEEE Trans. on Power Delivery, Vol. April 1997, pp. 901-907. Mwinyiwiwa, Z. Wolanski, B. T. DC _ Ooi and Y. Chen, “Multilevel Converters as Series VAR Compensators”, Record of IEEE PESC, 1997. [59] O. Pollakowski, H. Pouliquen and W. Schumacher, “State-Space Analysis of Diode- Clamped Multilevel Voltage Source EPE Conference, [60] G. Scheuer Inverters for Static Var Compensation”, Record of 1997. - and H. Stemmler, “Analysis of a Three Level VSI Neutral Point Control Fundamental Frequency Modulated Conference on AC and SVC Applications”, for Record of the 6th Intemational DC Transmission, London, March 1996, pp..303-310. _ [61] R. Rojas, T. Ohnishi and T. Suzuki, “An Improved Voltage Vector Control Method for Neutral-Point-Clamped Inverters”, IEEE Trans. on Power Electronics, Vol. 10, No. Nov., 1995, pp. óõó-672. 6, 146 [62] M. Koyama, H. Okayama, GTO Controlled Three Level “High Performance Vector T. Tsuchiya and Y. Kinpara, Inverter for Electric Traction”, Record of IPEC- Yokohama, 1995, pp. 766-771. [63] S. Ogasawara and H. Akagi, “Analysis and Variation of Neutral Point Potential Neutral-Point-Clamped Voltage Source PWM Inverters”, in Record of IEEE IAS, 1993, pp. 965-970. [64] C. Newton and M. Sumner, “Neutral Point Control for Multilevel Inverters: Theory, Design and Operational Limitations”, Record of IEEE IAS, 1997, pp. 1336-1343. [65] H. Okayama etc., “Large Capacity High Performance 3-Level Record of IEEE IAS, 1996, pp. 174-179. Steel Rolling Mill Drives”, W. McMurray, “Resonant Snubbers with [66] App., Vol. 29, No. [67] 2, GTO Inverter System for Auxiliary Switches”, IEEE Trans. on Ind. March/April 1993, pp. 355-362. H. Okayama, T. Tsuchiya and M. Kimata, “Novel Gate Power Supply Circuit Using GTO Valves”, Record of EPE G. L. Skibinski and D. M. Divan, “Design Methodology and Modeling of Low Snubber Capacitor Energy for Series Cormected Conference, 1997. [68] Inductance Bus Structures”, Record of EPE Conference, 1993, pp. 5.98-5.105. [69] E. Clavel, IGBT J. Roudet and Y. Marechal, “Design of a Commutation Cell of a High Power Inverter - The Contribution of the Simulation”, Record of IEEE IAS, 1997, pp. 1014-1021. [70] I-I. ' Miyazaki, H. Fukumoto, S. Sugiyama, M. Tachikawa and N. Azusawa, “Neutral- Point-Clamped Inverter with Parallel Driving of IGBTS for Industrial Applications”, Record of IEEE IAS, 1997, pp. 1292-1299. [71] T. Meynard and H. Foch, “Multi-Level Choppers Journal of EPE, Vol. 2, No. [72] F. Hamma, T. A. Meynard, 1, F. P. Carrère, T. Meynard and Leg”, Record of [74] T. A. High Voltage Applications”, March 1992, pp. 45-48. Tourkhani and P. Viarouge, “Characteristics and Design of Multilevel Choppers”, Record of IEEE [73] for J. PESC, 1995, pp. 1208-1214. P. Lavieville, EPE Conference, “4000V-300A Eight-Level IGBT Inverter 1995, pp. 1106-1111. Meynard, M. Fadel and N. Aouda, “Modeling of Multilevel Converters”, IEEE Trans. on Ind. Elec., Vol. 44, No. 3, June 1997, pp. 1-9. 147 IEEE A. Steimel, “Electric Railway Traction in Europe”, [75] Industry Applications J Magazine, Nov./Dec. 1996, pp. 7-17. W. McMurray, [76] Transistors Switching Converters”, IEEE July/August 1980, pp. 513-525. [77] Clamps “Selection of Snubbers and Trans. on to Optimize the Design of Ind. App., Vol. No. 4, _ M. H. Pong and R. D. Kackson, “Computer Aided Design of Power Snubber Circuits”, IEE Proceedings, Vol. 134, Part B, No. T. Undeland, F. Jenset, A. Steinbakk, T. [78] 16, Configuration for Both Power Transistors and Transistor Inverter March 1987, pp. 69-78. 2, Ronge and M. Hemes, “A Snubber GTO Inverters”, Record of IEEE PESC, 1984, pp. 42-53.' [79] R. Marquardt, “High Power ICE”, Record of EPE Conference, W. Runge and A. [80] GTO Steimel, Converters for The New German High Speed Train 1989, pp. 1.583-1.588. “Some Aspects of Circuit Design of High Power Converters”, Record of EPE Conference, 1989, pp. 1.555-1.560. [81] W. McMurray, “Efficient Snubbers for Voltage Source Power Electronics, Vol. [82] V. Gilblas and F. Favo, 2, No. 3, GTO V GTO Inverters”, IEEE Trans. on July 1987, pp. 264-272. “Low Losses New Snubber Circuit for High Power GTO Inverter Application: Tests and Results”, Record of EPE Conference, 1991, pp. 2-534-2.539. [83] W. W. Fuhrer, R. Marquardt and G. Papp, “Water Cooled for Electric Traction”, [84] M. GTO Converters Record of EPE Conference, 1993, pp. 5.294-5.298. Jung, “Improved Snubber for Passive Network”, Record of [85] High Power GTO Inverters with EPE Conference, Energy Recovery by Simple 1987, pp.15-20. A. Brambilla and E. Dallago, “Snubber Circuits and Losses of Voltage Source Inverters”, IEEE Trans. on Power Electronics, Vol. [86] B. Snyners, Ph. Lataire, G. Voltage Source _ 7, No. 1, GTO Jan. 1992, pp. 231-239. Maggeto, B. Detemrnerman, and J. M. Bodson, “Improved GTO Inverter with New Snubber Design”, Record of EPE Conference, 1987, pp. 31-36. [87] B. W. Williams, “High Voltage High Frequency Power Switching Transistor Module with Switching Aid Circuit Energy Recovery”, IEE Proceedings, Vol. 131, Part B, No. 1, Jan. 1984, pp. 7-12. [88] J A. . Tarfiq and Y. Shakweh, “New Snubber Energy Recovery Scheme 1 Traction Drive”, Record of IPEC-Yokohama, 1995, pp. 825-830. for High Power 148 [89] G. Fregien, H. G. Langer and H. C. Skudelny, GTO Inverter”, Record of IEEE 1988, pp. 498-505. “A New Snubber Energy Irokawa, T. Kitahara, F. Ichikawa and T. Nakajima, S. [90] PESC, “A Regenerative Snubber for A 200KVA Recovery Method for Voltage Source Self-Commutated Converters”, Record of IPEC- Yokohama, 1995, pp. 1572-1577. Scharnbock and K. Schiftner, [91] H. Recovery by an Active/Passive [92] J. “ “A New Snubber Circuit”, Holtz and S. Salama, “Megawatt for GTO Applications with Energy Record of EPE Conference, 1989, pp. 707-711. GTO Inverter with Three Level PWM Control and Regenerative Snubber Circuits”, Record of IEEE PESC, 1988, pp. 1263-1270. [93] J. D. Kim and Inverters”, [94] H. G. H. Cho, “A Regenerative Snubber for Three Level High Power EPE Conference, 1991, pp. 2.474-2.479. Tsuchiya, “A Novel Regenerative Snubber GTO Record of Okayama and T. Circuit for Three Level GTO Inverters”, Mitsubishi Electric ADVANCE, June 1997, pp. 24-26. [95] F. C. Zach, K. H. Kaiser, J. W. Kolar and F. J. Haselsteiner, “New Lossless Tum-on and Turn-off (Snubber) Networks for Inverters, Including Circuits for Blocking Voltage Limitation”, [96] A. IEEE Trans. on Power Electronics, Vol. Agbossou, I. Rasoanarivo and B. Davat, 1, No. “A Comparative 2, April 1986,- pp. 65-75. Study of High Power IGBT Model Behavior in Voltage Source Inverter”, Record of IEEE PESC, 1996, pp. 56-61. [97] A. Millner, S. Keir, J. Wen and R. Galante, “A High Power DC-AC Staggered Regenerative Recovery Circuit”, Record of IEEE [98] K. B. Hughes, C. D. Schauder, M. G. Gernhardt and E. Circuit Parameters on the Tum-off Spike of Large GTO J. APEC, Inverter with 1995, pp. 973-976. Stacey, “Effect of Snubber Thyristors”, Record of IEEE IAS, 1993, pp. 1280-1285. [99] M. A. E. Andersen, “Comparison of Three IGBT Inverters, One Hard-Switched and Two with Snubber Circuits Using A Minimum Number of Components”, Record of EPE Conference, 1993, pp. 3.306-3.311. [100] L. Fratelli and G. Giannini, “Dual-Voltage Power System in Heavy 1996, pp. 1414-1419. [101] I. Traction, Using High Power Converter for a Distributed High Voltage IGBTs”, Record of IEEE PESC, _ Barbi and A. Ferrari de Souza, “Comutação Suave”, Postgraduate Course-INEP- UFSC, 1995. [102] W. McMurray, “Power 149 Electronics in the 1990s”, Record of IEEE IECON, 1990, pp. 839-843. [103] R. L. Steigerwald, Power [104] J. “A Review of Soft Supplies”, Record of IEEE Petzoldt, S. IECON, J. G. Cho, J. Full Bridge DC 1995, pp.1-7. Bemet and T. Reimann, “Comparison of Power Converters with DC and AC Link”, Record of EPE Conference, [105] Switching Techniques in High Performance Sabate, G. Hua and 1993, pp. 3.189-3.195. F. C. Lee, “Zero Voltage and Zero Current Switching PWM Converter for High Power Applications”, Record of IEEE PESC, ` 1994, pp. 102-ros. [106] W. McMurray, ' _ “Thyristor Commutation in DC Choppers - A Comparative Study”, IEEE Trans. on Ind. App., Vol. 14, No. 6, Nov./Dec. 1978, pp. 547-557. [107] A. Abbondanti and P. Wood, “A Criterion for Performance Comparison Between High Power z I Inverter Circuits”, IEEE Trans. on Ind. App., Vol. 13, No. 2, March/April 1977, pp. 154-160. _ ' [108] G. K. Dubey, “Classification of Thyristor Commutation Methods”, IEEE Trans. on Ind. App., Vol. 19, No. 4, July/August 1983, pp. 600-606. [109] N. H. Kutkut, D. , M. Divan, D. W. Novotny and R. Marison, “Design Considerations and Topology Selection for IEEE PESC, A l20kW IGBT Converter for EV Charging”, Record of 1995, pp. 238-244. . [110] F. Forest, J Gonzalez and F. Costa, “Use of Soft-Switching Principles in . Inverters Design”, Record of EPE Conference, 1993, pp. 3.5-3.10. [111] D. M. Divan and G. PWM Voltage A Skibinski, “Zero-Switching-Loss Inverters for High-Power Applications”, Record of IEEE IAS, 1987, pp. 627-634. [112] M. H. Kheraluwala and D. M. Divan, “Delta Modulation Inverters”, [113] D. Strategies for Resonant Link Record of IEEE PESC, 1987, pp. 271-278. M. Divan, “Low-Stress Switching for Efficiency”, IEEE Spectrum, Dec. 1996, pp. 33-39. [114] I. T. Wallace, N. H. Kutkut, S. Bhattacharya, D. M. Divan and D. W. Novotny, “Inductor Design for High Power Applications with Broad Spectrum Excitation”, Record of IEEE PESC, 1995, pp. 1057-1063. [115] G. Venkataramanan and D. M. Divan, “Pulse Width Modulation with Resonant Link Converters”, Record of IEEE IAS,l990, pp. 1215-1221. DC 150 M. Divan and G. Venkataramanan, “Comparative Evaluation of Sofi Switching [116] D. Inverter Topologies”, Record of EPE Conference, 1991, pp. 2.013-2.018. [117] L. Malesani, P. Tenti, P. Tomasin and V. Toigo, “High Efficiency Quasi-Resonant PWM”, Record of IEEE APEC, Link Three Phase Power Inverter for Full-Range DC 1992, pp. 472-478. [118] S. Salama and Y. Tadros, “Novel Soft Switching Quasi-Resonant 3-Phase Inverter”, [119] F. Record of EPE Conference, 1995, pp. 2.95-2.99. Mendes de Seixas and D. J. C. Martins, “The ZVS-PWM Commutation Cell Applied DC-AC Converter”, IEEE Trans. on Power Electronics, Vol. to the IGBT 12, No. 4, July 1997, pp. 726-733. [120] H. Foch, Y. Cheron, M. Metz and T. Meynard, “Commutation Mechanism and Soft Commutation [121] P. P. Mok, R. in Static Converters”, Spee, H. J. Record of COBEP, 1991, pp. 338-346. Beukes and J. H. R. Enslin, “Control Complexities Related to A High Power Resonant Inverters”, Record of IEEE PESC, 1996, pp. 1040-1046. [122] G. Bingen, “Utilisation de Transistors a Fort Courant et Tension Elevee”, Record of EPE Conference, [123] H. J. Beukes, J. 1987, pp. 1.15-1.20. _ Converter Topologies in Utility Applications”, Record of 522. Inverter”, [125] H. IEEE PESC, 1995, pp. 517- ' W. De Doncker and [124] R. AC/AC H. R. Enslin and R. Spee, “Experimental Evaluation of J. J. P. Lyons, “Then Auxiliary Resonant Commutated Pole Record of IEEE IAS, 1990, pp. 1228-1235. Beukes, J. H. R. Enslin and R. Spee, “Integrated Active Snubber for High Power IGBT Modules”, Record of IEEE APEC, [126] K. Iida, T. Sakuma, A. Mechi, H. Conducting Inductance 1997, pp. 161-167. Matsuo and in the Auxiliary F. Kurokawa, “The Influence of the Resonant Commutated Pole Inverter”, Record of IEEE PESC, 1997, pp. 1238-1245. [127] F. F. Protiwa and J. KOB, “A 20 _ kVA Auxiliary Resonant Commutated Pole Converter- Design and Practical Experiences”, Record of EPE Conference, 1995, pp. 2.111-2.116. [128] F. R. Salberta, kHz J. S. Mayer and R. Auxiliary Resonant pp. 1246-1252. T. Cooley, “An Improved Control Strategy for a 50- Commutated Pole Converter”, Record of IEEE PESC, 1997, . 151 [129] H. Beukes, J. J. H. R. Enslin and R. Spee, “Performance of the Auxiliary Resonant Commutated Pole Converter in Converter Based Utility Devices”, Record of IEEE PESC, 1996, pp. 1033-1039. [130] I. Barbi and D. C. Martins, Additional “A True PWM Zero Voltage RMS Current Stress”, Record of IEEE PESC, Switching Pole with Very Low 1991, pp.261-267. kVA Soft Switching PWM Type AC Auxiliary Power Supply System of the Electric Railway Rolling Stock Using an Improved ARCP [131] H. Matsuo, K. Iida and Harada, “210 Three Phase Inverter with SI Thyristor Switches”, Record of IPEC-Yokohama, 1995, pp. 831-836. [132] R. L. Steigewvald, R. DC-DC Power No. 32, [133] J. P. W. De Doncker and M. H. Kheraluwala, “A Comparison of High- Soft-Switched Converter Topologies”, 5, Sep./Oct. 1996, pp. 1139-1145. IEEE Trans. on Ind. App., Vol. _ b Gegner and C. Q. Lee, “Zero-Voltage-Transition Converters Using an Inductor A Feedback Technique”, Record of IEEE APEC, 1994, pp. 862-868. [134] S. Frame, S. Dubovsky, Q. Li, D. Katsis, C. Cuadros, B. Borojevic and F. C. Lee, “Three-Phase Soft-Switching Inverters for Electric Vehicle Applications”, VPEC Seminar, 1995, pp. 45-52. [135] G. Moschopoulos, P. Jain and G. Joos, Converter”, Record of IEEE “A Novel Zero-Voltage Switched PWM BOOST PESC, 1995, pp. 694-700. [136] H. Stemrnler and T. Eilinger, “Spectral Analysis of the Sinusoidal , PWM with Variable Switching Frequency for Noise Reduction in Inverter-Fed Induction Motors”, Record of ' [137] IEEE PESC, J. Shen, J 1994, pp. 269-277. A . Taufiq and A. D. Characteristics of Traction No. 2, . Mansell, “Analytical Solution to Harmonics PWM Converters”, IEE Pro.-Elec. Power App., March 1997, pp. 158-168. [138] A. Weschta, W. Weberskirch, Vol. 144, ' . “Nonlinear Behavior of Voltage Source Inverters with Power Transistors”, Record of EPE Conference, 1989, pp. 533-537. [139] J. S. Lai, R. Evaluation of W. Young, a' G. W. Ott and J. W. McKeever, “Efficiency Modeling and Resonant Snubber Based Soft-Switching Inverter for Motor Drive Applications” , Record of IEEE PESC, 1995, pp. 943-949. [140] SEMIKRON Data-sheet, Oct. [141] IGBT Designer°s Manual, Intemational Rectifier, 1996, pp. B 6.17- B 6.22. 1996, pp. E105-E115. 152 - [142] 'W. Yi and G. H. Cho, “Novel Snubberless Three-Level GTO Inverter with Dual Quasi- Resonant DC-Link”, Record of IEEE PESC, 1991, pp.880-884. [143] J. G. Cho, W. Back, D. W. Yoo and J. Commutated Pole Inverter for » C. Y.VWon, “Three Level Auxiliary Resonant High Power Applications”, Record of IEEE PESC, 1996, H pp.1019-1026. [144] J QR. Pinheiro and Voltage [145] I. Barbi, “The Three Level PWM Converter-A New Concept in High DC to DC Conversion”, Record of IEEE IECON, H. Song, Inverter I. S. B. Yoo, B. S. Suh and D. S. Hyun, Topology for High Voltage 1992, pp. 173-178. “A Novel Three Level ZVS DC/DC Conversion with Balanced Voltage PWM Sharing 1 and Wider Load Range”, Record of IEEE IAS, 1996, pp. 973-979. [146] G. Botto, IGBT°s for M. and S. Tenconi, “Series Connected Soft Switched High Power, High Voltage Applications: Experirnental Results”, Record of IEEE PESC, [147] Carpita, E. Gilardi 1997. M. Delnnlow, K. Heumann, H. Jorgensen and G. in - SpieB, “Series Connection of IGBTS Resonant Converters”, Record of IPEC-Yokohama, 1995, pp. 1634-1638.