increasing the efficiency of grid-connected power-electronic...
TRANSCRIPT
Simon Ravyts
power-electronic invertersIncreasing the efficiency of grid-connected
Academic year 2015-2016Faculty of Engineering and ArchitectureChair: Prof. dr. ir. Jan MelkebeekDepartment of Electrical Energy, Systems and Automation
Master of Science in Electromechanical EngineeringMaster's dissertation submitted in order to obtain the academic degree of
Counsellor: Dimitar BozalakovSupervisors: Dr. ir. Bart Meersman, Prof. dr. ir. Lieven Vandevelde
Simon Ravyts
power-electronic invertersIncreasing the efficiency of grid-connected
Academic year 2015-2016Faculty of Engineering and ArchitectureChair: Prof. dr. ir. Jan MelkebeekDepartment of Electrical Energy, Systems and Automation
Master of Science in Electromechanical EngineeringMaster's dissertation submitted in order to obtain the academic degree of
Counsellor: Dimitar BozalakovSupervisors: Dr. ir. Bart Meersman, Prof. dr. ir. Lieven Vandevelde
iv
Permission to use
The author gives permission to make this master dissertation available for consultation
and to copy parts of this master dissertation for personal use. In the case of any other
use, the copyright terms have to be respected, in particular with regard to the obligation
to state expressly the source when quoting results from this master dissertation.
Simon Ravyts, 23 May 2016
v
vi Chapter 0. Permission to use
Preface
After six years of study, the end has finally arrived. It wasn’t always as easy but I am glad
that I haven’t given up. During these past years, I developed a passion for electrical power-
and electronic engineering. This thesis combines aspects of both fields and was therefore
perfectly suited for my preferences.
At first, I would like to thank Prof. dr. ir. Lieven Vandevelde for the opportunity to
carry out this research. I would also like to thank him for his interesting courses on the
construction of electrical machines and electrical grids. I would like to thank dr. ir. Bart
Meersman for his clear vision on this project and for proofreading all the chapters. His
managerial capabilities come in very handy for all the projects that he is involved in. But
the person whom I thank the most for the success of this project is ir. Dimitar Bozalakov.
Since his lab setup was right next to mine, I had the opportunity to overwhelm him with
practical questions about the design of power converters. Also the original subject was his
idea and he learned me a lot about power electronics in general. Thank you, Mitko.
Besides the academic work, I owe a lot to my parents. They were always there for me to
support me and they kept believing in me during all these years. Also my beloved girlfriend
Inneke deserves a very big ’Thank you’. She kept supporting me, even if she didn’t always
get the attention she deserved during the past years. After some long and exhausting days
in the lab, she was always able to make me smile again. Thank you, darling.
And last but not least, I wish to thank Luc Lasne, Professeur Agrege at the University of
Bordeaux. He awakened my passion for power electronics during my stay in La Rochelle,
which is now already two years ago. The way he teaches and explains is clear and precise.
Merci Luc.
Simon Ravyts, 23 May 2016
vii
viii Chapter 0. Preface
Summary
Increasing the efficiency of grid-connected
power-electronic inverters
Simon Ravyts
Master’s dissertation submitted in order to obtain the academic degree of Master of Science
in Electromechanical Engineering Supervisors: Prof. dr. ir. Lieven Vandevelde, Dr. ir
Bart Meersman Counsellor: Dimitar Bozalakov
Department of Electrical Energy, Systems and Automation
Chair: Prof. dr. ir. Jan Melkebeek
Faculty of Engineering and Architecture
Academic year: 2015-2016
The aim of this thesis is to investigate possible manners to increase the efficiency of grid-
connected inverters. This specific type of inverters can be used to connect distributed
generation units, such as PV panels, with the distribution grid. They are usually equipped
with a split DC bus such that the neutral can be connected with the midpoint of the DC
bus. This also enables the use of active filtering techniques. PV grid-connected invert-
ers work below their nominal operating point during a considerable amount of time. The
efficiency in this region can be much lower than the nominal or peak-efficiency. Possible
solutions to increase this efficiency are investigated and proposed. The focus will lie on
reducing the switching losses, since they are a large part of the losses. Switching losses
stem from the simultaneous occurence of both a high current and a high voltage at the
commutation instant. They can be effectively reduced by adding a soft switching auxiliary
circuit (SSAC). The working principle of such a SSAC is usually based on an LC resonance.
The circuit is active for only a small amount of time and has thus only a minor influence
on the main circuit. Several of these circuits will be simulated and experimentally tested.
Also the performance of Silicon Carbide components will be evaluated.
Keywords: grid-connected inverters, soft switching, silicon carbide, efficiency improve-
ments
ix
Increasing the efficiency of grid-connectedpower-electronic inverters
Simon Ravyts
Supervisor(s): Prof. dr. ir. Lieven Vandevelde, Dr. ir. Bart Meersman, Dimitar Bozalakov
Ghent University, Faculty of Engineering and Architecture,Departement of Electrical Energy, Systems and Automation
Academic year 2015-2016
Abstract—The use of grid-connected inverters to couple distributed gen-eration (DG) units to the distribution grid is on the rise. The possibilityto support and stabilize the distribution grid makes them an attractive so-lution for the distribution system operator (DSO) as the increased level ofrenewable energy sources can lead to over-voltages and voltage unbalance.The transient response of these inverters is determined by the switching fre-quency. A fast response is necessary for the inverters to work properly. Ahigh switching frequency however, leads to a low efficiency due to the in-creased switching losses. This article proposes several methods to increasethe efficiency of grid-connected inverters. Also the possibility to increasethe switching frequency will be evaluated.
Keywords—grid-connected inverters, soft switching, silicon carbide, effi-ciency improvements
I. INTRODUCTION
GRID-CONNECTED inverters are commonly used to coupleDG units, such as PV panels, with the distribution grid. A
growing interest is noticed in literature for installations that sup-port, stabilize and increase the power quality of the distributiongrid via appropriate control techniques [1], [2], [3]. For this pur-pose, the neutral point of the load needs to be connected to themidpoint of the DC bus. In this way, currents can be injected inthe neutral to reduce the zero-sequence component. The use ofgrid-connected inverters is thus on the rise. A typical efficiencycurve of such an inverter is shown in Figure 1. The efficiency isin general relatively high, except for the low power region. Theinverters however operate during an important part of the time inthis region because nominal power is only reached round noon,
Fig. 1: PV inverter - Efficiency curve
when the solar irradiation is at its maximum.Increasing the efficiency can be done in several ways but this
paper will focus on reducing the switching losses. Three kindsof losses are present at the level of the switches: conductionlosses, switching losses and driving losses. Conduction anddriving losses are normally rather limited in the case of IGBTs.The switching losses are the most important due to the IGBTcurrent tail. They can become a problem at higher switchingfrequencies.
Switching frequencies in the range of 16-20 kHz are gener-ally applied for IGBT inverters. Higher frequencies can causethermal problems because of the excessive heat that is beingdissipated inside the component, due to the switching losses.The use of a high frequency is however very attractive sincetransformers and filters can be sized smaller. This makes thecircuit smaller and lighter, which means that the power density(or the power-to-weight ratio) increases. Also the transient re-sponse and the output waveform quality of the inverter will im-prove. If higher switching frequencies are desired, includinga soft switching auxiliary circuit (SSAC) can be advantageous.The circuit aims to reduce the switching losses by using an LCresonance that influences the voltage or current during the com-mutation instants. In this way, more favorable switching loci areachieved. A graphical representation for the switching loci incase of hard and soft switching is given in Figure 2.
Fig. 2: Switching loci
Fig. 3: Midpoint clamped soft switching auxiliary circuit [7]
Different topologies for SSAC are already proposed in liter-ature [4], [5] [6]. They differ in complexity, number of com-ponents and working principle. The first SSAC placed the LCcircuit inside the main power path. In this way, series or parallelloaded inverters or converters were created. They were load andfrequency dependent, which made the control circuit very com-plex. A newer type of SSAC places the LC circuit outside themain power path. This decreases the conduction losses in theresonant elements. The circuit is only activated just before thecommutation instant of the main power switch, via an auxiliaryswitch. They are called Zero-Voltage-Transition (ZVT) or Zero-Current-Transition (ZCT) SSAC. An important criterion is thatregular PWM operation is not disturbed as the converter can stilloperate without the SSAC. The major drawback of these SSACis that they result in a higher cost and a higher control com-plexity due to the increased number of components. They alsoreduce the reliability of the system. However, the possible ef-ficiency improvements are high and makes them worthwhile tobe investigated. Also the system’s behavior towards EMI canbe improved because the slew rate of the voltage and current iscontrolled.
II. SIMULATIONS
In this particular case of grid-connected inverters, the pres-ence of the balanced midpoint was a motivation to investigatethe midpoint clamped soft switching auxiliary circuit (MPCSSAC), proposed in [7]. The circuit is shown in Figure 3.
First the circuit was simulated via PSpice. It has been appliedto a regular Buck converter and a Buck converter with split DCbus. The load is in this case connected to the midpoint of the DCbus, instead of to the negative bar of the DC bus. This is shownin Figure 4. The simulation was done with this topology becauseit is very close to the actual grid-connected inverter where theneutral point of the load is connected to the midpoint of the DCbus. The simulation settings are given in Table I. It was noticedthat some serious over voltages were present across the auxil-iary switch. A snubber circuit was added to solve this problem.The obtained waveforms for the voltage, current and power areshown in Figure 6 and Figure 7 when the SSAC is active at turn-on. One can see a clear improvement when the SSAC is applied.The switching loss at turn-on has completely disappeared whilethe turn-off loss decreased significantly. The efficiency for a softand a hard switched Buck converter with split DC bus are shownin Figure 5. An improvement of approximately 0,5% is visibleover the complete range. Only for δ = 55% the hard switched
Fig. 4: Buck converter with split DC bus
Simulation time step 10 nsDC bus 600 V
Max. ouput power 2,5 kWMax. output current 10 A
Switching frequency fs 20 kHzMain and aux. switch IRF450Output inductor Lf 2 mHOutput capacitor Cf 5 µF
Resonant inductor Lr 22 µHResonant capacitor Cr 11 nF
Table I: Simulation settings
converter performs better. This is because the converter operatesin discontinuous conduction mode (DCM), which means that theturn-on is already under zero current. Applying the SSAC in thiscase only introduces extra losses.
III. EXPERIMENTAL VERIFICATION
A. Midpoint Clamped SSAC
The simulation showed that indeed an efficiency improve-ment is possible and therefore a test setup was built. The cir-cuit was tested for both a regular Buck converter and a Buckconverter with split DC bus, as this topology is very close to anactual grid-connected inverter. The measurement setup is shownin Figure 8. Notice the bleeder resistor that is used to stabilizethe midpoint. Better alternatives exist for this purpose (see [8])but this solution was chosen for simplicity. The unbalance of themidpoint is a consequence of the current that gets injected dur-ing the turn-off time of the transistor. During the turn-off time,
Duty cycle δ [/]0,6 0,65 0,7 0,75 0,8 0,85 0,9 0,95
Eff
icie
ncy η [
/]
0,92
0,94
0,96
0,98
1
Hard switchedWith MPC SSAC
Fig. 5: Efficiency comparison in simulation
×10-53.5 4 4.5 5 5.5 6
Vol
tage
[V
]
0
500
×10-53.5 4 4.5 5 5.5 6
Cur
rent
[A
]
0
10
20
30
Time [s] ×10-53.5 4 4.5 5 5.5 6
Pow
er [
W]
0
5000
10000
Fig. 6: Simulation - Waveforms in the case of hard switching
×10-53.5 4 4.5 5 5.5 6
Vol
tage
[V
]
0200400600
×10-53.5 4 4.5 5 5.5 6
Cur
rent
[A
]
-505
1015
Time [s] ×10-53.5 4 4.5 5 5.5 6
Pow
er [
W]
-100
0
100
200
Fig. 7: Simulation - Waveforms in the case of soft switching
the current flows via the lower capacitor and the free-wheelingdiode back to the load. At first, the same combination of res-onant elements was used as in simulation, being Lr = 22 µHand Cr = 11nF. The used inductor was an industrial availableSMD type with shielded core. It was however noticed that theinductor became very hot and that the soft switching conditionswere not met anymore. It was concluded that an air coil wasmore appropriate. The best combination found was Lr = 2 µHand Cr = 3.3 nF. With this combination, the zero-voltage con-ditions were easily obtained and the current peak at turn-on waslimited. The measurement results that were obtained with thetest setup are shown in Figure 9 and Figure 10. The DC busvoltage is 600 V and the efficiency is measured for a switch-ing frequency of 20 and 30 kHz. The maximum output power
of the converter is 2,5 kW for an output current of 9,5 A, theused IGBT is a IRGP30B120KDE. The red lines represent theefficiency under hard switching and the blue lines represent theefficiency under soft switching. At 20 kHz, an efficiency im-provement of 2% is visible over the complete range when theMPC SSAC is active. At 30 kHz, the hard switched converterexperienced a thermal breakdown for a duty ratio of 80%. Whenthe MPC SSAC is active, the converter is able to span the entireoperating range with an efficiency that is higher than the hardswitched efficiency. This is an important result since also theuse of snubbers was considered. Snubbers however are not ableto increase the efficiency, they only divert the losses from theswitch towards the snubber network. This option was thereforenot further investigated. The measured waveforms under hard
Fig. 8: Measurement setup - Buck converter with split DC bus
Duty ratio δ [/]0,6 0,65 0,7 0,75 0,8 0,85 0,9 0,95
Eff
icie
ncy η
[/]
0,6
0,8
1
Hard switchedSoft switched
Fig. 9: Buck converter with split DC bus - 600 V, 20 kHz
and soft switching are given in Figure 11. The DC bus voltageis 600 V and the output current is approximately 6 A.
Turn-on: A strong overlap between the collector-emittervoltage and the collector current is present in hard-switching.The switching losses are thus very high. There is also a bigcurrent peak due to diode reverse recovery. Also notice thevery high dv/dt and di/dt. When the MPC SSAC is active, softswitching conditions are clearly present. The SSAC is activatedjust before the main pulse. This makes the voltage drop in acontrolled way. The dv/dt is much lower, which is beneficialfor the EMC. The current increases when the voltage is alreadyzero. This means that the turn-on switching loss is practicallyzero. Also the current peak is much lower.
Turn-off: In hard switching the IGBT current tail is visibleduring the turn-off time. It lasts for approximately 300 ns. Sincethe voltage increases much faster, the turn-off switching lossis very high. By applying the MPC SSAC, the collector cur-rent strongly decreases before the collector-emitter voltage rises.Small oscillations are however still visible. This means that theturn-off loss is not completely zero but it is certainly stronglydecreased.
The effectiveness of the MPC SSAC has thus been proven.The SSAC is activated just before and just after the turn-on andturn-off of the main pulse. The timing of these pulses withrespect to the main pulse should be as constant as possible to
Duty ratio δ [/]0,6 0,65 0,7 0,75 0,8 0,85 0,9 0,95
Eff
icie
ncy η
[/]
0,5
0,6
0,7
0,8
0,9
1
Hard switchedSoft switched
Fig. 10: Buck converter with split DC bus - 600 V, 30 kHz
achieve an easy control. Table II summarizes the duty cycle andthe phase shift of the turn-on and turn-off auxiliary pulses incase of the Buck converter with split DC bus, operating at 600V, 30 kHz. The duty ratio (δ) of the pulses is constant and equalto 1%. The time between the main and the auxiliary pulse isdenoted with Ton and Toff . From the table it is clear that thisadvancement is not constant. The timing of the on-pulse clearlyincreases for higher output power while the advancement of theoff-pulse decreases a little. During this investigation, the con-trol of the pulse width and timing has been done manually. Ifthe SSAC is implemented in a stand-alone version, a closed loopfeedback will be needed to check whether or not the soft switch-ing is achieved. This is again an increase in complexity. Furtherresearch is needed to check if this timing can be held constant,e.g. when other resonant elements are used. Another impor-tant remark is that the duty cycle of the SSAC is very low. Ascan be seen from Table II, it is only 1% for both the on or offpulse. The auxiliary circuit is thus applicable over a very wideoperating range.
B. Silicon Carbide
Another method to increase the efficiency of grid-connectedinverters is simply by using better components. Recently Sili-con Carbide (SiC) MOSFETS came into the market for a com-mercially acceptable price. These wide-bandgap semiconductor
×10-60 0.2 0.4 0.6 0.8 1
Vol
ts [
V]
0
500
Hard turn-on
Cur
rent
[A
]
0
20
×10-60.5 1 1.5 2 2.5
Vol
ts [
V]
0
500
Soft turn-on
Cur
rent
[A
]
0
10
20
Time [s] ×10-60 0.2 0.4 0.6 0.8 1
Vol
ts [
V]
0
500
Hard turn-off
Cur
rent
[A
]
0
5
Time [s] ×10-60 0.5 1 1.5 2
Vol
ts [
V]
0
500
Soft turn-off
Cur
rent
[A
]
0
5
Fig. 11: Comparison of the waveforms for the practical setup
δmain δon Ton δoff Toff Pout
[%] [%] [ns] [%] [ns] [W]55 1 267 1 68,5 6260 1 333 1 68,0 18265 1 333 1 68,8 35670 1 373 1 68,3 59575 1 400 1 62,5 88680 1 400 1 68,7 124185 1 427 1 69,5 167090 1 427 1 65,0 216195 1 427 1 65,8 2533
Table II: Duty cycle and shift of the auxiliary pulses for theSSAC
devices have superior properties when compared to normal Sil-icon (Si) devices. Both the conduction losses and the switchinglosses of SiC components are very low. They were comparedwith the regular Silicon (Si) IGBTs. The results are shown inFigure 12 and Figure 13. It can be seen that the efficiency ofthe hard switched Buck converter with split DC bus using SiCMOSFETS is even higher than the regular inverter under softswitching. At 20 kHz and low duty ratios, an efficiency im-provement of more than 10% is possible. At higher duty ratios,the efficiency improvement is lower but still an increase of 5%is visible. At 40 kHz, the SiC converter still obtains a very highefficiency over the complete range. The comparison with theMPC SSAC shows that the SiC components perform better andthat they are able to span the complete operating range. The SiIGBTs with MPC SSAC experienced a thermal breakdown atδ = 80%. The performance of the newer SiC MOSFETs is thusclearly superior over the older Si IGBTs. It was however noticedthat the switching transients are very fast and can be a problemwith regard to EMI. This however requires further research toinvestigate if snubbers could improve this behavior.
Duty ratio δ [/]0,6 0,65 0,7 0,75 0,8 0,85 0,9 0,95
Eff
icie
ncy η [
/]
0,6
0,8
1
Hard switched - normal IGBTSoft switched - normal IGBTHard switched - SiC MOSFET
Fig. 12: Buck converter with split DC bus - 600 V, 20 kHz
Duty ratio δ [/]0,6 0,65 0,7 0,75 0,8 0,85 0,9
Eff
icie
ncy η [
/]
0,6
0,8
1
Hard switched - SiC MOSFETSoft switched - normal IGBT
Fig. 13: Buck converter with split DC bus - 600 V, 40 kHz
IV. CONCLUSION
Different methods to increase the efficiency of grid-connectedinverters were proposed in this article. First an introductionabout SSAC was given. Simulations were carried out for a MPCSSAC since this circuit makes use of the stabilized midpoint thatis already present in grid-connected inverters. In simulation, anefficiency increase is possible and the waveforms at turn-on andturn- off were given. An experimental setup was build to val-idate the performance of the MPC SSAC. It was shown thatthis circuit increases the efficiency by effectively reducing the
switching losses. Applying the SSAC also means that higherswitching frequencies are attainable, which enables the use ofsmaller filter equipment. The drawback is the increased numberof components, which leads to higher costs, and the higher con-trol complexity. Another proposed alternative is the use of SiCcomponents. These components have very low switching lossesand conduction losses. This was again experimentally tested ina Buck converter with split DC bus. These components performeven better than the Si components in soft switching and do notneed any auxiliary circuits. It is expected that SiC devices willshape the future of power electronics since they are clearly themost appropriate choice when a high efficiency is wanted.
REFERENCES
[1] D. Bozalakov, T. Vandoorn, B. Meersman, C. Demoulias, and L. Vande-velde, “Voltage dip mitigation capabilities of three-phase damping controlstrategy,” Electric Power Systems Research, vol. 121, pp. 192–199, 2015.
[2] D. Bozalakov, T. Vandoorn, B. Meersman, G. Papagiannis, A. Chrysochos,and L. Vandevelde, “Damping-based droop control strategy allowing an in-creased penetration of renewable energy resources in low voltage grids,”IEEE Transactions on Power Delivery, vol. PP, no. 99, pp. 1–1, 2016.
[3] D. Bozalakov, T. Vandoorn, B. Meersman, and L. Vandevelde, “Overviewof increasing the penetration of renewable energy sources in the distribu-tion grid by developing control strategies and using ancillary services,” inProceedings of the IEEE Young Researchers Symposium, p. 5, EESA, 2014.
[4] G. Hua, C. Leu, Y. Jiang, and F. Lee, “Novel zero-voltage-transition pwmconverters,” Power Electronics, IEEE Transactions on, vol. 9, pp. 213–219,Mar 1994.
[5] C.-M. Wang, “Novel zero-voltage-transition pwm dc-dc converters,” Indus-trial Electronics, IEEE Transactions on, vol. 53, no. 1, pp. 254–262, 2006.
[6] J. He, N. Mohan, and B. Wold, “Zero-voltage-switching pwm inverter forhigh-frequency dc-ac power conversion,” IEEE Transactions on IndustryApplications, vol. 29, September/October 1993.
[7] C. M. De Oliveira Stein, H. A. Grundling, H. Pinheiro, J. R. Pinheiro, andH. L. Hey, “Zero-current and zero-voltage soft-transition commutation cellfor pwm inverters,” Power Electronics, IEEE Transactions on, vol. 19, no. 2,pp. 396–403, 2004.
[8] B. Meersman, Regeling van driefasige invertorgekoppelde decentrale gen-eratoren met betrekking tot de verbetering van de netkwaliteit. PhD thesis,University of Ghent, 2012.
Contents
Permission to use v
Preface vii
Summary ix
Extended Abstract xvi
Figures xxiii
Tables xxix
Nomenclature xxxi
1 Introduction 1
1.1 Grid-connected inverters . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Commutation and switching losses . . . . . . . . . . . . . . . . . . . . . . 4
1.3 Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
2 Literature review 7
2.1 Losses in hard switching . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
2.1.1 Power diodes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
xvii
xviii Contents
2.1.1.1 Conduction losses . . . . . . . . . . . . . . . . . . . . . . . 10
2.1.1.2 Switching losses . . . . . . . . . . . . . . . . . . . . . . . . 11
2.1.2 Power MOSFETs . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
2.1.2.1 Conduction losses . . . . . . . . . . . . . . . . . . . . . . . 13
2.1.2.2 Switching losses . . . . . . . . . . . . . . . . . . . . . . . . 14
2.1.3 IGBTs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
2.1.3.1 Conduction losses . . . . . . . . . . . . . . . . . . . . . . . 14
2.1.3.2 Switching losses . . . . . . . . . . . . . . . . . . . . . . . . 15
2.1.4 Switch cell . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
2.2 Driver circuits . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
2.3 Efficiency vs. switching frequency . . . . . . . . . . . . . . . . . . . . . . . 22
2.4 Inverters for grid connected power supplies . . . . . . . . . . . . . . . . . . 22
2.5 Enhancement of switching losses . . . . . . . . . . . . . . . . . . . . . . . . 26
2.5.1 RLC circuits . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
2.5.2 Silicon Carbide . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
2.5.3 Snubber circuits . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
2.5.4 Soft switching . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
2.5.4.1 Resonant DC link inverters . . . . . . . . . . . . . . . . . 31
2.5.4.2 Zero voltage and zero current switchings . . . . . . . . . . 32
2.5.4.3 Zero voltage and zero current transitions . . . . . . . . . . 34
Contents xix
2.6 Example: Losses of a hard switched inverter . . . . . . . . . . . . . . . . . 37
2.7 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
3 Simulations 43
3.1 Buck converter with split DC bus . . . . . . . . . . . . . . . . . . . . . . . 43
3.2 Simulation 1 - ZVT SSAC 1 . . . . . . . . . . . . . . . . . . . . . . . . . . 46
3.2.1 Components and simulation settings . . . . . . . . . . . . . . . . . 46
3.2.2 Buck converter . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
3.2.3 Buck converter with ZVT . . . . . . . . . . . . . . . . . . . . . . . 52
3.2.4 Full leg with ZVT . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
3.3 Simulation 2 - MPC SSAC . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
3.3.1 Buck converter . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
3.3.2 Buck converter with split DC bus . . . . . . . . . . . . . . . . . . . 62
3.4 Simulation 3 - PRDCL . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
3.5 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
4 Experimental verification 71
4.1 Setup . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
4.2 Measurement equipment . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
4.2.1 Voltage probes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
4.2.2 Current probes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
xx Contents
4.2.3 Other equipment . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
4.3 Switching loss measurements . . . . . . . . . . . . . . . . . . . . . . . . . . 74
4.4 Buck converter . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
4.4.1 Hard switched . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
4.4.2 Soft switch auxiliary circuit 1 . . . . . . . . . . . . . . . . . . . . . 77
4.4.3 Soft switch auxiliary circuit 2 . . . . . . . . . . . . . . . . . . . . . 80
4.5 Buck converter with split DC bus . . . . . . . . . . . . . . . . . . . . . . . 82
4.5.1 Comparison for a 400 V DC bus . . . . . . . . . . . . . . . . . . . . 82
4.5.2 Comparison for a 600 V DC bus . . . . . . . . . . . . . . . . . . . . 85
4.5.3 Waveform analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . 88
4.6 Silicon Carbide . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89
4.7 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91
5 Conclusion 93
A Matlab scripts 95
A.1 Calculation resonant parameters MPC SSAC . . . . . . . . . . . . . . . . . 95
A.2 Power losses - Buck converter . . . . . . . . . . . . . . . . . . . . . . . . . 96
A.3 Power losses - Inverter . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98
A.4 Determination of switching losses . . . . . . . . . . . . . . . . . . . . . . . 100
A.5 Calculation of the gate resistance . . . . . . . . . . . . . . . . . . . . . . . 104
Contents xxi
B Design considerations 107
B.1 Fundamentals of power electronics design . . . . . . . . . . . . . . . . . . . 107
B.1.1 Gate drive circuits . . . . . . . . . . . . . . . . . . . . . . . . . . . 107
B.2 Component selection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110
B.3 PCB design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111
B.4 PCB optimization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112
B.5 Thermal design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112
B.6 Control circuitry . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115
C DSP code 117
D PCB design 123
D.1 Schematics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123
D.2 Board lay-out . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 130
D.3 Pictures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 134
E Measurements 137
F Data sheets 151
Bibliography 161
xxii Contents
Figures
1.1 Installed worldwide capacity of PV in GW . . . . . . . . . . . . . . . . . . 2
1.2 Yearly worldwide number of PV installations . . . . . . . . . . . . . . . . . 3
1.3 PV inverter - Efficiency curve . . . . . . . . . . . . . . . . . . . . . . . . . 3
2.1 Switching transients and power losses for a component in isolation . . . . . 10
2.2 Diode . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
2.3 Reverse recovery of a diode . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
2.4 Symbols of commonly used controllable switches . . . . . . . . . . . . . . . 13
2.5 IGBT structure and tail current . . . . . . . . . . . . . . . . . . . . . . . . 16
2.6 Reference circuit for loss calculation . . . . . . . . . . . . . . . . . . . . . . 17
2.7 Waveforms during transition of switches . . . . . . . . . . . . . . . . . . . 18
2.8 Internal capacitances . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
2.9 IGBT switching behavior . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
2.10 IGBT with driver circuit, gate resistance and common emitter inductance . 21
2.11 Efficiency as a function of the switching frequency . . . . . . . . . . . . . 23
2.12 Inverter topologies for grid connection . . . . . . . . . . . . . . . . . . . . 25
2.13 Series RLC circuit . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
2.14 Step response of an RLC circuit . . . . . . . . . . . . . . . . . . . . . . . . 28
xxiii
xxiv Figures
2.15 Snubber circuits . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
2.16 Parallel resonant DC link inverter . . . . . . . . . . . . . . . . . . . . . . . 32
2.17 ZVS-PWM Buck converter . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
2.18 ZCS-PWM Buck converter . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
2.19 First implementation of a ZVT-PWM Buck converter [1] . . . . . . . . . . 35
2.20 Recently proposed soft switching auxiliary circuit [2] . . . . . . . . . . . . 36
2.21 Implementation the newly proposed SSAC in a Buck converter . . . . . . . 36
2.22 Proposed soft switching auxiliary circuit [3] . . . . . . . . . . . . . . . . . 37
2.23 Switching loci for different techniques . . . . . . . . . . . . . . . . . . . . . 40
3.1 Split DC bus topologies . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
3.2 Buck converter simulation model in SPICE . . . . . . . . . . . . . . . . . . 47
3.3 Input and output voltages . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
3.4 Inductor, diode and MOSFET currents . . . . . . . . . . . . . . . . . . . . 49
3.5 MOSFET and diode power consumption . . . . . . . . . . . . . . . . . . . 50
3.6 Reverse recovery current effects in diode and MOSFET . . . . . . . . . . . 50
3.7 Zoom at turn-on power consumption . . . . . . . . . . . . . . . . . . . . . 51
3.8 Zoom at turn-off power consumption . . . . . . . . . . . . . . . . . . . . . 51
3.9 Buck converter with soft switching cell . . . . . . . . . . . . . . . . . . . . 53
3.10 Main MOSFET power dissipation - 5 cycles . . . . . . . . . . . . . . . . . 55
Figures xxv
3.11 Zoom on main MOSFET power dissipation - 1 cycle . . . . . . . . . . . . . 55
3.12 Aux. MOSFET power dissipation - 5 cycles . . . . . . . . . . . . . . . . . 56
3.13 Zoom on aux. MOSFET power dissipation - 1 cycle . . . . . . . . . . . . . 56
3.14 Comparison of the efficiency as a function of δ - Buck converter . . . . . . 57
3.15 Spice circuit for a full leg . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
3.16 Graphical representation of the efficiency comparison - Full leg . . . . . . . 59
3.17 Proposed soft switching commutation cell [3] . . . . . . . . . . . . . . . . 61
3.18 Spice simulation circuit for a Buck with MPC SSAC . . . . . . . . . . . . 63
3.19 Efficiency comparison - Buck converter with and without ZVT circuit . . . 63
3.20 Efficiency comparison - Buck converter with split DC bus . . . . . . . . . . 64
3.21 Power consumption of the switch in hard switching . . . . . . . . . . . . . 65
3.22 Power consumption of the switch in soft switching . . . . . . . . . . . . . . 65
3.23 Parallel Resonant DC Link [4] . . . . . . . . . . . . . . . . . . . . . . . . 67
3.24 PRDCL simulation model in SPICE . . . . . . . . . . . . . . . . . . . . . . 68
3.25 Supplied inverter voltage . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
3.26 Current through the resonant inductor . . . . . . . . . . . . . . . . . . . . 69
4.1 Test setup . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
4.2 Comparison of different current measurement techniques . . . . . . . . . . 74
4.3 Definition of the switching losses [5] . . . . . . . . . . . . . . . . . . . . . . 75
xxvi Figures
4.4 Hard switched Buck converter - Measurement setup . . . . . . . . . . . . 76
4.5 Hard switched Buck converter - Efficiency comparison . . . . . . . . . . . 77
4.6 Hard switched Buck converter - Output power . . . . . . . . . . . . . . . 77
4.7 Soft switched Buck converter - Measurement setup . . . . . . . . . . . . . 78
4.8 Efficiency comparison - 20 kHz Buck converter . . . . . . . . . . . . . . . 79
4.9 Efficiency comparison - 40 kHz Buck converter . . . . . . . . . . . . . . . 80
4.10 Soft switching auxiliary circuit 2 . . . . . . . . . . . . . . . . . . . . . . . . 81
4.11 Voltage across diode D3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
4.12 Buck converter with split DC bus - Measurement setup . . . . . . . . . . . 83
4.13 Efficiency comparison Buck split DC bus at 400V - 20 kHz . . . . . . . . . 84
4.14 Efficiency comparison Buck split DC bus at 400V - 30 kHz . . . . . . . . . 85
4.15 Efficiency comparison Buck split DC bus at 600V - 20 kHz . . . . . . . . . 86
4.16 Efficiency comparison Buck split DC bus at 600V - 30 kHz . . . . . . . . . 87
4.17 Voltage and current waveforms . . . . . . . . . . . . . . . . . . . . . . . . 89
4.18 Efficiency comparison with SiC components - 20 kHz . . . . . . . . . . . . 90
4.19 Efficiency comparison with SiC components - 40 kHz . . . . . . . . . . . . 91
B.1 Recommended driver application circuit [6] . . . . . . . . . . . . . . . . . 113
B.2 Recommended application of the DESAT protection circuit [6] . . . . . . . 113
B.3 IGBT driving voltages (zoom) . . . . . . . . . . . . . . . . . . . . . . . . . 114
Figures xxvii
B.4 Thermal insulation tube . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115
B.5 The Digital Signal Processor . . . . . . . . . . . . . . . . . . . . . . . . . . 116
D.1 PCB layout - Buck . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 131
D.2 PCB layout - ZVT1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 132
D.3 PCB layout - ZVT2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133
D.4 Buck converter . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 135
D.5 Midpoint clamped SSAC . . . . . . . . . . . . . . . . . . . . . . . . . . . . 135
D.6 Soft switching auxiliary circuit 2 . . . . . . . . . . . . . . . . . . . . . . . . 136
xxviii Figures
Tables
2.1 Comparison of RDS,on . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
2.2 Most important IGBT specifications . . . . . . . . . . . . . . . . . . . . . 14
2.3 Comparison between Si and SiC components . . . . . . . . . . . . . . . . . 29
2.4 Comparison of possible techniques . . . . . . . . . . . . . . . . . . . . . . . 41
3.1 Component values for simulation . . . . . . . . . . . . . . . . . . . . . . . 47
3.2 Losses in a Buck converter . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
3.3 Resonant parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
3.4 Losses in a Buck converter with soft switching cell . . . . . . . . . . . . . . 56
3.5 Comparison of the efficiencies for a full leg with and without ZVT . . . . . 59
3.6 Peak current Ip in the MPC SSAC for different resonant elements . . . . . 62
4.1 Used components for the measurement setup . . . . . . . . . . . . . . . . . 76
4.2 Efficiency comparison for 20 and 30 kHz at 400 V . . . . . . . . . . . . . . 84
4.3 Efficiency comparison for 20 and 30 kHz at 600 V . . . . . . . . . . . . . . 86
4.4 Duty cycle and shift of the auxiliary pulses for the SSAC . . . . . . . . . . 88
4.5 Parameters of the SiC MOSFET . . . . . . . . . . . . . . . . . . . . . . . . 90
B.1 Influence of gate resistance on performance [7] . . . . . . . . . . . . . . . . 109
xxix
xxx Tables
E.1 Hard switched Buck - Measurements at 10 kHz . . . . . . . . . . . . . . . 138
E.2 Hard switched Buck - Measurements at 20 kHz . . . . . . . . . . . . . . . 139
E.3 Hard switched Buck - Measurements at 40 kHz . . . . . . . . . . . . . . . 140
E.4 Buck converter with MPC SSAC at turn-on - 20 kHz . . . . . . . . . . . . 141
E.5 Buck converter with MPC SSAC at turn-off - 20 kHz . . . . . . . . . . . . 141
E.6 Buck converter with MPC SSAC at turn-on and off - 20 kHz . . . . . . . . 142
E.7 Buck converter with MPC SSAC at turn-on and off - 40 kHz . . . . . . . . 143
E.8 Buck with split DC bus 400 V - 20 kHz - hard switched . . . . . . . . . . . 144
E.9 Buck with split DC bus 400 V - 30 kHz - hard switched . . . . . . . . . . . 144
E.10 Buck with split DC bus 400 V - 20 kHz - MPC SSAC . . . . . . . . . . . . 145
E.11 Buck with split DC bus 400 V - 30 kHz - MPC SSAC . . . . . . . . . . . . 146
E.12 Buck with split DC bus 600 V - 20 kHz - hard switched . . . . . . . . . . . 147
E.13 Buck with split DC bus 600 V - 30 kHz - hard switched . . . . . . . . . . . 147
E.14 Buck with split DC bus 600 V - 20 kHz - MPC SSAC . . . . . . . . . . . . 148
E.15 Buck with split DC bus 600 V - 30 kHz - MPC SSAC . . . . . . . . . . . . 149
E.16 Buck with split DC bus 600 V - 20 kHz - Silicon Carbide . . . . . . . . . . 150
E.17 Buck with split DC bus 600 V - 40 kHz - Silicon Carbide . . . . . . . . . . 150
Nomenclature
Abbreviations
CCM Continuous Conduction Mode
CSV Comma Separated Value
DCM Discontinuous Conduction Mode
DG Distributed Generation
DSP Digital Signal Processor
EMC Electro Magnetic Compatibility
EMI Electro Magnetic Interference
GaN Gallium Nitride
IGBT Insulated Gate Bipolar Transistor
MOSFET Meta-Oxide-Semiconductor Field-Effect Transistor
MPC MidPoint Clamped
PCB Printed Circuit Board
PRDCL Parallel Resonant DC Link
S Switch
SiC Silicon Carbide
SMD Surface Mount Device
SMT Surface Mount Technology
SPICE Simulation Program with Integrated Circuit Emphasis
SS Soft Switching
SSAC Soft Switching Auxiliary Circuit
VSI Voltage Source Inverter
ZCT Zero Current Transition
ZVT Zero Voltage Transition
xxxi
xxxii Chapter 0. Nomenclature
Symbols
C Capacitance F
E Energy J
f Frequency Hz
I Current A
L Inductance H
P Power W
R Resistance Ω
t Time s
T Period s
V Voltage V
Q Charge C
α Attenuation constant /
δ Duty ratio /
ω Angular frequency rad
ω0 Natural pulsation rad
η Efficiency /
ζ Damping factor /
Chapter 1
Introduction
1.1 Grid-connected inverters
Power electronic inverters and converters are being used more and more in our daily life and
in industry. Renewable energy sources such as photo-voltaic (PV) panels and wind turbines
are commonly connected via so called grid-connected inverters. The worldwide installed
capacity of PV panels (in GW) is shown in Fig. 1.1 1, while the number of PV installations
is shown in Fig. 1.2 2. A growing trend is clearly visible. All these installations require one
or more inverters. Besides this, also wind turbines are usually coupled via grid-connected
inverters. The use of these inverters is thus on the rise. Care should be taken since the
increased penetration of these distributed generation (DG) units can cause problems such
as over-voltages and voltage unbalance [8]. More recently, grid-connected inverters are
being used to stabilize, support and increase the power quality of the distribution grid. As
described in [9], voltage dips can be mitigated by using an efficient three-phase damping
control strategy. Other aspects such as efficient droop control strategies [10] and the
implementation of battery systems are still being investigated. The purpose of this thesis
however, is to investigate how the efficiency of grid-connected inverters can be improved.
A high efficiency for all these devices is crucial to obtain the maximum power output of
renewable energy sources. The efficiency that can be found in the data sheet is usually
determined under full power, or it is the peak efficiency. Very high efficiencies in the range
1http://solarcellcentral.com2http://www.solarpowereurope.org
1
2 Chapter 1. Introduction
Year2009 2010 2011 2012 2013 2014
Inst
alle
d ca
paci
ty [
GW
]
0
20
40
60
80
100
120
140
160
180
200
23.00
42.60
70.30
100.40
138.80
183.80
Fig. 1.1: Installed worldwide capacity of PV in GW
of 92 - 96 % can then be reached3. An important remark is that the inverters do not
operate under full power during most of the time. A household PV installation will only
deliver it’s nominal power around noon. In the morning and in the afternoon the output
power will be lower. The efficiency curve for a ’Sunnyboy 4000’ is shown in Fig. 1.3. One
can see that a peak efficiency of almost 97% is reached. The efficiency is remarkably lower
when the output power is lower than 1.5 kW. In this thesis it will be shown that efficiency
improvements are possible, especially in this region of low output power. This can be done
in different ways but the focus will lie on reducing the switching losses.
3http://gosolarcalifornia.ca.gov/equipment/inverters.php
1.1. Grid-connected inverters 3
Year2009 2010 2011 2012 2013
Num
ber
of in
stal
latio
ns
×104
0
0.5
1
1.5
2
2.5
3
3.5
4
7340
17107
30282 29865
37007
Fig. 1.2: Yearly worldwide number of PV installations
Fig. 1.3: PV inverter - Efficiency curve
4 Chapter 1. Introduction
1.2 Commutation and switching losses
Throughout the years, forced commutation of power electronic switches has always been
a point of concern. It dates back to the time that thyristors were used as controllable
switches. Techniques to turn the thyristors off usually involved an auxiliary LC circuit.
This circuit was able to reduce the current below a critical value, such that the thyristor
will turn off. In the late eighties of the previous century, power MOSFETs (Metal-Oxide-
Semiconductor Field-Effect Transistor) and IGBTs (Insulated Gate Bipolar Transistor)
came into the market. They are easily controlled and the turn-off circuits, that were
required to cancel the current, are not needed anymore. The simplicity of controlling
these switches makes that they are applicable in a wide range of devices. Nowadays, the
purpose of the research has shifted. The commutation is no longer a purpose on its own
but rather an opportunity to increase the efficiency. As will be discussed in later chapters,
a big share of the losses are related to the commutation, namely the switching losses.
They are directly proportional to the switching frequency and thus to the overall size and
weight of the converters. If they can be avoided or reduced, both the efficiency and the so
called power-to-weight ratio of the device can increase significantly. A possible method to
achieve this is via soft switching cells. A large number of soft commutation techniques is
already proposed in literature. As in the past, they are usually based on an LC resonance.
A handful of them will be discussed and further investigated in this thesis. However, it
is important to note that the current state-of-the-art solution is the use of wide band-
gap semiconductors such as Silicon Carbide (SiC) and Gallium Nitride (GaN). They have
a better performance compared to the regular Silicon components. Unfortunately, their
production process still contains some errors which makes them rather expensive. They will
be treated later on as it is expected that this technology will dominate the future of power
electronics. When they become affordable, soft commutation circuits will be superfluous.
1.3 Overview
The thesis is built up in the following way: First, this chapter gave a general introduction
to the subject and pointed out the problem. In Chapter 2 , a literature study will be made
in order to understand the origins of switching losses and to investigate how to minimize
them. Several types of soft switching auxiliary circuits (SSAC) will be discussed here. The
1.3. Overview 5
focus is on the number and type of used components, complexity, load-dependency etc. At
the end of the chapter, an overview of the different techniques is given. In Chapter 3, the
most interesting solutions of the literature study are simulated using PSpice. They will
be simulated for a Buck converter and a full leg topology. The waveforms and expected
efficiencies are then examined. An important part of the thesis deals with the experimental
validation of the simulations. This is described in Chapter 4. Also the used measurement
techniques and tools are discussed. Chapter 5 contains the conclusion and the possibilities
for future work.
6 Chapter 1. Introduction
Chapter 2
Literature review
2.1 Losses in hard switching
In this section, an overview of the losses in semiconductor switches will be given. The
focus will be on the switching losses since these can be reduced by using soft switching
auxiliary circuits (SSAC), which is the aim of this work. Firstly, hard switching power
losses are discussed in general and afterwards the specific properties of diodes, MOSFET’s
and IGBT’s and their combination will be discussed. In general, the power losses of a
component during one cycle are given by Eqn. 2.1.
P =1
T
∫ T
0
v(t)i(t)dt (2.1)
Where:
T is one switching period
v(t) is the voltage across the component during one period
i(t) is the current through the component during one period
Usually this integral is hard to solve, since the waveforms of the voltage nor the current are
analytically known. Therefore, a distinction is made between three kinds of power losses:
7
8 Chapter 2. Literature review
Off state losses
On state losses or conduction losses
Switching losses or commutation losses
The off state losses are usually negligible compared to the other two since the leakage
current is very small. Therefore, they will not be further discussed.
The conduction losses mainly depend on the on-state voltage of the switch. In power
electronics circuit analysis, the on-state voltage of switches is mostly neglected to under-
stand the working principle of a circuit. This is an appropriate approximation since this
voltage is mostly negligible compared to the in- and output voltages of the circuit. Anyhow,
for design aspects where losses need to be taken into account, they cannot be neglected
anymore. The on-state voltage is mainly determined by the type of power switch, especially
by its internal geometry and by the current flowing through the switch. The dissipated
energy over one period can be found directly from the current flowing through the switch
Ion(t), the voltage across it during the on-state Von(t) and the conduction time ton, Eqn.
2.2. The voltage waveform is usually quite constant while the switch is conducting, so
Von(t) = Von. The current can be constant or not. If not, the best way to calculate the
conduction losses is by breaking up the interval in smaller sub-intervals where the current
can be assumed constant. Afterwards the results have to be summed. This gives more
accurate results than taking the average or RMS-value of the current [11].
Pcond(t) =1
T
∫ ton
0
Von(t).Ion(t)dt (2.2)
The switching losses stem from the simultaneous occurrence of high voltages and cur-
rents during switching. The switching losses include both the turn-on and turn-off losses.
Compared to the conduction losses, they need to be treated from a dynamic point of view
since they only occur at the transitions of the switch (ON ↔ OFF). During these time
intervals, which are usually very small, in the order of nano- or microseconds, the current
increases while the voltage over the device decreases (or vice versa). These transitions are
shown in Fig. 2.1 together with the corresponding power loss. Notice that the transitions
are supposed linear. This is generally not the case. Every type of switch has its typical
waveforms. But for the sake of simplicity, they are mostly split up in linear subintervals.
2.1. Losses in hard switching 9
Also the circuit in which the component is used, influences the switching losses. For induc-
tive loads, a diode normally commutates together with the switch. As will be shown later
on, the diode characteristics will influence the power dissipation of the switch. If nothing is
done to prevent or reduce the switching losses, this is referred to as hard switching. Several
techniques exist to reduce the switching losses, this is called soft switching.
Switching losses are directly related to the switching frequency since they occur at every
switching instant. The losses are dissipated as heat. This means that the component will
get warmer for higher switching frequencies. If the frequency is too high, the dissipated
heat cannot be removed fast enough (think at the relatively small contact surface between
the component and the heat sink) and the thermal limitations of the component will be
exceeded and it will experience a thermal breakdown. Off course, this device failure has to
be avoided at all time. This is why the thermal design of a circuit is so important. This
will be discussed later on in section B.5. With respect to Fig. 2.1, the switching losses for
one period are given by Eqn. 2.3. This equation will be elaborated later on [12], [13], [14].
Pswitch =1
2T· VDC · IL · (ton + toff ) (2.3)
Where:
VDC is the DC bus voltage
IL is the load current
ton is the turn-on time of the device
toff is the turn-off time of the device
2.1.1 Power diodes
The symbol and V-I characteristic are given in Fig. 2.2. Note that the V-I characteristic
of power diodes has a more linear behavior in the first quadrant, compared to the more
exponential behavior of signal diodes. There is a small leakage current when the diode is
reversed biased. The power loss that corresponds to it is usually neglected since it is small
10 Chapter 2. Literature review
Fig. 2.1: Switching transients and power losses for a component in isolation
in comparison to conduction and switching losses. Also the structure of power diodes is
somehow different from that of signal diodes. For power diodes, a drift region is added in
between the P and N substrate. The doping level of the drift layer is much smaller than the
doping of the other two layers. It’s function is to absorb the depletion layer of the reverse
biased PN junction. By doing this, the reverse voltage rating of the diode will increase
[12]. However, we will not go into any further detail in these constructional aspects since
the working principle stays exactly the same as for signal diodes.
2.1.1.1 Conduction losses
The total voltage across a forward biased diode (VD)is the sum of the junction voltage (Vj)
and the drift region voltage (Vd). For practical purposes, a diode is usually modeled as a
voltage source Vs in combination with a dynamic resistance Ron, also shown in Fig. 2.2.
The voltage across the diode VD is then given by Eqn. 2.4 and the average conduction
losses over one cycle by Eqn. 2.5. In some cases, the data sheet only provides the forward
voltage drop across the diode, VD, the conduction losses are even easier to calculate, via
Eqn. 2.6.
2.1. Losses in hard switching 11
(a) Symbol
(b) I-V characteristic
Fig. 2.2: Diode
VD = Vs +RonID (2.4)
Pcond =1
T
∫ T
0
VDIDdt
= VsID,AV G +RonI2D,RMS
(2.5)
Pcond = VDID,AV G (2.6)
2.1.1.2 Switching losses
For low frequency applications, such as rectification of the grid voltages, diode switching
losses are often neglected. This is because the turn-on and turn-off times are extremely
small compared with the conduction time of the diode. For instance ton is in the order of
tens of nanoseconds, which is small compared to the period of the grid being 20 ms (for
f = 50 Hz). At higher frequencies the switching losses need to be included. Especially
12 Chapter 2. Literature review
Fig. 2.3: Reverse recovery of a diode
the turn-off needs special attention. An interesting phenomenon during turn-off that is
usually neglected in basic electronics courses is diode reverse recovery. Fig. 2.3 shows that
during a time trr, the reverse recovery time, a negative current flows through the diode.
This negative current is needed to reorganize the internal charge distribution, such that
the diode is able to block negative voltages again [12]. A low reverse recovery charge is
usually desired because the transistor losses can be reduced [15]. This will be explained
later on, in subsection 2.1.4.
2.1.2 Power MOSFETs
The symbol of an N-channel power MOSFET is shown in Fig. 2.4. MOSFETs are used
a lot in lower power applications. They can be switched on or off by applying a voltage
VGS between the gate and the source. Since they are voltage controlled, gate current will
only flow during the transitions to charge or discharge the internal gate capacitance. They
can be used for switching speeds up to 1 MHz. Their major drawback is the on-state
resistance RDS,on which increases rapidly for higher blocking voltage ratings, BVDSS. This
is why they are only used for voltages below 1 kV. Another important item is the inherent
’body diode’ of a MOSFET. This can be seen both as an advantage or a disadvantage.
It’s an advantage since no external anti-parallel diode needs to be added to the structure
2.1. Losses in hard switching 13
(a) MOSFET symbol (b) IGBT symbol
Fig. 2.4: Symbols of commonly used controllable switches
Serial number BVDSS (V) Id (A) RDS,on (Ω)
STD30NF03L 30 19 0,025
IRFR18N15DPbF 150 13 0,125
FDPF16N50 500 16 0,380
Table 2.1: Comparison of RDS,on
anymore. It can also be seen as a disadvantage since the body diode does not always have
the required properties (e.g. soft recovery)[16].
2.1.2.1 Conduction losses
When a MOSFET is conducting, it behaves as a resistor [13]. This resistor is usually
denoted as RDS,on and is given in the data sheet of the manufacturer. The conduction
losses are given by Eqn. 2.7. In Table 2.1, several examples of RDS,on are given. These
values can be found in the corresponding data sheet. Notice the strong relationship between
the blocking voltage BVDSS and RDS,on.
Pon = RDS,onI2D,RMS (2.7)
14 Chapter 2. Literature review
Serial number VCES (V) Ic (A) VCE,on (V)
ISL9V3040D3S 430 17 2,20
IRG4PH50UPbF 1200 24 2,78
IXGR16N170AH1 1700 16 6 5
Table 2.2: Most important IGBT specifications
2.1.2.2 Switching losses
The switching losses of MOSFETs are quite small since the turn-on and turn-off transients
are very fast. This means that ton and toff are very small, usually in the order of tens
of nanoseconds. However, the ideal switching transients will always be shorter than the
ones which are actually achieved. Therefore, the maximum data sheet parameters for ton
and toff should be used to give more realistic results [17]. As already mentioned in the
previous section, the switching losses do not only depend on the transistor itself. The circuit
in which it is placed and the diode that commutates together with it will have a significant
influence. This will be discussed in subsection 2.1.4. When no specific information about
the switching losses is given, Eqn. 2.3 can be used as a good approximation.
2.1.3 IGBTs
The symbol of an IGBT is shown in Fig. 2.4. IGBTs are the most used switching elements
in an intermediate power range (1→1000 kW). The state of the switch is determined by
the gate-emitter voltage VGE. Since they are voltage controlled, like MOSFETs, only a
small amount of power is required to switch them on or off. An IGBT’s on-state voltage
is rather small, like is the case for BJT’s. This can be seen from Table 2.2, summarizing
the most important parameters of several commercial IGBTs.
2.1.3.1 Conduction losses
The on-state voltage of an IGBT, VCE,on, is equal to the saturation voltage between the
collector and emitter and assumed constant when conducting [11]. This means that the
conduction losses are given by Eqn. 2.8.
2.1. Losses in hard switching 15
Pcond =1
T
∫ T
0
VCE,onIC
= VCE,onIC,AV G
(2.8)
2.1.3.2 Switching losses
Switching losses in IGBTs are rather high compared to diodes and MOSFETs. The major
contribution to the switching losses in an IGBT is due to a phenomenon which is called
the tail current of the IGBT. It only occurs at turn-off and is shown in Fig. 2.5. When the
device turns off, the collector current IC first decreases rapidly during time interval tf1.
This is called the MOSFET part of the turn-off. Afterwards, the decrease becomes much
smaller during a time interval tf2. This is called the BJT part of the turn-off. The length
of tf2 should be as short as possible since the voltage VCE is already at it’s blocking value,
which is usually quite high. This means that the power dissipation will be large during tf2.
The duration of the tail is usually in the range of 200 to 500 ns. The tail current cannot
be avoided since it is caused by minority carriers which are trapped in the base of the
BJT part of the IGBT, Fig. 2.5. They cannot be removed fast since there is no external
connection to this ’internal base’. The minority carriers have to recombine naturally. This
is a relatively slow process which causes the device to remain in his on-state for a longer
time. If temperature increases, the tail current will increase too [12], [18], [19].
In practice, hard switching losses are specified in the data sheet of the component. These
losses are usually referred to as Eon and Eoff and are valid under certain reference con-
ditions. If the reference conditions are not met, a linear interpolation should be applied
as described in [11]. It is also important to check whether or not the reverse recovery and
tail current losses are included. To find the power dissipation, one can simply multiply the
sum Eon + Eoff with the switching frequency, Eqn. 2.9.
Pswitch = fs · (Eon + Eoff ) (2.9)
16 Chapter 2. Literature review
(a) IGBT equivalent circuit (b) Tail current at IGBT turn-off
Fig. 2.5: IGBT structure and tail current
2.1.4 Switch cell
As mentioned before, the switching losses can be determined for a component as a sepa-
rate unit. However, it is more convenient to consider them for two commutating devices,
placed in a circuit. In power electronic circuits, the switched current mostly has an in-
ductive nature (e.g. Buck converter) and can therefore be considered as constant during
the switching interval. The most common and practical relevant case of a commutation
between a diode and a MOSFET/IGBT will be discussed here. This kind of commutation
is normally the one used for testing the transistor and is mostly given in the data sheet.
Consider the circuit and waveforms of Fig. 2.6. The current IL is assumed to be constant.
The transitions are assumed to be linear. Consider the following events[15]:
At first, the current IL flows entirely through the diode. The MOSFET is off.
At time t1, the MOSFET is turned on. The current IS linearly increases, while
ID linearly decreases, according to IL = IS + ID. The voltages VS and VD remain
constant, being VS = VDC and VD = VF . VF is the forward voltage drop across the
diode.
At time t2, ID becomes negative. This is due to the reverse recovery charge of the
diode. The maximum negative value of the current is called IRRM . Since IL = ID+IS
is still valid, this ’overshoot’ in the diode current is also visible in the MOSFET
current.
At t3, the diode starts to block and VD rises. VS falls during a time tB since VDC =
VD + VS
2.1. Losses in hard switching 17
Fig. 2.6: Reference circuit for loss calculation
Finally, at t4, the transient finishes. The current that was first going through the
diode has commutated to the MOSFET.
The switching losses over one period can be calculated as follows:
Pdiode =1
6TtBVDCIRRM (2.10)
Pswitch =1
2TVDCILtr + VDC(IL +
1
2IRRM)ta + VDC(
1
2IL +
1
3IRRM)tB (2.11)
18 Chapter 2. Literature review
(a) Diode voltage (b) Diode current
(c) Switch voltage (d) Switch current
Fig. 2.7: Waveforms during transition of switches
2.2. Driver circuits 19
The described process is called forced commutation. The most important aspect of this
process is that the reverse recovery does not only affect the diode but also the transistor.
Let us compare a normal and a soft recovery diode with the same trr. For a soft recovery
diode, tb > ta, and for a normal diode tb << ta. The switching losses in the diode itself
will be higher for the soft recovery diode, since they are proportional to tb. But more
importantly, the losses in the transistor will be much smaller. This means that the overall
system efficiency can be improved. Other advantages of soft switching diodes are the low
dv/dt during tb and the lower IRRM compared to regular diodes. This is beneficial for the
radiated EMI and ringing losses. [13],[15]
2.2 Driver circuits
In recent years, great improvements have been made in the design of driver circuits. When
the thyristor was still the most frequently used component, driver circuits consumed a lot
of power to turn on the device, since the gate was current controlled. To shut it down, an
auxiliary circuit was needed to force the main current to zero. Nowadays huge advances
are made with voltage controlled gates for both IGBTs and MOSFETs.
The driver circuit forms the interface between the power switches and the controller. It
provides the necessary power for switching the transistors. Mostly, galvanic insulation
between the controller and the power circuit is provided. This can be done by using
transformers, optocouplers or fiber optic cables. Since the driver circuit consumes a certain
amount of energy, this is also considered as a loss. Therefore, the driving losses should
be added to the conduction and switching losses to determine the overall losses. Usually
they are relatively small compared to the total power of the inverter. Some basic driver
considerations for MOSFETs and IGBTs will be considered here. The focus will lie on
IGBTs since they will be the main switches of the inverter and the MOSFETs will serve as
auxiliary switches. At turn-on, IGBTs and MOSFETs behave in almost exactly the same
way. At turn-off, the IGBT current has the typical current tailing problem which was
already mentioned in subsection 2.1.3. This phenomenon is not observed for MOSFETs.
The insulated gate of IGBTs and MOSFETs behaves like a capacitor. Therefore, for
driving power switches, the internal capacitances between the three terminals of the device
are important. They are shown for both an IGBT and a MOSFET in Fig. 2.8. These
20 Chapter 2. Literature review
devices are voltage driven, which basically means that they start switching when VGE
passes a certain threshold. An important aspect of the switching waveform is shown in
Fig. 2.9. One can see that the gate voltage exhibits a step. This occurs at turn-on and
turn-off. The gate voltage remains at the same level while VCE rises or falls. This effect
is due to the Miller capacitance CGC and the voltage threshold level is called the Miller
plateau. The Miller plateau should be as short as possible. This is because the switch
is operated in linear operation and not in switch mode during this event, which basically
means an increase in switching losses. The Miller effect can be strongly reduced or even
eliminated by using zero voltage switching (ZVS) as discussed in [20].
An appropriate design can speed up or slow down the switching times. But it is important
to notice that the IGBT tail current cannot be influenced by the driver. The tail current
losses are completely independent of the driver. Only the losses due to the dV/dt can be
reduced. The driver has to be capable of delivering high dV/dt rates and high peak current.
In this way, the needed charge for switching is delivered quickly and switching times can
be reduced. Usually the driver itself can be bought separately and only some external
components need to be added. The most important one is the gate drive resistance Rg as
shown in Fig. 2.10. A low Rg is beneficial since the gate will charge more quickly. The
switching event will be shorter and switching losses will be lower. However, eliminating
Rg is not possible. There is always a certain emitter inductance present in the PCB design
and even in the IGBT package itself. This emitter inductance is called Le. Together with
CGE, this forms an LC tank circuit. If no or very little gate resistance is present, the
circuit will start to oscillate when a voltage step is applied. The transistor will turn ON
and OFF several times, which will cause high losses. Another problem is that the voltage
across Le is subtracted during turn-on and added during turn-off. This slows down the
switching. Therefore, Le should be kept as low as possible by using wide PCB tracks and
locate the driver as close as possible to the IGBT gate. By doing so, Rg can be optimized
and be as low as possible. It is also important to mention that the optimum gate resistance
can be different for turn-on and turn-off. For high inductive loads, the turn on resistance
should be relatively high to reduce turn-on spiking. However, it is usually too high for fast
turn-off. To overcome this problem, the resistor is split up in 2 parts and an anti-parallel
diode is placed across one of them. This creates a short circuit over one of the resistors,
which speeds up the turn-off. [7], [18], [21], [22].
2.2. Driver circuits 21
(a) IGBT (b) MOSFET
Fig. 2.8: Internal capacitances
(a) Turn on (b) Turn off
Fig. 2.9: IGBT switching behavior
Fig. 2.10: IGBT with driver circuit, gate resistance and common emitter inductance
22 Chapter 2. Literature review
2.3 Efficiency vs. switching frequency
As seen in the previous paragraphs, the losses can be split up in different parts [19]:
1. Switching losses Psw
For Etot being the total energy lost in the switching transitions of one switching pe-
riod, the average switching losses Psw can be found by multiplying with the switching
frequency fs: Psw = Etot · fsw
2. Conduction losses Pcond
3. Driver losses Pdriver
The total losses Ploss are then given by:
Ploss = Etot · fs + Pcond + Pdriver (2.12)
One can see that the total losses are thus determined by a constant term and a term that
linearly increases with fs. The critical frequency fcrit is defined as the frequency at which
the switching losses are equal to the constant losses:
fcrit =Pcond + Pdriver
Etot(2.13)
Below fcrit, the losses are mainly determined by driver and conduction losses. Above fcrit,
the losses are mainly switching losses and the efficiency decreases rapidly with increasing
fs. This typical phenomenon is shown in Fig. 2.11 where fcrit = 14, 4 kHz.
2.4 Inverters for grid connected power supplies
Voltage source inverters (VSI) that are intended for motor drive applications usually only
have three legs and do not use any filters. The neutral point of the motor is not connected
since this could generate the flow of a zero-sequence current and the filtering is done by
2.4. Inverters for grid connected power supplies 23
Switching frequency fs [Hz]
102 103 104 105 106 107
Eff
icie
ncy η
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
fcrit
= 14.4kHz
Fig. 2.11: Efficiency as a function of the switching frequency
the inductive behavior of the motor. This basically means that the line current will be
approximately sinusoidal while the line voltages still contain lots of higher harmonics. The
same topology (three phase VSI) can be applied for grid connected inverters. However, in
many cases a slightly different version is used. Grid-connected inverters usually contain
four legs and a filter in between the grid and the inverter. This is because the neutral point
of the load needs to be connected. The advantage is that currents can be injected in the
neutral, which may be necessary if the zero-sequence component needs to be reduced. The
fourth leg can either contain two extra switches or two capacitors. These frequently used
topologies are shown in Fig. 2.12. The first one uses a fourth leg with two extra switches.
The inductor Ln is placed in between the midpoint and the neutral point of the grid for
filtering purposes. The second topology is called a ’split DC bus inverter’. The DC bus
is split up in two parts by using two capacitors in series. The neutral point of the grid
is connected to the midpoint. Notice that no neutral filter is used in this configuration
since the fourth leg does not participate in the switching. The split DC bus inverter is
usually preferred since only six instead of eight switches are used. Furthermore, it can be
considered as three half-bridge inverters which will be an advantage for the current control.
It’s major drawback is the DC bus voltage which has to be chosen a lot higher. Also the
required voltage control over the two capacitors is an undesired feature. [23]
The extra leg might seem to be a disadvantage at any time. This is not always the case.
24 Chapter 2. Literature review
Certain soft switching circuits also require a split DC bus. It is thus an advantage if this
midpoint is already present in the device.
2.4. Inverters for grid connected power supplies 25
(a) Using two extra switches
(b) Using a split DC bus
Fig. 2.12: Inverter topologies for grid connection
26 Chapter 2. Literature review
2.5 Enhancement of switching losses
The use of a high switching frequency can be very beneficial. It decreases the size of trans-
formers and filter components. This makes the circuit smaller and lighter, which means
that the power density increases. Also the transient response, which may be important
for control purposes, is faster. For VSI applications, the output waveform quality will
increase, which leads to fewer harmonics in the line current spectrum. The drawbacks
of a high switching frequency are the increased switching losses and an increased electro-
magnetic interference(EMI). In this part, possible solutions will be discussed that enhance
switching or reduce switching losses. A comparison between the different possibilities will
be made at the end of this section.
2.5.1 RLC circuits
Most soft switching applications rely on the effects that are caused by an LC resonance.
Therefore, the basics of RLC circuits will be reviewed. Especially the behavior of DC RLC
circuits is important since the resonance circuits are usually placed on the DC bus of the
inverter or converter.
Consider the series RLC circuit shown in Fig. 2.13 . The equation that describes this
system can be found via Kirchoff’s voltage law (KVL):
V = VR + VL + VC (2.14)
Fig. 2.13: Series RLC circuit
2.5. Enhancement of switching losses 27
Filling in the constitutive equations leads to:
V = Ri(t) + Ldi(t)
dt+
1
C
∫ t
0
i(t)dt (2.15)
Via differentiation of both parts, this can be rewritten as:
0 =d2i(t)
dt2+R
L
di(t)
dt+
1
LCi(t) (2.16)
This is a second-order linear differential equation with constant coefficients. Usually this
result is rewritten in a more general form:
0 =d2i(t)
dt2+ 2α
di
dt+ ω2
0i(t) (2.17)
In which:
α = R2L
is the attenuation constant
ω0 = 1√LC
is the natural pulsation of the system
The solution of this equation is given by:
i(t) = A1es1t + A2e
s2t (2.18)
Where A1, A2 are determined by the boundary conditions and where s1 and s2 are the
solutions of the characteristic equation:
0 = s2 + 2αs+ ω20 (2.19)
s1,2 = −α±√α2 − ω2
0 (2.20)
Depending on the values of R, L and C the system’s transient response will be different
and can be determined via ζ = αω0
= R2
√CL
, the damping factor:
ζ < 1: Under damped
ζ = 1: Critically damped
ζ > 1: Over damped
28 Chapter 2. Literature review
Time [s] ×10-40 1 2 3 4 5 6 7 8
Vol
tage
[V
] / C
urre
nt [
A]
-100
-80
-60
-40
-20
0
20
40
60
80
100
Current IVoltage V
L
Fig. 2.14: Step response of an RLC circuit
In our case, R will be usually very small since it is the Equivalent Series Resistance (ESR)
of both L and C. Therefore, the response will be under damped. This case is shown in
Fig. 2.14 where both the current i(t) (black) and the voltage across the capacitor vc(t)
(red) are shown for a step of 100 V. The parameters are: R = 0.5 Ω, L = 33 µH and C =
1 µF. Both waveforms are in anti-phase because VL = Ldidt
. The current eventually dies
out since the capacitor gets fully charged.
2.5.2 Silicon Carbide
A first possible method to increase the switching frequency, without having excessive losses,
is simply by using better components. Present day, Silicon (Si) is the most used semicon-
ductor material. A new upcoming and improved semiconductor is Silicon Carbide (SiC).
The benefits of SiC compared to Si are discussed below, [24], [25].
SiC devices have a lower on-resistance compared to Si devices. This results in lower
conduction losses.
SiC devices have higher reverse breakdown voltages.
2.5. Enhancement of switching losses 29
Product number Type Reverse voltage (VR) Forward current (IF ) Forward voltage drop (VF ) Price (¤)
DSR6V600D1 Ultrafast recovery rectifier diode - Si 600 6 3,0 0,56
SCS106AG Schottky barrier diode - SiC 600 6 1,5 9,49
Table 2.3: Comparison between Si and SiC components
SiC has a higher thermal conductivity.
The allowable junction temperatures of SiC are much higher. Today, SiC can operate
with temperatures up to 350 C, whereas Si is limited to 150 C.
The reverse recovery current of SiC devices is much lower compared to Si based
devices. This means that the switching losses are strongly reduced.
It may be clear from the previous arguments that in a near future, SiC will probably
replace Si as the standard semiconductor material used in power electronics applications.
It is estimated that a reduction of 40 % in weight and volume will be possible. Anyhow,
their biggest disadvantage is the very high price. An example is given in Table 2.3 where
the price of the SiC component is almost 17 times as high. This big price difference is due
to the high costs in the fabrication process of SiC. Present day techniques do not allow
a commercially acceptable price. Another disadvantage is the higher control complexity.
E.g. SiC MOSFETs need a +20V/-5V driving circuit. However, these constraints will be
eliminated over time, making SiC devices the future of power electronics. By using these
components, the need for auxiliary circuits will be eliminated.
2.5.3 Snubber circuits
In low power applications (< 100 W), snubbers are often used to provide damping , limit
dI/dt or dV/dt and ’reduce’ the switching losses. The term is put between brackets since
these circuits mostly only shift the problem. The switching losses are not dissipated in
the switch but are diverted to extra components which are placed in parallel to it. The
snubber circuit does not effectively reduce the switching losses. A possible snubber used for
diodes and IGBT’s is shown in Fig. 2.15. More complicated topologies exist but they are
not able to effectively increase the efficiency. For example, in [26], efficiency can increase
with 1% but only under full load conditions. We will not go in any further details since
this is clearly not an attractive solution.
30 Chapter 2. Literature review
(a) Snubber for diodes(b) Snubber for IGBTs
Fig. 2.15: Snubber circuits
2.5.4 Soft switching
Soft switching is a technique that shapes the voltage and current waveforms in such a way
that the switching transitions occur under favorable conditions. The device stresses are
strongly reduced by these zero voltage or zero current switchings. This purpose can be
achieved in many ways. Some of them will be more suited than others. Using soft switching
techniques is usually a must in high frequency applications, since otherwise the component
would heat up too much. The drawbacks differ from case to case but are mostly situated
in the following areas:
Higher conduction losses
More passive components, mainly diodes, capacitors and inductors
Extra active components
Higher circuit control complexity
Lower robustness, degraded reliability
Also the total cost of the circuit increases. However, if properly applied, the benefits are
considerable since the switching frequency can be increased:
2.5. Enhancement of switching losses 31
Lower switching losses
Lower switching stresses due to more favorable switching loci
Smaller filtering equipment
Smaller transformers
Smaller heat sinks
Less EMI due to lower dv/dt and di/dt
Lower acoustic noise
Since transformers, filters and heat sinks are usually quite large and heavy, these advance-
ments will lead to a lighter and more compact design. This enables a higher power density
and a higher efficiency [12].
Most soft switching circuits are designed for PWM DC-DC converters. The topologies
that exist for inverters are rather limited. But, this will not be a problem. As will be
explained in more detail in Chapter 3, the split DC link inverter can be thought of as six
buck converters in parallel / anti-parallel. This allows us to use the same soft switching
topologies for the DC-AC inverter as the ones that are used for DC-DC converters.
2.5.4.1 Resonant DC link inverters
Resonant DC link inverters shift the resonant circuit from the inverter bridge to the DC bus
[27]. Series Resonant DC Link (SRDCL) inverters were the first ones being used. They
employed a resonant inductor-capacitor circuit between the DC source and the inverter
bridge to supply a resonating voltage across the inverter. Their major drawback is that
the resonating voltage can exceed twice the DC bus voltage. This means that the blocking
voltage of the switches needs to be increased. Several circuits exist to overcome this problem
but they all suffer from reliability problems [28], [29]. Also a special control strategy is
needed to establish the same initial conditions for each switching cycle. Another possible
type are the parallel resonant DC link inverters (PRDCL), Fig. 2.16. Several designs of
this type are proposed and investigated in [4] and [30]. However, also the PRDCL inverters
have some important drawbacks:
32 Chapter 2. Literature review
Fig. 2.16: Parallel resonant DC link inverter
Three extra switches and three extra diodes. The power rating of these auxiliary
components is comparable to the power rating of the main switches. This means a
significant increase in the overall cost of the inverter.
Increased complexity. Current control is needed before the resonant cycle can be
initiated.
Very high conduction losses in the resonant link.
A decrease of the fundamental output voltage due to the notches in the line-to-line
voltages. This is because the resonance takes a finite amount of time.
Due to these limitations, resonant DC link inverters have never known a real breakthrough
and will probably never be a commercial success. [4], [31]
2.5.4.2 Zero voltage and zero current switchings
Resonant inverters place LC networks in between the load and the switch network. These
are called ’tank networks’. Series LC, parallel LC and LCC filters are all possible. By
2.5. Enhancement of switching losses 33
Fig. 2.17: ZVS-PWM Buck converter
adding them to the circuit, series or parallel loaded resonant converters are created. They
provide an oscillating load voltage and/or current which leads to zero voltage or zero
current switchings (ZVS/ZCS). A possible implementation of a ZVS-PWM converter is
shown in Fig. 2.17. ZVS is preferred for power MOSFETs that are used at high operating
frequencies. This is because ZVS eliminates the capacitive turn-on loss and it reduces the
turn-off switching loss by slowing down the voltage rise and reducing the overlap between
the switch voltage and switch current [32]. This circuit achieves soft switching for both
transistors and allows a bidirectional power flow. However, the peak current through the
transistor increases. Because of this, conduction losses can increase up to 40%. Another
drawback is that the frequency needs to be changed in order to achieve the ZVS over a wide
load range. This makes the optimization of filters, transformers etc. more difficult. Due
to these limitations, the ZVS-PWM converters are only used in some specific applications.
For example, the full bridge ZVS-PWM converter where no auxiliary switch is needed.
For IGBTs, zero current switching (ZCS) is more effective for reducing switching losses at
high switching frequencies. This is because the ZCS forces the current to zero before the
switch voltage rises. A possible topology for a Buck converter is shown in Fig. 2.18. It
achieves ZCS for the transistor and ZCS for the diode. But several drawbacks arise. The
voltage stress of the diode doubles. Due to parasitic ringing, circulating energy is present,
which increases the conduction losses. The zero current switching is also very sensitive to
changes of the line voltage and of the load. [32]
34 Chapter 2. Literature review
Fig. 2.18: ZCS-PWM Buck converter
2.5.4.3 Zero voltage and zero current transitions
Zero voltage and zero current transition (ZVT/ZCT) converters were developed to minimize
the disadvantages of ZVS and ZCS converters. In the previous type, the resonant element
is placed in the main power path. This leads to the aforementioned shortcomings such
as circulating energy, additional voltage stresses and dependency on line voltage and load
current. If the resonant elements are placed outside the main power path, the auxiliary
switch can be sized much smaller. This resonance network is called a soft switching auxiliary
circuit (SSAC) and is placed in shunt across the switch. The SSAC is only active during
a small time interval, when it creates the transition conditions for the main switch. A
possible implementation of a ZVT SSAC is shown in Fig. 2.19, applied to a Buck converter
[1]. The resonance is created by closing the auxiliary switch Sa. An important feature is
that, after the transient has finished, the circuit goes back to its normal operating mode.
This means that a regular PWM control can still be applied for the main switch Sm. The
key features are summarized below:
Soft switching of both the main switch and the diode. The auxiliary switch is hard
switched. However, the power rating of the auxiliary switch is relatively small com-
pared to the power rating of the main switch. This means that the overall switching
losses of the converter still decrease.
Low switching stresses on the switches since the switching loci are very beneficial.
2.5. Enhancement of switching losses 35
Fig. 2.19: First implementation of a ZVT-PWM Buck converter [1]
Soft switching is achieved over almost the whole load range of the inverter. Also the
dependency on the line voltage is very low.
This SSAC was created in 1993. In the meanwhile, other circuits have emerged that solved
the shortcomings of the aforementioned topology. An interesting circuit is proposed in
[2]. The circuit and its implementation in a Buck converter are shown in Fig. 2.20 and
Fig. 2.21. In this SSAC, all active and passive semiconductor devices turn on and off with
ZVS. The auxiliary switch is operated under ZCS conditions. There are no additional
voltage or current stresses for the main switch. The circuit uses two switches, four diodes,
two resonant inductors and two resonant capacitors. The auxiliary switch has a small
power rating compared to the main switch. This SSAC is clearly one of the best possible
solutions for reducing the switching losses. Therefore, the cell will be discussed more in
detail in Chapter 3.
However, also other ZVT SSAC are available. Fig. 2.22 shows a possible configuration that
is used for inverters, proposed in [3]. It is called a MidPoint Clamped (MPC) SSAC since
a split DC bus is required for its operation. This midpoint is usually a disadvantage for
36 Chapter 2. Literature review
Fig. 2.20: Recently proposed soft switching auxiliary circuit [2]
Fig. 2.21: Implementation the newly proposed SSAC in a Buck converter
2.6. Example: Losses of a hard switched inverter 37
Fig. 2.22: Proposed soft switching auxiliary circuit [3]
VSI that are intended for motor applications. For grid-connected inverters, the midpoint
is already available since it is needed for voltage unbalance compensation. Notice that only
a small amount of components is needed (two switches, two diodes, two inductors and two
capacitors). The circuit can be used for both the top and bottom switch of an inverter leg.
For three-phase applications, three SSAC need to be applied with a separate control. This
means that the control is relatively easy. This topology also seems an attractive solution
and will therefor be further developed in Chapter 3.
2.6 Example: Losses of a hard switched inverter
In this section, the losses of a hard switched three-phase inverter are estimated. The
following items are given:
IGBT and free-wheeling diode in one package, IRGP30B120KD-E
DC bus voltage, VDC = 850V
Peak output power per phase, Pmax = 4kW
Minimum RMS voltage of the grid, Vmin = 207V
Switching frequency, fs = 20kHz
38 Chapter 2. Literature review
Unity power factor
From the modulation strategy, the maximum duty cycle per period is δmax = 0, 7
and thus m = 0,7
Only the losses for one transistor of one leg will be calculated. At the end, the result can
be extended for three legs. The IGBT data sheet can be found in Appendix F. First, the
maximum RMS collector current needs to be determined from the maximum output power
and the minimum output voltage:
Ic,RMS,max =PmaxVmin
=4000W
207V= 19, 32A
The total switching losses for a junction temperature of 125 C can be found from the data
sheet: Etot = 4436µJ . This value already includes the diode reverse recovery and tail
current losses. However, this value is specified for a collector current Ic = 25A and a DC
bus voltage of VDC = 600V . So it needs to be recalculated for this application. A linear
approximation is used:
Etot,new = Etot,old ·V DC
VDC,datasheet· Ic,RMS,max
Ic,datasheet
= 4436µJ · 850V
600V· 19, 32A
25A
= 4857µJ
By multiplying this result with the switching frequency fs, the switching losses are deter-
mined:
Pswitching = Etot,new · fs = 4857µJ · 20kHz = 97, 1W
For calculating the conduction losses we refer to [33] and [34], where a formula is given to
determine the conduction losses of inverters based on the output current and modulation
coefficient under the following assumptions:
2.7. Conclusion 39
Linear modulation of the inverter (0 5 m 5 1)
Switching times are neglected
Constant junction temperatures (valid for a fundamental output frequency of 50 Hz)
Ripple of the output current Io is negligible
Linear characteristics of the components in the forward conduction region: VCE =
VCE,0 +RCE · IC (IGBT) and VD = VD,0 +RD · ID (diode)
Pcond,diode =1
2(VD,0 ·
Io,pπ
+RD ·I2o,p4
)−m · cos(φ) · (VD,0 ·Io,p8
+1
3π·RD · I2o,p)
= 3.22 W
Pcond,IGBT =1
2(VCE0 ·
Io,pπ
+RCE ·I2o,p4
) +m · cos(φ) · (VCE0 ·Io,p8
+1
3π·RCE · I2o,p)
= 17.54 W
This leads to a total power dissipation of Ptotal = 117.86 W. This is approximately equal
to the maximum power dissipation of the casing, being 120 W at 100 C, as taken from
the data sheet. Notice that the switching losses account for 82,3 % of the total losses. If
they could be eliminated or strongly reduced, the total power dissipation will be more than
halved. Or, another approach is to reduce them while increasing the switching frequency.
This means that the heat sink stays at the same size but the filter components can be
reduced.
2.7 Conclusion
As can be seen from the example, switching losses are the main part of the losses in an
inverter. They increase linearly with the switching frequency. The switching frequency is
usually chosen as high as possible. This is done to reduce the size and weight of inductors
40 Chapter 2. Literature review
Fig. 2.23: Switching loci for different techniques
and transformers such that the power density of the device can be increased. This trade-
off between switching losses and switching frequency is one of the major concerns for
the design of a power electronic circuit. It has been shown that the type of switch, the
switching frequency, the freewheeling diode and the driver determine the total switching
losses. To reduce the switching losses, several methods such as snubbers, Silicon Carbide
components and soft switching techniques have been proposed. The switching loci for
these methods are shown in Fig. 2.23 . One can see that the soft switching techniques
have the best switching loci, which is translated in low switching stresses and good EMI
performance. Table 2.4 gives a comparison of the different techniques, based on several
important criteria such as the number of components, the cost and the complexity of the
circuit. Therefrom, it should be clear that, to the present day, soft switching techniques are
still the most valuable solution. Their greatest disadvantage is the relatively high number
of components that is needed per switch. However, it is possible that in the near future,
SiC components become cheaper and therefore more interesting to use.
Also notice that list of described soft switching techniques is not meant to be exhaustive.
Several other possibilities that exist are described in [27] but only the most interesting ones
for this specific case are presented here.
2.7. Conclusion 41
SiC Snubber RDCL ZVS/ZCS ZVT 1 ZVT 2 MPC
Effectiveness ++ −− + ++ + ++ ++
N of comp. ++ - + + −− −− +
Cost −− + + + - −− +/-
SS for all comp. NA −− - - - ++ ++
Control complexity - + - + - - -
Switching loci - - + + ++ ++ ++
Load dependency ++ + - −− ++ ++ ++
Influence on regular circuit NA + - −− ++ ++ ++
Power rating aux. switches NA NA High High Low Low Low
Table 2.4: Comparison of possible techniques
42 Chapter 2. Literature review
Chapter 3
Simulations
In this chapter, the ZVT circuits that have been discussed in Chapter 2 will be further
elaborated. Three simulations have been done and will be discussed in detail. This discus-
sion is important for the further development since only the most interesting soft switching
auxiliary circuits (SSAC) will be experimentally tested.
3.1 Buck converter with split DC bus
The proposed examples until now were all adopted for Buck converters. This has been done
intentionally since a three-phase inverter is in fact an extension where Buck converters are
placed in parallel/ anti-parallel. In this section, a Buck converter with split DC bus will
be analyzed. The transfer function of this converter is different from the transfer function
of a regular Buck converter. Therefore, we will derive it first. The relation between the
in- and output voltages of a regular Buck converter can be easily found by integrating the
voltage over the inductor over one switching period, T . Since the converter is analyzed in
steady state, this integral equals zero.
∫ T
0
vLdt =
∫ ton
0
vLdt+
∫ T
ton
vLdt = 0 (3.1)
43
44 Chapter 3. Simulations
After developing this equation, the well known result for a Buck converter is found, Eqn.
3.2 where δ is the duty ratio, defined as ton/T
VoutVin
= δ (3.2)
The behavior is different in case of the split DC bus, when compared to a regular Buck
converter. Consider the circuit of Fig. 3.1. The following assumptions are made:
The voltage across the two capacitors is evenly distributed and constant
The output voltage is assumed constant
The transistor and the diode are ideal switches
The converter is analyzed in steady state
The transfer function can again be found by integrating the inductor voltage over one
switching period.
∫ T
0
vLdt = 0 =
∫ ton
0
vLdt+
∫ T
ton
vLdt
=
∫ ton
0
(Vin − Vout)dt+
∫ T
ton
(−Vin − Vout)dt
= (Vin − Vout)ton − (Vin + Vout)(T − ton)
(3.3)
So the transfer function is given by Eqn. 3.4.
VoutVin
= 2δ − 1 (3.4)
This result has been verified by simulations in PSPICE. Only a slight difference (1%) in
the output voltage was observed due to the losses in the diode and in the transistor and
the above derivation is thus valid. The importance of this adapted circuit, stems from the
fact that it is the closest converter possible to the inverters with split DC bus and neutral
connection, as described in Chapter 2. The analogy is also shown in Fig. 3.1. If a SSAC
3.1. Buck converter with split DC bus 45
works correctly in case of the Buck converter with split DC bus, we also know that it will
work in case of a grid-connected inverter with split DC bus. A positive consequence is that
the simulations become less complex. The circuit will be tested for different duty ratios in
between 0,55 ≤ δ ≤ 0,95.
(a) Buck converter with split DC bus
(b) Three-phase inverter with split DC bus
Fig. 3.1: Split DC bus topologies
46 Chapter 3. Simulations
3.2 Simulation 1 - ZVT SSAC 1
This section contains the results of the simulations of the different topologies with and
without ZVT. The circuits are implemented in PSpice via Orcad Capture. The post-
processing has been done via Matlab and Microsoft Excel. It is important to notice that
the purpose of the simulations is a qualitative approach of the problem. It has to be shown
that ZVT conditions occur and that, despite the increased number of components, the
total efficiency of the system increases. It is not the aim to predict the exact efficiency
improvements. This is also not possible since the components that will be used are not yet
determined.
Also notice that the simulations use MOSFETs as main power switches, where the actual
design of an inverter can be with MOSFETs or IGBTs. A positive consequence is the
increased simulation speed. An IGBT is a more complex component in SPICE compared
to a MOSFET and as a consequence it slows down the entire simulation process. As
mentioned before, the turn-on behavior of MOSFETs and IGBTs is comparable, where
as the turn-off of IGBTs is usually more energy consuming due to the IGBT tail current
problem.
3.2.1 Components and simulation settings
Throughout all simulations, the same components are used such that the results can be
easily compared. They are given in Table 3.1. It is also important to notice that the passive
components are assumed ideal. This means that an inductance will be modeled as a simple
inductance, without any parasitic capacitance or resistance. They are assumed linear, so
the saturation and hysteresis are neglected. The same holds for the capacitors that will be
modeled without parasitic inductance or resistance. This is off course not entirely correct.
However, in this first design stage, the components that will be used are not selected yet,
it is best practice to exclude them from the simulation because it is impossible to allocate
an appropriate value. The main contribution will be a decrease in efficiency due to the
resistance. It is however possible that they are included in later simulations when the
appropriate components have been chosen.
The simulation itself is always a transient analysis with a maximum step size of 10 ns.
3.2. Simulation 1 - ZVT SSAC 1 47
Input voltage Vin 425 V
MOSFET IRF450
Diode MUR450
Filter inductance Lf 2 mH
Filter capacitance Cf 5 µF
Gate resistance Rg 10 Ω
Driving voltage 15 V
Switching frequency fs 50 kHz
Table 3.1: Component values for simulation
Fig. 3.2: Buck converter simulation model in SPICE
This is small enough compared to a switching frequency fs of 50 kHz which corresponds
to a period Ts of 20 µs The efficiency analysis is done via Matlab by using trapezoidal
integration over an integer multiple of switching periods.
3.2.2 Buck converter
At first, a regular Buck converter is simulated. The circuit is shown in Fig. 3.2, and the
voltages, currents and power consumption over 3 periods for a duty cycle δ = 60% are
shown in Fig. 3.3, Fig. 3.4, Fig. 3.5, respectively. The results are as expected. The input
48 Chapter 3. Simulations
voltage is constant, while the output voltage contains a small ripple of approximately 1 V.
This is due to the finite output capacitance. The inductor current contains a DC value of
approximately 10 A and a ripple of approximately 1 A peak-to-peak.
The inductor current rises during 60 % of the time. This is when the MOSFET is on. The
inductor current then equals the MOSFET current. While the MOSFET is off (during
40% of the time), the inductor current equals the diode current. The most important phe-
nomenon for our case arises when the MOSFET is turned on. One can then observe a peak
in the MOSFET current as shown in Fig. 3.6. This is due to the diode reverse recovery.
The diode current becomes negative and has a peak of −76 A. When going back to zero,
an overshoot of 48 A occurs before the current goes back to zero. These waveforms are also
found in the MOSFET current since the currents are related via IL = ID + IM . Where
IL can be assumed constant since the time interval of the reverse recovery is very short.
However, these large currents will lead to a lot of dissipation as can be seen in the Fig. 3.5
which shows the power dissipation at the same instant for the diode and the MOSFET.
One can clearly see the peaks in the power dissipation during turn-on and turn-off of the
MOSFET. These peaks are related to the switching losses of the device and are clearly
much higher than the conduction losses. From the figure, it can be seen that the conduction
losses of the MOSFET are equal to 29 W and the conduction losses of the diode equal 14 W.
Fig. 3.7 and Fig. 3.8 show the turn-on and turn-off in more detail. The turn-on has a
peak power consumption of almost 40 kW for the MOSFET and a peak of 3 kW for the
diode. However, this phenomenon takes less than 0.1 µs which means that the total energy
dissipation is rather limited. The turn-off power dissipation in the MOSFET contains a
peak value of about 4.5 kW while no peak is observed for the diode. So the total energy
dissipation of one switching action is rather limited due to the very small time intervals
that are involved. However, it may be clear that the occurrence of these phenomena is
directly related to the switching frequency. A higher switching frequency thus means more
switching losses, which deteriorates the efficiency.
3.2. Simulation 1 - ZVT SSAC 1 49
×10-35 5.01 5.02 5.03 5.04 5.05 5.06
Vol
ts [
V]
424
424.5
425
425.5
426Input voltage V
in
Time [s] ×10-35 5.01 5.02 5.03 5.04 5.05 5.06
252
253
254
255
256
257Output voltage V
out
Fig. 3.3: Input and output voltages
×10-35 5.01 5.02 5.03 5.04 5.05 5.06
8
10
12Inductor current I
L
×10-35 5.01 5.02 5.03 5.04 5.05 5.06
Cur
rent
[A
]
0
5
10
Diode current ID
Time [s] ×10-35 5.01 5.02 5.03 5.04 5.05 5.06
0
5
10
Mosfet current IM
Fig. 3.4: Inductor, diode and MOSFET currents
50 Chapter 3. Simulations
×10-35 5.01 5.02 5.03 5.04 5.05 5.06
Pow
er c
onsu
mpt
ion
[W]
×104
-2
0
2
4Power consumption - Mosfet
Time [s] ×10-35 5.01 5.02 5.03 5.04 5.05 5.06
-10000
-5000
0
5000Power consumption Diode
X: 0.00502Y: 3.656e+04
X: 0.005032Y: 4392X: 0.005004
Y: 29.08X: 0.005014Y: 0.1708
X: 0.005004Y: 5.344e-06
X: 0.005015Y: 13.94
Fig. 3.5: MOSFET and diode power consumption
×10-35.02 5.0200 5.0200 5.0200 5.0200 5.0200 5.0201 5.0201 5.0201 5.0201 5.0201
Cur
rent
[A
] -100
-50
0
50
100Diode current I
D
Time [s] ×10-35.02 5.0200 5.0200 5.0200 5.0200 5.0200 5.0201 5.0201 5.0201 5.0201 5.0201
-100
0
100Mosfet current I
M
Fig. 3.6: Reverse recovery current effects in diode and MOSFET
3.2. Simulation 1 - ZVT SSAC 1 51
×10-35.02 5.0200 5.0200 5.0200 5.0200 5.0200 5.0201 5.0201 5.0201 5.0201 5.0201
×104
-2
0
2
4Power consumption - Mosfet
Time [s] ×10-35.02 5.0200 5.0200 5.0200 5.0200 5.0200 5.0201 5.0201 5.0201 5.0201 5.0201
Pow
er C
onsu
mpt
ion
[W]
-10000
-5000
0
5000Power consumption - Diode
Fig. 3.7: Zoom at turn-on power consumption
Time [s] ×10-35.05 5.0505 5.051 5.0515 5.052 5.0525 5.053 5.0535 5.054 5.0545 5.055
Pow
er C
onsu
mpt
ion
[W]
-5
0
5
10
15Power consumption - Diode ×10-3
5.05 5.0505 5.051 5.0515 5.052 5.0525 5.053 5.0535 5.054 5.0545 5.0550
2000
4000
6000Power consumption - Mosfet
Fig. 3.8: Zoom at turn-off power consumption
52 Chapter 3. Simulations
δ [/] Win [J] Wout [J] WMOS [J] WD [J] η [/]
0,1 0,0884 0,0775 0,0094 0,0015 0,8768
0,3 0,6902 0,6586 0,0281 0,0036 0,9541
0,5 1,8619 1,804 0,0529 0,0050 0,9689
0,7 3,6028 3,5122 0,0884 0,0025 0,9748
0,9 5,9072 5,7714 0,1387 -0,0029 0,9770
1 7,1409 7,0585 0,0828 2,46E-08 0,9885
Table 3.2: Losses in a Buck converter
The losses in the different components, together with the total efficiency η are summarized
in Table 3.2 for different values of δ. The different losses are always measured over 1 ms.
Notice the negative value for WD that is probably a consequence of numerical errors during
the simulation. The abbreviations have the following meaning:
Duty cycle δ
Input energy Win
Output energy Wout
MOSFET energy losses WMOS
Diode energy losses WD
Total efficiency of the converter η
3.2.3 Buck converter with ZVT
The next circuit that will be simulated is a Buck converter with soft switching cell. The
aim is to reduce the switching losses. The SPICE circuit is shown in Fig. 3.9. At first, the
parameters of the resonant network being Lr1, Lr2, Cr1 and Cr2 will be calculated using the
following equations:
nC =Cr2Cr1
(3.5)
3.2. Simulation 1 - ZVT SSAC 1 53
Fig. 3.9: Buck converter with soft switching cell
nL =Lr2Lr1
(3.6)
Lr1 =Z0
2πfr(3.7)
Cr1 =Lr1Z2
0
(3.8)
where Lr1 Resonant inductance 1
Lr2 Resonant inductance 2
Cr1 Resonant capacitance 1
Cr2 Resonant capacitance 2
nL Inductance ratio
nC Capacitance ratio
fr Resonant frequency
Z0 Characteristic impedance
54 Chapter 3. Simulations
Z0 40 Ω
fr 229.7 kHz
Lr1 30 µH
Lr2 10 µH
Cr1 16 nF
Cr2 2.2 nF
Ton,aux 2.176 µs
Tdelay,main 1.1 µs
Table 3.3: Resonant parameters
To easily obtain soft switching conditions, [2] proposes to use Z0 = 40 Ω and fr = fs/0.35.
These parameters can however be tweaked via simulation for our purposes. For a first try-
out, the values as proposed by [2] will be used. Also the timing between the two on-signals
of the converter are important. The auxiliary switch is started first and has a pulse-width
determined by the LC resonant parameters Lr1 = 30 µH and Cr1 = 16 nF. The resonant
frequency is then given by:
fr =1
2π√Lr1 · Cr1
= 229.7 kHz (3.9)
The auxiliary switch is only ON during one half of the resonant period. The pulse-width
of the aux. switch then becomes:
Ton,aux =1
2fr= 2.176 µs (3.10)
The on-signal of the main switch is chosen such that the aux. pulse is symmetrical about
the starting pulse of the main switch. The delay of the main switch with respect to the
aux. switch is then given by:
Tdelay,main =Ton,aux
2= 1.1 µs (3.11)
The above values are summarized in Table 3.3. They are plugged into the simulation
file and the results showing the power consumption are plotted in Fig. 3.10, Fig. 3.11,
Fig. 3.12, Fig. 3.13. By looking at the power consumption of the main MOSFET, one
can conclude that the purpose of soft switching has been achieved. Under hard switching
conditions (Fig. 3.5), the turn-on power consumption peaked to 36 kW and the turn-off to
4.4 kW. When the SSAC is applied (Fig. 3.10), the turn-on peak has totally disappeared
3.2. Simulation 1 - ZVT SSAC 1 55
Time [s] ×10-34 4.01 4.02 4.03 4.04 4.05 4.06 4.07 4.08 4.09 4.1
Pow
er [W
]
-300
-200
-100
0
100
200
300
400
500
Fig. 3.10: Main MOSFET power dissipation - 5 cycles
Time [s] ×10-34 4.002 4.004 4.006 4.008 4.01 4.012 4.014 4.016 4.018 4.02
Pow
er [W
]
-300
-200
-100
0
100
200
300
400
500
X: 0.004008Y: 36
X: 0.004013Y: 431.7
X: 0.004016Y: 0.1708
Fig. 3.11: Zoom on main MOSFET power dissipation - 1 cycle
while the turn-off peak is reduced to 0.45 kW. This is clearly a big improvement. However,
the major drawback is the increased number of components, that also add to the total
losses. As shown in Fig. 3.12, the switching losses of the auxiliary MOSFET are not
negligible. They have a peak of 2 kW. So the critical question is: Has the total efficiency
of the converter been improved or not? The efficiency of the regular Buck converter was
already calculated in the previous section. The same calculation has been done for the
Buck converter with soft switching cell and the results are summarized for different duty
ratios (δ) in Table 3.4. A graphical representation of the results is given in Fig. 3.14. It is
clear that, even with the extra losses in the soft switching cell, the total efficiency of the
converter rises. The biggest improvements are noticeable for lower duty ratios, whilst the
efficiencies converge for higher duty ratios.
56 Chapter 3. Simulations
Time [s] ×10-34 4.01 4.02 4.03 4.04 4.05 4.06 4.07 4.08 4.09 4.1
Pow
er [W
]
-1000
-500
0
500
1000
1500
2000
Fig. 3.12: Aux. MOSFET power dissipation - 5 cycles
Time [s] ×10-34 4.002 4.004 4.006 4.008 4.01 4.012 4.014 4.016 4.018 4.02
Pow
er [W
]
-1000
-500
0
500
1000
1500
2000
X: 0.004014Y: -0.07919
X: 0.004002Y: 1887
Fig. 3.13: Zoom on aux. MOSFET power dissipation - 1 cycle
δ [/] Win [J] Wout [J] WMOS,main [J] WMOS,aux [J] η [/]
0,1 0,2075 0,1944 0,0013 0,0074 0,9369
0,3 0,9284 0,9038 0,0042 0,0123 0,9735
0,5 2,2116 2,1682 0,0130 0,0204 0,9804
0,7 4,0463 3,9733 0,0310 0,0314 0,9819
0,9 6,4149 6,3004 0,0625 0,0432 0,9821
1 7,1423 7,0605 0,0808 0,0004 0,9885
Table 3.4: Losses in a Buck converter with soft switching cell
3.2. Simulation 1 - ZVT SSAC 1 57
Duty ratio δ [/]0,2 0,3 0,4 0,5 0,6 0,7 0,8 0,9 1
Eff
icie
ncy η
[/]
0.86
0.88
0.9
0.92
0.94
0.96
0.98
1
Buck
Buck with ZVT
Fig. 3.14: Comparison of the efficiency as a function of δ - Buck converter
3.2.4 Full leg with ZVT
The original field of application of the used soft switching cell are DC/DC converters (Buck,
Boost, . . . ). However, the intention of this thesis is to use the cell in an inverter (DC/AC).
Therefore, the soft switching cell needs to be applied for both the upper and lower tran-
sistor in a full leg configuration. The obtained circuit is shown in Fig. 3.15, where now
the Buck converter is made between the mid-point and the positive DC voltage such that
the circuit can be tested for the lower transistor. The results are similar compared to the
previous case and are summarized in Table 3.5 and Fig. 3.16. This is an important result
since it means that the auxiliary circuit can be placed next to a leg of an inverter to obtain
soft switching conditions.
Also the possibility to merge the two cells of Fig. 3.15 was investigated. This however
leads to a lower efficiency increase and even an efficiency decrease in some regions. This
possibility is therefor not further investigated.
58 Chapter 3. Simulations
Fig. 3.15: Spice circuit for a full leg
3.2. Simulation 1 - ZVT SSAC 1 59
Without ZVT With ZVT
δ [/] Win [J] Wout [J] η [/] Win [J] Wout [J] η [/]
0,1 0,1369 0,1062 0,7760 0,2383 0,2204 0,9249
0,3 0,7214 0,6779 0,9396 0,9553 0,9263 0,9695
0,5 1,8776 1,8147 0,9665 2,2331 2,1847 0,9783
0,7 3,6034 3,5125 0,9748 4,0647 3,9860 0,9806
0,9 5,8986 5,7656 0,9775 6,4154 6,2959 0,9814
1 7,1417 7,0585 0,9884 7,1426 7,0605 0,9885
Table 3.5: Comparison of the efficiencies for a full leg with and without ZVT
Duty ratio δ [/]0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Eff
icie
ncy η
[/]
0.75
0.8
0.85
0.9
0.95
1
Buck
Buck with ZVT
Fig. 3.16: Graphical representation of the efficiency comparison - Full leg
60 Chapter 3. Simulations
3.3 Simulation 2 - MPC SSAC
The second circuit that has been simulated is shown in Fig. 3.17. It is based on a topology
presented by [3] and is called a midpoint clamped (MPC) SSAC. As with the previous
solution, also this commutation cell is placed outside the main power path. A second
advantage is the applicability to inverters that use the capacitor midpoint, such as grid-
connected split DC bus inverters.
The design procedure to calculate the resonant elements is given in [3] and is used as a
starting point for optimization. The following parameters are defined:
Ipk is the current peak diverted from main switch to the auxiliary circuit
Ipk =E
2√
2Z(3.12)
Io,pk is the output current peak
Io,pk =
√2PoVo
(1 + ∆I) (3.13)
A parameter k is defined such that:
k =IpkIo,pk
; k ≥ 1 (3.14)
A practical value is k = 1, 1
E is the DC input voltage
Po is the output power (of one leg)
Vo is the RMS output voltage
∆I is the output current ripple
fs is the switching frequency
Z is the characteristic impedance
3.3. Simulation 2 - MPC SSAC 61
Fig. 3.17: Proposed soft switching commutation cell [3]
didt
is the rate of current change during turn-off of the main diodes:
di
dt=
Io,pkω√2 arcsin( 1
2k)
(3.15)
A practical value, suggested in [3], is 80 A/µs.
ω is the resonant tank frequency. It is determined via the didt
:
ω =didt
√2 arcsin( 1
2k)
Io,pk(3.16)
Lr and Cr are the resonant elements.
Lr =Z
ω(3.17)
Cr =1
Zω(3.18)
The above procedure has been implemented in a Matlab script that can be found in sec-
tion A.1. The input parameters are: E = 800 V, Po = 3 kW, Vo = 250 V, fs = 20 kHz,
fo = 50 Hz. The outcome is Lr = 5 µH and Cr = 30 nF.
The above result was used in simulation. However, the peak output current turned out to
be relatively high. Therefore, some other LC combinations that result in (approximately)
the same resonant frequency were tested. It turned out that the resonant current strongly
decreases in case of a higher characteristic impedance Z. This current should be as low as
possible in order to have low conduction losses and low circulating energy in the SSAC.
The results are given in Table 3.6.
62 Chapter 3. Simulations
Lr [µH] Cr [nF] Z[Ω] f [kHz] Ip [A]
5 30 12,9 401 50
8 20 20 398 40
33 16 45,4 219 30
12 12 31,6 419 30
18 8 47,4 419 18
Table 3.6: Peak current Ip in the MPC SSAC for different resonant elements
3.3.1 Buck converter
The first circuit that will be simulated is again a Buck converter. This means that the
load is not connected to the midpoint (see Fig. 3.17) but to the negative bar of the DC
bus. The circuit is shown in Fig. 3.18. The DC bus voltage is now chosen to be 600 V
because two identical power supplies are needed and the available EELAB power supplies
only go up to 300V. Also notice that two snubber circuits are added to the switches of
the auxiliary circuit. This was deemed necessary because some serious over-voltages were
observed during simulation. They dissipate the circulating energy that is already present
in the SSAC and thus do not introduce extra losses. The results are quite promising. In
Fig. 3.19, a strong efficiency increase is visible for lower duty ratios. For δ = 10%, an
improvement of more than 30 % is visible. The MPC SSAC is thus very effective since the
switching losses decrease significantly.
3.3.2 Buck converter with split DC bus
The next topology to be tested is again the buck converter with split DC bus as the behavior
of this converter is closest to the grid-connected inverter with a capacitor midpoint. The
results are shown in Fig. 3.20. The maximum output power is 2,5 kW for an output
current of 10 A. The DC bus voltage is 600 V. An efficiency improvement is visible over
the complete operating range. The improvement is however rather small, about 0,5%.
Only at δ = 55 % the hard switched efficiency is higher. This is because the circuit
operates in Discontinuous Conduction Mode (DCM). The turn-on then already occurs at
zero-current, including the MPC SSAC is thus not effective in this case. Fig. 3.21 and
Fig. 3.22 show the drain-source voltage, the drain current and the power dissipation of
3.3. Simulation 2 - MPC SSAC 63
Fig. 3.18: Spice simulation circuit for a Buck with MPC SSAC
Duty cycle0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9
Effi
cien
cy
0.5
0.55
0.6
0.65
0.7
0.75
0.8
0.85
0.9
0.95
1
Without ZVTWith ZVT
Fig. 3.19: Efficiency comparison - Buck converter with and without ZVT circuit
64 Chapter 3. Simulations
Duty cycle δ [/]0,6 0,65 0,7 0,75 0,8 0,85 0,9 0,95
Eff
icie
ncy η [
/]
0,92
0,94
0,96
0,98
1
Hard switchedWith MPC SSAC
Fig. 3.20: Efficiency comparison - Buck converter with split DC bus
the MOSFET in case of hard switching and when the MPC SSAC is applied at turn-
on. In hard switching, very high peaks in the power (up to 11 kW) are visible because
the voltage and current waveforms overlap. With the MPC SSAC, these peaks disappear
almost completely. Also the dv/dt and the current slope during the commutations is much
lower, which has a positive influence on the EMC. The MPC SSAC is thus very effective
to reduce the switching losses.
3.3. Simulation 2 - MPC SSAC 65
×10-53.5 4 4.5 5 5.5 6
Vol
tage
[V
]
0
500
×10-53.5 4 4.5 5 5.5 6
Cur
rent
[A
]
0
10
20
30
Time [s] ×10-53.5 4 4.5 5 5.5 6
Pow
er [
W]
0
5000
10000
Fig. 3.21: Power consumption of the switch in hard switching
×10-53.5 4 4.5 5 5.5 6
Vol
tage
[V
]
0200400600
×10-53.5 4 4.5 5 5.5 6
Cur
rent
[A
]
-505
1015
Time [s] ×10-53.5 4 4.5 5 5.5 6
Pow
er [
W]
-100
0
100
200
Fig. 3.22: Power consumption of the switch in soft switching
66 Chapter 3. Simulations
3.4 Simulation 3 - PRDCL
The third topology that has been simulated is a Parallel Resonant DC Link (PRDCL).
This resonant circuit is placed in between the DC voltage source and the inverter as shown
in Fig. 3.23. Different implementations exist but we have chosen for the one presented
in [4]. The SPICE simulation model of this circuit is shown in Fig. 3.24. The required
number of auxiliary components is low and the voltage stress of the auxiliary components
is limited to the DC bus voltage. This is an advantage over other possible circuits where
the voltage stress may be twice the DC bus voltage. A disadvantage is that the on-time
of the auxiliary circuit is dependent on the inverter current. Current feedback is thus
required. Notice that, for the simulation, the inverter legs are replaced by an RL circuit
that represents the current that is drawn.
As shown in Fig. 3.25, the voltage that is supplied to the inverter is drawn to zero during
some finite amounts of time. The length of the zero-voltage period can be controlled via the
auxiliary switches. During this period, the main IGBTs need to be turned on. Although
this may seem an attractive solution, several concerns may arise:
The required resonant current for this circuit is plotted in Fig. 3.26. The peak value
equals almost 35 A. This very high value obviously corresponds to a high power
loss and means that the auxiliary components need to be sized quite big which also
increases the total costs. This resonant current can never be really low because it
always needs to be bigger than the current going to the inverter. In the case of a 12
kW inverter (4 kW per phase) and a DC bus of 800 V, the DC input current equals
IDC = 12 kW800V
= 15 A.
The DC bus voltage drops for a large amount of time. The consequence is a decrease
in the fundamental output voltage and extra harmonics because of the notches.
The resonant circuit can be tuned such that the voltage decreases faster. This high
dV/dt may cause EMI.
The control must be adapted such that the IGBTs are all turned on in the same
period. In this case only one resonance is needed for the three phases together. This
means an increased control complexity.
3.4. Simulation 3 - PRDCL 67
Fig. 3.23: Parallel Resonant DC Link [4]
The output current must be measured to ensure an adequate turn-on time of the
device.
These drawbacks have made that this solution was not further elaborated.
68 Chapter 3. Simulations
Fig. 3.24: PRDCL simulation model in SPICE
Time [s] ×10-41.5 1.6 1.7 1.8 1.9 2 2.1 2.2 2.3 2.4 2.5
Vol
tage
[V
]
0
50
100
150
200
250
300
350
400
450
Fig. 3.25: Supplied inverter voltage
3.5. Conclusion 69
Time [s] ×10-41.5 1.6 1.7 1.8 1.9 2 2.1 2.2 2.3 2.4 2.5
Cur
rent
[A]
-5
0
5
10
15
20
25
30
35
Fig. 3.26: Current through the resonant inductor
3.5 Conclusion
The modeling and simulation of several resonant circuits was discussed throughout this
chapter. From the advantages and disadvantages that followed from the simulation output,
results that the first and second SSAC give the best results. In both cases an efficiency
improvement is possible. Therefore, they will be examined further on and an experimental
setup will be built to verify the simulations. The disadvantages of PRDCL circuit are too
important and it will not be further investigated.
70 Chapter 3. Simulations
Chapter 4
Experimental verification
This chapter describes the different measurements and results that have been carried out
in order to verify the simulation results of Chapter 3. A hard switched converter will be
compared to two soft switching alternatives. Also a hard and soft switched converter using
Si IGBTs will be compared to a hard switched converter that is build with SiC MOSFETs.
4.1 Setup
A test setup has been built in order to verify the simulation results. One inverter leg
and the two auxiliary circuits have been made via Altium Designer 14.3.11. The design
procedure of the different boards is described in Appendix B. Also the choice of the
different components is discussed in this appendix. A picture of the test set-up is shown
in Fig. 4.1.
71
72 Chapter 4. Experimental verification
Fig. 4.1: Test setup
4.2 Measurement equipment
4.2.1 Voltage probes
Since the applied voltages that will be used are higher than 50 V, special care must be
taken to measure the voltages. Both the time delay as the bandwidth of the probes is an
important factor because of the very high dv/dt, possible overshoots and oscillations that
are present in power electronic converters. Regular Tektronix voltage probes (P2220) are
used when the applied voltages are below 300 V. When the applied voltages are higher,
the better insulated P5120 will be used for signals up to 1000 V. Both probes have a
bandwidth of 200 MHz. Also the probes presented in [35] were tested since they have a
high bandwidth(10 MHz). However, the time delay introduced by this equipment was too
high. This means that the voltage waveform will have a certain phase shift with respect to
the original signal. Power measurements will thus not be accurate and therefor they are
not used.
4.2. Measurement equipment 73
4.2.2 Current probes
A precise current measurement is needed such that the power losses can be estimated
appropriately. Three types of current measurement were tested:
Current sense resistor - VISHAY LVR05 0.1 Ω
The current sense resistor is cheap and reliable. The current signal will also have
the same delay as the voltage signal if it is measured by the same voltage probe. Its
major drawback is the limited current that is allowed. For a regular 5 W resistor, the
maximum RMS current is only 7.07 A. Another drawback is that it is intrusive.
Current probe - Fluke 80I-110S
The current probe has the advantage of being non-intrusive but the drawback is
the limited bandwidth of only 100 kHz and the introduced delay as will be shown
further on. The consequence of the delay is that the current measurement needs to be
time-shifted during post-processing to correspond correctly with the voltage signal.
High frequency current transformer - EELAB design
This current transformer has the same advantages as the current probe but a higher
bandwidth. The design is based on an adapted version of the current transformer
that can be found in [36]. The device is able to measure both AC as DC current whilst
the previous design was only intended for AC current measurements. It saturates
above 10 A.
The obtained waveforms are shown in Fig. 4.2. The current sense resistor and the current
transformer clearly give the best results. The signal of the current probe is delayed with
approximately 1.5 µs, which is too much for our purposes. Also the diode reverse recovery
is not present in the current probe signal. This makes it an inappropriate candidate. The
current transformer is found to be most suitable and will be used during the rest of the
experiments.
4.2.3 Other equipment
The probes will be connected to an oscilloscope with storage capabilities. This way the
data can be saved in CSV format for further processing in Matlab. The used oscilloscope is
74 Chapter 4. Experimental verification
×10-50 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
Cur
rent
[A
]
-10
0
10Current sense resistor
×10-50 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
Cur
rent
[A
]
-505
1015
Current transformer
Time [s] ×10-50 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5C
urre
nt [
A]
-20246
Fluke current probe
Fig. 4.2: Comparison of different current measurement techniques
a Tektronix TPS2014 with isolated channels. Besides the equipment that is used to capture
the voltage and current waveforms, also the input and output power needs to be measured
with a high accuracy. This is done via true RMS digital multi-meters such as the Fluke 177
and the Fluke 179 in combination with shunt resistors (1mV/1A). The multi-meters are
first calibrated with a high precision tabletop multi-meter, namely the RIGOL DM3068.
4.3 Switching loss measurements
Measuring switching losses is not as straight forward as it seems to be. A variety of
standards [37] are followed by the different manufacturers. For comparative purposes,
the same method as the manufacturer needs to be chosen. In this case the measurement
method is given in the data sheet of our IGBT modules [5] and is shown in Fig. 4.3. The
turn-on losses are measured by integrating the power loss from 10% Ic to 5% Vce. The
turn-off losses are measured via integrating from 10% Vce to 5% Ic. This strategy has also
been followed in this thesis and is implemented in a Matlab script that can be found in
section A.4. This tool was developed such that an analytic comparison of the switching
losses with and without SSAC can be made. However, it will be shown later on that
the integration limits that are used in the definition of the switching losses switch sides.
4.4. Buck converter 75
(a) Turn on
(b) Turn off
Fig. 4.3: Definition of the switching losses [5]
At turn-on, the collector-emitter voltage decreases before the collector current rises. At
turn-off, the current decreases before the voltage increases. The behavior is thus totally
different when a SSAC is applied. This basically means that the integration would yield
a negative number. This is off course a worthless outcome. It means that the definition
of switching losses under hard switching is useless for soft switching. The definition or
suggestions for another standard falls beyond the scope of this thesis. This topic is treated
in more detail in [38] and [39]. The developed MATLAB script is however included since
it can be a valuable tool if switching losses need to be estimated under hard switching.
4.4 Buck converter
4.4.1 Hard switched
At first, the efficiency of a hard switched Buck converter is measured. The system is shown
in Fig. 4.4. The used components are described in Table 4.1. Notice that only the top
76 Chapter 4. Experimental verification
Fig. 4.4: Hard switched Buck converter - Measurement setup
Component Value
C1 - Electrolytic capacitor 1000 µF
C2, C3 - Film capacitor 1 µF
S1, S2 - IGBT IRG4PH40UD
R1, R2 - Shunt resistor 1 mΩ
Lf - Inductor 2.2 mH
C4 - Film capacitor 5 µF
Table 4.1: Used components for the measurement setup
IGBT is used for switching. The gate of the bottom IGBT is connected to the emitter
such that it is always off and thus works as a free-wheeling diode. The resistors R1 and
R2 are used for measuring the input and output current.
The results for different switching frequencies are shown in Fig. 4.5. The measured effi-
ciency shows very good correspondence with the results of the simulations of Chapter 3.
As expected, the efficiency drops in case of a higher switching frequency. This is explained
by the higher switching losses (Wsw ∼ fs). One can also see that the efficiency is the
lowest for low duty cycles because the relative importance of the switching losses is higher
compared to the transferred power. Also notice that the 40 kHz measurement stops at δ =
65% because the IGBTs failed during the experiment because of an excessive temperature
increase. The output power is shown in Fig. 4.6. A clear quadratic dependence is present
since Pout ∼ V 2out and Vout ∼ δ. In the following sections, the aim will be to increase the
efficiency by using soft switching auxiliary circuits (SSAC) that reduce the switching losses.
4.4. Buck converter 77
Duty ratio δ [/]0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9
Eff
icie
ncy η
[/]
82
84
86
88
90
92
94
96
98
100
10 kHz20 kHz40 kHz
Fig. 4.5: Hard switched Buck converter - Efficiency comparison
Duty ratio δ [/]0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9
Out
put p
ower
[W
]
0
500
1000
1500
2000
2500
300010 kHz20 kHz40 kHz
Fig. 4.6: Hard switched Buck converter - Output power
4.4.2 Soft switch auxiliary circuit 1
First, an experiment has been done with the midpoint clamped (MPC) SSAC, as described
in [3]. The measurement setup is shown in Fig. 4.7. This auxiliary circuit requires a split
78 Chapter 4. Experimental verification
Fig. 4.7: Soft switched Buck converter - Measurement setup
DC bus to operate. This has been achieved by using a second power source that stabilizes
the midpoint. The resistors R1, R2 and R3 are used to measure the input and output
currents. The circuit was first tested for a DC bus of 200 V to check the correct behavior
and to test different resonant parameters (Lr and Cr). The first set of parameters is Lr
= 22 µH (industrial VISHAY SMD shielded core) and Cr = 11 nF. It was however noticed
that the inductor became very hot and that the soft switching conditions were not met
anymore. It was concluded that an air coil would be more appropriate. Several inductance
values ( 1.3 µH, 2 µH and 8 µH) were fabricated and tested. Also multiple capacitors were
compared (Cr = 3.3 nF, 6.8 nF, 10 nF, 33 nF). The best results were obtained for the
combination: Lr = 2 µH and Cr = 3.3 nF. This combination was chosen because:
Zero voltage conditions are easily obtained
Current peak at turn-on is limited
Time delay of the main pulse is relatively constant
Now, the efficiency of a hard switched and a soft switched Buck converter, operating
at 20 kHz and 300 V is compared. The experimentally obtained results can be found in
Appendix E. Fig. 4.8 summarizes the results. The SSAC can be applied at turn-on, turn-
off or both. One can see that in all three cases a strong improvement is realized over the
complete range. The effect is most visible for lower duty cycles, e.g. an efficiency increase
4.4. Buck converter 79
Duty ratio δ0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9
Eff
icie
ncy η
0,84
0,88
0,92
0,96
1
Hard switchedSoft switched @ onSoft switched @ offSoft switched @ on + off
Fig. 4.8: Efficiency comparison - 20 kHz Buck converter
of about 6% for δ = 0,1. This is because the switching losses become more important when
the power output is lower (Pout ∼ δ2). Briefly, their relative importance is higher. A small
difference is visible when the SSAC is applied at turn-on or turn-off. The efficiency is higher
when the turn-on losses are reduced. This might not seem in accordance to intuition since,
generally spoken, the turn-off losses of IGBTs are higher due to the current tail. However,
the current tail depends on both the amplitude of the current as on the temperature of
the IGBT while the turn-on losses mainly depend on the diode reverse recovery, which is
rather constant.
Also the efficiency of a hard switched and a soft switched Buck converter, now operating
at 40 kHz and 300 V is investigated. The results are shown in Fig. 4.9. Notice that the
hard switched inverter experienced a thermal breakdown at δ = 70%, as already discussed
in subsection 4.4.1. By applying the SSAC, full range of operation is achieved and the
efficiency is increased. This means that the SSAC expands the application capabilities of
the converter while maintaining a high efficiency. This is an important conclusion since
also snubbers could have been used to extend this range. They however will never improve
the efficiency of the device.
80 Chapter 4. Experimental verification
Duty ratio δ0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9
Eff
icie
ncy η
0,82
0,86
0,9
0,94
0,98
BuckBuck ZVT
Fig. 4.9: Efficiency comparison - 40 kHz Buck converter
4.4.3 Soft switch auxiliary circuit 2
Not only the previous ’midpoint clamped’ SSAC had promising simulation results. It was
also proved that the SSAC shown in Fig. 4.10 is able to reduce the switching losses. An
impression of the PCB that was built can be found in Appendix D, Fig. D.6. The SSAC
was first tested for low voltages to see if it was able to provide the ZVT conditions. During
these tests, it was noticed that some serious over-voltages were present in the device.
The measured voltage across diode D3 is shown in Fig. 4.11. A voltage peak of 240 V is
present when the DC bus voltage is only 10 V. When the DC bus voltage increased to
50 V, sparks were observed that destroyed the transistor and other equipment. A possible
solution would be to add over-voltage snubbers. This option is not chosen because the
SSAC already contains a lot of components, especially when compared to the previous
SSAC. It was concluded that this circuit is not suitable for our purposes since it would
gain too much complexity. This possibility is therefor not further elaborated.
4.4. Buck converter 81
Fig. 4.10: Soft switching auxiliary circuit 2
Time [µs]0 1.0 2.0 3.0 4.0 5.0
Vol
ts [V
]
-50
0
50
100
150
200
250
Fig. 4.11: Voltage across diode D3
82 Chapter 4. Experimental verification
4.5 Buck converter with split DC bus
This section explains the results that were obtained by including the MPC SSAC to a Buck
converter with a split DC bus. This topology was theoretically discussed in Chapter 3.
The experimental setup for both hard and soft switching is shown in Fig. 4.12. Notice the
bleeder resistor R4 that is used to stabilize the midpoint of the DC bus. Better developed
systems for this purpose are described in [23]. They are more energy-efficient and can be
used for grid-connected inverters that use the midpoint of the DC bus. For this thesis,
the solution with a bleeder resistor is preferred because of the simplicity and because the
power losses that correspond to it can be easily measured. This is important since they
are needed to calculate the efficiency. R1, R2, R3 and R5 are used to measure the input
and output currents.
4.5.1 Comparison for a 400 V DC bus
At first, the circuit is tested for a DC bus of 400 V. The used IGBTs are the IRG4PH40UD,
the data sheet can be found in Appendix F. The results for switching frequencies of 20 and
30 kHz are shown in Fig. 4.13 and Fig. 4.14, respectively. Also a measurement at 40 kHz
was done but the converter failed in both hard as soft switching. This result is therefor not
included. The maximum output power of the converter is 1,7 kW, for a current of 9,5 A.
Especially at 20kHz, very high efficiency improvements are achieved in the lower current
region, for 0, 55 ≤ δ ≤ 0, 7. The average improvement in this region is roughly 8,6 %. For
30 kHz the improvement is rather constant over the complete operating area, it is in the
range of 1 to 3 %. This result is summarized in Table 4.2. It might seem counterintuitive
that the efficiency improvement at 30 kHz is lower compared to the 20 kHz case since the
switching losses are higher at 30 kHz. It should however be noted that the SSAC also
needs to be operated at 30 kHz. This means that the SSAC will consume more energy
since the conduction and switching losses of this device are not completely negligible. A
trade-off thus exists between the energy that is saved in the converter and the energy that
is consumed by the SSAC. As long as this balance is positive, the efficiency will increase
but the increase is not constant as the losses in the SSAC become more dominant.
4.5. Buck converter with split DC bus 83
(a) Hard switched
(b) Soft switched
Fig. 4.12: Buck converter with split DC bus - Measurement setup
84 Chapter 4. Experimental verification
δ [%] 55 60 65 70 75 80 85 90 95
20 kHz
ηhard 75,5 74,1 81,7 90,6 92,6 94,1 94,5 95,0 95,0
ηsoft 80,9 90,2 92,0 93,4 94,5 95,0 95,5 95,5 96,1
∆η 5,4 16,1 10,3 2,9 1,9 0,9 0,9 0,5 1,2
30 kHz
ηhard 74,0 85,9 89,6 91,8 92,7 93,9 94,0 93,9 93,7
ηsoft 76,9 88,6 92,0 94,0 94,9 94,8 95,1 95,1 95,1
∆η 2,8 2,7 2,4 2,2 2,1 0,9 1,1 1,3 1,4
Table 4.2: Efficiency comparison for 20 and 30 kHz at 400 V
Duty ratio δ [/]0,6 0,65 0,7 0,75 0,8 0,85 0,9 0,95
Effic
ienc
y η
[/]
0,7
0,8
0,9
1
Hard switched
Soft switched
Fig. 4.13: Efficiency comparison Buck split DC bus at 400V - 20 kHz
4.5. Buck converter with split DC bus 85
Duty cycle δ [/]0,6 0,65 0,7 0,75 0,8 0,85 0,9 0,95
Effic
ienc
y η
[/]
0,7
0,8
0,9
1
Hard switched
Soft switched
Fig. 4.14: Efficiency comparison Buck split DC bus at 400V - 30 kHz
4.5.2 Comparison for a 600 V DC bus
After the successful tests at 400 V, it was decided to increase the DC bus voltage level to
600V. The used IGBTs are IRGP30B120KD-E, the data sheet can be found in Appendix F.
Due to the higher voltage, the maximum output power of the converter increases to 2,5
kW for a current of 9,5 A. The measurement setup remains the same as before.
The outcome of the experiments is shown in Fig. 4.15 and Fig. 4.16 for switching frequencies
of 20 and 30 kHz, respectively. A summary of the efficiency increase is given in Table 4.3
Similar results as in the previous paragraph are found. The improvement is again the
highest in the region of 0,65 ≤ δ ≤ 0,75. The efficiency increase is then always more
than 2 %. If we take a closer look at the 30 kHz results, one can see that the hard
switching measurements stop at δ = 0, 8. This is because the IGBT experienced a thermal
breakdown. Both the efficiency as the operating range increases when the MPC SSAC is
included.
The effectiveness of the MPC SSAC has been proven in the previous paragraph. The SSAC
is activated just before the turn-on and turn-off of the main pulse. The timing of these
pulses with respect to the main pulse should be as constant as possible to achieve an easy
control. Table 4.4 summarizes the duty cycle and the phase shift of the turn-on and turn-
off auxiliary pulses in case of the Buck converter with split DC bus, operating at 600 V, 30
86 Chapter 4. Experimental verification
δ [%] 55 60 65 70 75 80 85 90 95
20 kHz
ηhard [%] 65,0 77,3 86,1 89,3 91,2 92,5 93,3 93,9 93,6
ηsoft [%] 67,1 82,6 89,1 92,2 93,5 94,5 95,0 95,2 94,2
∆η [%] 2,1 5,3 3,0 2,9 2,3 2,0 1,7 1,3 0,7
30 kHz
ηhard [%] 52,9 72,1 81,7 85,6 87,9 88,9 / / /
ηsoft [%] 63,0 80,8 87,3 90,5 92,0 92,9 93,7 94,7 92,7
∆η [%] 10,1 8,6 5,6 4,9 4,0 4,0 / / /
Table 4.3: Efficiency comparison for 20 and 30 kHz at 600 V
Duty ratio δ [/]0,6 0,65 0,7 0,75 0,8 0,85 0,9 0,95
Effic
ienc
y η
[/]
0,6
0,7
0,8
0,9
1
Hard switched
Soft switched
Fig. 4.15: Efficiency comparison Buck split DC bus at 600V - 20 kHz
4.5. Buck converter with split DC bus 87
Duty ratio δ [/]0,6 0,65 0,7 0,75 0,8 0,85 0,9 0,95
Effic
ienc
y η
[/]
0,5
0,6
0,7
0,8
0,9
1
Hard switched
Soft switched
Fig. 4.16: Efficiency comparison Buck split DC bus at 600V - 30 kHz
kHz. The duty ratio (δ) of the pulses is constant and equal to 1%. The time between the
main and the auxiliary pulse is denoted with Ton and Toff . From the table it is clear that
this advancement is not constant. The timing of the on-pulse clearly increases for higher
output power while the advancement of the off-pulse decreases a little. For this thesis, the
control of the pulse width and timing has been done manually. If the SSAC is implemented
in a stand-alone version, a closed loop feedback will be needed to check whether or not
the soft switching is achieved. This is again an increase in complexity. Further research is
needed to check if this timing can be held constant, e.g. when other resonant elements are
used. Another important remark is that the duty cycle of the SSAC is very low. As can
be seen from Table 4.4, it is only 1% for both the on or off pulse. The auxiliary circuit is
thus applicable over a very wide operating range.
88 Chapter 4. Experimental verification
δmain δon Ton δoff Toff Pout
[%] [%] [ns] [%] [ns] [W]
55 1 267 1 68,5 62
60 1 333 1 68,0 182
65 1 333 1 68,8 356
70 1 373 1 68,3 595
75 1 400 1 62,5 886
80 1 400 1 68,7 1241
85 1 427 1 69,5 1670
90 1 427 1 65,0 2161
95 1 427 1 65,8 2533
Table 4.4: Duty cycle and shift of the auxiliary pulses for the SSAC
4.5.3 Waveform analysis
This section discusses the different voltage and current waveforms that occur at turn-on
and turn-off. They are plotted in Fig. 4.17. Notice that the timescale is in microseconds.
The DC bus voltage is 600 V and the output current is approximately 6 A. The behavior
of the circuit is clearly very different when the MPC SSAC is active.
Turn-on
A strong overlap between the collector-emitter voltage and the collector current is
present in hard-switching. The switching losses are thus very high. There is also a
big current peak due to diode reverse recovery. Also notice the very high dv/dt and
di/dt. When the MPC SSAC is active, soft switching conditions are clearly present.
The SSAC is activated just before the main pulse. This makes the voltage drop in
a controlled way. The dv/dt is much lower, which is beneficial for the EMC. The
current increases when the voltage is already zero. This means that the turn-on
switching loss is practically zero. Also the current peak is much lower.
Turn-off
In hard switching the IGBT current tail is visible during the turn-off time. It lasts for
approximately 300 ns. Since the voltage increases much faster, the turn-off switching
loss is very high. By applying the MPC SSAC, the collector current strongly decreases
before the collector-emitter voltage rises. Small oscillations are however still visible.
4.6. Silicon Carbide 89
×10-60 0.2 0.4 0.6 0.8 1
Vol
ts [
V]
0
500
Hard turn-on
Cur
rent
[A
]
0
20
×10-60.5 1 1.5 2 2.5
Vol
ts [
V]
0
500
Soft turn-on
Cur
rent
[A
]
0
10
20
Time [s] ×10-60 0.2 0.4 0.6 0.8 1
Vol
ts [
V]
0
500
Hard turn-off
Cur
rent
[A
]0
5
Time [s] ×10-60 0.5 1 1.5 2
Vol
ts [
V]
0
500
Soft turn-off
Cur
rent
[A
]
0
5
Fig. 4.17: Voltage and current waveforms
This means that the turn-off loss is not completely zero but it is certainly strongly
decreased.
4.6 Silicon Carbide
This section is dedicated to the results that were obtained with Silicon Carbide (SiC)
components. A SiC Power MOSFET was selected and the results will be compared to the
hard switching and soft switching measurements with IGBTs of the previous sections. The
selected component is a CREE C2M0080120D. The most relevant part of the data sheet
can be found in Appendix F. Some interesting parameters are summarized in Table 4.5.
Notice the combination of a very high drain-source breakdown voltage (VDS,BR) with a
very low on-state resistance (RDS,on). The RDS,on would be much higher when a regular
Silicon MOSFET would have been used.
The SiC MOSFET is immediately applied in a Buck converter with split DC bus, as this is
of highest interest for this thesis. The same PCB and drivers are used. The only difference
is the used DC-DC converter that is needed to supply the isolated driving voltages. SiC
90 Chapter 4. Experimental verification
Model VDS RDS,on ID,100
C2M0080120D 1200 V 80 mΩ 20 A
Table 4.5: Parameters of the SiC MOSFET
Duty ratio δ [/]0,6 0,65 0,7 0,75 0,8 0,85 0,9 0,95
Effic
ienc
y η
[/]
0,6
0,7
0,8
0,9
1
Hard switched - normal IGBT
Soft switched - normal IGBT
Hard switched - SiC MOSFET
Fig. 4.18: Efficiency comparison with SiC components - 20 kHz
components need higher driving voltages, as already mentioned in Chapter 2. The SiC
MOSFET is supplied with +20/-5 V.
Measurements at 20 and 40 kHz are done and the results are plotted in Fig. 4.18 and
Fig. 4.19. At 20 kHz, a very clear improvement is visible. The SiC components (hard
switched!) perform much better than the hard or soft switched IGBTs. The efficiency
improvement ranges from 3 to 18 % when compared with hard switched IGBTs and it
ranges from 2 - 16 % when compared to soft switched IGBTs. This is a clear benefit as no
SSAC is needed to obtain this very high efficiency under hard switching. Fig. 4.19 shows
the efficiency at a switching frequency of 40 kHz. Again a very high efficiency is obtained
and the component is able to work in the entire operating area. This was not the case for
the soft switched IGBT that experienced a thermal breakdown at δ = 0, 75. When a high
efficiency in combination with a high switching frequency is required, SiC components are
clearly the best choice.
4.7. Conclusions 91
Duty ratio δ [/]0,6 0,65 0,7 0,75 0,8 0,85 0,9
Effic
ienc
y η
[/]
0,6
0,7
0,8
0,9
1
Hard switched - SiC MOSFET
Soft switched - normal IGBT
Fig. 4.19: Efficiency comparison with SiC components - 40 kHz
4.7 Conclusions
Two different SSAC were developed and tested. The first ZVT SSAC dealt with very high
over voltages. A possible solution is to add a snubber circuit. This was not done since
the SSAC already contained a high amount of components. This possibility was thus not
further developed.
The MPC SSAC showed its effectiveness on a Buck converter and was thus implemented
on a Buck converter with split DC bus. Experiments were done for a DC bus of 300, 400
and 600V and switching frequencies of 10, 20, 30 and 40 kHz. The SSAC is best perform-
ing at higher switching frequencies, when the switching losses are dominant. In this case,
an efficiency increase of about 10% is possible for low duty ratios, but the efficiency also
increases slightly for high output powers. This is clearly an advantage since the SSAC will
be applied to grid-connected inverters. These inverters are commonly connected with PV
panels, which means that they are only operating at full power when the solar irradiation
is at its maximum, at noon. During an important part of their lifetime, they will operate
below their nominal power. The efficiency is usually lower in this case, as was discussed
in Chapter 1. A possibility is thus to activate the MPC SSAC only in a certain operating
range, where the efficiency of an inverter is usually rather low. In case of high switching
frequencies, regular IGBTs usually experience a thermal breakdown because of the high
92 Chapter 4. Experimental verification
switching losses. This is not the case when the MPC SSAC is used. It increases the ef-
ficiency and prolongs the operating range. It has also been shown that the SSAC is only
active during a limited amount of time (δ = 1%). This means that it is applicable over a
wide range. The biggest concern is the shift of the pulse, which is not constant but varies
with the output power.
Also the recently commercially available SiC components were compared with the proposed
soft switching solutions. These components are rather expensive but it has been shown
that they perform much better. The efficiency of the SiC MOSFET in hard switching is
higher than the Si IGBT under soft switching. The additional cost can thus be justified
since no SSAC is needed. This solution is thus clearly more attractive and it is expected
that the future of power electronics will be dominated by these components.
Chapter 5
Conclusion
This chapter intends to give an overview of the work that has been done and the results
that have been achieved with the experimental setup. Also the possibilities for future
research are discussed.
The efficiency of a grid-connected inverter is rather high. Efficiencies above 90% are gen-
erally obtained. Only at low output power, the efficiency is rather low. Three types of
losses are present at the level of the switches: switching losses, conduction losses and driver
losses. It has been shown that the switching losses play a dominant role when IGBTs are
applied, because of the specific current tailing problem. An efficiency increase would thus
be possible if the switched losses can be omitted or reduced. This can be achieved by
using soft switching auxiliary circuits (SSAC). Different SSAC are proposed in literature.
A careful selection was made and an overview of their working principle and their topology
was presented. Three of these SSAC were simulated via PSpice to further investigate the
typical behavior of the circuit. Two of them were deemed suitable and an experimental
setup with a regular Buck converter and a Buck converter with split DC bus was realized.
Measurements have shown that the first SSAC was not an appropriate candidate since it
suffered from severe over-voltages. The second topology, being the MPC SSAC, was shown
to be very effective. The comparison of voltage and current waveforms under hard and soft
switching showed that the switching losses are virtually eliminated. The efficiency increase
that corresponds to it was in the range of 2-10% for switching frequencies of 20 and 30 kHz
and a DC bus voltage of 400 and 600 V. The efficiency over the complete operating area
improves but the effect is usually more visible in the lower power region. A possibility is
93
94 Chapter 5. Conclusion
thus to apply the SSAC only in this region where the SSAC is most effective.
Another alternative, besides the use of SSAC, is changing the Si IGBTs to the newer SiC
MOSFETs. SiC is a wide-bandgap semiconductor material and has superior properties
compared to Si. Both conduction and switching losses are remarkably lower. A converter
that uses SiC MOSFETs was compared to a soft switching Si IGBT converter. Measure-
ments for a DC bus of 600 V at a switching frequency of 20 and 40 kHz were carried
out. The efficiency of this device under hard switching is significantly higher than the soft
switched alternative. Switching frequencies of 40 kHz are proven to be possible without
any auxiliary components. A growing interest in this semiconductor material is noticed
in literature. It is therefore expected that SiC and GaN will shape the future of power
electronics as the potential for improvement is much higher compared to regular Si.
Future work
The MPC SSAC was tested for a regular Buck converter and a Buck converter with split
DC bus. It is thus expected that the SSAC will also perform correctly when it is applied
in a grid-connected inverter. There was unfortunately not enough time left at the end of
the research to build and troubleshoot an inverter PCB. Also an appropriate and flexible
control algorithm needs to be written in order to carry out enough tests.
Appendix A
Matlab scripts
A.1 Calculation resonant parameters MPC SSAC
1 %--------------------------------------------------------------
2 %Matlab script to determine the resonant elements of a MPC SSAC
3 %Simon Ravyts and Dimitar Bozalakov
4 %April 2016
5 %---------------------------------------------------------------
6
7 E = 800; %DC bus voltage
8 V o = 250; %RMS output voltage
9 P o = 3000; %Output power
10 Io pk max = 13.7; %Output peak current
11 Io pk min = 10.96;
12 delta I = Io pk max-Io pk min;
13 k = 1.095;
14 f s = 20000; %Switching frequency
15 dIdt = 80/1E-6; %Rate of change of the current
16
17 Io pk = sqrt(2)*P o*(1+delta I/Io pk max)/V o;
18 Z = (V o*E)/(4*k*P o*(1+delta I/Io pk max)); % Characteristic Impedance
19 I pk = E/(2*sqrt(2)*Z); %Peak current diverted from the main switch
20
21 omega = (dIdt *sqrt(2)*asin(1/(2*k)))/Io pk; %Angular frequency
95
96 Appendix A. Matlab scripts
22
23 Lr = Z/omega; %Henry
24 Cr = 1/(Z*omega); %Farad
A.2 Power losses - Buck converter
1 %-------------------------------------------------------------------------
2 %Calculation of the losses in case of a 4kW Buck converter with IGBTs
3 %Author: Simon Ravyts
4 %Date: March 2016
5 %-------------------------------------------------------------------------
6 clc
7 clear all
8
9 %% Input parameters converter
10 %--------------------------
11
12 Vin = 300; %volts
13 Vout = 150;
14 delta = Vout/Vin; %duty ratio t on/T
15 f s = 40000; %Switching frequency Hz
16 Pout = 1650; %Output power in watt
17 Rload = Voutˆ2/Pout; %Load resistance ohm
18
19 %% Component electrical properties - Assume linear approximation
20 %-------------------------------------------------------------
21 %Vce0, Rce, Vd0 and Rd have beel found via the figures in the data sheet
22
23 %IGBT
24 E sw tot = 7.04E-3; %Total switching loss, needs to be rescaled
25 Vce0 = 1.1946; % Volts, Vce = Vce0 + Rce * Ic
26 Rce = 0.0663;
27
28 %Diode
29 Vd0 = 1.59; % Volts, Vd = Vd0 + Rd * Id
30 Rd = 0.175;
31
32
33 %% Thermal properties casing/component
34 %------------------------------------
A.2. Power losses - Buck converter 97
35
36 P d maxi = 65; % Maximum power dissipation in watts at Tc = 100C
37 R JC T = 0.77; % C/W Junction-Casing Transistor
38 R JC D = 1.7; % C/W Junction-Casing Diode
39 R CS = 0.24; % C/W Case to Sink, greased surface
40 R SA = 0.29; % C/W
41
42 T J maxi = 100; %Maximum junction temperature in C
43 %T C = 100; %Casing temperature in C
44
45 %% Average and RMS currents
46 %--------------------------
47
48 %Assume that the output current is DC such that the ripple can be neglected
49 %The current passes through the transistor (T) during delta, through the
50 %diode(D) during (1-delta)
51
52 I out = Vout/ Rload;
53 I T avg = delta * I out;
54 I T rms = sqrt(delta) * I out;
55 I D avg = (1-delta)*I out;
56 I D rms = sqrt(1-delta) * I out;
57
58 %% Losses
59 %--------
60
61 E sw new = E sw tot * (I T rms/21) * (Vin/800); %Rescaled switching loss
62
63 P T cond = Rce * I T rmsˆ2 + Vce0 * I T avg; %Conduction loss
64 P T sw = E sw new .* f s; %Switching loss
65 P T total = P T cond + P T sw;
66
67 P D cond = Rd * I D rmsˆ2 + Vd0 * I D avg;
68
69 P tot = P T cond + P T sw + P D cond;
70
71 P out = Rload * I outˆ2;
72
73 Efficiency = P out./(P out + P tot);
74
75
76 %Heatsink - Temperature rise
77 %---------------------------
98 Appendix A. Matlab scripts
78
79 T A = 60 %Ambient temperature in C
80
81 %Transistor temperature increase
82 T S T = T A + R SA*P T total %Sink temperature transistor
83 T C T = T S T + R CS*P T total %Casing temperature transistor
84 T J T = T C T + R JC T*P T total %Junction temperature transistor
A.3 Power losses - Inverter
1 %-------------------------------------------------------------------------
2 %Calculation of the losses in case of a 4kW inverter using IGBTs
3 %Based on SEMIKRON Application Note AN-8005
4 %Author: Simon Ravyts
5 %Date: April 2016
6 %-------------------------------------------------------------------------
7 clc
8 clear all
9
10 %% Input parameters inverter
11 %--------------------------
12
13 Vin = 800; %DC bus voltage
14 Vout = 230; %Output RMS voltage
15 Pout = 4000; %Output power in watt
16 cos phi = 1; %Cos phi of the load
17 Iout RMS = Pout/(sqrt(3)*Vout*cos phi); %Output RMS current
18 Iout P = sqrt(2)*Iout RMS; %Peak output current
19 f s = 20000; %Switching frequency Hz
20 m = 0.7; %Modulation coeffcient
21
22 %% Component electrical properties - Assume linear approximation
23 %-------------------------------------------------------------
24 %Vce0, Rce, Vd0 and Rd have been found via the figures in the data sheet
25
26 %IGBT
27 E sw tot = 7.04E-3; %Total switching loss, needs to be rescaled
28 Vce0 = 1.1946; % Volts, Vce = Vce0 + Rce * Ic
29 Rce = 0.0663;
30
A.3. Power losses - Inverter 99
31 %Diode
32 Vd0 = 1.59; % Volts, Vd = Vd0 + Rd * Id
33 Rd = 0.175;
34
35
36 %% Thermal properties casing/component
37 %------------------------------------
38
39 P d maxi = 65; % Maximum power dissipation in watts at Tc = 100C
40 R JC T = 0.77; % C/W Junction-Casing Transistor
41 R JC D = 1.7; % C/W Junction-Casing Diode
42 R CS = 0.24; % C/W Case to Sink, greased surface
43 R SA = 0.29; % C/W
44
45 T J maxi = 100; %Maximum junction temperature in C
46 %T C = 100; %Casing temperature in C
47
48 %% Losses (of one IGBT)
49 %----------------------
50
51 %Switching losses
52 E sw new = E sw tot * (Iout RMS/21) * (Vin/800);
53 P SW IGBT = E sw new .* f s;
54
55 %Conduction losses
56 P C DIODE = 1/2 * (Vd0*Iout P/pi + Rd/4 * Iout Pˆ2) - m * cos phi * (Vd0 ...
* Iout P/8 + Rd * Iout Pˆ2 / (3*pi));
57 P C IGBT = 1/2 * (Vce0 * Iout P/pi + Rce/4*Iout Pˆ2) + m * cos phi ...
*(Vce0*Iout P/8 + Rce*Iout Pˆ2/(3*pi));
58
59 P tot = P SW IGBT + P C DIODE + P C IGBT;
60
61 %Temperature increase
62 %---------------------------
63
64 T A = 60 %Ambient temperature in C
65
66 T S = T A + R SA*P tot %Sink temperature
67 T C = T S + R CS*P tot %Casing temperature
68 T J = T C + R JC T*P tot %Junction temperature
100 Appendix A. Matlab scripts
A.4 Determination of switching losses
1 %---------------------------------------------------------------------
2 %Script to determine the switching losses of an IGBT, based on the same
3 %technique as used in the datasheet
4 %Author: Simon Ravyts
5 %Date: April 2016
6 %----------------------------------------------------------------------
7
8 clc
9 clear all
10
11 %First read in the CSV files of the ossciloscope using a function provided
12 %by Tektronix.
13 %CH1 <-> Vce Collector-Emitter voltage
14 %CH2 <-> Ic Collector current
15 %CH3 <-> Vge Gate-Emitter voltage (=driving voltage)
16
17 data Vce = read tektronix csv('F0050CH1.CSV');
18 data Ic = read tektronix csv('F0050CH2.CSV');
19 data Vge = read tektronix csv('F0050CH3.CSV');
20
21 %Exctract the useful data from the CSV files(time, voltages, currents)
22
23 t = data Vce.time;
24 %The time instances are symmetrical (t=0 in the middle), such as displayed
25 %on an oscilloscope. We want to start at t=0
26 t = t - min(t);
27
28 Vce = data Vce.values;
29 Ic = data Ic.values / 100; %Because the settings for the current are not ...
correct
30 Vdr = data Vge.values*5; %Since the home-made passive probe has an ...
attenuation of X50
31
32 Ts = data Vce.Sample Interval; %Sample period
33 Ns = data Vce.Record Length; %Number of samples - usually 2500
34
35 %% Display the values that are read in one figure using subplots
36
A.4. Determination of switching losses 101
37 figure
38 subplot(3,1,1)
39 plot(t,Vdr);
40 ylabel('Volts [V]')
41 title('Gate-Emitter voltage - Driving signal')
42 grid on
43 subplot(3,1,2)
44 plot(t, Vce );
45 title('Collector-Emitter Voltage')
46 ylabel('Volts [V]')
47 grid on
48 subplot(3,1,3)
49 plot(t, Ic);
50 xlabel('Time [s]')
51 ylabel('Current [A]')
52 title('Collector current')
53 grid on
54
55 %% Determine the instantaneous power dissipation
56 P = Vce.*Ic;
57 figure
58 plot(t,P)
59 xlabel('Time [s]')
60 ylabel('Power [VA]')
61 title('IGBT power consumption')
62 grid on
63
64 %If needed, the result can be filtered using median or average filtering
65 %P filt = medfilt1(P,10);
66 %figure
67 %plot(t, P filt)
68
69 %% Split up the results
70 %This part is only necessary if the scope captures one entire waveform. In
71 %that case, the first part of the measurement is used for the turn-on and
72 %the second part for the turn-off. However, it's better to leave this part
73 %out and capture only one of the two events
74
75 t 1 = t(1:1250);
76 t 2 = t(1251:2500);
77
78 Ic 1 = Ic(1:1250);
79 Ic 2 = Ic(1251:2500);
102 Appendix A. Matlab scripts
80
81 Vce 1 = Vce(1:1250);
82 Vce 2 = Vce(1251:2500);
83
84 %% Determine the integration limits for 'E off', the turn-off energy
85 %The procedure is based on how the switching losses are specified in the ...
data sheet
86 %E off -> Integrate from 10%Vce to 5%Ic
87 %E on -> Integrate from 10%Ic to 5%Vce
88
89 %Determine 100%Vce using the final part of the values (last 200 entries)
90 Vce 100p = mean(Vce 2(1050:1250));
91 Vce 10p = 0.1 * Vce 100p;
92
93 %Make an array 'A' that has a constant value, equal to 10%Vce
94 A = ones(1250,1);
95 A(1:1250) = Vce 10p;
96
97 %Use the function 'curveintersect' to find the time values corresponding to
98 %the first integration limit(=int start). The int start corresponds to the
99 %place in the array where the value can be found
100 [X1,Y1] = curveintersect(t 2,Vce 2,t 2,A);
101 int start 1 = round(X1(1)/ Ts - 1250);
102
103 %Plot the functions for visual inspection of the intersection points that
104 %are determined (= red dots)
105 figure
106 plot(t 2,Vce 2,'k',t 2,A,'b',X1,Y1,'ro')
107 xlabel('Time [s]')
108 ylabel('Volts [V]')
109 title('E-off: Intersection points (red dots) to find "int-start" ')
110 grid on
111
112 %Now the above procedure is repeated to determine the 2nd intersection
113 %point/ integration limit. The only difference is that there are usually
114 %multiple intersections due to the oscillations in the current so we can
115 %choose in fact between one of these values. A good choice might be
116 %comparing the first and last intersection.
117 Ic 100p = mean(Ic 2(1:200));
118 Ic 5p = 0.05*Ic 100p;
119 B = ones(1250,1);
120 B(1:1250) = Ic 5p;
121 [X2,Y2] = curveintersect(t 2,Ic 2,t 2,B);
A.4. Determination of switching losses 103
122 int stop 1 = round(X2(end)/Ts - 1250);
123 figure
124 plot(t 2,Ic 2,'k',t 2,B,'b',X2,Y2,'ro')
125 xlabel('Time [s]')
126 ylabel('Current [A]')
127 title('E-off: Intersection points (red dots) to find "int-stop" ')
128 grid on
129
130 %Both integration limits are now determined. The 'trapz' function can be
131 %used for trapezoidal integration. However, 'trapz' works for the entire
132 %array so we need to extract the interesting part of the array first.
133 t off = t 2(int start 1:int stop 1);
134 P off = Vce 2(int start 1:int stop 1).*Ic 2(int start 1:int stop 1);
135 W off = Ts*trapz(P off);
136
137 figure
138 plot(t off,P off,'r',t off,W off,'o')
139 xlabel('Time [s]')
140 ylabel('Power [VA]')
141 title('Turn off power loss')
142 grid on
143
144 %% The same method will be used to determine E on, the turn on energy
145 %E on -> Integrate from 10%Ic to 5%Vce
146 %Note that sometimes the same names are used!
147
148 %int start
149 Ic 100p = mean(Ic 1(1050:1250));
150 Ic 10p = 0.1*Ic 100p;
151 C = ones(1250,1);
152 C(1:1250) = Ic 10p;
153 [X3,Y3] = curveintersect(t 1,Ic 1,t 1,C);
154 int start 2 = round(X3(1)/Ts);
155 figure
156 plot(t 1,Ic 1,'k',t 1,C,'b',X3,Y3,'ro')
157 xlabel('Time [s]')
158 ylabel('Current [A]')
159 title('E-on: Intersection points (red dots) to find "int-start" ')
160 grid on
161
162 %int stop
163 Vce 100p = mean(Vce 1(1:200));
164 Vce 5p = 0.05 * Vce 100p;
104 Appendix A. Matlab scripts
165 D = ones(1250,1);
166 D(1:1250) = Vce 5p;
167 [X4,Y4] = curveintersect(t 1,Vce 1,t 1,D);
168 int stop 2 = round(X4(1)/ Ts);
169 figure
170 plot(t 1,Vce 1,'k',t 1,D,'b',X4,Y4,'ro')
171 xlabel('Time [s]')
172 ylabel('Volts [V]')
173 title('E-on: Intersection points (red dots) to find "int-stop" ')
174 grid on
175
176 t on = t 1(int start 2:int stop 2);
177 P on = Vce 1(int start 2:int stop 2).*Ic 1(int start 2:int stop 2);
178 W on = Ts*trapz(P on);
179 figure
180 plot(t on,P on,'r',t on,W on,'o')
181 xlabel('Time [s]')
182 ylabel('Power [VA]')
183 title('Turn on power loss')
184 grid on
185
186
187 %% Comparison
188 %The obtained results can be compared to the data sheet values if they are
189 %corrected via linear interpolation
190
191 E on ds = 1.80E-3;
192 E off ds = 1.93E-3;
193 Vcc ds = 800;
194 Ic ds = 21;
195
196 E on ds conv = E on ds * (Vce 100p/Vcc ds) * (Ic 100p/Ic ds);
197 E off ds conv = E off ds * (Vce 100p/Vcc ds) * (Ic 100p/Ic ds);
198
199 %The comparison should lead to values close to 1
200 comp on = W on/E on ds conv;
201 comp off = W off/E off ds conv;
A.5 Calculation of the gate resistance
A.5. Calculation of the gate resistance 105
1 %-------------------------------------------------------------------------
2 %Matlab script to calculate the required value of R gate, based on the
3 %Avago ACPL-337J data sheet
4 %Author: Simon Ravyts
5 %Date: March 2016
6 %-------------------------------------------------------------------------
7
8 I max = 4; %Ampere
9 V on = 15; %Volt
10 V off = 8.7; %Volt
11 R DS OH min = 0.5; %ohm, min resistance of the top (high side) transistor
12 R DS OL min = 0.2; %ohm, min resistance of the bot (low side) transistor
13
14 Rg min1 = V on/I max - R DS OH min;
15 Rg min2 = V off/I max - R DS OL min;
16
17 Rg = max(Rg min1,Rg min2);
18 %The gate resistance must be higher than the highest of both values.
19 %However, also the total power dissipation of the IC needs to be checked.
20
21 %Led power dissipation
22 duty cycle = 0.9; % Max duty cycle of the switches
23 I f = 16E-3; %Amps, max led current
24 V f = 1.95; %volts, max voltage drop over led
25 P E = duty cycle * I f * V f ;
26
27 %Input IC power dissipation
28 I cc1 = 6E-3; %Max input current
29 V cc1 = 5.5; %Volts, recommended max input voltage
30 P IN = I cc1*V cc1;
31
32 %Output IC power dissipation
33 V cc2 = 15;
34 V ee2 = -8.7;
35 I cc2 = 7.5E-3;
36 Qg = 130E-9; % From IGBT datasheet
37 f s = 50E3; % Switching frequency
38 R DS OH max = 4.5;
39 R DS OL max = 3.6;
40 P O = I cc2*(V cc2 - V ee2) + V cc2*Qg*f s*R DS OH max/(R DS OH max + ...
Rg)/2 + V cc2*Qg*f s*R DS OL max/(R DS OL max + Rg)/2;
41
42 P tot = P E + P IN + P O;
106 Appendix A. Matlab scripts
43
44 disp('If P tot is lower than 600mW, the selected Rg is appropriate')
Appendix B
Design considerations
This chapter will handle the basics of designing an IGBT or MOSFET inverter. At first,
the theoretical background and practical implementation of gate drivers and protection
circuits will be treated. Thereafter, an overview of the used components will be given,
together with their relevant features. This is important to understand the next part, that
deals with the lay-out of the circuit.
B.1 Fundamentals of power electronics design
B.1.1 Gate drive circuits
Gate drive circuits have already been briefly discussed in section 2.2 but we will go into
some more detail in this chapter.
The drive circuit is considered as the interface between the control circuit and the power
switch itself. Its main purpose is to turn the device ON and OFF. Although this may seem
simple, a lot of constraints regarding costs, safety and performance need to be taken into
account [12] :
Price A complex design with additional features and thus more components will be
107
108 Appendix B. Design considerations
more costly while a low-cost solution is usually preferred
Responsiveness The component passes through the active region during switching. In
this zone, the power dissipation is very high. So it is important that the driver is able
to switch the device ON and OFF very rapidly to limit the switching losses. Therefore,
the power and even more the peak current of the drivers is a very important aspect.
This may seem odd since MOSFETs and IGBTs are essentially voltage controlled. But
at turn-ON, a current is needed to charge the gate capacitance such that the component
starts conducting.
Bipolar output Drivers with bipolar output supply the power switch with a positive
voltage to turn it on and with a negative voltage to turn it off. The use of such a driver
is usually preferred over a unipolar output (switches between a positive voltage and
GND) because it decreases the turn-OFF time of the device. Another advantage is that
the circuit is less sensitive to switching transients caused by other components. These
transients may lead to unwanted ON ↔ OFF oscillations that cause additional losses.
Blanking time A blanking time is usually needed to prevent shoot-through of the leg.
This may happen during the small time interval when one of the switches turns OFF
while the other one turns ON.
Emitter inductance There is always a small amount of emitter inductance present in
the circuit. At first, the device package contains some inductance. This is usually rather
limited and can be decreased when the manufacturer foresees two leads for the emitter.
One for the main power path, the other for the driving signal. The second cause of
emitter inductance are the PCB tracks between the driver and the component. They
should be kept as short as possible, which means that the driver needs to be placed as
close as possible to the switch.
Gate resistance This is the resistor that needs to be applied between the gate of the
IGBT or MOSFET and the driving circuit. A low value is preferable for fast response
and low power losses. However, if the gate resistance is too low, the EMI performance
usually deteriorates because of the high di/dt and dv/dt and possible ringing. An
indication of the required value is normally given in the data sheet. This value usually
needs to be tweaked for optimal performance since the gate resistor Rg together with
the gate-emitter capacitance CGE and the emitter inductance Le form an RLC circuit
that may start ringing when a voltage from the driver is applied. Rg should be chosen
B.1. Fundamentals of power electronics design 109
Rg Rg ton toff Eon Eoff
Turn-on peak current dv/dt di/dt
EMI noise
Table B.1: Influence of gate resistance on performance [7]
such that enough damping is present in the circuit. Table B.1 summarizes the most
important tendencies when the resistance is in- or decreased [7].
Electrical isolation The GND of the control circuit usually differs from the GND of the
driver. Each driver has a separate ground that needs to be connected with the emitter of
the IGBT since the applied driving signals need to be referred to this potential and not
to the common GND. This isolation is usually achieved by transformers, optocouplers
or fiber optics.
Over-current protection Over-currents may arise when the load is too high or when
a short circuit is present. The current can be measured using an external sensor but
usually a more elegant method is preferred. The collector-emitter voltage VCE is mea-
sured and compared with the nominal ON-state voltage. When the current increases,
also VCE will increase. If the voltage exceeds a certain, user-defined limit, the driver
stops working and gives an error. This method is usually referred to as desaturation
protection.
Under-voltage protection In power electronics, the MOSFETs or IGBTs are used in
switch mode. This means that the component is completely ON or completely OFF and
is done by applying a sufficient amount of voltage. If the voltage to switch the device
falls below a certain threshold, the component will be used in active mode. The losses
in active mode are much higher compared to switch mode. It is therefore recommended
to detect an under-voltage condition and consequently shut down the driver since the
low voltage may be a consequence of faulty conditions.
110 Appendix B. Design considerations
B.2 Component selection
The selection of the relevant components is discussed here together with their important
features. Both a regular Buck converter and a Buck converter with split DC bus will be
built using mostly the same components.
Main switches IGBTs with integrated diodes will be used as the main switches: Inter-
national Rectifier IRG4PH40UDPbF and also IRGP30B120KD-E. The relevant part of
the data sheet can be found in Appendix F. To make a comparison with better compo-
nents, also SiC transistors will be used: CREE C2M0080120D SiC MOSFETs (RDS,on
= 80 mΩ)
Auxiliary switches The auxiliary switches that have been selected are both regular
MOSFETs as SiC MOSFETs: CREE C2M0280120D (RDS,on = 280 mΩ)
Diodes The free-wheeling diodes of the Buck converter/ inverter are integrated in the
main switches. For the soft cell, Silicon Carbide Schottky diodes will be used since they
can be used for high frequencies, have very low switching losses and switch very fast:
CREE C4D02120A.
Inductances Several inductances will be tested ant their performance will be compared.
The main issue is here the linearity of the inductor over a broad current range and the
maximum heat dissipation capabilities. If the industrial available components are not
suitable, air coils can be used as an alternative. They have the advantage that they do
not saturate.
General purpose power supply The board will be supplied via an external 5V power
supply: Myrra 47200.
Driver power supplies The driving voltages of the IGBTs are +15/-8.7V. The driving
voltages of the SiC transistor are slightly different: +20/-5V They will be fed by Murata
MGJ2 isolated DC/DC converters.
Gate drivers For fast switching, a high quality driver is needed. Therefore, the Avago
ACPL-337J has been selected. It contains an integrated desaturation protection, active
Miller clamping and error status feedback.
B.3. PCB design 111
B.3 PCB design
The software that has been used for the design of the boards is Altium Designer 14.3.11.
Since I was personally not able to work with this software, a considerable amount of time
went to training via instruction videos and reading the provided tutorials of the software.
The design of a PCB consists of different steps. First, a schematic of the circuit must be
made. To do so, the different items that will be used must be available in the ’component
library’. This means that both a schematic symbol and a ’footprint’ need to exist. If they
are not available, which is usually the case, they must be custom made. If the schematic is
complete and all the components have been assigned the correct footprints, the schematic
can be transferred to the PCB lay-out. In PCB lay-out one must first define the board’s
dimensions. Then, the components can be placed on the board and the routing can be
done. This process has usually some internal feedback in it, which makes that it is ran
through several times for perfection.
Component selection has already been discussed in a previous section. However, a lot of
components/ICs still need some extra auxiliary parts to work properly (e.g. filter capac-
itors for the drivers). SMD components were chosen for this purpose. This way, a high
component density (components per unit are) and low parasitics are achieved. They can
be placed on both sides of the PCB and they are usually cheaper. Also the traces can be
made shorter, which again decreases the parasitic resistance and inductance. So different
advantages are present when SMD components are chosen. The only disadvantage is that
they also need to be soldered by hand, which needs some experience.
The PCBs have been fabricated in the workshop of EELAB. This leaded to some constraints
since no high-end, automated machines are used:
The board can only be single- or double-layered.
The minimum trace width is 0.3 mm.
The drilling is done manually.
The components must be soldered by hand.
For the eventual PCB lay-out and schematics, we refer to Appendix D.
112 Appendix B. Design considerations
B.4 PCB optimization
During the initial design stage, the guidelines of the data sheets of the components have
been followed. However it has been found, via experimental verification, that the proposed
solutions still needed some optimization.
For the driving circuit, an Avago ACPL-337J IC is used. The recommended application
circuit can be found in the data sheet and is shown in Fig. B.1 and Fig. B.2. The involved
circuit is used to protect the IGBT and must intervene when the collector current is too
high. This is sensed via the collector-emitter voltage VCE by reasoning that an over-current
will also lead to an increased voltage across the terminals of the device. The protection
must trip at a current of w 21 A. The circuit however intervened too quickly. Therefore, it
needed to be tuned such that it acts only above this current level. This was made possible
by using a combination of a Zenerdiode and two regular diodes, namely the combination
of BZX79C2V4, BYV26, 1N4148. With this configuration the protection triggers for a
current in the range of 21-23 A.
For the first tests, normal cables were used to transmit the signal from the DSP to the
driver. This however leaded to some disturbances because the signal was influenced by
the EMI of the power cables (high dv/dt and di/dt). This problem was solved by using
shielded cables.
Also the gate resistor needed to be determined. This can be done using the equations in
the driver data sheet [6] where it is also important to control if the power dissipation of
the driver is not too high. This calculation has been done via a Matlab script that can be
found in Appendix A.5. The result is Rg = 3.9 Ω. To verify this result, the driving voltage
is plotted in Fig. B.3. One can see that the behavior is as expected, without any ringing.
So the calculated Rg is valid.
B.5 Thermal design
The switching and conduction losses are dissipated inside the IGBT. This leads to a tem-
perature increase of the package. The component will be destroyed if a certain critical
B.5. Thermal design 113
Fig. B.1: Recommended driver application circuit [6]
Fig. B.2: Recommended application of the DESAT protection circuit [6]
114 Appendix B. Design considerations
×10-70 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
Vol
ts [V
]
-10
-5
0
5
10
15
20Turn-on
Time [s] ×10-70 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
Vol
ts [V
]
-15
-10
-5
0
5
10
15Turn-off
Fig. B.3: IGBT driving voltages (zoom)
temperature is reached. For most components this is Tmax = 150 C. To extend the ca-
pabilities of the switches, they are usually mounted on a heat sink using thermal grease.
This is done to maximize the heat transfer towards the heat sink. Another important
aspect is that the transistor needs to be isolated from the heat sink. This is achieved using
thermal insulators as shown in Fig. B.4. The reliability of the entire system depends on
an adequate thermal design.
The losses of the IGBTs can be calculated as described in Chapter 2. The Matlab script
that has been used for this purpose can be found in section A.2. Both the losses of a Buck
converter operating at 4 kW and an inverter with an output of 4 kW have been calculated.
This information together with the chosen heat sink leads to the temperature increase
of the device. The heat sink was chosen rather large such that the components would
not experience a thermal breakdown during the tests. The Matlab script can be adapted
easily to the appropriate input parameters when for example the input voltage or switching
frequency is altered.
B.6. Control circuitry 115
Fig. B.4: Thermal insulation tube
B.6 Control circuitry
The control signals that are supplied to the boards are generated via a Digital Signal
Processor (DSP) of Texas Instruments, namely the TMS320F28335, 32 bit digital signal
processor with a clock frequency of 150 MHz, in combination with the Peripheral Explorer
Board. Pictures of both are shown in Fig. B.5. The DSP was programmed via the program
’Code Composer Studio’, version 6.1.2 . These programs are written in the C language
and can be found under Appendix C.
116 Appendix B. Design considerations
Fig. B.5: The Digital Signal Processor
Appendix C
DSP code
This appendix contains the program, written in C, that was used to deliver the signal
to the gate drivers. It was uploaded via Code Composer Studio 6.1.2 towards the TI
TMS320F28335. The main purpose was to produce three square waves (PWM) with a
variable frequency and variable duty cycle. Also the timing needed to be easily changeable.
The first square wave is the main PWM signal to the converter while the other two are the
pulses for the auxiliary circuit. They need to be adjusted such that they start just before
the begin or the end of the main pulse. These two signals are then combined via a logic
OR gate (7432) and then fed to the driver of the auxiliary circuit.
1 //Created on: 24 mrt. 2016
2 //Author: simon
3 //Based on examples provided by Texas Instruments
4
5 #include "DSP2833x Device.h"
6
7 // external function prototypes
8 extern void InitSysCtrl(void);
9 extern void InitPieCtrl(void);
10 extern void InitPieVectTable(void);
11 extern void InitCpuTimers(void);
12 extern void ConfigCpuTimer(struct CPUTIMER VARS *, float, float);
13
14 // Prototype statements for functions found within this file.
15 void Gpio select(void);
117
118 Appendix C. DSP code
16 void Setup ePWM(void);
17 interrupt void cpu timer0 isr(void);
18
19 float duty 1 = 50.0;
20 float duty 2 = 10.0;
21 float duty 3 = 10.0;
22 float duty 4 = 10.0;
23
24 float set 1 = 0.0; //Pulse width of PWM 1A
25 float set 2 = 0.0; // Pulse width of PWM 2A
26 float set 3 = 0.0; //Phase shift of 2 relative to 1 75 =1us; 150 =2us; ...
225 =3us
27 float set 4 = 0.0; //Phase shift of 2 relative to 1 75 =1us; 150 =2us; ...
225 =3us
28 float set 5 = 0.0;
29 float set 6 = 1;
30
31 int set TBPRD = 3750; // 3750 = 20kHz, /2 for 40kHz, /2.5 for 50kHz
32
33 //###########################################################################
34 // main code
35 //###########################################################################
36 void main(void)
37 38 int counter=0; // binary counter for digital output
39
40 InitSysCtrl(); // Basic Core Init from DSP2833x SysCtrl.c
41
42 EALLOW;
43 SysCtrlRegs.WDCR= 0x00AF; // Re-enable the watchdog
44 EDIS; // 0x00AF to NOT disable the Watchdog, ...
Prescaler = 64
45
46 DINT; // Disable all interrupts
47
48 Gpio select(); // GPIO9, GPIO11, GPIO34 and GPIO49 as output
49 // to 4 LEDs at Peripheral Explorer)
50
51 Setup ePWM(); // init of ePWM1A
52
53 InitPieCtrl(); // basic setup of PIE table; from ...
DSP2833x PieCtrl.c
54
119
55 InitPieVectTable(); // default ISR's in PIE
56
57 EALLOW;
58 PieVectTable.TINT0 = &cpu timer0 isr;
59 EDIS;
60
61 InitCpuTimers(); // basic setup CPU Timer0, 1 and 2
62
63 ConfigCpuTimer(&CpuTimer0,150,100);
64
65 PieCtrlRegs.PIEIER1.bit.INTx7 = 1;
66
67 IER |=1;68
69 EINT;
70 ERTM;
71
72 CpuTimer0Regs.TCR.bit.TSS = 0; // start timer0
73
74 while(1)
75 76 while(CpuTimer0.InterruptCount == 0);
77 CpuTimer0.InterruptCount = 0;
78
79 EALLOW;
80 SysCtrlRegs.WDKEY = 0x55; // service WD #1
81 EDIS;
82 counter++;
83
84 set 1 = duty 1 * set TBPRD *0.01; //Duty cycle is ...
converted to a value between 0 and TBPRD
85 set 2 = duty 2 * set TBPRD * 0.01;
86 //set 3 = duty 3 * set TBPRD * 0.01; //If a symmetrical ...
pulse is wanted
87 set 4 = duty 4 * set TBPRD * 0.01;
88 if (set 6 == 1) //set 5 and set 6 ...
are used to position the pulse at turn-off
89 set 5 = 3750 - set 1 + 5;90
91
92 EPwm1Regs.CMPA.half.CMPA = set 1;
93 EPwm2Regs.CMPB = set 2;
94 EPwm2Regs.TBPHS.half.TBPHS = set 3;
120 Appendix C. DSP code
95 EPwm3Regs.CMPB = set 4;
96 EPwm3Regs.TBPHS.half.TBPHS = set 5;
97
98 99
100
101 void Gpio select(void)
102 103 EALLOW;
104 GpioCtrlRegs.GPAMUX1.all = 0; // GPIO15 ... GPIO0 = General ...
Puropse I/O
105 GpioCtrlRegs.GPAMUX1.bit.GPIO0 = 1; // ePWM1A active
106 GpioCtrlRegs.GPAMUX1.bit.GPIO2 = 1; // ePWM2A active
107 GpioCtrlRegs.GPAMUX1.bit.GPIO4 = 1; // ePWM3A active
108
109
110 GpioCtrlRegs.GPAMUX2.all = 0; // GPIO31 ... GPIO16 = General ...
Purpose I/O
111 GpioCtrlRegs.GPBMUX1.all = 0; // GPIO47 ... GPIO32 = General ...
Purpose I/O
112 GpioCtrlRegs.GPBMUX2.all = 0; // GPIO63 ... GPIO48 = General ...
Purpose I/O
113 GpioCtrlRegs.GPCMUX1.all = 0; // GPIO79 ... GPIO64 = General ...
Purpose I/O
114 GpioCtrlRegs.GPCMUX2.all = 0; // GPIO87 ... GPIO80 = General ...
Purpose I/O
115
116 GpioCtrlRegs.GPADIR.all = 0;
117 GpioCtrlRegs.GPADIR.bit.GPIO9 = 1; // peripheral explorer: LED ...
LD1 at GPIO9
118 GpioCtrlRegs.GPADIR.bit.GPIO11 = 1; // peripheral explorer: LED ...
LD2 at GPIO11
119
120 GpioCtrlRegs.GPAMUX2.bit.GPIO17 = 0; // ePWM3A active
121 GpioDataRegs.GPASET.bit.GPIO17 = 1; // ePWM3A active
122
123 GpioCtrlRegs.GPBDIR.all = 0; // GPIO63-32 as inputs
124 GpioCtrlRegs.GPBDIR.bit.GPIO34 = 1; // peripheral explorer: LED ...
LD3 at GPIO34
125 GpioCtrlRegs.GPBDIR.bit.GPIO49 = 1; // peripheral explorer: LED LD4 ...
at GPIO49
126 GpioDataRegs.GPBSET.bit.GPIO49 = 1;
127
121
128
129 GpioCtrlRegs.GPCDIR.all = 0; // GPIO87-64 as inputs
130 EDIS;
131 132
133 void Setup ePWM(void)
134 135 EPwm1Regs.TBCTL.bit.CLKDIV = 0; // CLKDIV = 1
136 EPwm1Regs.TBCTL.bit.HSPCLKDIV = 1; // HSPCLKDIV = 2
137 EPwm1Regs.TBCTL.bit.CTRMODE = 0; // up mode - DUs enkel omhoog ...
tellen!!!
138 EPwm1Regs.AQCTLA.all = 18; // ZRO = set, PRD = clear
139 //EPwm1Regs.AQCTLA.all = 96;
140 EPwm1Regs.TBPRD = set TBPRD; // 20KHz - PWM signal - Geen ...
1/2 nodig als we enkel omhoog counten!
141 EPwm1Regs.TBCTL.bit.SYNCOSEL = 1; //Generate a signal if CTR=0, ...
needed for the phase shift
142
143 //EPwm1Regs.CMPA.half.CMPA = set;
144
145 EPwm2Regs.TBCTL.bit.CLKDIV = 0; // CLKDIV = 1
146 EPwm2Regs.TBCTL.bit.HSPCLKDIV = 1; // HSPCLKDIV = 2
147 EPwm2Regs.TBCTL.bit.CTRMODE = 0; // up mode
148 EPwm2Regs.AQCTLA.all = 258; // ZRO = set, PRD = clear THIS ...
LINE IS MEANINGLESS!!! SHOULD BE 18 OR 06
149 EPwm2Regs.TBPRD = set TBPRD; // 20KHz - PWM signal
150
151 EPwm2Regs.TBCTL.bit.PHSEN = 1; // Set phase enable
152 EPwm2Regs.TBCTL.bit.SYNCOSEL = 0; //Sync Out Select: SYNCIN = SYNCOUT
153 //EPwm2Regs.TBPHS.half.TBPHS = 100; //Phase shift of 2 relative to 1
154 //EPwm2Regs.CMPB = 500; //dutycycle
155
156 EPwm3Regs.TBCTL.bit.CLKDIV = 0; // CLKDIV = 1
157 EPwm3Regs.TBCTL.bit.HSPCLKDIV = 1; // HSPCLKDIV = 2
158 EPwm3Regs.TBCTL.bit.CTRMODE = 0; // up mode
159 EPwm3Regs.AQCTLA.all = 258; // ZRO = set, PRD = clear
160 EPwm3Regs.TBPRD = set TBPRD; // 20KHz - PWM signal
161 EPwm3Regs.TBCTL.bit.SYNCOSEL = 0; //Sync Out Select: SYNCIN = SYNCOUT
162 EPwm3Regs.TBCTL.bit.PHSEN = 1;
163 EPwm3Regs.TBPHS.half.TBPHS = set 5; // set your phase shift ...
of the aux pulse
164
165
122 Appendix C. DSP code
166
167 interrupt void cpu timer0 isr(void)
168 169 CpuTimer0.InterruptCount++;
170 EALLOW;
171 SysCtrlRegs.WDKEY = 0xAA; // service WD #2
172 EDIS;
173 PieCtrlRegs.PIEACK.all = PIEACK GROUP1;
174 175 //===========================================================================
176 // End of SourceCode.
177 //===========================================================================
Appendix D
PCB design
In this section, the different schematics and board lay-outs for the converter and auxiliary
circuits can be found. They are made with Altium Designer 14.3.11
D.1 Schematics
123
1
1
2
2
3
3
4
4
D D
C C
B B
A A
Title
Number RevisionSize
A4
Date: 25-2-2016 Sheet ofFile: C:\Users\..\5V Power supply.SchDoc Drawn By:
1
2 3
4ACN
ACL -Vout
+VoutD27
ECE Series
12
P3
Header 2
C31100nF
C32100nF
C3010uF
R33
100
R34
100
+5V
GND_P
1
2
7
6
5
MGJ2
+Vin
-Vin-Vout
0V
+Vout
DCDC1
Murata MGJ2 Series
+15V_TOP_M
-5V_TOP_M
GND_TOP_MC34100nF
C35100nF
C33100nF
C36100nF
+5V
GND_P
L5
22uH
L6
22uH
1
2
7
6
5
MGJ2
+Vin
-Vin-Vout
0V
+Vout
DCDC2
Murata MGJ2 Series
C39100nF
+5V
GND_BOT_M
+15V_BOT_M
-5V_BOT_M
C37100nF
C40100nF
GND_P
10mH
L7
10mH
10mH
L8
10mH
C38100nF
1
2
7
6
5
+Vin
-Vin-Vout
0V
+VoutDCDC3
Murata MGJ2 Series
C42100nF
C41100nF
C44100nF
GND_TOP_A
+15V_TOP_A
-5V_TOP_A
+5V
GND_P
1
2
7
6
5
+Vin
-Vin-Vout
0V
+VoutDCDC4
Murata MGJ2 Series
C47100nF
+5V
GND_P
C45100nF
C48100nF
+15V_BOT_A
-5V_BOT_A
GND_BOT_A
C43100nF
C46100nF
10mH
L9
10mH
10mH
L10
10mH
10mH
L11
10mH
10mH
L12
10mH
AC1
AC2
12
P2
Header 2
12
P4
Header 2
12
P6
Header 2
12
P10
Header 2
12
P5
Header 2
12
P7
Header 2
12
P8
Header 2
12
P9
Header 2
12
P?
Header 2
12
P?
Header 2
1
1
2
2
3
3
4
4
D D
C C
B B
A A
Title
Number RevisionSize
A4
Date: 25-2-2016 Sheet ofFile: C:\Users\..\Buck.SchDoc Drawn By:
Q1IGBT-N
Q2IGBT-N
R10.1
R90.1
1
2
3
4
5
6
7
8 9
10
11
12
13
14
15
16ACPL-337J
Vee1
Vin+
Vcc1
Vleddr
Uvlo
Fault
Anode
Cathode
Vee2
Vled
Desat
Ve
Vcc2
Vout
Vclamp
Vee2
IC1
ACPL-337J
R6
5
R5 5
R14
5
R13 5
1
2
3
4
5
6
7
8 9
10
11
12
13
14
15
16ACPL-337J
Vee1
Vin+
Vcc1
Vleddr
Uvlo
Fault
Anode
Cathode
Vee2
Vled
Desat
Ve
Vcc2
Vout
Vclamp
Vee2
IC2
ACPL-337J
+Vbus
-Vbus
D5 D Schottky
D10 D Schottky
C11uF
150R7
R8150
C21uF
C31uF
C61uF
Cblank1220pF
R2
1K
D1Diode
+15V_TOP_M
-5V_TOP_M
GND_TOP_M
GND_TOP_M
D3
D Zener
D4
D SchottkyR410K
R310K
C4330pFC5
330pF
GND_1
J1
Socket
J5
SocketGND_BOT_M
C81uF
C91uF
C101uF
-5V_BOT_M
+15V_BOT_M
Cblank2220pF
R10
1K
D9
D Schottky
D8
D Zener
D7
Diode
GND_BOT_MR1110K
R1210K
100pF
C11Cap
100pF
C12Cap
R15
150
R16
150
C7
1uF
GND_1
J3
Socket
J2
Socket
J4
Socket
1 23 45 67 89 1011 1213 1415 1617 1819 20
P1
Header 10X2
+5V
GND_1
FAIL
FAIL
M_D_TOP
GND_1GND_1
M_D_TOPA_D_TOP
A_D_BOTM_D_BOT
FAIL
M_D_BOT
FAIL
FAIL
GND_1+5V
1 2
3 4
G
R
D2
HSMF-C155
1 2
3 4
G
R
D6
HSMF-C155
-5V_TOP_M
-5V_BOT_M
-5V_BOT_M
-5V_TOP_M
1
1
2
2
3
3
4
4
D D
C C
B B
A A
Title
Number RevisionSize
A4
Date: 25-2-2016 Sheet ofFile: C:\Users\..\Soft_cell.SchDoc Drawn By:
1
2
3
4
5
6
7
8 9
10
11
12
13
14
15
16Vee1
Vin+
Vcc1
Vleddr
Uvlo
Fault
Anode
Cathode
Vee2
Vled
Desat
Ve
Vcc2
Vout
Vclamp
Vee2
IC3
ACPL-337J
1
2
3
4
5
6
7
8 9
10
11
12
13
14
15
16Vee1
Vin+
Vcc1
Vleddr
Uvlo
Fault
Anode
Cathode
Vee2
Vled
Desat
Ve
Vcc2
Vout
Vclamp
Vee2
IC4
ACPL-337J
Q3MOSFET-N
Q4MOSFET-N
D15Diode
D23Diode
D19
Diode
D18
Diode
C152.2nF
C262.2nF
C22
16nF
C17
15nF
L1
30uH
L2
10uH
L3
10uH
L4
30uH
D16Diode
D20Diode
J6
Socket
J11
Socket
J9
Socket
J7
Socket
R2710K
R2810K
C28330pF
C29330pF
R31
150
R32
150
C23
1uF
R1810K
R1910K
C19330pF
C20330pF
R22
150
R24
150
C13
1uF
GND_P
+5VGND_P
+5VGND_P
R305
R295
D26 D Schottky
C241uF
C251uF
C271uF
+15V_BOT_A
Cblank4220pF
D25
D Schottky
D24
D Zener
GND_BOT_A
R23
0.1
R25
0.1
-5V_BOT_AGND_BOT_A
R26
1K
R215
R205
D17 D Schottky
C141uF
C161uF
C181uF
+15V_TOP_A
Cblank3220pF
D14
D Schottky
D13
D Zener
GND_TOP_A
R17
1K
-5V_TOP_A
D22
Diode
D12
Diode
A_D_TOP
A_D_BOT
FAIL
FAIL
FAIL
FAIL
GND_P
GND_TOP_A
1 2
3 4
G
R
D11
HSMF-C155
1 2
3 4
G
R
D21
HSMF-C155
-5V_TOP_A
-5V_BOT_A
1
1
2
2
3
3
4
4
D D
C C
B B
A A
Title
Number RevisionSize
A4
Date: 28-4-2016 Sheet ofFile: C:\Users\..\Soft_cell_one.SchDoc Drawn By:
1
2
3
4
5
6
7
8 9
10
11
12
13
14
15
16Vee1
Vin+
Vcc1
Vleddr
Uvlo
Fault
Anode
Cathode
Vee2
Vled
Desat
Ve
Vcc2
Vout
Vclamp
Vee2
IC3
ACPL-337J
Q3MOSFET-N
C17
15nF
L1
30uH
L2
10uH
J9
Socket
J11
Socket
J10
Socket
R1810K
R1910K
C19330pF
C20330pF
R22
150
R24
150
C13
1uF
+5VGND_P
R23
0.1
R215
R205
D17 D Schottky
C141uF
C161uF
C181uF
+15V_TOP_A
Cblank3220pF
D14
D Schottky
D13
D Zener
GND_TOP_A
R17
1K
-5V_TOP_A
D12
Diode
A_D_TOP
FAIL
FAIL
GND_P
GND_TOP_A
1 2
3 4
G
R
D11
HSMF-C155
-5V_TOP_A
G3
P9
Mounting pad
P10
Mounting pad
P11
Mounting pad
P7
Mounting pad
P8
Mounting pad
D15Diode_SR
D16
Diode_SR
D18Diode_SR
J16
Socket
J17
Socket
J18
Socket
+DCbus
-DCbus
Midpoint
1234
P33
Header 4
FAIL+5VA_D_TOP
GND_P
J30
Socket
1
1
2
2
3
3
4
4
D D
C C
B B
A A
Title
Number RevisionSize
A4
Date: 28-4-2016 Sheet ofFile: C:\Users\..\T_type_schema.SchDoc Drawn By:
1
2
3
4
5
6
7
8 9
10
11
12
13
14
15
16Vee1
Vin+
Vcc1
Vleddr
Uvlo
Fault
Anode
Cathode
Vee2
Vled
Desat
Ve
Vcc2
Vout
Vclamp
Vee2
IC1
ACPL-337J
R210K
R310K
C5330pF
C6330pF
R6
150
R7
150
C1
1uF
+5VGND_P
R55
R45
C21uF
C31uF
C41uF
+15V_A1
Cblank1220pF
D4
D Schottky
D3
D Zener
SOURCE_A1
R1
1K
-5V_A1
D2
Diode
Pulse_A1
FAIL
FAIL
GND_P
1 2
3 4
G
R
D1
HSMF-C155
-5V_A1
1
2
3
4
5
6
7
8 9
10
11
12
13
14
15
16Vee1
Vin+
Vcc1
Vleddr
Uvlo
Fault
Anode
Cathode
Vee2
Vled
Desat
Ve
Vcc2
Vout
Vclamp
Vee2
IC2
ACPL-337J
R910K
R1010K
C11330pF
C12330pF
R13
150
R14
150
C7
1uF
+5VGND_P
R125
R115
C81uF
C91uF
C101uF
+15V_A2
Cblank2220pF
D10
D Schottky
D9
D Zener
SOURCE_A2
R8
1K
-5V_A2
D8
Diode
Pulse_A2
FAIL
FAIL
GND_P
1 2
3 4
G
R
D7
HSMF-C155
-5V_A2
Lr118uH
Lr218uH
Cr18nF
Cr28nF
J_Midpoint_SwitchesSocket
J_Midpoint_CapSocket
GATE_A1
GATE_A1
GATE_A2
GATE_A2
SOURCE_A2
DRAIN_A1
DRAIN_A2
DRAIN_A2
R150.1 J1
Socket
D11Diode
R17100K
C141nF
D5Diode
DRAIN_A1
SOURCE_A1C131nF
R16100K Q_A1
MOSFET-N
Q_A2MOSFET-N
P1
Mounting pad
P2
Mounting pad
P3
Mounting pad
P4
Mounting pad
1234
P5
Header 4GND_P
Pulse_A1Pulse_A2
FAIL
1
1
2
2
3
3
4
4
D D
C C
B B
A A
Title
Number RevisionSize
A4
Date: 28-4-2016 Sheet ofFile: C:\Users\..\T_type_powersupplies.SchDoc Drawn By:
1
2
7
6
5
MGJ2
+Vin
-Vin-Vout
0V
+Vout
DCDC1
Murata MGJ2 Series
+15V_A1
-5V_A1
SOURCE_A1C31100nF
C32100nF
C30100nF
C33100nF
+5V
GND_P
L3
22uH
L4
22uH
1
2
7
6
5
MGJ2
+Vin
-Vin-Vout
0V
+Vout
DCDC2
Murata MGJ2 Series
C36100nF
+5V+15V_A2
-5V_A2
C34100nF
C37100nF
GND_P
L5
22uH
L6
22uH
C35100nF
12
P23
Header 2
12
P24
Header 2
SOURCE_A2
A1
A2
A3
A4
130 Appendix D. PCB design
D.2 Board lay-out
D.2. Board lay-out 131
Fig. D.1: PCB layout - Buck
132 Appendix D. PCB design
Fig. D.2: PCB layout - ZVT1
D.2. Board lay-out 133
Fig. D.3: PCB layout - ZVT2
134 Appendix D. PCB design
D.3 Pictures
D.3. Pictures 135
Fig. D.4: Buck converter
Fig. D.5: Midpoint clamped SSAC
136 Appendix D. PCB design
Fig. D.6: Soft switching auxiliary circuit 2
Appendix E
Measurements
137
138 Appendix E. Measurements
The measurement data that was obtained during the experiments can be found in this
appendix.
δ Vin Iin Pin Vout Iout Pout η
[/] [V] [A] [W] [V] [A] [W] [%]
0, 10 300, 00 0, 11 33, 00 28, 98 1, 04 30, 14 91, 33
0, 15 300, 00 0, 24 72, 00 43, 60 1, 57 68, 45 95, 07
0, 20 300, 00 0, 43 129, 00 58, 40 2, 10 122, 64 95, 07
0, 25 300, 00 0, 66 198, 00 72, 70 2, 62 190, 47 96, 20
0, 30 300, 00 0, 96 288, 00 87, 40 3, 15 275, 31 95, 59
0, 35 300, 00 1, 30 390, 00 102, 10 3, 68 375, 73 96, 34
0, 40 300, 00 1, 70 510, 00 116, 80 4, 21 491, 73 96, 42
0, 45 300, 00 2, 15 645, 00 131, 50 4, 74 623, 31 96, 64
0, 50 300, 00 2, 64 792, 00 146, 00 5, 26 767, 96 96, 96
0, 55 300, 00 3, 20 960, 00 160, 70 5, 79 930, 45 96, 92
0, 60 300, 00 3, 80 1140, 00 175, 40 6, 32 1108, 53 97, 24
0, 65 300, 00 4, 47 1341, 00 190, 10 6, 85 1302, 19 97, 11
0, 70 300, 00 5, 19 1557, 00 205, 20 7, 40 1518, 48 97, 53
0, 75 300, 00 5, 96 1788, 00 220, 00 7, 94 1746, 80 97, 70
0, 80 300, 00 6, 79 2037, 00 234, 90 8, 47 1989, 60 97, 67
0, 85 300, 00 7, 68 2304, 00 249, 70 9, 02 2252, 29 97, 76
0, 90 300, 00 8, 62 2586, 00 264, 70 9, 55 2527, 89 97, 75
Table E.1: Hard switched Buck - Measurements at 10 kHz
139
δ Vin Iin Pin Vout Iout Pout η
[/] [V] [A] [W] [V] [A] [W] [%]
0, 10 300, 00 0, 12 36, 00 29, 10 1, 05 30, 56 84, 88
0, 15 300, 00 0, 25 75, 00 43, 80 1, 58 69, 20 92, 27
0, 20 300, 00 0, 43 129, 00 58, 10 2, 09 121, 43 94, 13
0, 25 300, 00 0, 67 201, 00 72, 60 2, 61 189, 49 94, 27
0, 30 300, 00 0, 96 288, 00 87, 30 3, 15 275, 00 95, 48
0, 35 300, 00 1, 31 393, 00 102, 20 3, 69 377, 12 95, 96
0, 40 300, 00 1, 71 513, 00 116, 90 4, 22 493, 32 96, 16
0, 45 300, 00 2, 16 648, 00 131, 50 4, 74 623, 31 96, 19
0, 50 300, 00 2, 66 798, 00 146, 30 5, 28 772, 46 96, 80
0, 55 300, 00 3, 22 966, 00 161, 20 5, 81 936, 57 96, 95
0, 60 300, 00 3, 84 1152, 00 175, 90 6, 35 1116, 97 96, 96
0, 65 300, 00 4, 51 1353, 00 190, 70 6, 88 1312, 02 96, 97
0, 70 300, 00 5, 23 1569, 00 205, 50 7, 42 1524, 81 97, 18
0, 75 300, 00 6, 02 1806, 00 220, 30 7, 96 1753, 59 97, 10
0, 80 300, 00 6, 86 2058, 00 235, 20 8, 50 1999, 20 97, 14
0, 85 300, 00 7, 76 2328, 00 250, 10 9, 04 2260, 90 97, 12
0, 90 300, 00 8, 74 2622, 00 265, 10 9, 58 2539, 66 96, 86
Table E.2: Hard switched Buck - Measurements at 20 kHz
140 Appendix E. Measurements
δ Vin Iin Pin Vout Iout Pout η
[/] [V] [A] [W] [V] [A] [W] [%]
0, 10 300, 20 0, 14 42, 03 29, 20 1, 20 35, 04 83, 37
0, 15 300, 20 0, 28 84, 06 42, 80 1, 70 72, 76 86, 56
0, 20 300, 20 0, 46 138, 09 57, 20 2, 19 125, 27 90, 71
0, 25 300, 20 0, 71 213, 14 71, 90 2, 72 195, 57 91, 75
0, 30 300, 20 1, 01 303, 20 86, 80 3, 26 282, 97 93, 33
0, 35 300, 20 1, 37 411, 27 101, 60 3, 81 387, 10 94, 12
0, 40 300, 20 1, 79 537, 36 116, 50 4, 37 509, 11 94, 74
0, 45 300, 20 2, 26 678, 45 131, 10 4, 92 645, 01 95, 07
0, 50 300, 20 2, 78 834, 56 145, 90 5, 48 799, 53 95, 80
0, 55 300, 20 3, 37 1011, 67 160, 70 6, 03 969, 02 95, 78
0, 60 300, 20 4, 02 1206, 80 175, 60 6, 59 1157, 20 95, 89
0, 65 300, 20 4, 74 1422, 95 190, 50 7, 15 1362, 08 95, 72
Table E.3: Hard switched Buck - Measurements at 40 kHz
141
δmain δon Ton Vin,1 Iin,1 Vin,2 Iin,2 Pin Vout Iout Pout η
[%] [%] [s] [V] [A] [V] [A] [W] [V] [A] [W] [/]
10 1 3, 33E − 07 300, 1 0, 13 150, 1 0, 01 40, 51 31, 7 1, 14 36, 14 0, 89
20 1 3, 33E − 07 300, 1 0, 45 150, 1 0, 01 136, 55 60, 3 2, 17 130, 85 0, 96
30 1 3, 33E − 07 300, 1 0, 97 150, 1 0, 04 297, 10 89, 1 3, 21 286, 01 0, 96
40 1 4, 00E − 07 300, 1 1, 72 150, 1 0, 05 523, 68 118, 6 4, 28 507, 61 0, 97
50 1 4, 00E − 07 300, 1 2, 68 150, 1 0, 06 813, 27 148 5, 34 790, 32 0, 97
60 1 4, 27E − 07 300, 1 3, 86 150, 1 0, 07 1168, 89 177, 5 6, 41 1137, 78 0, 97
70 1 4, 53E − 07 300, 1 5, 26 150, 1 0, 08 1590, 53 207 7, 49 1550, 43 0, 97
80 1 4, 67E − 07 300, 1 6, 88 150, 1 0, 09 2078, 20 236, 5 8, 55 2022, 08 0, 97
90 1 5, 07E − 07 300, 1 8, 71 150, 1 0, 10 2628, 88 266, 1 9, 63 2562, 54 0, 97
Table E.4: Buck converter with MPC SSAC at turn-on - 20 kHz
δmain δoff Toff Vin,1 Iin,1 Vin,2 Iin,2 Pin Vout Iout Pout η
[%] [%] [s] [V] [A] [V] [A] [W] [V] [A] [W] [/]
10 0, 5 6, 67E − 08 300, 1 0, 13 150, 1 0, 01 40, 51 31, 1 1, 12 34, 83 0, 86
20 0, 5 6, 67E − 08 300, 1 0, 44 150, 1 0, 01 133, 55 59, 6 2, 15 128, 14 0, 96
30 0, 5 6, 67E − 08 300, 1 0, 98 150, 1 0, 01 295, 60 88, 8 3, 20 284, 16 0, 96
40 0, 5 6, 67E − 08 300, 1 1, 73 150, 1 0, 02 522, 18 118, 2 4, 26 503, 53 0, 96
50 0, 5 6, 67E − 08 300, 1 2, 69 150, 1 0, 02 810, 27 147, 7 5, 33 787, 24 0, 97
60 0, 5 6, 67E − 08 300, 1 3, 87 150, 1 0, 03 1165, 89 177, 2 6, 39 1132, 31 0, 97
70 0, 5 6, 67E − 08 300, 1 5, 27 150, 1 0, 03 1586, 03 206, 7 7, 46 1541, 98 0, 97
80 0, 5 6, 67E − 08 300, 1 6, 89 150, 1 0, 04 2073, 69 236, 3 8, 54 2018, 00 0, 97
90 0, 5 6, 67E − 08 300, 1 8, 72 150, 1 0, 04 2622, 88 265, 9 9, 61 2555, 30 0, 97
Table E.5: Buck converter with MPC SSAC at turn-off - 20 kHz
142 Appendix E. Measurements
δmain
δon
Ton
δoff
Toff
Vin,1
Iin,1
Vin,2
Iin,2
Pin
Vout
Iout
Pout
η
[%]
[%]
[s][%
][s]
[V]
[A]
[V]
[A]
[W]
[V]
[A]
[W]
[/]
101
3,33E−
070,5
6,67E−
08300,1
0,12150,1
0,0239,01
321,15
36,800,94
201
3,33E−
070,5
6,67E−
08300,1
0,44150,1
0,03136,55
60,72,19
132,930,97
301
3,33E−
070,5
6,67E−
08300,1
0,97150,1
0,05298,60
89,53,23
289,090,97
401
3,33E−
070,5
6,67E−
08300,1
1,71150,1
0,06522,18
118,64,28
507,610,97
501
3,33E−
070,5
6,67E−
08300,1
2,67150,1
0,07811,77
147,95,34
789,790,97
601
4,00E−
070,5
6,67E−
08300,1
3,85150,1
0,091168,89
177,76,41
1139,060,97
701
4,00E−
070,5
6,67E−
08300,1
5,24150,1
0,111589,04
207,17,47
1547,040,97
801
4,27E−
070,5
6,67E−
08300,1
6,85150,1
0,122073,70
236,78,56
2026,150,98
901
4,67E−
070,5
6,67E−
08300,1
8,70150,1
0,142631,88
266,39,64
2567,130,98
Tab
leE
.6:B
uck
converter
with
MP
CSSA
Cat
turn
-onan
doff
-20
kH
z
143
δ main
δ on
Ton
δ off
Toff
Vin,1
I in,1
Vin,2
I in,2
Pin
Vout
I out
Pout
η
[%]
[%]
[s]
[%]
[s]
[V]
[A]
[V]
[A]
[W]
[V]
[A]
[W]
[/]
101
3,33E−
071
6,67E−
0830
0,1
0,15
150,
10,
0451,0
235,7
01,
2845,7
00,
90
201
3,33E−
071
6,67E−
0830
0,1
0,47
150,
10,
0715
1,55
63,4
02,
2814
4,55
0,95
301
3,33E−
071
6,67E−
0830
0,1
1,00
150,
10,
1031
5,11
91,8
03,
3030
2,94
0,96
401
4,00E−
071
6,67E−
0830
0,1
1,78
150,
10,
1355
3,69
121,
704,
3953
4,26
0,96
501
4,67E−
071
6,67E−
0830
0,1
2,76
150,
10,
1785
3,79
151,
205,
4682
5,55
0,97
601
4,67E−
071
6,67E−
0830
0,1
3,94
150,
10,
2012
12,4
118
0,40
6,52
1176,2
10,
97
701
4,67E−
071
6,67E−
0830
0,1
5,34
150,
10,
2416
38,5
620
9,60
7,58
1588,7
70,
97
801
4,67E−
071
6,67E−
0830
0,1
6,97
150,
10,
2721
32,2
223
9,00
8,65
2067,3
50,
97
901
4,67E−
071
6,67E−
0830
0,1
8,82
150,
10,
2926
90,4
126
8,60
9,73
2613,4
80,
97
Tab
leE
.7:
Buck
conve
rter
wit
hM
PC
SSA
Cat
turn
-on
and
off-
40kH
z
144 Appendix E. Measurements
δ Vin,1 Iin,1 Vin,2 Iin,2 Pin VR IR PR Vout Iout Pout η
[%] [V] [A] [V] [A] [W] [V] [A] [W] [V] [A] [W] [/]
55 200 0, 58 200, 1 1, 09 334, 11 200, 1 1, 53 306, 15 20, 1 1, 05 21, 11 0, 75
60 200 1, 37 200, 1 1, 20 514, 12 200, 1 1, 93 386, 19 42, 5 2, 23 94, 78 0, 74
65 200 1, 98 200, 1 1, 02 600, 10 200, 1 1, 92 384, 19 58, 4 3, 02 176, 37 0, 82
70 200 2, 86 200, 1 0, 87 746, 09 200, 1 1, 97 394, 20 78, 9 4, 04 318, 76 0, 91
75 200 3, 86 200, 1 0, 81 934, 08 200, 1 2, 02 404, 20 96, 6 5, 08 490, 73 0, 93
80 200 4, 94 200, 1 0, 87 1162, 09 200, 1 2, 03 406, 20 116, 2 6, 12 711, 14 0, 94
85 200 6, 15 200, 1 1, 03 1436, 10 200, 1 2, 03 406, 20 135, 8 7, 17 973, 69 0, 95
90 200 7, 53 200, 1 1, 31 1768, 13 200, 1 2, 03 406, 20 156, 4 8, 27 1293, 43 0, 95
95 200 8, 97 200, 1 1, 70 2134, 17 200, 1 2, 03 406, 20 176, 1 9, 32 1641, 25 0, 95
Table E.8: Buck with split DC bus 400 V - 20 kHz - hard switched
δ Vin,1 Iin,1 Vin,2 Iin,2 Pin VR IR PR Vout Iout Pout η
[%] [V] [A] [V] [A] [W] [V] [A] [W] [V] [A] [W] [/]
55 200 0, 58 200, 1 1, 58 432, 15 200, 1 2, 02 404, 20 19, 9 1, 04 20, 69 0, 74
60 200 1, 23 200, 1 1, 25 496, 12 200, 1 2, 02 404, 20 39, 3 2, 01 78, 99 0, 85
65 200 2 200, 1 1 600, 10 200, 1 2, 02 404, 20 57, 9 3, 03 175, 43 0, 89
70 200 2, 87 200, 1 0, 86 746, 08 200, 1 2, 02 404, 20 77, 3 4, 06 313, 83 0, 91
75 200 3, 87 200, 1 0, 81 936, 08 200, 1 2, 02 404, 20 96, 9 5, 09 493, 22 0, 92
80 200 4, 96 200, 1 0, 87 1166, 08 200, 1 2, 02 404, 20 116, 5 6, 14 715, 31 0, 93
85 200 6, 18 200, 1 1, 04 1444, 10 200, 1 2, 02 404, 20 136, 2 7, 18 977, 91 0, 94
90 200 7, 5 200, 1 1, 32 1764, 13 200, 1 2, 02 404, 20 155, 5 8, 21 1276, 65 0, 93
95 200 8, 97 200, 1 1, 7 2134, 17 200, 1 2, 02 404, 20 175, 4 9, 24 1620, 69 0, 93
Table E.9: Buck with split DC bus 400 V - 30 kHz - hard switched
145
δ main
δ on
Ton
δ off
Toff
Vin,1
I in,1
Vin,2
I in,2
Pin
VR
I RPR
Vout
I out
Pout
η
[%]
[%]
[s]
[%]
[s]
[V]
[A]
[V]
[A]
[W]
[V]
[A]
[W]
[V]
[A]
[W]
[/]
551
2,67E−
071
6,67E−
0820
00,
6720
0,1
0,62
258,
0620
0,1
1,12
224,
1122,9
1,20
27,4
80,
81
601
2,67E−
071
6,67E−
0820
01,
3320
0,1
0,30
326,
0320
0,1
1,12
224,
1141,8
2,20
91,9
60,
90
651
3,33E−
071
6,67E−
0820
02,
1220
0,1
0,08
440,
0120
0,1
1,12
224,
1161,5
3,23
198,
650,
92
701
3,33E−
071
1,33E−
0720
03,
0020
0,1
0,87
774,
0920
0,1
2,03
406,
2080,7
4,26
343,
780,
93
751
4,00E−
071
6,67E−
0820
03,
9820
0,1
0,84
964,
0820
0,1
2,02
404,
2010
05,
2952
9,00
0,94
800,
84,
00E−
071
1,52E−
2120
05,
0620
0,1
0,92
1196,0
920
0,1
2,03
406,
2011
9,1
6,30
750,
330,
95
850,
84,
67E−
070,
71,
00E−
0720
06,
3020
0,1
1,11
1482,1
120
0,1
2,03
406,
2013
9,4
7,37
1027,3
80,
95
900,
84,
67E−
071
6,67E−
0820
07,
5820
0,1
1,39
1794,1
420
0,1
2,03
406,
2015
8,2
8,38
1325,7
20,
96
951
4,67E−
071
7,33E−
0820
08,
7220
0,1
1,70
2084,1
720
0,1
2,03
406,
2017
4,6
9,24
1613,3
00,
96
Tab
leE
.10:
Buck
wit
hsp
lit
DC
bus
400
V-
20kH
z-
MP
CSSA
C
146 Appendix E. Measurements
δmain
δon
Ton
δoff
Toff
Vin,1
Iin,1
Vin,2
Iin,2
Pin
VR
IR
PR
Vout
Iout
Pout
η
[%]
[%]
[s][%
][s]
[V]
[A]
[V]
[A]
[W]
[V]
[A]
[W]
[V]
[A]
[W]
[/]
551,5
2,67E−
071,5
6,05E−
08200
0,72200,1
1,51446,15
200,12,02
404,2024,8
1,3032,24
0,77
601,5
3,33E−
071,5
6,13E−
08200
1,30200,1
1,26512,13
200,12,02
404,2044,7
2,1495,66
0,89
651,5
3,33E−
071,5
6,22E−
08200
2,00200,1
1,08616,11
200,12,02
404,2063,7
3,06194,92
0,92
701,5
3,33E−
071,5
1,02E−
07200
3,04200,1
0,88784,09
200,12,02
404,2082,5
4,33357,23
0,94
751,5
3,73E−
071,5
1,03E−
07200
4,04200,1
0,87982,09
200,12,02
404,20102,1
5,37548,28
0,95
801,5
3,73E−
071,5
6,33E−
08200
5,12200,1
0,961216,10
200,12,02
404,20120,8
6,37769,50
0,95
851,5
3,73E−
071,3
6,55E−
08200
6,31200,1
1,161494,12
200,12,02
404,20140,1
7,401036,74
0,95
901,5
4,40E−
071,3
6,63E−
08200
7,66200,1
1,471826,15
200,12,02
404,20159,9
8,461352,75
0,95
951,2
4,40E−
071
6,72E−
08200
9,09200,1
1,882194,19
200,12,02
404,20179,2
9,501702,40
0,95
Tab
leE
.11:B
uck
with
split
DC
bus
400V
-30
kH
z-
MP
CSSA
C
147
δ Vin,1 Iin,1 Vin,2 Iin,2 Pin VR IR PR Vout Iout Pout η
[%] [V] [A] [V] [A] [W] [V] [A] [W] [V] [A] [W] [/]
55 300 0, 88 300 1, 80 804, 0 300 2, 37 711, 0 41, 1 1, 47 60, 42 0, 65
60 300 1, 38 300 1, 57 885, 0 300 2, 37 711, 0 61, 4 2, 19 134, 47 0, 77
65 300 2, 19 300 1, 32 1053, 0 300 2, 37 711, 0 90, 6 3, 25 294, 45 0, 86
70 300 3, 12 300 1, 18 1290, 0 300 2, 37 711, 0 120 4, 31 517, 20 0, 89
75 300 4, 15 300 1, 15 1590, 0 300 2, 37 711, 0 149, 3 5, 37 801, 74 0, 91
80 300 5, 29 300 1, 23 1956, 0 300 2, 37 711, 0 178, 9 6, 44 1152, 12 0, 93
85 300 6, 55 300 1, 41 2388, 0 300 2, 37 711, 0 208, 3 7, 51 1564, 33 0, 93
90 300 7, 91 300 1, 70 2883, 0 300 2, 37 711, 0 237, 8 8, 58 2040, 32 0, 94
95 300 9, 23 300 2, 08 3393, 0 300 2, 37 711, 0 263, 6 9, 52 2509, 47 0, 94
Table E.12: Buck with split DC bus 600 V - 20 kHz - hard switched
δ Vin,1 Iin,1 Vin,2 Iin,2 Pin VR IR PR Vout Iout Pout η
[%] [V] [A] [V] [A] [W] [V] [A] [W] [V] [A] [W] [/]
55 300 0, 73 300 1, 88 783, 0 300 2, 35 705, 0 34, 1 1, 21 41, 3 0, 53
60 300 1, 44 300 1, 56 900, 0 300 2, 35 705, 0 62, 8 2, 24 140, 7 0, 72
65 300 2, 26 300 1, 33 1077, 0 300 2, 36 708, 0 91, 9 3, 28 301, 4 0, 82
70 300 3, 20 300 1, 21 1323, 0 300 2, 36 708, 0 121, 3 4, 34 526, 4 0, 86
75 300 4, 25 300 1, 20 1635, 0 300 2, 36 708, 0 150, 7 5, 41 815, 3 0, 88
80 300 5, 44 300 1, 31 2025, 0 300 2, 36 708, 0 180, 5 6, 49 1171, 4 0, 89
Table E.13: Buck with split DC bus 600 V - 30 kHz - hard switched
148 Appendix E. Measurements
δmain
δon
Ton
δoff
Toff
Vin,1
Iin,1
Vin,2
Iin,2
Pin
VR
IR
PR
Vout
Iout
Pout
η
[%]
[%]
[s][%
][s]
[V]
[A]
[V]
[A]
[W]
[V]
[A]
[W]
[V]
[A]
[W]
[/]
551
2,67E−
071
6,67E−
08300
0,83300
1,80789,00
3002,35
705,0039,4
1,4356,34
0,67
601
2,67E−
071
6,67E−
08300
1,47300
1,51894,00
3002,35
705,0065,6
2,38156,13
0,83
651
3,33E−
071
6,67E−
08300
2,29300
1,291074,00
3002,35
705,0095,3
3,45328,79
0,89
701
3,33E−
071
1,33E−
07300
3,17300
1,181305,00
3002,36
708,00123,4
4,46550,36
0,92
751
4,00E−
071
6,67E−
08300
4,24300
1,171623,00
3002,36
708,00153,9
5,56855,68
0,94
800,8
4,00E−
071
1,52E−
21300
5,35300
1,261983,00
3002,36
708,00182,6
6,601205,16
0,95
850,8
4,00E−
070,8
6,67E−
08300
6,58300
1,462412,00
3002,36
708,00211,6
7,651618,74
0,95
900,8
4,67E−
070,8
6,67E−
08300
7,97300
1,782925,00
3002,36
708,00241,5
8,742110,71
0,95
951
4,67E−
071
7,33E−
08300
9,23300
2,143411,00
3002,37
711,00265
9,602544,00
0,94
Tab
leE
.14:B
uck
with
split
DC
bus
600V
-20
kH
z-
MP
CSSA
C
149
δ main
δ on
Ton
δ off
Toff
Vin,1
I in,1
Vin,2
I in,2
Pin
VR
I RPR
Vout
I out
Pout
η
[%]
[%]
[s]
[%]
[s]
[V]
[A]
[V]
[A]
[W]
[V]
[A]
[W]
[V]
[A]
[W]
[/]
551
2,67E−
071
6,85E−
0830
0,0
0,89
300
1,79
804,
030
02,
3570
5,0
41,3
1,51
62,3
60,
63
601
3,33E−
071
6,80E−
0830
0,0
1,61
300
1,49
930,
030
02,
3570
5,0
70,7
2,57
181,
700,
81
651
3,33E−
071
6,88E−
0830
0,0
2,41
300
1,30
1113,0
300
2,35
705,
099,2
3,59
356,
130,
87
701
3,73E−
071
6,83E−
0830
0,0
3,33
300
1,21
1362,0
300
2,35
705,
012
8,2
4,64
594,
850,
91
751
4,00E−
071
6,25E−
0830
0,0
4,34
300
1,22
1668,0
300
2,35
705,
015
6,5
5,66
885,
790,
92
801
4,00E−
071
6,87E−
0830
0,0
5,47
300
1,33
2040,0
300
2,35
705,
018
5,2
6,70
1240,8
40,
93
851
4,27E−
071
6,95E−
0830
0,0
6,74
300
1,55
2487,0
300
2,35
705,
021
4,9
7,77
1669,7
70,
94
901
4,27E−
071
6,50E−
0830
0,0
8,09
300
1,87
2988,0
300
2,35
705,
024
4,2
8,85
2161,1
70,
95
951
4,27E−
071
6,58E−
0830
0,0
9,21
300
2,25
3438,0
300
2,35
705,
026
4,4
9,58
2532,9
50,
93
Tab
leE
.15:
Buck
wit
hsp
lit
DC
bus
600
V-
30kH
z-
MP
CSSA
C
150 Appendix E. Measurements
δ Vin,1 Iin,1 Vin,2 Iin,2 Pin VR IR PR Vout Iout Pout η
[%] [V] [A] [V] [A] [W] [V] [A] [W] [V] [A] [W] [/]
55 300 0, 84 300 1, 74 774, 0 300 2, 34 702, 0 40, 8 1, 47 60, 0 0, 83
60 300 1, 29 300 1, 53 846, 0 300 2, 34 702, 0 59, 2 2, 13 126, 1 0, 88
65 300 2, 10 300 1, 26 1008, 0 300 2, 34 702, 0 88, 7 3, 20 283, 8 0, 93
70 300 3, 01 300 1, 10 1233, 0 300 2, 34 702, 0 118, 2 4, 27 504, 7 0, 95
75 300 4, 02 300 1, 05 1521, 0 300 2, 34 702, 0 147, 7 5, 34 788, 7 0, 96
80 300 5, 16 300 1, 11 1881, 0 300 2, 34 702, 0 177, 5 6, 42 1139, 6 0, 97
85 300 6, 40 300 1, 27 2301, 0 300 2, 33 699, 0 207, 3 7, 49 1552, 7 0, 97
90 300 7, 77 300 1, 54 2793, 0 300 2, 33 699, 0 237, 2 8, 59 2037, 5 0, 97
95 300 9, 20 300 1, 93 3339, 0 300 2, 33 699, 0 265, 3 9, 64 2557, 5 0, 97
Table E.16: Buck with split DC bus 600 V - 20 kHz - Silicon Carbide
δ Vin,1 Iin,1 Vin,2 Iin,2 Pin VR IR PR Vout Iout Pout η
[%] [V] [A] [V] [A] [W] [V] [A] [W] [V] [A] [W] [/]
55 300 0, 65 300 1, 87 756 300 2, 34 702 32, 0 1, 17 37, 4 0, 69
60 300 1, 34 300 1, 52 858 300 2, 34 702 60, 9 2, 20 134, 0 0, 86
65 300 2, 15 300 1, 26 1023 300 2, 34 702 90, 2 3, 25 293, 2 0, 91
70 300 3, 06 300 1, 12 1254 300 2, 34 702 119, 7 4, 31 515, 9 0, 93
75 300 4, 09 300 1, 06 1545 300 2, 34 702 149, 4 5, 38 803, 8 0, 95
80 300 5, 23 300 1, 13 1908 300 2, 33 699 179, 0 6, 45 1154, 6 0, 95
85 300 6, 48 300 1, 30 2334 300 2, 33 699 208, 7 7, 54 1573, 6 0, 96
90 300 7, 85 300 1, 57 2826 300 2, 32 696 238, 5 8, 64 2060, 6 0, 97
Table E.17: Buck with split DC bus 600 V - 40 kHz - Silicon Carbide
Appendix F
Data sheets
In this appendix, the most relevant parts of the data sheets are published. Not all the
used components are discussed. Two IGBTs and a SiC Power MOSFET are included since
their characteristics are important in the framework of this thesis.
151
Features
E
G
n-channel
C
=
!"#$%&' ( )* ($%&'+ ,-.,/ ++(*+ )$*'+"012
* )* ($%&'*(),-.,/ $%&'345(*6))
7/+89:;;1
TO-247AC
!
" # $%& # ' ( ) #* (+, ) - . / " +, "( $ 0" +, "( $ 0 1$2 3 4 $5
4 $67 -8*0-*8*0 97 9
+:60-+- * ;7<8*=< 9
>
5θ 2 . ??? ??? *@@
5θ 2 ( ??? ??? *@ A>5θ 4676 7 ??? *! ???
5θ 2' ;6 B$ ??? ??? !> >C ??? 08*9 ??? 8D9
www.irf.com 1
<+
2 www.irf.com
E .C89 F #0 - 'E . C89 F - ! 4)*#
E .C89 F G !! 1(B F !0 F
5 F - F '6# 177(B F G@ 65Ω ) F ! -0 B HH 14C& F *# F % %B*
1774C& F *G- F 2 4)*G66# 4C& F -*@- !*0
1(B F ! F 64)*6# 5 F - F '6#
177(B F ! F 65Ω ) F F B HH
4C& F @*! F 2 % %B*& F - F I + 7 $
$$ F # F 1$$ F F $) - 4)*@
5% 7$ F # F J*+ID (5% 5%B F 0- G 4)*
F 0 0 !#*' ("5% 5%B F !* #* ' 4)*
F 0* E (5% 5%BC F ! -# 4)*
F -- ## 0A'AK A ("57)75%B F -- F 'AK 4)*
( F # F @
! "#$
F F 6K'
ΔΔ $77*7 F *!- F A 6* ' 4 F *!- -* '
F *G@ F !' 4)*6F *!@ F '6
.C C -* F 0* 6K'ΔAΔ $77*7C C F F A 6K'
) 0 ! F 4 6' L. F F K' 6
F F 66 ()($ F *0 -*- #*' 4)*-
F *! -* #*'6 . & F F / ' /
% ! "#$
INSULATED GATE BIPOLAR TRANSISTOR WITHULTRAFAST SOFT RECOVERY DIODE
Features
Benefits
Absolute Maximum Ratings
Thermal Resistance
Parameter Min. Typ. Max. UnitsRθJC Junction-to-Case - IGBT ––– ––– 0.42RθJC Junction-to-Case - Diode ––– ––– 0.83 °C/WRθCS Case-to-Sink, flat, greased surface ––– 0.24 –––RθJA Junction-to-Ambient, typical socket mount ––– ––– 40Wt Weight ––– 6 (0.21) ––– g (oz)ZθJC Transient Thermal Impedance Junction-to-Case (Fig.24)
E
G
C
IRGP30B120KD-EMotor Control Co-Pack IGBT
TO-247AD
N-channel
www.irf.com 1
Parameter Max. UnitsVCES Collector-to-Emitter Breakdown Voltage 1200 VIC @ TC = 25°C Continuous Collector Current (Fig.1) 60IC @ TC = 100°C Continuous Collector Current (Fig.1) 30ICM Pulsed Collector Current (Fig.3, Fig. CT.5) 120ILM Clamped Inductive Load Current(Fig.4, Fig. CT.2) 120 AIF @ TC = 100°C Diode Continuous Forward Current 30IFM Diode Maximum Forward Current 120VGE Gate-to-Emitter Voltage ± 20 VPD @ TC = 25°C Maximum Power Dissipation (Fig.2) 300PD @ TC = 100°C Maximum Power Dissipation (Fig.2) 120TJ Operating Junction and -55 to + 150TSTG Storage Temperature Range
Soldering Temperature, for 10 seconds 300, (0.063 in. (1.6mm) from case)°C
Mounting Torque, 6-32 or M3 screw. 10 lbf•in (1.1N•m)
• Low VCE(on) Non Punch Through (NPT) Technology • Low Diode VF (1.76V Typical @ 25A & 25°C) • 10 μs Short Circuit Capability • Square RBSOA • Ultrasoft Diode Recovery Characteristics • Positive VCE(on) Temperature Coefficient • Extended Lead TO-247AD Package
• Benchmark Efficiency for Motor Control Applications • Rugged Transient Performance • Low EMI • Significantly Less Snubber Required • Excellent Current Sharing in Parallel Operation • Longer leads for Easier Mounting
VCES = 1200V
VCE(on) typ. = 2.28V
VGE = 15V, IC = 25A, 25°C
IRGP30B120KD-E
2 www.irf.com
Electrical Characteristics @ TJ = 25°C (unless otherwise specified)Parameter Min. Typ. Max. Units Conditions Fig.
V(BR)CES Collector-to-Emitter Breakdown Voltage 1200 V VGE = 0V,Ic =250 μA
ΔV(BR)CES / ΔTj Temperature Coeff. of Breakdown Voltage +1.2 V/°C VGE = 0V, Ic = 1 mA ( 25 -125 oC )
2.28 2.48 IC = 25A, VGE = 15V 5, 6
Collector-to-Emitter Saturation 2.46 2.66 IC = 30A, VGE = 15V 7, 9
VCE(on) Voltage 3.43 4.00 V IC = 60A, VGE = 15V 10
2.74 3.10 IC = 25A, VGE = 15V, TJ = 125°C 11
2.98 3.35 IC = 30A, VGE = 15V, TJ = 125°C
VGE(th) Gate Threshold Voltage 4.0 5.0 6.0 V VCE = VGE, IC = 250 μA 9,10,11,12
ΔVGE(th) / ΔTj Temperature Coeff. of Threshold Voltage - 1.2 mV/oC VCE = VGE, IC = 1 mA ( 25 -125 oC )
gfe Forward Transconductance 14.8 16.9 19.0 S VCE = 50V, IC = 25A, PW=80μs
250 VGE = 0V,VCE = 1200V
ICES Zero Gate Voltage Collector Current 325 675 μA VGE = 0v, VCE = 1200V, TJ =125°C
2000 VGE = 0v, VCE = 1200V, TJ =150°C
1.76 2.06 IC = 25A
VFM Diode Forward Voltage Drop 1.86 2.17 V IC = 30A 8
1.87 2.18 IC = 25A, TJ = 125°C
2.01 2.40 IC = 30A, TJ = 125°C
IGES Gate-to-Emitter Leakage Current ±100 nA VGE = ±20V
Switching Characteristics @ TJ = 25°C (unless otherwise specified)Parameter Min. Typ. Max. Units Conditions Fig.
Qg Total Gate charge (turn-on) 169 254 IC = 25A 23
Qge Gate - Emitter Charge (turn-on) 19 29 nC VCC =600V CT 1
Qgc Gate - Collector Charge (turn-on) 82 123 VGE = 15V
Eon Turn-On Switching Loss 1066 1250 IC = 25A, VCC = 600V CT 4
Eoff Turn-Off Switching Loss 1493 1800 μJ VGE = 15V, Rg = 5Ω, L=200μH WF1
Etot Total Switching Loss 2559 3050 TJ = 25oC, Energy losses include tail and diode reverse recovery
WF2
Eon Turn-on Switching Loss 1660 1856 Ic =25A, VCC=600V 13, 15
Eoff Turn-off Switching Loss 2118 2580 μJ VGE = 15V, Rg = 5Ω, L=200μH CT 4
Etot Total Switching Loss 3778 4436 TJ = 125oC, Energy losses include tail and diode reverse recovery
WF1 & 2
td(on) Turn - on delay time 50 65 Ic =25A, VCC=600V 14, 16
tr Rise time 25 35 ns VGE = 15V, Rg = 5Ω, L=200μH CT 4
td(off) Turn - off delay time 210 230 TJ = 125oC, WF1
tf Fall time 60 75 WF2
Cies Input Capacitance 2200 VGE = 0V
Coes Output Capacitance 210 pF VCC = 30V 22
Cres Reverse Transfer Capacitance 85 f = 1.0 MHz
TJ =150oC, Ic = 120A 4
RBSOA Reverse bias safe operating area FULL SQUARE VCC = 1000V, VP = 1200V CT 2
Rg = 5Ω, VGE = +15V to 0 V
TJ = 150oC CT 3
SCSOA Short Circuit Safe Operating Area 10 ---- ---- μs VCC = 900V,VP = 1200V WF4
Rg = 5Ω, VGE = +15V to 0 V
Erec Reverse recovery energy of the diode 1820 2400 μJ TJ = 125oC 17,18,19
trr Diode Reverse recovery time 300 ns VCC = 600V, Ic = 25A 20, 21
Irr Peak Reverse Recovery Current 34 38 A VGE = 15V, Rg = 5Ω, L=200μH CT 4, WF3
Le Internal Emitter Inductance 13 nH Measured 5 mm from the package.
1 C2M0080120D Rev. A
C2M0080120DSilicon Carbide Power MOSFET Z-FET
TM MOSFET
N-Channel Enhancement Mode Features
• High Speed Switching with Low Capacitances• High Blocking Voltage with Low RDS(on)• Easy to Parallel and Simple to Drive• Avalanche Ruggedness• Resistant to Latch-Up• Halogen Free, RoHS Compliant
Benefits
• HigherSystemEfficiency• Reduced Cooling Requirements• Increased System Switching Frequency
Applications
• Solar Inverters • High Voltage DC/DC Converters• Motor Drives• Switch Mode Power Supplies• UPS
Package
TO-247-3
Part Number Package
C2M0080120D TO-247-3
VDS 1200 V
ID @ 25˚C 31.6 A
RDS(on) 80 mΩ
Maximum Ratings (TC=25˚Cunlessotherwisespecified)
Symbol Parameter Value Unit Test Conditions Note
IDS (DC) Continuous Drain Current31.6
AVGS@20 V, TC =25˚C Fig. 16
20 VGS@20 V, TC =100˚C
IDS (pulse) Pulsed Drain Current 80 APulse width tP = 50 μs
duty limited by Tjmax, TC =25˚C
VGS Gate Source Voltage -10/+25 V
PtotPower Dissipation 208 W TC=25˚C Fig. 15
TJ , TstgOperating Junction and Storage Temperature -55 to
+150 ˚C
TLSolder Temperature 260 ˚C 1.6mm (0.063”) from case for 10s
Md Mounting Torque 18.8
Nmlbf-in M3 or 6-32 screw
2 C2M0080120D Rev. A
Electrical Characteristics (TC=25˚Cunlessotherwisespecified)
Symbol Parameter Min. Typ. Max. Unit Test Conditions Note
V(BR)DSS Drain-Source Breakdown Voltage 1200 V VGS = 0 V, ID=100μA
VGS(th)Gate Threshold Voltage
1.7 2.2V
VDS = 10V, ID = 1 mA
Fig. 83.2 VDS = 10V, ID = 10 mA
1.2 1.7V
VDS = 10V, ID = 1 mA, TJ = 150ºC
TBD VDS = 10V, ID = 10 mA, TJ = 150ºC
IDSS Zero Gate Voltage Drain Current1 100
μAVDS = 1200 V, VGS = 0 V
10 250 VDS = 1200 V, VGS = 0 V TJ = 150ºC
IGSS Gate-Source Leakage Current 0.25 μA VGS = 20 V, VDS = 0 V
RDS(on) Drain-Source On-State Resistance80 98
mΩVGS = 20 V, ID = 20 A
Fig. 6150 208 VGS = 20 V, ID = 20A, TJ = 150ºC
gfs Transconductance9.8
SVDS= 20 V, IDS= 20 A
Fig. 48.5 VDS= 20 V, IDS= 20 A, TJ = 150ºC
Ciss Input Capacitance 950
pFVGS = 0 V
VDS = 1000 V
f = 1 MHz
VAC = 25 mV
Fig. 13, 14
Coss Output Capacitance 80
Crss Reverse Transfer Capacitance 6.5
Eoss Coss Stored Energy 40 μJ Fig. 12
td(on)v Turn-On Delay Time 12.0
ns
VDD = 800 V, VGS = 0/20 V
ID = 20 A
RG(ext)=0Ω,RL=40Ω
Timing relative to VDS
Fig. 20tfv Fall Time 18.4
td(off)v Turn-Off Delay Time 23.2
trv Rise Time 13.6
RG Internal Gate Resistance 4.6 Ω f = 1 MHz, VAC = 25 mV
Built-in SiC Body Diode Characteristics
Symbol Parameter Typ. Max. Unit Test Conditions Note
VSD Diode Forward Voltage3.3
VVGS = -5 V, IF=10 A, TJ = 25 ºC
3.1 VGS = -2 V, IF=10 A, TJ = 25 ºC
trr Reverse Recovery Time 40 ns VGS = -5 V, IF=20 A, TJ = 25 ºCVR = 800 V, diF/dt= 350A/μs
Qrr Reverse Recovery Charge 165 nC
Irrm Peak Reverse Recovery Current 6.4 A
Thermal Characteristics
Symbol Parameter Typ. Max. Unit Test Conditions Note
RθJC Thermal Resistance from Junction to Case 0.60 0.65
K/W Fig. 17RθCS Case to Sink, w/ Thermal Compound TBD
RθJA Thermal Resistance From Junction to Ambient 40
Gate Charge Characteristics
Symbol Parameter Typ. Max. Unit Test Conditions Note
Qgs Gate to Source Charge 10.8
nCVDS = 800 V, VGS = 0/20 VID =20 APer JEDEC24 pg 27
Fig. 28Qgd Gate to Drain Charge 18.0
Qg Gate Charge Total 49.2
158 Appendix F. Data sheets
Bibliography
[1] G. Hua, C. Leu, Y. Jiang, and F. Lee, “Novel zero-voltage-transition pwm converters,”
Power Electronics, IEEE Transactions on, vol. 9, pp. 213–219, Mar 1994.
[2] C.-M. Wang, “Novel zero-voltage-transition pwm dc-dc converters,” Industrial Elec-
tronics, IEEE Transactions on, vol. 53, no. 1, pp. 254–262, 2006.
[3] C. M. De Oliveira Stein, H. A. Grundling, H. Pinheiro, J. R. Pinheiro, and H. L. Hey,
“Zero-current and zero-voltage soft-transition commutation cell for pwm inverters,”
Power Electronics, IEEE Transactions on, vol. 19, no. 2, pp. 396–403, 2004.
[4] M. Krogemann, The Parallel Resonant DC Link Inverter-A Soft Switching Inverter
Topology with PWM Capability. PhD thesis, University of Nottingham, 1997.
[5] I. Rectifier, “Irg4ph40udpbf - ultrafast copack igbt data sheet,” May 2011.
[6] A. Technologies, Avago ACPL-337J 4.0Amp gate drive optocoupler. Avago, 2014.
[7] M. Hermwille, IGBT Driver Calculation. SEMIKRON, 2007.
[8] D. Bozalakov, T. Vandoorn, B. Meersman, G. Papagiannis, A. Chrysochos, and
L. Vandevelde, “Damping-based droop control strategy allowing an increased pen-
etration of renewable energy resources in low voltage grids,” IEEE Transactions on
Power Delivery, vol. PP, no. 99, pp. 1–1, 2016.
[9] D. Bozalakov, T. Vandoorn, B. Meersman, C. Demoulias, and L. Vandevelde, “Voltage
dip mitigation capabilities of three-phase damping control strategy,” Electric Power
Systems Research, vol. 121, pp. 192–199, 2015.
[10] D. Bozalakov, T. Vandoorn, B. Meersman, and L. Vandevelde, “Overview of increasing
the penetration of renewable energy sources in the distribution grid by developing
control strategies and using ancillary services,” in Proceedings of the IEEE Young
Researchers Symposium, p. 5, EESA, 2014.
[11] International Rectifier, IGBT Characteristics, 2012.
[12] N. Mohan and T. M. Undeland, Power electronics: converters, applications, and de-
sign. John Wiley, 2003.
159
160 Bibliography
[13] L. Lasne, Electronique de puissance. Dunod, 2003.
[14] J. Pollefliet, Vermogenselektronica. Academia Press, 2003.
[15] P. Haaf and J. Harper, Understanding diode reverse recovery and its effect on switching
losses. FairchildSemiconductor Europe, 2007.
[16] A. van den Bossche, “Power electronics,” 2015. Course notes Ughent.
[17] J. Brown, Power MOSFET basics: Understanding gate charge and using it to assess
switching performance. Vishay Siliconix, 2004.
[18] B. Maurice and L. Wuidart, Drive circuits for power MOSFETs and IGBTs. STMi-
croelectronics, 1999.
[19] R. W. Erickson and D. Maksimovic, Fundamentals of Power Electronics, Second Edi-
tion. Springer, 2014.
[20] C. Basso, “Get rid of the miller effect with zero voltage switching,” November 2004.
[21] “The function of gate resistors in gate drive circuits.” http://irf.custhelp.com/
app/answers/detail/a_id/215. Accessed: 2015-10-24.
[22] ON Semiconductor, IGBT Gate Drive Considerations, 2011.
[23] B. Meersman, Regeling van driefasige invertorgekoppelde decentrale generatoren met
betrekking tot de verbetering van de netkwaliteit. PhD thesis, University of Ghent,
2012.
[24] B. Ozpineci, System Impact of Silicon Carbide Power Electronics on Hybrid Electric
Vehicle Applications. PhD thesis, University of Tennessee, 2002.
[25] A. Elasser and T. P. Chow, “Silicon carbide benefits and advantages for power elec-
tronics circuits and systems,” Proceedings of the IEEE, vol. 90, no. 6, pp. 969–986,
2002.
[26] A. Isurin and A. Cook, “Passive soft-switching snubber circuit with energy recovery,”
in Annual IEEE Conference on Applied Power Electronics Conference and Exposition
(APEC), 2008.
[27] M. Bellar, T. Wu, A. Tchamdjou, J. Mahdavi, and M. Ehsani, “A review of soft-
switched dc-ac converters,” Industry Applications, IEEE Transactions on, vol. 34,
pp. 847–860, July 1998.
Bibliography 161
[28] J.-S. Lai, “Resonant snubber-based soft-switching inverters for electric propulsion
drives,” IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, vol. 44, Febru-
ary 1997.
[29] D. M. Divan, “The resonant dc link converter-a new concept in static power conver-
sion,” IEEE Transactions on Industry Applications, vol. 25, March/April 1989.
[30] J. He, N. Mohan, and B. Wold, “Zero-voltage-switching pwm inverter for high-
frequency dc-ac power conversion,” IEEE Transactions on Industry Applications,
vol. 29, September/October 1993.
[31] J.-S. Lai, “Resonant snubber-based soft-switching inverters for electric propulsion
drives,” Industrial Electronics, IEEE Transactions on, vol. 44, pp. 71–80, Feb 1997.
[32] G. Hua and F. Lee, “Soft-switching techniques in pwm converters,” Industrial Elec-
tronics, IEEE Transactions on, vol. 42, pp. 595–603, Dec 1995.
[33] B. Backlund, Applying IGBTs - AN 5SYA 2053-04. ABB, 2004.
[34] F. Sargos, IGBT Power Electronics Teaching System Principle for sizing power con-
verters AN-8005. Semikron, 2008.
[35] A. van den Bossche and D. Bozalakov, “Two channel high voltage differential probe
for power electronics applications,” pp. 1–6, 2013.
[36] A. Van den Bossche and V. C. Valchev, Inductors and Transformers for Power Elec-
tronics. CRC Press, 2005.
[37] D. U. Nicolai and D. A. Wintrich, AN1403: Determining switching losses of Semikron
IGBT modules. Semikron, 2014.
[38] G. Ortiz, H. Uemura, D. Bortis, J. W. Kolar, and O. Apeldoorn, “Modeling of soft-
switching losses of igbts in high-power high-efficiency dual-active-bridge dc/dc con-
verters,” Electron Devices, IEEE Transactions on, vol. 60, no. 2, pp. 587–597, 2013.
[39] M. C. Cavalcanti, E. R. Da Silva, D. Boroyevich, W. Dong, and C. B. Jacobina, “A
feasible loss model for igbt in soft-switching inverters,” in Power Electronics Specialist
Conference, 2003. PESC’03. 2003 IEEE 34th Annual, vol. 4, pp. 1845–1850, IEEE,
2003.