Enhancing power density and efficiency of variable speed drives with 1200V SiC T-MOSFET

Transcription

1 Enhancing power density and efficiency of variable speed drives with 12V SiC T-MOSFET Benjamin Sahan, Infineon Technologies AG, Germany, Anastasia Brodt, Infineon Technologies AG, Germany, Daniel Heer, Infineon Technologies AG, Germany, Ulrich Schwarzer, Infineon Technologies AG, Germany, Maximilian Slawinski, Infineon Technologies AG, Germany, Tim Villbusch, Infineon Technologies AG, Germany, Klaus Vogel, Infineon Technologies AG, Germany, Abstract Employing SiC T-MOSFETs to variable speed drives leads to a significant reduction in power losses even though dv/dt is limited to 5 kv/µs in order to use standard motors. This loss reduction enables lower system costs by reducing the cooling effort and thus downsizing the heatsink as well as lower operating costs due to higher power conversion efficiency. For a 22 kw inverter the heatsink volume could be reduced by more than 5 % and the calculated energy savings are more than 1 kwh per year for a typical load profile of a pump. The calculations were verified by an experimental inverter test. 1. Introduction More than 5 % of the electric energy generation in Europe is consumed by industrial motors [1]. Variable speed drives (VSD) enable considerable power savings compared to fixed speed operation with mechanical throttling. On the other hand the application of an inverter creates initial investment costs, space requirements and cooling efforts compared to a direct connection of the motor to the grid. As silicon based power semiconductors are expected to saturate in terms of electrical performance and power density, power switches based on wideband gap materials such as silicon carbide (SiC) are becoming the focus of interest for several applications to enhance power density and efficiency [2], [3], [4]. In this paper the usage of the novel CoolSiC Trench-MOSFET (T-MOSFET) was investigated and discussed for a 22 kw variable speed drive. The switching transistor voltage slope was limited to 5 kv/µs in order to use standard motors as described in [2]. The following application parameters for typical variable speed drives (VSD) were used: Table 1: Virtual inverter parameters Parameter Nom. ambient temperature range Max. heat sink temperature Nom. DC voltage Nom. RMS output current Nom. motor power Nom. RMS motor voltage Value T a, nom = 5 C T h,max = 9 C V DC = 7 V I N,out = 44 A P N = 22 kw V N = 4 V Nom. power factor cos(φ) N =.85 Nom. output frequency Max. switching slope (.1V DC -.9V DC ) f out,n = 5 Hz dv/dt < 5 kv/µs Most VSDs are equipped with a diode rectifier as depicted in Fig 1. For reasons of simplicity this paper focuses on the inverter stage only. Hence there will be a direct comparison between Si IGBT and SiC T-MOSFET. Fig. 1. Sketch of a diode rectifier (with optional chokes) and the investigated SiC T-MOSFET inverter 196

2 Three half-bridge modules in Easy1B package were used for the CoolSiC MOSFET (FF11MR12W1M1_ENG, nominal R DS,on = 11mΩ). The IGBT reference module was the FS1R12KT4 in the Econo2 package. Both modules have a comparable chip current rating of 1 A. 2. Properties of SiC T-MOSFETs 2.1. General properties A novel 12 V SiC T-MOSFET has been introduced in [5] featuring high switching performance, low specific R DS,on combined with a highly reliable gate oxide. Compared to other SiC based transistor solutions, the presented CoolSiC MOSFET copes with standard and proven +15 V/-5 V gate drive circuitries wellknown from Si IGBT based converter set-ups. There is no need of developing novel and special gate drivers. Furthermore, the internal body diode of the CoolSiC MOSFET can be operated as freewheeling diode (FWD) in the half-bridge with interlock time. The interlock times for SiC are typically a few hundred ns long and therefore the higher forward voltage drop can be accepted. Further advantages of SiC T-MOSFETs in comparison to recent Si IGBTs are the electrical characteristics. Due to the unipolar structure of a MOSFET, no minority charge carriers are involved during switching processes. This leads to significantly reduced switching losses of both transistor and diode. Additionally the absence of a knee voltage together with a low on-state resistance results in strongly reduced conduction losses, especially at partial load. Calculation of reduced inverter power losses and dimensioning of a down-sized heatsink using CoolSiC MOSFETs will be shown below Adjustable switching speed dv/dt Motors supplied with typical PWM voltage signals from VSDs will experience higher stress in their isolation systems as well as in its bearings due the pulse rise time (t r ) or equivalently dv/dt of PWM [2]. pulses. The use of long motor cables leads to even higher peak voltages at the motor terminals. Therefore, motor manufacturers recommend a maximum dv/dt of approximately 5kV/µs for 4V motors. If the switching speed of the power semiconductors is exceeding the recommended limit, measures are needed to slow the devices down. For Trench-IGBTs the switching speed dv/dt during turn-on rises with decreasing load current and junction temperature. A common way to select a proper gate resistance regarding the switching speed is therefore the consideration of the turn-on event at one tenth of the nominal module current and a junction temperature of 25 C. The switching speed during turn-off rises with increasing load current and therefore turn-off gate resistance is determined at the nominal current. Fig. 2 shows the switching speed and the switching losses of the IGBT reference module FS1R12KT4 in dependency of the gate resistance. The dv/dt was measured directly at the output terminals of a laboratory test inverter. Swichting energies [mj] R G [Ω] E on E off E rec dv/dt on dv/dt off Limit Fig. 2. dv/dt of FS1R12KT4 at 25 C and 1 A (turnon) and 1 A (turn-off) and switching energies at 15 C, 6 V, 1 A in dependency of R G with V GE =15V/-5V A valuable advantage of the CoolSiC MOSFET is the controllability of the switching speed during turn-on as well as turn-off event by simply adjusting the gate resistance R G, as demonstrated in [5]. Fig. 3 shows the switching speed in dependency of the gate resistance and the switching losses of the chosen SiC T- MOSFET module. Again, the dv/dt was measured directly at the output terminals of a laboratory test inverter. The switching losses were measured in a double pulse test. It should be noted that all measurements were based on engineering samples (ENG) dv/dt [kv/µs] 197

3 Swichting energies [mj] R G [Ω] E on E off dv/dt on dv/dt off Limit Fig. 3. dv/dt of FF11MR12W1M1_ENG at 15 C and 1 A (turn-on) and 1 A (turn-off) and switching energies at 15 C, 6 V, 1 A in dependency of R G (preliminary) with V GS =15V/-5V To ensure a fair comparison of the two device technologies in inverters for standard industrial motors, in the first step the gate resistances were adjusted to achieve a maximum switching speed of 5 kv/µs under the discussed conditions. For the reference IGBT module the R G,on = 1.6 Ω for turn-on and R G,off = 4 Ω for turn-off was chosen. For the intrinsically faster SiC T-MOSFET module higher gate resistances of R G,on = 15 Ω and R G,off = 3 Ω were needed. 3. Power losses and heatsink selection 3.1. Power loss calculation The power losses per switch shown in Fig. 4 were calculated considering the discussed gate resistance values. The calculations were based on measured switching energies in a standard double pulse test. Even at the same dv/dt of 5 kv/µs the strongly decelerated CoolSiC MOSFET had more than 5% lower switching losses compared to the IGBT as can be seen from Fig. 2, 3 and Fig. 4. This is due to the fact that the MOSFET has no tail current and the diode has no reverse recovery current. The conduction losses were also reduced by around 4 % due to the lower on- state resistance R DS, on compared to the saturation voltage of the Si IGBT and Si FWD, leading to an overall reduction in inverter power losses by 53 % at a PWM switching frequency of 4 khz and 58 % at 8 khz. dv/dt [kv/µs] Power losses switch [W] R h,max thha,max TMOSFET 4kHz vjop T = IGBT 4kHz TMOSFET 8kHz h,max loss T P loss a thjh IGBT 8kHz Pcond,D Psw,D Pcond,T Psw,T Fig. 4: Loss calculation per switch P loss at nominal output current at 4 khz and 8 khz SPWM at T a = 5 C 3.2. Heatsink selection In the vast majority of low voltage drives forcedair cooled aluminum heatsinks are employed. A maximal thermal resistance R thha,max of heatsink to ambient can be approximated with the following equation using the power losses given in Fig. 4: The maximum heatsink temperature was limited to 9 C. Guidelines for maximum heatsink temperatures are given in standard EN If such limitation does not apply the maximum stationary heatsink temperature T h,max can be easily determined by: T = T P R where T vjop is the maximum junction temperature of the transistor or diode under switching conditions and P loss is the loss of a single transistor and diode. Using the sum of power losses from Fig. 4 the maximum thermal resistance for 4 khz operation was R thha,max_igbt =.11 K/W for the Si IGBT and R thha,max_sic =.24 K/W for the SiC T-MOSFET. A sketch of the arrangement of the modules can be found in Fig. 5. It should be noted that the three SiC half-bridge modules were arranged in a row to allow symmetrical current flow [1]. Table 2 shows the parameters of the selected heatsinks. Off-the-shelf extruded heatsinks with hollow fins and integrated fans were selected. It can be observed from the datasheet values R thha,ds [6] LA6 fitted quite well to the calculated R thha,max criteria for the SiC inverter and LA14 matched with the calculated R thha,max for the Si inverter. The next smaller heatsink LA 9 did not fit. 198

4 4. Experimental results Fig. 5: Top view arrangement of power modules and heatsinks; left) SiC T-MOSFET inverter right) Si IGBT inverter As the fans have standardized dimensions (4, 6, 8, 12mm) the heatsink does not match exactly to the calculated value and thus the volume does not scale linearly with the value of R thha. Afterwards the R thha for the selected heatsinks were measured with power modules mounted on top Heatsink temperature To verify the simulated results a laboratory test inverter was build up. Fig. 6 shows the setup of the heatsinks with mounted power modules on top. The IGBT inverter was operated with the gate driver EiceDRIVER 1EDI6I12AF and the SiC T-MOSFET inverter used the driver 1EDI6N12AF. The inverters were run in a back-to-back configuration which allows operation at different power factors and output voltages. As the influence of these two parameters on the power loss and junction temperature of the SiC T- MOSFET is almost negligible the SiC inverter was used as a load and the IGBT inverter as device under test (DUT). Table 2: Selected heatsink parameters, measured R thha and R thha,ds from heatsink datasheet (volume without fan) Heatsink LA6 with MOSFET 62x74x17 LA 9 with IGBT 8x83x15 LA14 with IGBT 12x12x15 R thha,ds [K/W] R thha [K/W] Volume [L] The heatsink temperatures were measured directly underneath the two hottest chips with a drilling of 1 mm below the surface of the heatsink. A conventional Typ K thermoelement was used. Impressing a defined DC power into the module and measuring the temperature difference between heatsink and ambient the effective R thha could be determined. As can be seen in table 2 this value is actually better than the datasheet value. All in all, it is obvious that the heatsink volume can be significantly reduced by using SiC T-MOSFETs. Fig. 6: Setup of the heatsinks and power modules The heatsink temperature was measured underneath the hottest chip during inverter operation. The transient heatsink temperatures for the nominal operating condition can be found in Fig. 7. Even though the size of heatsink of the SiC inverter was much smaller, it was still overdimensioned since the temperature was even lower compared to the heatsink temperature of the Si inverter due to drastically reduced losses. Looking at the transients the time constant τ thha of the heatsink is a product of the R thha and the thermal capacitance C thha which is proportional to the weight of the heatsink. 199

5 Fig. 7: Measurement of heatsink temperatures T h, T a = 24 C, f sw = 8 khz, I Nout = 44 A, cos(φ) =.85 The SiC inverter has a much lower weight but the value of R thha is also higher. Therefore the time constants are similar Junction temperature To measure the junction temperature during inverter operation modules without silicone gel and black colored DCBs were prepared. A small hole was cut into the module and the PCB to measure the chip temperature with an IR camera. Special care needs to be taken with EasyPACK modules due to the pressure distribution. Moreover, due to the absence of the gel the DC voltage must not exceed certain voltages. An operation at 4 V DC and 7 khz was chosen to have comparable switching losses as in nominal operation at 7 V and 4 khz. Fig. 8 shows the measured and simulated stationary junction temperatures for the CoolSiC MOSFET and Si IGBT in dependency of the output rms current. The measurement was done at the hottest chip in the system. As the test was conducted at room temperature the maximum junction temperature at T a = 5 C will be around 25-3 C higher. As can be seen from the IR picture in Fig. 8 the chip temperature at the nominal output current of 44A was 6 C for the CoolSiC MOSFET and 75 C for the Si IGBT both at ambient room temperature Power losses The measurement of power losses was based on electro-thermal properties of the inverter similar to the method in [9]. Fig. 8: Junction temperature measurements with V DC = 4 V, 7 khz, T a = 24 C Using the measured thermal resistance of heatsink to ambient R thha as described before, the losses during inverter operation could be determined using the measured heatsink temperature below the chip and ambient temperature. Fig. 9 displays the measured power losses at T a = 24 C and the simulated ones. It can be concluded that loss simulation and measurement match very well. Ploss [W] IGBT measured IGBT simulated MOSFET measured MOSFET simulated 4 8 fsw [khz] Fig. 9: Loss comparison at nominal output current at 4 khz and 8 khz SPWM at T a = 24 C 5. Efficiency 5.1. Standard EN For VSD the regulations of the Ecodesign directive in Europe adopts the requirements of the EN standard. This standard defines the efficiency classes for motor systems [2]. The concepts introduced are: The Complete Drive Module (CDM) is consisting of the VSD and the other components installed at the mains supply side, such as line inductors and 2

6 EMC filters. At the motor side, output inductors, du/dt filters and motor cables are considered. The EN defines the different efficiency classes to be used for a CDM. Regarding losses, a reference CDM (RCDM) is considered. These losses are taken at the operation point 9 %, 5 % and % of rated output frequency (alternatively 5 Hz) and 1 %, 5 % and 25 % of the torque-producing current (see Fig. 11) where the testing current can differ from the rated output current of the actual inverter. To achieve the higher efficiency class IE2, the CDM needs to have 25 % lower losses compared to the RCDM IE1. As this paper focuses on the power losses of inverter stage only a classification was out of scope. But the operating points defined in EN were taken as a reference for the following energy saving calculations Measurements The measured savings in dependency of the relative output current and frequency are plotted in Fig 1. If a drive would run one year at nominal torque and nominal speed the energy savings would be 2 W*876 h = 1752 kwh per year. On the opposite, if the drive runs at lowest speed and lowest torque savings would be 5 W * 876 h = 438 kwh p.a. Torque prodcing current Output current 1% 75% 5% 25% % % 25% 5% 75% 9% 1% Speed 1% 75% 5% 25% % % 25% 5% 75% 1% Torque producing current Pump Fig. 11: Centrifugal pump torque speed characteristic and measurement points according to EN kW class Fig. 12: Relation between torque producing current and output current for 22 kw inverter according to EN Fig. 1: Measured savings for the SiC inverter at different operating points according to EN Pumps and fans are based on the centrifugal working principle. The resulting square torquespeed curve leads to a drastic reduction of active power at lower speeds. Fig. 11 shows the most relevant operating points for a pump (highlighted dots). However, due to the high reactive magnetizing current needed for standard asynchronous motors the inverter current can still reach more than 3 % of the nominal current even at low torque as can be seen from Fig 12. EN defines the relation between torque producing current and output current according to Fig. 12. Combining Fig. 11 and Fig. 12 the overall energy saving can be approximated using different flow rate profiles according to [7]. The results are given in Table 3. While for IGBTs the losses are almost proportional to the current, thanks to its linear onstate characteristic, the on state losses of SiC T- MOSFETs reduce with the power of two. In particular regarding partial load efficiency this gain can be directly transferred into lower operating costs. The calculated energy savings were between 1489 kwh and 882 kwh per year depending on the load profile. That equals about 149 to 82 for typical industrial consumers (.1 /kwh). Assuming cabinet based converters the amount would even increase if the energy consumption of the air conditioning of the control room would be also considered [8]. Next to the reduction in heatsink size this leads to a further decrease in system cost when using SiC switches. 21

7 6. Summary Several advantages by the use of SiC based switches instead of Si based IGBT devices in variable speed drive converters have been demonstrated within this work. Due to the lower static and dynamic losses of CoolSiC MOSFET devices a power loss reduction of more than 5 % compared to a Si IGBT solution at the same dv/dt and power rating could be achieved. Operating two laboratory test inverters in a backto-back operation could not only verify the results but also demonstrated a good matching between simulations and measurements. This significant loss reduction was not only a consequence of the loss reduction at full load conditions but in particular of the superior electrical performance of SiC T-MOSFET based devices at partial load conditions. It has been shown that SiC T-MOSFET based VSD would not only save operational costs due to less power consumption during operation. The possibility of reducing the volume and weight of the converters heatsink is potentially reducing system costs and leading to a higher system power density. This paper is showing a clear guide how to use the superior electrical performance of CoolSiC MOSFETs in VSD to achieve better system performance and economics. [4] M. Slawinski, T. Villbusch, D. Heer, M. Buschkuehle "Demonstration of superior SiC MOSFET Module performance within a Buck- Boost Conversion System, PCIM 216 [5] D. Heer, D. Domes, and D. Peters Switching performance of a 12 V SiC-Trench-MOSFET in a low-power module, PCIM Europe 216 [6] [7] [8] U. Jansen, U. Schwarzer Thermische Auslegung von Umrichtern und Schaltschränken in der Antriebstechnik, Elektronik Power Juni 215 [9] J. Rabkowski, D. Peftitsis, and H. P. Nee, Design steps towards a 4-kVA SiC inverter with an efficiency exceeding 99.5 %, in Proc. 27th Annu. IEEE Applied Power Electronics Conference, 212 [1] R. Bayerer, D.Domes Power circuit design for clean switching, CIPS 21 Acknowledgements The authors would like to thank J. Baurichter, C. Backhaus, M. Buschkühle, A. Herbrandt, D. Hoffmann, A. Lenze, U. Jansen, M. Prell, A. Schmal, C. Urban, O. Wette for their valuable advice and support. References [1] Study for an update of the Ecodesign Working Plan Amended Ecodesign Working Plan for the European Commission, Brüssel/Delft, 211 [2] K. Vogel, A. Brodt, A. Rossa Improve the efficiency in AC-Drives: New semiconductor solutions and their challenges, EEMODS 216 [3] M. Schulz et al Pushing Power Density Limits using SiC-JFet-based Matrix Converter, PCIM

8 Appendix Table 3: Measured operating points for a virtual 22 kw pump inverter Speed Torque I out [A] cos(φ) V ph [V] P Igbt [W] P SiC [W] Savings[W] Eta IGBT Eta SiC 5 % 25 % % % 75 % (9 %) 5 % % % 9 % 1 % % % Table 4: Typical pump load profiles [7] Flow rate/speed Time occurrence Savings [kwh] p.a. Profile A Profile B Profile C Profile A Profile B Profile C -5 % 37.5 % 1 % 7 % % 25 % 15 % 15 % % 37.5 % 75 % 15 % Sum