Comparative performance of SiC 1200V MOSFETs vs. comparably-rated silicon MOSFETs and IGBTs at 30kHz and 100kHz+ using a SEPIC demonstration
Comparing the performance of 1200V silicon carbide (SiC) MOSFETs with 1200V silicon MOSFETs and IGBTs at high frequency is critical to establish the difference in device losses and power-handling capabilities between the technologies. Using a demonstration platform that can be set to emulate a given set of parameters (e.g., switch voltage, current, frequency and case temperature) that the devices will face at high power under hard-switched conditions, the difference in losses between the device technologies can be characterized by comparing DC input power.
By Bob Callanan (Applications Manager, SiC Power Devices) and Julius Rice (Applications Engineer), Cree, Inc.
Because it is not practical to build separate platforms (retrofitted with SiC devices) for each potential application, (including grid connected solar inverters, power factor correction circuits, motor drives, interruptible power systems, etc.), it was necessary to design a demonstration platform to overcome this limitation. A transformerless DCDC converter platform with a single ground-referenced switch affords the capability of recirculating the load current back to the input link, eliminating the inherent limitations of a transformer-based platform (prohibitively high losses through proximity and skin effects of the transformer windings at high frequencies).
The single-ended primary inductor converter (SEPIC) provides the ability to buck/boost without inverting the output voltage, thus making it an ideal platform to compare relative device performance between SiC MOSFETs and comparably-rated conventional silicon MOSFETs and IGBTs.
Comparison of the devices under test
The comparison included the best available examples of a 1200V silicon MOSFET (Microsemi APT28M120B2)1, the highest current 1200V silicon MOSFET in a TO-247 package available (its maximum current rating at 100°C is similar to that of the SiC MOSFET); a 1200V silicon IGBT (Infineon IGW40N120H3)2, a 1200V/40A trench and field-stop device with a forward voltage of 20A, closely matching the SiC MOSFET; and finally, the 1200V SiC MOSFET (Cree’s CMF20120D)3, which has the lowest gate charge with a gate voltage f 20V. All devices were housed in TO-247 plastic packages (see Table 1 for comparative specifications).
SEPIC demonstrator design platform
The SEPIC topology chosen for the demonstration platform is a simple design with a buck/boost characteristic, allowing the converter’s output current to be re-circulated back to the input while the device under test (the switch) is operating at a duty cycle slightly higher than 50%. SEPIC converter topology is not widely used for high power operation (largely due to high switch stress); however, it is an ideal platform for this demonstration. In this converter, the voltage across the switch is twice the input voltage, and the current through the switch is twice the output current. Thus, the input DC supply has to provide just half of the desired switch voltage. Furthermore, because the switch is referenced to ground, precise measurements of voltage and current are greatly simplified.
As shown in Figure 1, the SEPIC platform schematic consists of a switch (the device under test, or DUT); a diode (D1); a blocking capacitor (C1); output capacitor (C2); and two inductors (L1 and L2). The output current is fed back into the input source (VIN) as shown by the arrows. The controls consist of a simple peak current mode controller, which is fed from a current transformer to sample the drain of the device under test. Input for the current mode controller is supplied from an error amplifier used to regulate the (re-circulated) output current; which in turn is measured by a Hall-effect sensor. Diagnostics include a high frequency current-viewing resistor to observe the DUT current and a Kelvin-connected voltage probe to measure the DUT voltage. The gate driver is isolated to prevent a ground loop being formed around the currentviewing resistor.
SEPIC Demonstration hardware
The SEPIC hardware is shown in Figure 2. A lower rack assembly houses the high voltage and logic power supplies along with the SEPIC inductors. The remainder of the converter and control board are located above the rack assembly in an acrylic enclosure. A monitor displays voltage and current waveforms for the device under test, while four digital meters display input voltage, total system loss, delivered current and delivered power.
The SEPIC demonstration unit facilitates the ability to change out devices under test by mounting the switching devices on separate connectorized daughter-boards with heat sinks, and then connecting them to the main power board.
Comparison results at 30kHz
The first results from the SEPIC demonstrator plot the input power vs. the switch current for each of the switches under test. Since they are tested under identical conditions with identical circuits, the difference in input power between switches is a direct measurement of switching losses.
To simplify the comparative data, the input power vs. peak switch current for the CMF20120D SiC MOSFET was used as a baseline, and this data was subtracted from the input power vs. peak power switch current of the other devices. The result shows the increased switch losses of one device under test vs. another.
A plot of the system input power loss relative to the SiC MOSFET vs. peak switch current is presented in Figure 3. The input voltage to the SEPIC demonstrator was 400V, resulting in a switch voltage of 800V. The test was terminated when the thermal design limits were exceeded.
The silicon carbide MOSFET (CMF20120D) achieves the highest switch current (25A) with delivered power of 5kW. The silicon IGBT IGW40N120H3) reaches a maximum switch current of 12A, for delivered power of 2.4kW, with the device’s higher switching losses being the limiting factor. The silicon MOSFET (APT28M120B3) reaches a maximum switch current of just 10A, or delivered power of 2kW; however, the device’s higher conduction losses limit its efficiency and maximum current.
The test results also demonstrate the devices’ efficiency instead of relative power loss. The switches under evaluation all have fairly low losses, while the inductors as part of the SEPIC demonstrator have significant (but predictable) I2R losses. Thus, while it is difficult to see performance differences between switches in a straight efficiency plot using the combined input and output power, this can be mitigated by subtracting inductor losses from the total system loss.
A plot of switch efficiency vs. peak switch current is show in Figure 4, which clearly shows the SiC MOSFET is the highest efficiency switch that can deliver 5kW of power. The silicon IGBT delivers less than half the power (2.4kW), and has efficiency approximately 1.2% lower than the SiC MOSFET. The silicon MOSFET, while its peak efficiency is only about 0.25% lower than the SiC device, delivers the least amount of power (2kW), and because of its high conduction losses, its efficiency drops rapidly as current is increased.
Comparison results at 100kHz
Testing at 100kHz compared only the SiC MOSFET and the silicon MOSFET, since the silicon IGBT switching losses were so high that hard-switched operation at this frequency was impractical.
Input power losses (system-wide) relative to the SiC MOSFET vs. peak switch current is presented in Figure 5. Once again, input voltage was 400V, resulting in a switch voltage of 800V, and the test was terminated when thermal design limits were exceeded. The results demonstrate that at the higher frequency, the SiC MOSFET achieved the highest switch current (17A) with delivered power of 3.6kW, while the silicon MOSFET reached a maximum switch current of only 8.5A with delivered power of just 1.7kW. It should be noted that below 1.5A, the silicon device exhibited lower losses than the SiC MOSFET (indicated by the negative relative loss in Figure 5), which is attributable to the silicon MOSFET’s higher transconductance. However, the silicon device’s higher on-resistance causes its conductance losses to increase dramatically as the current increases, thus causing higher power dissipation than the SiC device for the majority of the test currents.
The efficiency measured at 100kHz is shown in Figure 6. Once again, the magnetics losses have been subtracted to make it easier to compare switching device performance. The results demonstrate that the SiC MOSFET achieves the highest overall efficiency, with delivered power of 3.6kW at an 18A peak current. The silicon MOSFET reaches a maximum of only 8.5A with delivered power of 1.7kW, and it does not achieve higher overall efficiency than the SiC MOSFET (except when current is less than 2A).
Loss performance and power handling capability of Cree’s silicon carbide 1200V MOSFET was compared with an Infineon 1200V silicon IGBT and a Microsemi 1200V silicon MOSFET, using a SEPIC converter demonstration platform at high power under hard-switched conditions to emulate actual power applications. Tests conducted at 30kHz and 100kHz with 800V across the switching devices demonstrate that the SiC MOSFET can process twice the current with approximately 40W lower loss than either of the silicon devices.
2) a) http://www.infineon.com/dgdl/DS_IG40N120H3_1_1_final.pdf
3) Cree CMF20120D datasheet link: www.cree.com/products/pdf/CMF20120D.pdf