ON Semiconductor Corporation (ON) Earnings Call Transcript & Summary

Good morning, everyone, and welcome to today's webinar on novel system-level simulators for power application in industrial and automotive domain. I'm Kyli Miller with onsemi, and I'll be your moderator today. In today's webinar, we'll introduce you to various simulation tools. We'll see how we qualify and calibrate these new online tools. We'll do a comparison between simulation results and actual measurements. Finally, we'll review a design and simulation with the new silicon carbide MOSFET generation to analyze and benefit of this new generation before the conclusion. At the end of the webinar, we'll be holding a Q&A session to answer any questions you may have. You can type your questions into the Ask Questions box on your right. A recording of the webinar will be shared with you via e-mail. Now let's meet today's presenter. Didier Balocco is EMEA Power Solution Group Business Marketing Engineer. He has extensive experience in power electronics, including DC/DC and AC/AC converters design and industrial applications. He published more than 10 papers on power electronics and holds 1 patent. Now without further ado, let's get started with our webinar. Didier, the floor is yours.

Thank you, Kyli. Thank you for the introduction. So let's start this webinar and see how we have calibrate and validate this new online simulator. So before, this is the agenda. We will do an introduction about this new Elite Power Simulator and the PLECS Model Generator and compare that to SPICE. Then we will use the traction inverter tester that we have developed to analyze and qualify the tool. We will do some -- I will describe the tester setup itself, describe the measurement and the calibration we have done, and then compare that to the Elite Power Simulator result. So the second example, we will use the 25 fast EV charger demo. I will describe the demo, compare the measurement with the tool we have developed and with the Elite Power Simulator. And before conclusion, I will do -- I will show you the results with the new generation, our new M3S silicon carbide. So let's look at what is this new tool, what I call a 2-in-1. It includes the Elite Power Simulator and the Self-Service PLECS Model Generator. So the tool is, let's say, can be explained like this. On the left side, we have the Self-Service PLECS Model Generator that communicate, the arrow in the middle, with our Elite Power Simulator. So the left side produce a PLECS Model that we can send to the simulator itself, the topology simulator itself to get the result of this topology. So the left part of the tool is based on our unique expertise in SPICE modeling that we call physical and scalable modeling, and it's a SPICE engine that generates the PLECS model. Then when you send the model to PLECS, it's the PLECS online engine that generate the results for the topology you are analyzing. Sometimes instead of saying Self-Service PLECS Model Generator, I may use the acronym SSPMG. So SSPMG refer to the left side of the tool. So our model are very unique in the sense that SPICE has been created in the '70s and mainly made for designing ICs where people use lateral MOSFET. In power, we use vertical MOSFET, and we have a technique like trench-gate or field-stop player or p-pillar or -- and MOSFET are vertical. So for that, we have created a new set of equations modeling each zone in the MOSFET, like the trench gate if we use that, the field-stop player, and so on, and relate those equation also to the layout of the die and connect those equations to the electrical schematic we see on the top right of this slide. And doing so, we can reproduce the performances and the exact performances of the device. And it makes much more sense with silicon carbide, that is a new material, not having the same behavior as silicon. So we say it's physical because it includes the physical equation of the matter. And scalable because we have -- we create one core model. And when we need to have the 22 medium, for example, or the 40 or the 60 medium MOSFET, we just give the size of the die to that model. So we scale somehow the model. Those models are then verified in our lab to be sure that they fit the measurement. So you can see some comparison between measurement and simulation results on the bottom of the slide. If I start from the left to the right, on the left -- the extreme left, I have the, for example, the On region. And you can see that on this On region, as the gate voltage increase, the current increase, but also the self-heating increase. And then -- and the slope of the On region become more and more smoother as we increase the gate voltage because of the self-heating. On the middle of the graph, you can see the output capacitor, the reverse or middle capacitor and the input capacitor. When the depletion zone reach a new doping zone, this is the case with our MOSFET, we have 2 doping zones, the epi and the substrate, there's a sudden change of the capacitor that occur around 10 volt. And you can see that this sudden change is perfectly captured by the model. But those models are not only made to extract parameter from data sheet, reproduce the data sheet, they are made to be used in a real application, in a real system. This is what you see on the right graph where you see a turn-On of one silicon carbide MOSFET, you see on dark-blue the current. And so the over-current turn-On lead to the reverse recovery of the MOSFET in [ average ] structure, for example, for here. You see in red the drain voltage, you see that the drain voltage going down slowly and then very fast when the drop in voltage reach the zone where the capacitor change, changing the doping zone. This is captured by the model. And you can see that on the pink, we have the pink waveform, we also capture the effect of the gate impedance network with a deep of the gate voltage when we cross the Miller Plateau. So the -- actually, PLECS model are obtained by measurement or picking the data on the data sheet. But as you may know, making measurement takes time. So taking -- making a PLECS model by measurement will take time. And actually, we have very few points in the data sheet. So we will have a very weak model. So to solve that, we, when we try to approach this issue, we are facing 4 major problems that we want to solve. First of all, the switching losses are dependent on the parasitic of the measurement set. In our measurement test setup in our lab, we have almost perfect [ meter ] and perfect capacitor and perfect layout. Connection and switching energy data are very limited in the data sheet of what we do in our lab. And so this makes very inaccurate interpolation and extrapolation in PLECS. Losses data are based on the nominal process. And as we all know, when we manufacture something, there's variation. And switching loss data are valid for our switching only and not valid for self-switching. People think in self-switching there's no losses. Yes, losses are very low but not 0. It's like when you charge a battery, you put an energy in, but you never get out the same amount of energy. There are some losses in this one trick due to the resistor and the parasitic element in the path. So it's the same here. When you change -- you exchange the energy between 2 capacitors, this is the purpose of self-switching, you have resistor in the path that creates losses. And we will capture those losses. So that's why onsemi came with this Self-Service PLECS Modern Generator to provide -- to solve those 4 issues. And you get a model that will be really adequate for your board and your application, and not using a model that will give you reason that you will not switch. You can use the model generator at onsemi.com/self-plecs-generator. You can see here some examples where the switching losses are different than the application. Those switching losses can be lower, but also most of the case higher than the one you see in the data sheet. This is mainly due to the parasitic surrounding the [ active ] device, and this is what we have tried to capture with this schematic you see on the left where you have to enter many parameters to generate the model. But I really encourage you, and this is what we will demonstrate in that session, to create your own model and to enter those parasitic even if you have just, let's say, a rough idea of the value, enter this value, or if you want to use a maximum, to see what will be the maximum impact of those parameters. So this is also an example of the PLECS model. PLECS model, I didn't say that, is table of value where PLECS look at and try to guess what will be the value at the point is actually operating. So for that, it does some interpolation, if it's in between 2 points, this is on the left side, on the bottom, or extrapolation, if it's outside the point it has. And then try to guess, but you see that any interpolation or extrapolation will give an error. While if you increase the number of points, and this is easy to do in simulation, we can change the temperature immediately. We don't have to wait to make a temperature extraction. So we'll get much more accurate interpolation and extrapolation without almost error. And this is between the 2 graphs on the right, graph obtained with the data sheet, the PLECS model, where you have only 2 temperature value. And the graph on the bottom side obtained with this new model generator, you can see how dense and how many points are available. And so the simulation will be more accurate. The Elite Power Simulator is just a topology simulator and it has a very large range of topology for DC/DC isolated or nonisolated, AC/DC and DC/AC single phase, 3-phase. In total, we have more than 30 topologies that cover the major application in industrial like DC fast charging, UPS, energy storage, solar inverter. And automotive like OBC and traction inverter. It can be used in any domain, in fact. We have included what we call corner simulation. It means that we provide you the min and max model and not only the minimum that was an issue with PLECS model actually. We solved that by providing you this extreme corner simulation capability. You can upload those custom model that you have created in the model generator inside the tool. We have modeled the soft switching also losses because of the model, including the soft switching losses. It's, in particular, very important for topologies like DC/DC, LLC or CLLC or dual-active bridge or the old phase shift full bridge. We have included features to explore and plot the losses of the terminal impedance and so on, and it's very flexible and fast to design or to evaluate the topology. Here, we only evaluate the losses in our device, in the active device. If you want to have a complete picture of the system, you have also on your side to evaluate the losses of the passive design. You can find this tool at onsemi.com/elite-power-simulator. Let's see now how we have verified the tool with our traction inverter tester. So this is a picture of the setup. You can see various equipment to make measurements. And in the middle of the equipment, inside the equipment, you have the module we are testing. And the testing tool behaves like a motor simulator where we have many probing to measure each device. So the measurement device we use for some kind of Kelvin connection. And to characterize the voltage drop, we use the current source and we measure the voltage on the MOSFET itself or on the body diode to make the body diode current test. And this has been done with multiple current value and at multiple temperature. Then we also do IC characterization to capture the losses. And for capturing the losses as we have an average, we will measure the high side and the low side because you will see that low side and high side do not operate the same way and do not have the same performance, the same losses, depending on the, let's say, on the operating point. And this measurement includes the parasitic of the tester that we have characterized also. Just for your information, this is the way that IEC standard defined turn-On losses measurement, is between the drain -- the gate voltage going up to the gate -- the drain voltage -- sorry, I will repeat. It's from the gate voltage going up to the drain voltage going down. So we measure between 10% of the gate voltage to 2% of the drain voltage. For turn-Off, we measure between 99% of the gate voltage going down to the current going down to 2%. This is made to be compatible with agilities. Normally, with both sets, the drain current and the gate voltage, let's say, cross each other at almost the same time. You don't see any big difference between the voltage rising and the current falling down. In IGBT, we have the tail current that make this current falling edge a little bit smoother and lasting longer. That's why it's important for IGBT to capture the effect of this gate current, and the definition is not the same for turn-On and turn-Off losses. Then we have calibrated the heat sink attached to the measurement setup by injecting a current inside the body diode that is easy to characterize and measure temperature and junction temperature inside the module to calibrate the heat sink. Here, the heat sink is made with water plate, so with a coolant flow inside the water plate. And you can see that we match quite well the measurement and the result in terms of junction temperature inside the 2. We can see also the coupling in terms of thermal between the high side and the low side. You see that we have a very low coupling between the 2. So it will make easier to extract the losses for each device because the coupling effect will be low. And then we have created a complete first-time network of the heat sink used. So the water pipe coolant, and we have used for network -- a faster network to model this heat sink. Then we have created an internal tool for simulating the M-Track module losses where we -- when we want to measure to stimulate some point, we do some kind of loop inside the tool until we get to the point that the reference TG is equal to the average TG with less than 10% error. And after that, we can calculate the maximum TG and the losses when we have the same temperature. For that, we have a place inside the tool, the various switching sequence. So we have a sequence where the high side is on and the low side is on. And we can see that, for example, in the backstage mode for the traction inverter, because the high side switch [ by the inverter ], you can see on the rectangle one that the current flow for the drain, why indicates 3, the rectangle with 3. You see that the current is flowing through the body diode so we don't have the same losses. And so depending on the switching sequence, we will use various value that we have measured on the sequence for. For example, you have the current flowing directly inside the channel of the MOSFET. So all of that is taken into account by our simulator. And as you have seen in the beginning, we have measure on both the high side and low side of the device in order to capture those various operating mode of the MOSFET. So this is, for example, one result of this internal tool, and we can compare also the losses between IGBTs and silicon carbide. And this module was in silicon carbide. And we have obtained that for this operating point, [ 400-volt in 300-amp ] in the model. We have less losses with silicon carbide than IGBT. Then we have also here a comparison between measurement and this internal tool about the TG. Here you see that we have a pretty good match between the TG obtained by simulation with this tool we have created and the measurement. Then we have introduced the new tool, the Elite Power Simulator and the model generator. So we have used and configured those tools to simulate this application that was 400-volt in DC, 300 in MS -- RMS current, with the open 95 modulation index using space vector modulation with switching frequency of, I think, it's not 120. It's 10 kilowatts. I made a mistake here, and it's 120 as output frequency. So 10 rpm. Then with that, we obtained losses with the nominal value that we have in the data sheet, about 760 watts, and an average junction temperature of 78 BUC. If now I use a model that has been obtained with the model generator where I have included all the layout parasitic that we have on the tester, on this best traction tester that we use, as a reference. And also the model capacitor that, here, it's model emulator, which we still have a capacitor. So I obtained losses of 952 volts and an average junction temperature of 91 BUC. Here, I have also introduced the thermal impedance of the heat sink I have used for the tester -- that is used for the test. So here, this is the result of the internal simulation tool. We have 938 volts and an average temperature of 82 degrees. And now we shall make the comparison between the results I have obtained. The column number 1 is our internal tool based on measurement and that has been calibrated. The column number 2 is the Elite Power Simulator, I would say, stand-alone with the nominal data, the data of the data sheet. And the column 3 is the Elite Power Simulator with a custom PLECS model that has been obtained with this new onsemi Self-Service PLECS Modern Generator, including the layout parasitic to the heat sink tag, the real heat sink and the motor winding capacitor. And you can see that at the end, we have a good match between our internal tool based on measurement and the third column that is the -- our Elite Power Simulator with the SSPMG model. And we have 1-degree difference and less than -- or around 10 to 20 difference in term of losses. So let's do now a second example, a second verification with our 25-kilowatt fast DC charger demo. So this is a summary of the fast DC charger demo. It's a combination of a PFC board and a DC/DC isolated board. We have the 3-phase AC in for Europe and U.S. cover. We can cover both grid frequency, 50 and 60 hertz. We decided to have a -- now put voltage going from 200-volt to 1,000-volt with maximum 25-kilowatt or maximum 50-amp to avoid having high current at low output voltage. Then we have included the protection features that improve output like overvoltage protection, overcurrent short circuit, indoor voltage protection. We have also included and used the DESAT function of the [ driver ] to monitor the devices. And we have good communication and so on, and the product has been made to run up to 40 degrees. So the demo consists in 2 boards, as I have said. Here, you see the 2 boards, with, on the left, the 6-pack boost AC/DC active front-end, taking the AC in, creating this DC 800-volt boost in the middle. And then on the right, we have the DC/DC dual active bridge isolated that convert this 800-volt boost to 200 to 1,000-volt DC boost to charge the battery. This is the internal block diagram of the 6-pack boost. What is interesting here is the 3 module that we have used, the 3 10 medium first generation of silicon carbide technology from onsemi and surrounding those 3 modules, we have a gate driver of voltage and current sensor, and many things like filter, boost inductors and so on. And then we have the control board. The DC/DC schematic block diagram is a little bit similar, except the topology that changed now that is a dual-active bridge, isolated, and we use now 4 modules, 2 at the primary site, 2 at the secondary site. And for the, let's say, the surroundings, the measurement, the gate drive and the control, we have used almost the same part in a different number because we have more -- one more module to drive here. So let's now analyze the results that we can obtain with our internal tool. So we have run our internal tool based on measurement to evaluate the losses of this 25 kilowatts. And this is the orange curve that you have -- you see on this slide. And I have plot in shaded gray, light gray, sorry, the plus 5 and plus/minus 5 limit and on dark gray, the plus and minus 10 limit in this way -- plus or minus 10% limit in this way. When I will superimpose the measurement curve, I will see how I'm comparing to this curve. So let's now superimpose the measurement. This is the measurement we obtain. You see we have quite some noise in the measurement. But in average, let's say, we are in this range of plus or minus 10%. If I do an interpolation here, I use a very simple, in Excel, exponential interpolation. You can see that the exponential interpolation is fitting quite well with the simulated nominal or the internal tool simulation. And this curve stay in the range of plus or minus 5%. We may get perhaps better interpolation with another method. So now if from this graph, I superimpose the Elite Power Simulator results with the data sheet value of the losses, I see a very optimistic result where this result is about 20% less than the result obtained by our simulation tool based on measurement. But now I superimpose in green the simulation obtained by the Elite Power Simulator with a model, a PLECS model, not anymore the nominal value, but the PLECS model, including the surroundings or the parasitic or the various elements that are inside the PC board, or also the characteristic of the boost inductor, the coupling capacitor and so on, all those characteristic has been included in that model. And you can see now with the green curve that the model, the simulation results are matching quite well our internal tool. And so the interpolate curve that we have created to smooth a little bit the measurement error. So now let's -- looking forward, we have introduced our third generation of silicon carbide, and let's compare with the first generation. The first generation result is the dash line on top. You can see that, using the third generation, either 8 medium to 10 medium or 15 medium, I can get -- I can get lower losses. And I can see that with 8 medium, I get lower losses; with 10, I get even more. And with 15 medium is the best candidate for this application in terms of losses. And so it means that we don't need to use for 25-kilowatt silicon carbide as low as 10 medium, we can use 15 medium and we will get much better performances. But this 15 medium was not available at the beginning of this development. So that's why we have used the 10 medium. But if you do this kind of design, you can use now the 15 medium in our third generation that is available. So key takeaway for this seminar. We have demonstrated the accuracy of this new online tool that onsemi provide with this combination of the Self-Service PLECS Model Generator and the Elite Power Simulator. I can simulate and reproduce the losses of our traction inverter test setup or the 25-kilowatt fast charger demo we have created. We have seen also the importance of including the surroundings or the parasitic or the PCB, your real PCB, with your real inductor and your real capacitor, and all the parasitic included in those components because they are generating losses, extra amount of losses, that we can capture with the model generator. To use those 2 online tools, you only need a MyON account. No licenses needed. You can use the tool as much as you want. We have included in our -- in this tool our new SiC portfolio. We have discrete diode, discrete MOSFET and module available. And on the topology simulator, we cover all types of topology: AC/DC with single-phase or 3-phase input; DC/DC isolated or nonisolated; and single or 3-phase tool or 3-level DC/AC inverter. Also at the front end, we recover also 2-level and 3-level active front. So this is the end of my presentation. I thank you for the attention. And Kyli, we will go for the Q&A session.

All right. Wonderful. Thank you, Didier, for this excellent presentation. Before we do start our live Q&A session, I would now like to ask for our attendees to fill our survey. I'm centering it on screen now. I'd also like to point out that you can find today's slide deck in the related context box on your left.

And now let's get started with the Q&A session. All right. First question that we have is, is there a plan to include IGBT in the Elite Power Simulator?

Yes. There's a plan to include IGBT. This plan is almost finished. So I hope that by the end of the -- you can use IGBTs also in the Elite Power Simulator and also in the self-service model generator. And to continue on that, we will include also the -- our new generation of low and medium-voltage MOSFET, the T10. In the next year, or early next year, we will also include this technology because they are also used in combination with silicon carbide. In some case, like when you do a DC/DC converter from high voltage to low voltage in a car application, you could use the T10 technology. And so that's why we will also include the T10 inside the tool.

All right. Next question. Is it possible to use the generated PLECS model on PLECS computer version or only on PLECS online?

It's possible too when you upload the model generated by the Self-Service PLECS Model Generator to online to include this model inside the PLECS standalone, the PLECS you have on your computer. Because if you have a dedicated topology, for example, that is not covered by our Elite Power Simulator, you may want to know also the losses of our device. So that's why the model is compatible with both platforms, the online and the off-line.

All right. And then one last question for now, I believe. For the SPICE model, where are they available? And is there some help to use one?

Okay. Yes, the SPICE model are available on our website. You can download them. The easy way is to go to the product page you want to use. And down on the technical information on the product page, you will find the simulation model. Most of the time, they are available for SPICE and SIMetrix. And then to use them, we have many video tutorial and application, not available. Some of them are very interesting and, in particular, the one described how to set up a SPICE simulator to use our model. So it explains step-by-step how to use that. Some describe how in detail what I have explained roughly in one slide at the beginning. We have obtained those models, we do. And we have also many tools, maybe 3 or 4 now, application notes giving example on how to extract parameters from the data sheet and to make a data sheet somehow that fit the operating point you are using. Because you may not use the same, let's say, drain voltage or drain current that we used in the data sheet to extract these parameters. So you can extract your own parameters. And then in those applications, you have some of the example of AC/DC, DC/DC, or example on how to extract the junction temperature and so on, how to use anything in SPICE and so on. All of that is available online, and we can send the link to the attendees. I think we have a PDF file with all the links available.

All right. Thank you very much. All right. Thank you again. And these are all the questions we have time for today. And on behalf of onsemi, I would like to thank everyone for attending. I wish you the nicest of your day.