Below is the text version of the webinar titled "Opportunities for Wide Bandgap Semiconductor Power Electronics for Hydrogen and Fuel Cell Applications," originally presented on October 21, 2014. In addition to this text version of the audio, you can access the presentation slides.
Amit Talapatra:
—recorded, so a recording, along with slides, will be posted to our website in about ten days, and we will send out an email once those are posted to our website.
All participants in this webinar are on mute, so please submit your questions via the question function, and we will cover those during the Q&A at the end of the presentation. Since we have multiple speakers today, please indicate who your question is for when submitting your questions.
[Next slide]
You can see the question box shown here. It will look like this. And please check back to our website for future webinars. We host them monthly.
Also, I encourage everyone to sign up for our newsletter that we send out monthly, and you can sign up for that on our website. With that, I'd like to hand it off to Eric Miller. Eric is the technology manager for hydrogen production in the Fuel Cell Technologies Office. Eric?
[Next slide]
Eric Miller:
Thanks, Amit. Welcome, all, to the webinar. The objective for today's webinar is to bring together experts in hydrogen and fuel cell technologies together with the wide bandgap semiconductor industry. We think there's some interesting opportunities here for mutual benefit. And take a look at our first slide. We've seen that there is an emergence of fuel cell and hydrogen technologies. It's an exciting time for us. But we're seeing that this has happened in part due to the research and development efforts through the Fuel Cell Technologies Office to reduce costs in all the components, such as fuel cells and electrolyzers.
I really encourage everyone to investigate the program and look into these reductions and these great accomplishments. The website is shown at the top of this slide. For those not familiar with the program, I encourage you to do so. Amit, could you go to the next slide, please? Thanks.
[Next slide]
As a part of our program, we use extensive techno-economic analysis to identify where the costs are and where cost reductions are possible. We see that both in the fuel cell industry and in the electrolyzer industry, that there's been significant progress. However, at the systems level, there's still quite a bit of room for improvement, particularly in the balance of system and balance of plant components, such as power electronics.
Anecdotally, I've spent some time talking to the hybrid fuel cell bus people over in Switzerland, and they're excited because the fuel cells last forever. However, their power electronics in terms of the DC/DC converters tend to fail quite—unfortunately, too often. So we know there's an opportunity here for the communities to get together to look at alternative technologies that could address this issue, and I think what I'm going to do is I'll hand this over to the experts in the power electronics industry.
We're very fortunate to have with us leading experts in wide bandgap semiconductor power electronics from DOE and from Cree. You'll be talking to them and they'll be talking with you.
[Next slide]
But first, since it's their show, without further ado, I'd like to introduce Dr. Anant Agarwal, who is the senior advisor at DOE to the Wide Bandgap Initiative, who's leading our institute efforts, and leading this field. Thanks, Anant.
Anant Agarwal:
Thank you, Eric. Next slide, please.
[Next slide]
So I'm going to use the next three slides to introduce wide bandgap to you and what it does for us, and how we are funding it. So basically, there are two wide bandgap semiconductors that are most popular right now and mature, and those are silicon carbide and gallium nitride.
And as you see on the top of the screen, they have wider energy bandgap than silicon, which is 1.1 eV. And because they have wider bandgap, they have—they can operate at higher temperatures compared to silicon, and they have high breakdown electric fields.
So as I show in that little cartoon there, generally silicon has both types of carriers, electrons and holes flowing through the semiconductor for devices, whereas in wide bandgap semiconductors, we can use only the electrons, and therefore use only the unipolar devices, and that makes them faster than silicon, much faster than silicon. Amit, can you—yeah.
So basically, what we see on the left hand side in the green box is that wide bandgap semiconductors have high temperature, high voltage, and high frequency attributes. And the question is, can we turn them into power electronics, which is more efficient, smaller, and cheaper? So that's what we want to do. Next, please.
[Next slide]
So the two semiconductors we talked about, silicon carbide and gallium nitride—and often people think they are competing with each other, but that's actually not the case. They have their own voltage range where they're applicable. So for example, gallium nitride semiconductors are useful from 200 volts to 900 volts. An ideal application for gallium nitride semiconductors is power supplies such as laptop power adaptors, server power supplies, solar converters up to about 10 kilowatts. Next.
So silicon carbide is useful from 900 volts all the way to 15 kV, using only the unipolar devices—remember, which use only one of the carriers, electrons in this case. And actually, if you use bipolar devices, we can go all the way up to 40 kV, but today, we will talk about 900 volt to 15 kV devices. And there we can use these semiconductors such as in solar inverters larger than 10 kilowatts, even megawatt-class central solar inverters, could be fuel cell inverters, certainly automotive inverters and quick chargers, traction applications. One of the big applications is medium voltage motor control and grid. Next, please.
[Next slide]
So what we are doing—what we are doing at DOE is funding Power America Institute. It's a manufacturing institute, and it's being funded at NC State University. We are still in negotiations, so we cannot talk too much about it. But we have 20-plus partners from universities, national labs, and small and large industry. Cree actually is a very important partner of this institute, so we are happy to have them today.
The idea is to capture manufacturing leadership, both in devices and power electronics. So as you see there, there are three elements to it: the commercial foundry, where the purpose is to reduce the cost of these devices, and I'm sure Cree will talk about it. The middle one is the advanced power modules, where we basically put these chips into the modules and work on them, so then they can be put into power electronics. And the last piece is power electronics, where we want to show that we can use wide bandgap to reduce the system size and weight by at least 50 percent, and improve efficiency at the same time.
So here—Amit, can you hit the next? Yeah. So here, we really want our graduate students in universities to use wide bandgap semiconductor in building power electronics inverters, and I think that is the best investment we can make, because tomorrow, they will be the leaders in the industry.
So with that, let me introduce our colleagues at Cree.
[Next slide]
So we're really pleased to have Dr. John Palmour and Dr. Jeff Casady from Cree to talk about silicon carbide power devices and future outlook. And John Palmour is the CTO and a founding member of Cree. He's also an IEEE fellow, and actually has been the—is probably one of the only persons in the world who has pushed silicon carbide from its very infancy to now. Jeff is a senior program manager in silicon carbide power devices at Cree, and he's involved in development of new applications for silicon carbide power electronics, such as fuel cell inverters. So now I'll hand it over to John and Jeff at Cree.
John Palmour:
Thank you very much, Anant. This is John Palmour, and as Anant said, I'm CTO for power and RF here at Cree. I just wanted to quickly introduce myself, and in the interest of simplicity, we're going to have Jeff Casady give the presentation materials, and I'm going to be here for moral support and to help answer questions. So with that, I will hand it over to Jeff Casady, who's going to present on the opportunities for power electronics for hydrogen and fuel cell applications.
Anant did a very nice job of introducing why wide bandgap is of interest, and we'll show you some examples of devices, et cetera, and try to focus a little more on the applications that we're currently going into. The big markets for silicon carbide today are in server power supplies and solar inverters and industrial power supplies for industrial equipment. And with that, I'll hand it over to Jeff.
Jeff Casady:
Thank you. As John said, we'll start out really talking about the silicon carbide power product we have at Cree, and as well as the applications that they're going into right now.
[Next slide]
And if you look at the slide that's out now, just kind of to give you some flavor, if you're not familiar with silicon carbide, we've got over 90 products in the field. It started out heavily with the Diodes back in 2001, 2002. There are 70 Diodes, ranging from 650 or 600 volts, excuse me, up to 1,700 volts currently.
And those have been in the field for quite some time. We continue to grow that line. But more recently, within the last five years, we've now added power MOSFETs and also Power Modules, and you can see that we're growing those product lines as well. So it's becoming a pretty well-balanced portfolio of products that are available so that people have a lot in the toolkit they can design with. And we'll see a lot of these—it's not just Cree, but other vendors, we'll continue to see product proliferation at a pretty rapid pace right now, because it's hitting a lot of major markets, where it just makes so much more sense now to go with silicon carbide and lower the BOM costs than it does to stay with silicon. So the next slide, please.
[Next slide]
This is just as an example of as we expand the portfolio, we're going up and down in voltage and current. This is the first all silicon carbide 1,700 volt module. It's a half-bridge power module in a 62 millimeter platform. It was just released last month, but these are all fully available from a number of catalogs and traditional distributors, and it comes with a gate driver and all the things that you would expect with a silicon module. This one's about 250 amps at temperature. And this is being targeted for a lot of the solar applications, especially some of the more central inverter applications with solar. Next slide.
[Next slide]
One advantage we have then, too, with the products being out for now over a decade, you can start to get a lot of field reliability. And one measure of reliability is failure in time. It's a pretty standard industry nomenclature. And what we have now is showing over a trillion hours of field data on our products. Most of that's been the diodes, because they've been out the longest. But you can see the FIT rate for any of the products on the diodes is much less than one, and typical for silicon it would be five to ten in this voltage range, and can be actually higher in some cases. So we're running at least ten times lower field rate failures, FIT rate, than the typical silicon components.
And if you look at the—I think if you toggle the next—next, you can see the MOSFET there circled. The MOSFET, since it's been out less than five years, has less, but that's kind of rapidly expanding now as the volumes continue to ramp up. But we now have about 1.3 billion hours with the MOSFET. We haven't had any field failures yet, but if we assume that there are two failures reported, that would give us a FIT rate of three, which again is still well in line with what you'd expect, actually much less than the silicon.
John Palmour:
That's actually recording or assuming two fails for the first generation MOSFET and two fails for the second generation MOSFET.
Jeff Casady:
So total of four fails.
John Palmour:
Yep.
Jeff Casady:
And I would say just as a note on these, we often assume more failures than what we actually get back, so these are conservative numbers. The next slide, please.
[Next slide]
One of the things also that comes with having the products out, is you see the cost reduction. With any new technology, when it's first introduced, often the price is high. But what we see with silicon carbide, as the volumes have picked up, and our learning rate has picked up, it follows really just what you'd see in any standard industry curve. The costs come down typically about 85 percent from what the introduction costs are in a fairly short time, and you can see the first products out were the 600 volt Schottkys, then the 1,200 volt Schottkys, then the 1,200 volt MOSFETs, and each time we've introduced a product, that slope has been steeper, so the price has come down more quickly.
And also, you can see underlying that this 2002 through 2016—so we've gone from 3-inch wafers back in 2002 and 2003 to 4-inch today, and we're on the cusp of going to 150 millimeter in the future. And each time you grow the wafer diameter, you get a pretty good cost reduction as well. Next slide.
[Next slide]
So now we will switch and just really talk about some of the portfolio applications, where things are today, and how people are using them in the industry. So next slide, please.
[Next slide]
Obviously not every customer will do a press release with us, but we do have a few that have done that, and so just walking through some public available press releases, you can see solar inverters, as John mentioned, is one of the applications that's really picked up the use of our silicon carbide MOSFETs very quickly. They were already using the diodes. Now the MOSFETs are being used by quite a number of customers. Here, we have press release examples from Delta last year, 2013, and Sanix, the Japanese company, in 2014.
And you can see the Delta was an 11 and 12 kilowatt design. The Sanix was a 9.9 kilowatt three-phase. And the theme is very similar for both these customers, or any customer in solar. They're looking to get better efficiency, but the real driver is cost. So they're in a sense using the efficiency by going up to higher switching frequencies and keeping the efficiency high, and when you can do that with silicon carbide, that allows you to take the cost, size, and weight out of your passive elements, like your magnetics, your heat sinks, and your enclosures.
So the end result is you can get lower losses, lower costs, and better performance. And there's a quote there from the Sanix—I think this was the general manager from the press release, and you can see his quote there. He got 30 percent lower losses, and they were able to simplify their topology, and they're actually capturing market share. And this is a pretty common thread you'll see with most of the customers adopting the silicon carbide. Next slide, please.
[Next slide]
So this will be a similar value proposition, but now with a different market. This is induction heating power supply. This is using—the previous examples were actually with our 1,200 volt MOSFETs, discrete. This one was using a 1,200 volt MOSFET—this customer is using a 1,200 volt MOSFET and a half-bridge power module, 300 amp, 62 millimeter half-bridge power module. And the bottom line is when they did this, they were able to go to a 2.5 times lower part count, still maintain or actually improve their efficiency, so they reduced the power losses and passed along the cost of ownership reduction to their customers.
And you can see the quote there from the R&D manager. This is EFD. They're one of the top two induction heating customers in the world. And being able to drop this all-silicon-carbide module in, they were able to achieve 99 percent efficiency, and as I said, simplify the part count. Reliability is a huge deal to them, so 2.5 times lower part count, and each part obviously has an implied failure rate, that greatly increases the system reliability, and so they're able to get a better product, it's simpler, it's more reliable, more efficient, and these are the things our customers care about. Next slide, please.
[Next slide]
This is a different one. Again, this one happens to be for hybrid electric buses. And in particular in China, this is a fast-growing market for HEV buses, due to the pollution and that they're trying to get those pollution levels down. And this is a customer who approached us. They did a press release with us last year. The customer's name is Shinry, and what they had was basically for an HEV bus, it's basically an auxiliary power supply. It's a DC/DC topology with a 750 volt DC in and a 27 volt DC output voltage.
And they had a traditional silicon-based approach, and that's the box you see on the left with the three fans, fairly large. They redesigned it with our silicon carbide MOSFET. It's a 1,200 volt, basically 20 amp MOSFET in a TO-247. And when they were able to do that, you can see the box on the right, the benefits are listed at the bottom of the slide. They got 8 point increase in efficiency, 88 to 96 percent. The size of the box went down by about 25 percent, the weight of the box went down by 60 percent, and they were able to eliminate the three cooling fans. You can see those are gone. So the audible noise and the kind of reliability concerns with that went away.
So they were able to eliminate all these components and actually get a cheaper box, and it's, again, helped them really differentiate in the market. And they were kind of a small company in a very crowded market, and are doing pretty well with this product. They've told us their market share is continuing to increase with this product. Next slide, please.
[Next slide]
This is another example, which actually our applications team here at Cree has done, working in concert with a few different customers, and we've actually turned this into kind of a demo board. There's an evaluation you can see in the—under the title, there's a blue box with Cree eval P/N there. That's the part number for this board, so you can actually order this for evaluation.
And this is something that we had people in EV chargers, customers in EV chargers, and also telecom power suppliers approaching us, and they basically wanted a more compact, lighter weight, more efficient box, and also they did not want to obviously compromise on efficiency or reliability. So it's again the same thing, where we can move to higher frequencies and take advantage of the very efficient silicon carbide component.
And in a nutshell, what they've done is they've gone from a three level silicon solution in a resonant topology and moved it to a two level, and they've taken the frequency from 120 to 260 kilohertz, and in doing that, you can see the benefits there in the box. They've cut the part count in half, so they've got a much simpler architecture now. The power density is 35 watts per cubic inch, and yet the efficiency's actually gone up from 97.8 to 98.1.
And so to our knowledge, it's not possible to do that in silicon. At least, that's what the customers tell us that evaluated it. And we do see a lot of customers using this topology—they may not use this exact, obviously, but they do this as a first step, and then they take their product and do some derivative design off of this. And this was published at PCIM.
[Next slide]
There's a—the next slide, we'll kind of go back to solar for a moment and go a little more in depth. That's fine. Go ahead and move to the next slide.
This is the—a boost stage for a solar inverter, and we'll go through more of the details, but before we get into it, basically, when you look at this box, if you're designing a boost stage from 10 to 50 kilowatts, what we've found, and at least our experience with customers has been that when they use the siliconized BTs, the 1,200 volt components, and then they use our silicon carbide, they can get size, weight down, they can get the cost of the build materials down, and improve the efficiency. And these are kind of rough numbers in the table below. This is hardware we built, just kind of to demo that, but we've seen this evaluated by multiple customers, and a lot of customers have done something very similar to what we're showing here.
And it can be pretty significant, if you look at the—basically size and weight going down by 50, 60 percent. Build materials cost—that's always tough to gauge because so many customers have different cost structures, but it's generally in the 10 to 20 percent range. And then the losses in temperature go down as well. So if we go to the next slide.
[Next slide]
Where this is being used is in solar, and one specific example would be for rooftop solar, and this is an area where our applications team here at Cree has been working with some market leaders in the PV inverter industry, and trying to really understand what the challenges are that this industry faces, and even their end customers, and work to come up with a solution that works for everyone.
And the basic need for rooftop solar is they really want a lightweight, high power density, three phase inverter that's cheap, it's reliable, it doesn't take up much space, because the solar panels have gotten so cheap now that you really just want to cover the roof with these panels and not have inverters taking up a lot of that space. So they'd like the power density to be at least above one kilowatt per kilogram. So if you go to the next slide.
[Next slide]
This is one example of what's used today in a 50 kilowatt inverter. This happens to be from a manufacturer called Kaco, but we've looked at a lot of different systems from suppliers, and this is pretty typical.
And the numbers in red there obviously jump out to you. It's a 50 kilowatt system. The CEC efficiency, which if you're not familiar with solar, that's California efficiency. It's a weighted efficiency number. It's 97.5 percent. It's got a 480 volt AC output, and the operating temperature range is de-rated to about 45 C. That's 173 kilograms in that box for the 50 kilowatts.
John Palmour:
Yeah. I want to stress, this is based on silicon technology.
Jeff Casady:
Right. This is the—
John Palmour:
The silicon version.
Jeff Casady:
—typical products that are being used today. And we'll come back to this in a minute with the silicon carbide version. So if we go to the next slide, this is what we can go to with silicon carbide.
[Next slide]
The numbers in green are the ones that have actually improved by going to silicon carbide. Especially note, though, the 50 kilowatts. That's not changed any. But if you look at that CEC efficiency number, it's improved. The peak efficiency's gone up. But the real focus here wasn't to improve efficiency as much as it was to get the power density.
And if you go down to the weight number, the 173 kilograms in that standard silicon version can now be 50 kilograms, and this is a box here on the right showing that. And also notice the volume is about half, 50 percent of what the volume was with the existing system. If we go to the next slide, we're going to go into a little bit more of the details in this design.
[Next slide]
Next.
[Next slide]
So you have a two input inverter. It goes through a boost stage, and this is all silicon carbide, just to kind of keep in mind. So this was our design in that 50 kilowatt box. It goes through an MPPT boost at 75 kilohertz. You have a 825 DC output voltage going into a three phase inverter. So the boost stage is running at 75 kilohertz. Typically in silica you run this at 20 kilohertz. The three phase inverter, which is the box there on the right, it's running at 48 kilohertz, and again, that would typically be 20 kilohertz or maybe even less in silicon. And that would be the output.
And so this is the system that we've built. This is kind of the schematic overview. Next slide.
[Next slide]
This is the schematic on the PV inverter, just showing again now the schematic details behind the boost stage and the inverter stage. It's a T type inverter, interleaved boost, and this is all using discrete products. You can see the boost there is using two 1,200 volt 20 amp MOSFETs in parallel per phase leg, to the basically 10 amp, 1,200 volt diodes as well.
The T type inverter uses the larger MOSFET, but it's still in the TO-247 discrete. It's a 1,200 volt, 25 milliohm MOSFET, two of them in parallel, so it's about 60 amps per MOSFET, so 120 amps total there, when you put the two in parallel. And then the T branch just has another 1,200 volt MOSFET.
[Next slide]
This is the photograph of the completed—the box on the left with the 50 kilowatts. Here you can show a wave form, here at 47.8 kilowatts, and with the output voltage shown, and again, the DC link voltage at 850, as we described earlier. And John, is there any comments you want to make about—
John Mookken:
The only thing I can say is that we would have gone to 50 kilowatts, but we were source limited. That was the only reason we didn't go all the way to 50.
Jeff Casady:
You can still see there's space to optimize inside the box if we wanted to. The biggest volume is taken up by the DC link capacitors.
John Mookken:
Yeah. I'd also point out that this is the 50 kilogram, so it's achieved using film capacitors for the DC link, which was done for reliability reasons. But typically, they use electrolytic caps here, and if you used electrolytics, that weight would be even lower. It would weigh less, and it would cost less.
John Palmour:
So by the way, this is John Palmour. The John speaking was not John Palmour. It was John Mookken, applications engineer here at Cree, who designed this box.
Jeff Casady:
OK. Next slide.
[Next slide]
And this—even though we've taken the power density up to about one kilowatt per kilogram, and that weight of that box, if you remember, it comes down from 173 to 50 kilograms, the efficiency is still better, even though we're at one-third the weight. And here's some numbers. Just looking at the solid lines here are the Cree box, the numbers we're getting, which are approaching 98.6, I believe, peak, and the Kaco silicon box is the dashed line. And we're showing these, if you can't read it on your computer, it's the different voltage levels—I mean, yeah, 850, 760, 480, trying to keep the comparisons the same for all of them. And you can see this goes from 50 kilowatts all the way back down to about 5 kilowatts output power. So it's pretty consistent across the whole power range. OK. The next slide.
[Next slide]
This is the PCB for the 50 kilowatt evaluation unit, just to show you a photograph. It's really nice and clean. John, is there any comments you want to say here?
John Mookken:
We were able to put the whole thing on a single PCB board, including the magnetics, as you can see there, and the only reason we could do that was because we're switching at 75 kilohertz, so each phase leg is switching at 75 kilohertz, so the effective output ripple is 4 times that, so you need very little output capacitance to maintain the output voltage. So yeah, the board by itself, the complete assembly weighs seven kilograms.
Jeff Casady:
For the boost.
John Mookken:
And that cannot possibly be done with silicon. No.
Jeff Casady:
OK. Next slide.
[Next slide]
And these are just the specifications. This is available as an evaluation platform. At least, I think we've loaned it to a few customers for evaluation.
John Mookken:
Yes, we have, about three or four.
Jeff Casady:
And I think we have three or four now, for those people interested. We've primarily worked with the solar inverter market, as you can tell from this slide.
[Next slide]
OK. And this is just more on the—it's kind of the boost features. We don't—I think for the sake of time, to give—we can keep going.
[Next slide]
And this is the measured versus calculated efficiency. Again, everything lined up really well for the boost stage.
[Next slide]
The thermal advantages are pretty striking. I think people know they get more—or better efficiency, excuse me, with the silicon carbide, but when you translate to what that means thermally, it can be pretty impressive. And these are the thermal images showing—the upper left is our 1,200 volt MOSFET, and then the middle upper photograph is our 1,200 volt diode, and then you see the boost inductor here in the lower right. The boost inductor is sitting around 70 degrees C. The peak temperature there in the MOSFETs is around 92 to 95 degrees C. There's two in parallel there. And the diodes you can see are running 65 to 67 degrees C.
And this is—of course, make—there's a note there at the bottom that says after full load, 30 minutes of operation with a fan, and ambient is 25 C. And by the way, we've done measurements with silicon parts, and they've obviously run much hotter when you get to lower efficiency.
[Next slide]
And this is—for the boost stage, this is kind of the evaluation information, the part numbers and the cost and all that. But we provide full—that CAD model schematic layout file, just to help people—because if you're not familiar with designing with silicon carbide MOSFETs, if you're used to siliconized BTs, this can help you get there much faster, in our opinion, and that's the feedback we've had from other customers. Next slide.
[Next slide]
This is going—so that was the boost stage, now a little more in detail, the inverter stage. Go to the next slide.
[Next slide]
Just this is kind of more of a mindset. In silicon, at least for solar, and I'm assuming it's that way for some of the systems you may be familiar with, the topology is really optimized for the efficiency you need at whatever frequency you're operating. So a two level topology, if you look at this graph, the dashed lines are silicon, and what we're showing here is a two level, three level neutral point clamp, and a three level T type.
And if you look at the dashed lines first, look at this green dashed line, that's a two level silicon topology. So what that's saying is to keep your efficiency at acceptable levels in the marketplace, which in solar is certainly above 98 percent, you're really limited to about 20 kilohertz for two level. And to get better efficiency in silicon, you go to a three level neutral point clamp or a three level T type.
In silicon carbide, because the semiconductor is so efficient, the power losses are about ten times lower switching losses. You really don't have to choose based on this anymore. The two solid lines are silicon carbide. We have a three level T type in the red solid line. We have a two level in the blue solid line. And you can see up to 50 kilohertz, you're running around 99 percent or above efficiency, whichever topology you go with.
So then it becomes more what's better for your application. Is it T type, maybe to focus on lower harmonics in the output, or is it two level because of simplicity? It really opens up more tradeoffs for the designer. And we've seen I think for the solar customers that we talked to that they now are going with different topologies, based on things other than the efficiency constraints they have with silicon.
And the important thing is either one gets you up to the 50 kilohertz or 48 kilohertz you need to take advantage of the cost reduction in your magnetics and your [inaudible]. Next slide.
[Next slide]
This is—we went with a T type. We easily could have gone with a two level. I think there's actually some debate on which one is better. We just had to pick one. So for our inverter's results that we showed, it was a simple T type, three level, and we used all silicon carbide discretes here in this version. Next slide.
[Next slide]
And so that kind of wraps up the existing products and what we've done, and we just want to finish it with a few slides on kind of where do we go from here, and that's based on the future product.
[Next slide]
The two things before we talk about that we have to keep in mind is because the efficiency and reliability or ruggedness, excuse me, of the silicon carbide is so much better than silicon, you don't typically do—drop in the same ratings in silicon carbide as what you have in silicon. And this bar chart here, what we're showing here is the comparison of 1,200 volt power modules, both half-bridge.
On the right is the silicon IGBT, and the ratings on it are 600 amps, 1,200 volts. I believe this is the H4. And it's running at three kilohertz in a half-bridge topology, same inverter. And the bar graph on the left is a 300 amp silicon carbide MOSFET, half-bridge, 1,200 volt, but instead of running at 3 kilohertz, running at 10 kilohertz.
And the interesting thing to note is we all know that silicon carbide's more efficient, but when we see real results, that's pretty surprising. You don't run a silicon 600 amp IGBT module at 600 amps, because the losses are just too great, and thermally, that module couldn't handle it. But you can run the silicon carbide MOSFET much closer to the rated current, because the efficiency is higher. It doesn't have to deal with the thermal issues. So even though we've tripled the frequency, or over-tripled, we actually have lower losses in the silicon carbide power module, even though it's got half the rated current that the silicon IGBT power module has.
So this is often kind of a stumbling block for people who are, again, used to IGBTs and silicon, and want to start looking at silicon carbide. Naturally, the first thing you ask for is what's the same rating component that I have now. But that's normally not what you want to do.
John Palmour:
Typically, our customers are using anywhere from half to as little as a third the number of amps to get the same amount of work done as what they were using in silicon IGBT modules.
Jeff Casady:
OK.
John Palmour:
Because so many amps are thrown into the IGBT modules to keep the losses down and spread them out.
Jeff Casady:
So the next slide.
[Next slide]
So the current's not often the same. Usually, we see people using less current rating in the silicon carbide than in the silicon, but the same could be true in the voltage. This is something that's still under investigation. But medium voltage, as Anant mentioned earlier, is a very interesting application space for silicon carbide, where a lot of benefit can be derived. Silicon, say 6.5 kV IGBTs, normally they're rated—they're de-rated, excuse me, from the rated voltage based on the FIT rate, and for medium voltage, kind of cosmic ray, a FIT rate of 100 is the maximum a customer can allow. They'd really like 10, but they'll allow up to 100, usually, for the customers we've spoken with.
And in silicon, this is literally a textbook curve. If you look at the FIT rate, which is the Y axis, versus the voltage for a 6.5 kV commercial silicon IGBT module and just plot those, you can see that FIT rate, that failure rate, does not cross 100 till it's down at 3,600 volts. So for a customer planning to operate at 3,600 volts, and they have a maximum FIT rate of 100, that means they have to purchase a 6.5 kV rated silicon IGBT.
Silicon carbide, we have ongoing investigations from customers who've looked at our 1,200 volt silicon carbide MOSFET, and they've found that up to basically 50 times, if not higher, better cosmic ray resistance for silicon carbide than silicon. Basically, the problem they're having is they cannot get the silicon carbide to fail at 1,200 volts in our cosmic ray. So there could be a substantial advantage for silicon carbide in this rating for voltage as well, and this is something we're investigating. Next slide.
[Next slide]
John Palmour:
Just to stress, for those who aren't familiar with cosmic rays, this is not talking about a space-based application. Cosmic rays are one of the leadings causes of failure in these high voltage systems. This is cosmic rays that you get hit with at sea level. The higher you go in elevation, if you're up at 3,000 meters and higher, the failure rate increases.
Jeff Casady:
Right. Basically, it's for large power systems, it's a big issue. So where we're going, if you go to the next slide, then—is looking at these bigger power systems in the future.
[Next slide]
So for medium voltage, there are applications in rail, grid-tied solar, grid-tied wind, HVDC, offshore wind, and grid—anything grid-tied power distribution. And basically, people can use almost solid state transformers and things like that. It's already being used in Japan in some rail systems now, so it is coming pretty quickly, this medium voltage. And so this is where we see a huge advantage in the future. The next slide, please.
[Next slide]
And what the customers so far that we've engaged with—and they've been market leaders in these different fields—basically, it sounds a lot like what we just went over in the lower voltage. They're looking for a simpler topology that's more reliable and it's cheaper than what they're using today. And this is an example from Fraunhofer Institute, a publication they did this year, using our 10 kV silicon carbide MOSFET. But the image on the left, the schematics, we have the box schematic, is a traditional system. We have a low voltage PV system with DC power, inverts to AC, and then goes through a traditional transformer and gets put up on the grid, 10 to 20 kV, and then onto the AC distribution network from there.
The size of these systems is generally about two megawatts each based on the transformer limit. If you go to the figure on the right with silicon carbide, you really eliminate the traditional low frequency transformer, so a lot of times this is referred to as transformer-less. There's still a transformer there, but it's no longer the transformer that you're used to seeing.
John Palmour:
It's a very small high frequency transformer.
Jeff Casady:
Right. So you have, on the figure on the right, the same solar panel, but now it's a DC/DC converter with 1,500 volts in, 3.5 kV out, and that's medium voltage. And then it can go into, for example, where—there's a big push for a lot of applications now to look at DC distributed power. That can go directly onto the DC distribution network. Or if you need to go back to the grid, you can do a medium voltage inverter. The size of that now can be anything that you want it to be, because it becomes solid state. It can be very small. It can be very large. And that gives a lot of flexibility to the designers.
So the advantages are they could take this sub-unit power rating anywhere from a few kilowatts—to like small points of load, if you want a lot of them—or greater than two megawatts if you have a large power system. You can have—the power cabling now can be run at higher voltage, so that's going to cut your losses, also make the cables cheaper because they're smaller and lighter, not carrying as much current. And you can eliminate this really large, heavy, costly transformer. So this allows you to bring medium voltage really within very close proximity to whatever your point of load is, whether that's a tool inside a factory, a data center, what have you. So a lot of advantages there, and you can reduce the number of system components. You can build in reliability in different ways.
So we think this is going to be pretty revolutionary. At least, that's what we're hearing from a lot of these customers. If you go to the next slide.
[Next slide]
Just to kind of finish, this is kind of the building block. It's kind of a crude prototype, but basically, what you need then is DC/DC converters and inverters at medium voltage. And this is an example of one using our 10 kV MOSFETs where they took a 30 kilowatt DC converter, 3.5 kV input, 8.5 kV output, 98.5 percent efficiency. It was switched to 8 kilohertz, which would be 15 times higher than you could do with silicon. So very compact. It still has the transformer there, but that's a small one. That's still most of the volume there. This is only 36 by 30 centimeters. And this would enable this "transformer-less" power distribution to the grid or to the DC distribution network, and that's where we think the future is going.
So I think that's all the slides we have today, and that gives us about 10 to 12 minutes I think left.
[Next slide]
Eric Miller:
Well, thank you. Thanks to Jeff and John and John. I really appreciate the presentation. You did a great job, and we do have a number of questions. We'll try to get to as many as possible. I believe we'll get—Alli will—someone will tell you what we'll do with the ones we can't get to. We'll get them answered offline after the webinar.
But let's get started with some of the questions that have come in from the viewers already, and I think the first one we'll address is probably on everyone's mind. It's related to specifically slide 21, slide 22, but the gist of it is, what is the cost of the silicon product shown on slide 21 compared to the cost of the silicon carbide product on slide 22? And I'll open the floor to Cree on this.
John Palmour:
All right, we're—OK. So the—remember, this is the cost on number one, I mean that's a typical silicon system. I can't give you exact cost numbers on that design.
Jeff Casady:
Well, we—
John Mookken:
Yeah, I would say our target was to keep the efficiency and cost of the system the same while we reduce the weight.
John Palmour:
Yeah. That's what we were working towards.
John Mookken:
And given that we've only built the prototype, and we've done some pricing exercises, and we think that's realistic.
Jeff Casady:
So we think it's the same price, the silicon—
John Mookken:
Right.
John Palmour:
The cost of the silicon carbide component is definitely more than the cost of the IGBT, the silicon IGBT. But you save a lot of money on the balance of the cost of the system. You save money on the inductors. So typically, where we are winning in solar inverters is because we actually make the system cheaper, because they're saving money on the magnetics and heat sinks.
John Mookken:
Yeah. One of the things with silicon carbide you always have to keep in mind is you have to keep the bigger picture in mind when you're looking at cost. You can't look at component to component. You also can't look at even system to system. You have to look at higher than that. What is the labor involved in installing an inverter that weighs 50 kilos? One of the reasons the 50 kilos came in as a target was it's because U.S. labor laws—
John Palmour:
OSHA requirements.
John Mookken:
Right. It allows two people to carry the inverter to the roof versus using a mechanical device to carry it up to the roof. So there are benefits in system level installation. There's cost reduction in labor, which was the reason why we came up with the 50 kilogram target.
John Palmour:
Right. You don't have to bring in a crane—
John Mookken:
Right.
John Palmour:
—if it's 50 kg or less.
Jeff Casady:
Right. So we think the system is roughly the same price, but you get all these other cost benefits on top of that. But we don't make inverters, so we can't tell you what the—that's kind of the industry. But I think the industry is running $0.15 to $0.20 per watt, something like that. There's a—people that are familiar with solar inverters kind of know what they run.
John Mookken:
Yeah. If you look at it as an installation, total installed cost, you'll definitely see that it's much cheaper if you go with silicon carbide.
Jeff Casady:
OK. Next question.
Eric Miller:
Great. And I'm going to follow up on that a little—yeah, let's follow up on that a little bit with cost, because you also gave a good example how, for example, a DC/DC converter on the hybrid bus had a lot of benefits in terms of the new technology. I guess back to the cost, again, it's an emerging technology, so if you—cost versus price is a little different, right? So if you had to buy that off the shelf, would it be the same as you would be buying a silicon system at this time, or is it—is there prospects for reducing the costs, so that—clearly, there's a cost benefit, in addition to the performance benefits in such a system as the buses or others.
John Palmour:
Well, that's basically what I was saying. Most of our design wins are actually because we're bringing a cost benefit, and that helps our customers be more competitive in the market.
Jeff Casady:
And these customers have told us they're gaining market share by using these designs.
John Mookken:
Yeah. So we try to be cost neutral at the least, and bring in a whole lot of other performance benefits as an added.
Anant Agarwal:
Hi. This is Anant from DOE. Also, I think within the institute that we talked about, I think the cost of wide bandgap devices will come down in the next three to five years, and so that will actually make the silicon carbide system cheaper than the silicon system.
Eric Miller:
That's good news.
Anant Agarwal:
So that will also help. And Cree is at the forefront of all this, because they have six-inch silicon carbide wafers.
Eric Miller:
Great. All right. So that's good news. Thank you.
Anant Agarwal:
Yes.
Eric Miller:
All right. We have more questions. I think we have some more time. Let's get to the real technical detail one. Automotive fuel cells operate at substantially lower voltage than you're talking about. Would the gallium nitride be better for a system operating at 200 volts? For example, a fuel cell at 0.6 volts per cell running at—with about 300 cells?
John Palmour:
So—go ahead.
John Mookken:
Go ahead, John.
John Palmour:
I was going to say, so typically, first off, that would be the line voltage, so typically, you would use something—a part a good bit higher. So that'd probably be in the 600 volt range. The answer is you could use—at 600 volts, you can use silicon super junction MOSFETs, gallium nitride. There's not a 600 volt silicon carbide MOSFET available yet. It doesn't mean there can't be, but there's three pretty competitive technologies in that voltage range, silicon being one.
We tend to focus on the higher voltages because we have—the higher in voltage you go, the more benefit we have over silicon, so we don't really compete at that 600 volt device level yet.
Eric Miller:
OK.
Anant Agarwal:
And this is Anant. I think gallium nitride would be very competitive at 200 volts, but right now, I think we're limited to 10 to 20 kilowatt type of power levels. In future, I think we can—we could do 50 to 60 kilowatts, so then it becomes useful for light duty automotive vehicles.
Eric Miller:
OK. Very good. All right. A few more questions. I think we have a few more minutes. Are there thermal concerns with using silicon carbide devices, and if so, what are—what methods are being used to research and reduce thermal losses?
John Palmour:
There really shouldn't be much in the way of thermal concerns. One is because the increased efficiencies that we have, quite often, we're actually running lower temperatures than you are in silicon, because they're just so efficient. Now we keep wanting to shrink the die to reduce cost, et cetera, so it's—silicon carbide in reality should be able to operate at much higher temperatures than silicon, and we have some customers that buy our silicon carbide die and put them in metal packages and rate them for well in excess of 200 C, I think 215 Celsius, which is higher typically than silicon.
So there's really not a big issue with temperature. The big issue is trading off efficiency versus temperature versus cost, because that's the trade space. You can let it run at higher temperatures, but nature dictates in any semiconductor your losses, particularly conduction losses, will increase as you go up in temperature. So the question is where is that sweet spot where you can minimize how much silicon carbide you have to buy, but not let the temperature get too high, because it gets more resistive, which means you have to add more silicon carbide.
Eric Miller:
And a little bit of a follow-up to that question. We have the main failure mechanism for silicon devices has been wire bonds and delaminations from the substrate. Silicon carbide running at higher temperatures will have higher fatigue, so how can the unit be more reliable at these assumed higher temperature cycles, due to operating at higher junction temperatures?
John Palmour:
The—well, what that refers to is mostly in power modules, and that is the most common failure in power modules. That's very much a function of the cycle and the delta T that you operate at. So we've done a fair amount of testing. Basically, there's no difference between silicon carbide and silicon in terms of the fatigue you see in the wire bonds and die attach, et cetera. We've done—two hundred thousand cycles or something like that?
Jeff Casady:
Yeah, it's all the same, and kind of—
John Palmour:
Basically acts the same.
Jeff Casady:
To kind of reiterate John's point, too, I think just to be clear, most of the applications that we've shown, we did not show a thermal comparison of silicon versus silicon carbide, but we've done that internally, and we generally are seeing 20, 30 degrees C cooler on silicon carbide, because the efficiency is just better.
John Palmour:
Right. So the question is—the question is a good one, but typically, if you ran it under the same conditions, you actually wouldn't see as much of a delta T during those cycles. So—
Eric Miller:
OK. Yeah, but—
John Palmour:
You can try to push it up to higher temperatures, and that—it's a trade space.
Jeff Casady:
The only way you can—
Eric Miller:
Could you clarify for the audience what the actual operating temperature is—on one of the slides, there was a question that it was specified as only 35 degrees, which seems low. So I think it depends on the application and installation, but maybe can you just clarify what operating temperature we're talking about?
John Palmour:
Well, we've—that was I think just a circuit demonstration at room temperature, how high something was. But the die are fully rated up to 150 C, and we're looking at going to 175 C right now, at the die level.
John Mookken:
Yeah, we've done a test where we replaced a 50 amp IGBT with a 50 amp silicon carbide MOSFET under exactly the same conditions, and we saw a 40 degree drop in the heat sink temperature that we were measuring. I mean, that just makes sense. We just dissipate less power.
Jeff Casady:
The only way you can heat the silicon carbide die is just keep pushing the frequency up and up and up and up.
John Mookken:
Right. So then it's up to the designer to know what to do with that delta of 40 degrees C. What do you do with it? Do you decrease the amount of silicon carbide, or do you decrease the heat sink?
John Palmour:
Get rid of fans?
John Mookken:
Yeah.
John Palmour:
That's—again, design [inaudible] is what it offers.
Eric Miller:
Good. Thank you. We've got about two minutes left. I want to try to—maybe we'll just introduce these questions and take them offline. One of the features I think the hydrogen and fuel cell community is interested in is the response and the turn down time, and the fact that you can run at high efficiencies over quite a range, up to the maximum power rated point. That is very important when you're putting it on renewables or you're trying to run it with variable loads and variable responses. Are there other failure mechanisms, even though you're getting higher efficiencies across the board, up through the maximum efficiencies? Are there other failure mechanisms that occur when you're running at lower power ratings?
John Palmour:
No. I would say our field data that we showed would indicate the answer is definitely not, because we get far fewer fails than we do in silicon.
Eric Miller:
That's great.
John Palmour:
And a lot of the device hours are in things like server power supplies, which run 24 hours a day, but not always at full power. Actually, I forget what a typical number was. They kind of average maybe one-third load or something like that. So we have tons of field hours that say that it's not a problem.
Eric Miller:
Terrific. And we're going to get off in a second here. Let me just throw something out at you, and maybe we can take this offline. In terms of techno-economic analyses, have you worked with the solar companies and others to identify what the cost savings are in the large-scale installations, for many reasons, including the reduction of the transformers?
John Palmour:
I think—
Jeff Casady:
The short answer is yes, but I don't know—we have NDAs with a lot of these guys, too.
John Palmour:
The short answer—
Eric Miller:
That's fine.
John Palmour:
—is yes, there is a—part of the issue is even though it's understood that over the long term, the higher efficiency, et cetera, has a certain value, typically, for a big solar farm installation, you'd be surprised how much they don't look at that, because there's a capital budget up front that they want to stick with. So up front cost is still an issue.
Eric Miller:
OK. Good. I think that's great. I think we've just run out of time, but I'm going to open it back to Amit, and—for final closing remarks, and take it from there. Thank you all again for such a good presentation. We really appreciate it.
John Palmour:
You're welcome.
Jeff Casady:
Thank you.
Amit Talapatra:
We will have any unanswered questions submitted to our speakers, and we'll post the responses to those on our website. I'd just like to take a moment to thank today's speakers, and to also thank all of our participants, especially for submitting questions today. A recording of today's webinar as well as a copy of the slides will be posted to our website within ten days. Thank you.
Anant Agarwal:
Thank you.