 So, hi everyone, I'm Professor Juan Ribas, my research group here in Stamford works on power electronics, and I joined the university back in 2014. So before I joined the university, actually Stamford didn't have a power electronics research group at all. So I got a chance to start it, and actually that presents some challenges, but also like immense opportunities, that like particularly gives me the chance to do whatever I want, like I don't have to be bound to like somebody else's program or anything. So that has given me and my students like a lot of freedom to experiment and particularly find applications. So I'm an application-driven engineer that like the most fulfilling part of like designing or building something is when actually has an application. A lot of respect for the theoreticians out there, but like in my group, we really try to like drive things home and use. So with that, I'm just going to start like I, please ask that at any point you have any questions. I just hope to jump in. It's okay. I don't mind like interrupting and try to keep it as informal as possible, like call me Juan, and happy to answer any questions you have at any time. So with that. So what we do is power electronics. And most of the time, like my experience, like when I start talking about power electronics to people is that they think power supplies, which to a certain extent is true. So we design power supplies. But when people think about power supply, usually the first one that comes to mind is like the one that is attached to your cell phone, to your laptop. And what I want to kind of convey, the message that I want to convey, it goes a lot more than that. So power converters are important because like a very large fraction of power electronics in general is important nowadays, because a very large fraction of all the energy, all the energy produced throughout the world passes at some point through some form of power converter. So it's important to make them efficient because like we don't want to waste energy, of course. And also it's important to make them dense. So in my field, one of the most important metrics that there is is power density, like how small can you make a power converter for a given power. And this is important because when there's many applications that are limited by the size of the power supply. For example, if the power supply that drives the motor in your electric car is too heavy or big, you can perhaps not go as fast or as fast as far as you would like. And similarly, like if the power supply of your laptop was 30 pounds, you probably wouldn't take it with you on a trip or you're going to have to pay extra. So but at this point, I just wish we could go everyone on any trip. So in and in generally when we think about power electronics, we have few variables that we can use to make the power supply more compact. So in the ingredients to make any power supply are semiconductor power semiconductor switches and passive components, usually in doctors and capacitors. And usually you have to set an arrangement of these devices on top of a printed circuit board that like connect them together and you have kind of a working converter. And the semiconductor switches, we operate them as their name says, as switches, we turn them on and off. And it happens that like the faster you can turn them on and off, the smaller you can make your power supply. So so there is an interest of trying to go as fast as possible in switching frequency so that you can like reduce them like as much as you can. But very soon, you find that this is there's restrictions to how fast in frequency you can go. And in big part, this is due to switching losses. So these switches are non-instantaneous. They take some time to turn on and turn off and that leads to loss. So there's a limit on how fast you can turn them on and off. So for example, this is an example like here in the bottom. You can see a paper that was published 10 years ago now and shows how when you design a converter, all of these are about the same power. But when you design a converter that is operating at 72 kilohertz of switching frequency, the power density that you're able to achieve for this design is about four and a half kilowatt per liter. So if you drastically increase the frequency to 250 kilohertz, it is noticeable that like your power density more than doubles to about 10 kilowatts per year, which is great. Now you have a much more compact system. So then the authors here double the switching frequency again to 500 kilohertz and the power density like now goes all the way to 13 kilowatt per liter, which is great. But like you can notice that like the increasing frequency from this first step to the second and from the second to the third, like the increases in power density are starting to diminish. Like notice that like if I were to double the switching frequency again, which is not an easy task, it's particularly like difficult considering parasitics and other situations. Like if you were to double the switching frequency from 500 kilohertz to 1 megahertz, notice that the power density only goes from 13 to 14 kilohertz per liter. So like a double in switching frequency does not necessarily translate in doubling of your power density. And that means that we're reaching a point of diminishing returns. And like there are multiple reasons why this is happening. And one that is very common to think of is again switching losses as you turn on and off the switches faster. They dissipate more and more heat. When you dissipate more heat, you need to add larger heat sinks to address the issues of like increased temperatures. But another important component is losses in mathematics. So as I mentioned, one of the components that we have in this print circuit board are inductors. And these inductors usually consist of a magnetic core around which you wind a wire. It turns out that that magnetic coil is prone to losses. And particularly, those losses increase rapidly with frequency. So this is an example of perhaps one of the best materials out there available for operation at higher frequencies. And like what we find is that like if you were to operate this magnetic material at a given magnetic flux density with just one of the parameters that we operated at, say at a frequency of 2 megahertz, you have this much core losses per unit volume. But I'm wearing this kind of purple coat. But if we were to design the same magnetic material to operate 10 times faster at a higher frequency of 20 megahertz, you'll find that the power density, the power loss density in the material increases by more than 100. So that is not good. So what that means is that like again, you increase your switching frequency by a certain amount and your losses in your magnetic component increase even more, which means that things are going to get even hotter and then you need to do something to maintain the temperature in acceptable levels. And that usually magnetic materials is particularly difficult to cool them. So generally what we do is we just make them bigger. So like if you make them bigger, you can operate at a lower magnetic field density, which means that it gets less hot, but you have a larger component. So these losses in magnetic materials is really like one of the most important why it's difficult to keep increasing the switching frequency and we try to make power converters smaller. And turns out that we end up finding ourselves in this green curve here. As I mentioned, when we are trying to design a converter and make them more compact, we increase the switching frequency. And when we do that, we have improvements in power density. But then when we reach a frequency of few hundreds of kilohertz, then we reach this point of diminishing return. And then when we reach this frequency past one megahertz or so, we'll find that the power density actually start going in the opposite direction, which means that things are starting to get so hot that you have to like severely derail the devices and make them much bigger. So that kind of like would normally indicate that this is the end of the road of using switching frequency as a variable to make things smaller, unless there are dramatic changes in technology or properties of materials or whatnot. But so one of the fundamental properties that we deal with in power electronics, and I mentioned that here, is that the value of the inductors and capacitors that you use, this is fundamental, the values of the inductors and capacitor that you use vary inversely with the switching frequency, which means that like if for some reason, like if we were to design a converter and higher and higher frequency, the values of those inductors become smaller. If you design it conventionally, the losses in those components would force you to have to make them bigger in volume, but not in value. But but if we're clever about this, like one of the things that we can do is start making these inductors without using any magnetic core. So as I mentioned, like the reason this magnetic components become like lossy is because losses in the magnetic material. So what if we just eliminate the magnetic material all together? And this is possible to do what it comes with costs. And particularly, I mean, there's a reason we generally people want to use magnetic cores because that means that you can make a large volume inductor in a small volume. But it's possible that like if the value of the inductor that you need is small enough that you can make it out of a simple air core. So have no magnetic core all together. And that will penalize you in power density, but it will completely eliminate this loss mechanism, which means that like now you can go back using the switching frequency as a means to improve the power density. But what it's necessary to do is that like in order because you're going to lose a lot of power density by virtue of not having a magnetic core in order to gain back what you gave up by eliminating the magnetic core. Now you need to operate at a much, much higher frequency instead of operating at like say one megahertz. In order for this to make sense, we need to be operating at frequencies of 10 megahertz, 20 megahertz and beyond. And this is extremely challenging to do from an engineering perspective. But a lot of gives us a lot of research opportunities. And that's how we work. Particularly if we are able to make our inductors out of like Primo, just a piece of wire, it means that we can also make them out of pieces of like metal on our printed circuit boards itself. So like instead of having physical components that we add on top of a switching of a piece of printed circuit board, we can now engineer the printed circuit board to have traces that perform the function of that inductor. And we can like if we're clever and I don't know if you guys are familiar with like printed circuit board technology, but it is possible to make printed circuit board with multiple layers of metal in inner layers. So it would be possible to make all these inductors, capacitors and transformers within the inner layers of a very thin printed circuit board in a way that like they have a function. So think about like kind of like what happened. So like we go from having a printed circuit board that is nothing more than a glorified connector that lets you connect the components that you put on top to each other to design a printed circuit board in which every trace matters. Every trace that you add is a circuit component that has a function. And that function is to help you improve power density and make it easy to manufacture because making a printed circuit board is at large volumes is incredibly low cost nowadays. So but in order to be able to do that, we have to do things have to be very clever about how we operate because like as I mentioned, we also have switching losses in the semiconductor. So this is a little like a very common circuit, a common circuit that we use in power electronics. And again, I'm just going to show the voltage and current across across this semiconductor. Ideally, we want them to operate at a switch. That means that it's either fully on and there's no voltage across it, but on the current or it's fully off and that there's no current and on the voltage across. And if you take the product of voltage and current across the terminals of that device, you can see how much power you are losing in that device. So ideally, if it's an ideal switch, you would experience no loss at all. And ideally, if all your components have no parasitic losses, you could make something that is 100% efficient. But because of the non-idealities of the semiconductors, the fact that they take their switch time to turn on and off, you'll find that you always have instant in time every time you turn off and off the switch that you have an overlap of voltage and current in the semiconductor. And that leads to instantaneous switching loss. So you're going to have this blip of instantaneous switching loss every time you turn on and off the switch. And you notice that if you were to double the number of times you turn on and off the switch in a given period, that means that you would have more and more of this switching loss at the terminals here, which leads to a loss component that is proportional to frequency. If you were to increase the switching frequency 10 times, your switching losses would increase by 10 times. And that loss is something that you have to address as heat coming out of your power supply. So again, this is one of the reasons we are limiting how fast we can go and switch in frequency. So one of the things that you can do, though, is use better materials. Use the most advanced of the best new family of semiconductors that are called wide-band gap semiconductors that uses new material like gallium nitride or silicon carbide that allows you to make semiconductor switches that are 1,000 times faster. So hopefully with that, you can reduce the switching losses 1,000 times and everything will be great. But it turns out that that is not the case at all. And so for example, here is a plot that shows the heat density that you would need to extract from a power semiconductor when you're operating at a very common 400 volts voltage. And you're operating at a frequency in this range. So you can see that even when operating at a frequency of a few hundreds of kilohertz, let's say 100 kilohertz, the heat density that you need to be able to extract, like the heat that passes through your semiconductor that you need to be able to extract, it's comparable to the heat density that you encounter in a fully working CPU. And you know that the CPU nowadays are incredibly hot. And they need massive heat sinks if you were to open your desktop or laptop. You'll see that you need to have this super complicated heat sink to extract it. But now look what happens if you start operating at a frequency, as I mentioned, that I want to operate at in order to eliminate magnetic codes altogether. I'm proposing operating at frequencies of 20, 30, 40 megahertz. So you can see that if you just work to do this, the heat density is that you need to extract of your semiconductor probably means that you won't be able to do this. It means that this is going to get so ridiculously hot that something is going to melt and surely they will. So what gives? So what it means is that we need to completely change the way we design power converters. Particularly, we need to, as I always would for something great we do in academia, is make things more complicated so that we can add more inductors and capacitors in the network to make the converter operating what it's called a resonant mode, which means that these inductors and capacitors set up conditions that you can solve using differential equations in a way that we make a circuit that when we turn off the switch, the voltage across the semiconductor will be determined by the other components in the network. And this component will act to pretty much make the voltage ring in this specific way, like have a ring to it in a way that the voltage across the semiconductor would naturally become zero by the time you have to turn the switch on again. So what that means is that if we are able to make the voltage across semiconductor go to zero before current starts conducting through it again, that means that we can, by operating resonantly, we can completely eliminate this overlap of voltage in current across the semiconductor that was limiting us in operation. So by operating things this way, we can, for the most part, completely eliminate switching losses, which allow us to operate at this tenths of megahertz of frequency that I'm proposing. So let's assume that we're able to do this, that we know how to design these converts, and we do, to a large extent. So we can now operate at these frequencies of tenths of megahertz. So now we're in that frequency range. Now we can completely eliminate magnetic cores. So now, how do we make this inductor? How do they look like? Well, they look something like this. So now we can use traces on a printed circuit board. We can make this funky little spiral to make it work like an inductor, which is OK. Similarly, we can use a little piece of plastic and just wind a wire around it and set up some inductance. But we actually prefer to have things a little more engineered. So what we can do is making a solenoid board, making it go around a printed circuit board and make it bite its own tail, if you think about that. And the reason we do that is that this structure allows us to confine the magnetic field generated by the current traveling to this component to be confined to within the volume of the toroid. And this is actually currently shown here by using an FEM simulation. We can see the magnetic field external to these three structures. So it becomes evident that on top of the toroid, you have the least amount of magnetic field external to the structure. And that is important because you don't want to have magnetic fields inadvertently reaching other places and contributing to electromagnetic interference. So this is kind of like the stuff that we've done in my lab in the past few years, so that we are designing printed circuit board and all the components in it, and we have demonstrated it effectively. So this is an example of a converter that delivers about 300 watts that is operating at a frequency of 27 megahertz, in which all the inductors and capacitors are implemented using simple traces on a printed circuit board. So by doing a sophisticated but not impossible layout on a printed circuit board, you can deliver hundreds of watts in something that pretty much has no external components, which means that I think that has an opportunity to make things lower cost and better. And there's still a lot of things that we need to deal with, we need to deal with external magnetic fields and whatnot. But if you have questions later or want more details, you can reach out. So it turns out that this is not ideal by no means. So for example, the one H1 of this inductor that I show here, this is kind of like what you have in mind when you design an inductor. So you provide a structuring with the current travels around and establish a magnetic field and produce some inductance. But in practice, that's not what you have. This is what you get. Here's on the right is like if we were to remove all the green FR4, like the fiberglass mesh, the metal that we are left with is this theoretical structure and doesn't look like what we had in mind. And this is far from ideal for many reasons. First of all, you know that currents don't like to travel in. They always look the path for this resistance. And having 90 degrees sharp edges at every corner is something that they don't favor. So similarly, when you connect the top and bottom layers of your PCB to make this toroid, you cannot do that with a solid connection. Instead, you have to connect a series of vias that connect the top and bottom layer. And that leaves a lot of empty space between them that increases, makes the path for the current to have higher resistance. But also, it makes opportunities for the magnetic fields to leak out. And again, that's something that we don't like. So we start looking around and we realize that people had actually described that perhaps the current flowing in a square cross section, this is what you would get if you were to slide this toroid, the rectangular cross section of a toroid is not the best to conduct current in these structures. So someone did actually, someone calculated what would be the optimal cross sections that you would need to have in order to have an inductor that has minimum loss. So what we realized though is that we were in a point that we can leverage technology in different ways. So if you ever had a class on electromagnetics, there is a concept that you perhaps have heard that is called skin depth, skin effect, which indicates that currents do not flow. They don't like to flow. AC currents, alternating currents, don't like to flow deep into a metal. They actually get constrained to a very thin layer at the surface of the metal, which means that and how deep they can travel within the metal depends on the frequency that they're operating at. And as you go higher in frequency, this skin effect becomes shallower and shallower, which means that at the frequency that we're proposing operating like 10 minutes or so, the thickness in which the current travels from the surface of the conductor is only a few tens of microns, which means that if you were to make this out of metal, you're actually not even using most of the metal. You're only using a tiny, tiny little layer, a very thin layer at the surface of this structure. So we thought perhaps we can truly print that. So that's kind of what we did. So we computed, calculated what this optimal process for an inductor would be. And then we put it inside a computer. So we develop a hijacked program that is normally used to make cell phone cases in Java. So they use them to make cell phone cases like we use it to design inductors in a parametric way. So we can change all the parameters that we want. And then we take this, we generate a 3D file that we can import into FEM software to study the magnetic fields in the structure. Now we can predict the value of the inductance and quality that we would get. And let's say we get from this particular inductor, we give these parameter values. Then we can proceed and just 3D print it. So we can get a low cost 3D printed and 3D printed as a scaffold. And again, this is not conductive. So we still need to do, but it's very lightweight when we use a low cost 3D printer. And then we can have like a simple process to electroplate. So that we can electroplate it like we perhaps 20, 30 microns of copper and top and have a component that is incredibly lightweight because mostly there's no metal, just like a tiny little layer on the surface. And we're able to obtain the inductance and quality factors that we need at a very, very low weight. And like, I'm not gonna go into the business of designing inductors only, like I'm not that interested in that. I'm more interested and see if we can design full-fledged converters this way. So one of the things that we've been trying to do like in the past few years, is see if we can come up with the structures that we can just 3D print and in one shot have all the inductors that we need to make pretty much print them like pancakes. So that we can use a low cost 3D printer printing all the scaffolds with all the components that we need and have a working converter. So that's what we did. So like this particular converter is a resonant converter that operates at 27 megahertz. The scaffold, the plastic scaffold that has three inductors that like different values all connected together. The scaffold itself only weights about one gram. And then we can like plate it and then add a semiconductor somewhere in here. And we end up with a power converter that like now when we implemented using the PCB this description that I had before used to weight 10 grams which is still very little. But like now we're able to shed half of the weight and bring them down into the five grams that we have. And you may wonder like, why is he building this weird looking converters for? This is actually made to strike a plasma. So like this power supply delivers radio frequencies power about like 50 watts of it and is able to ionize novel gas in this case is argon. And like we use it to strike a plasma. It's super bright like you can barely see this like the program doesn't do justice of how bright this is. And this is done with something that is operates very efficiently and weights very little. And you may ask me like, okay and why is he using the plasma for? Well, it turns out that we use them for making rockets. We're trying, we're started a collaboration with like Australia National University that we're trying to develop satellite thrusters like plasma thrusters to be able to keep a small satellites in orbit longer. So like nowadays if you happen to have an extra few hundred thousand dollars now it's actually possible for you to send a satellite into lower orbit. You can like either participate in some like academic programs with NASA or you can just do it on your own with like if you have a body in SpaceX and whatnot and you can convince them to send your CubeSat to the International Space Station at that point like an astronaut space station at some point we put him in a catapult launcher do they have in the space station and they just like send your satellite in a slingshot to go around the earth in a lower orbit. And that's great. But that satellite will only last about a month because at the altitude where the space station lives there's still some friction and drag and like after several orbits your small CubeSat start losing a velocity and then just become a night shooting star. So there's interest that like if you were to send assets into space to make say like an internet connectivity or if you want to have like a system of sensors to be able to monitor things around the earth you would like to have those satellites last longer. And in order to counteract the effects of drag which you need is a rocket but you can imagine that making something that is very compact turning into a rocket is really hard. So normally people then use chemical rockets instead they use plasma rockets and the Australian National University Development one but they didn't have a power supply to power it so that's what we did. So we put together one of these power supply that we designed. Originally we started with a 3D printed one but then we realized that it had some issues with like surviving vibration forces in practice. So we went back to build it using a proof circuit board because it gave us a much more robust solid frame. So that's what we did. So we put all the components together we packaged together. We put it in a mock-up CubeSat. So this is the CubeSat measure 10 centimeters on the side. Our friends in ANU put together the propellant distribution network like system and we show it in a vacuum chamber. So that we're able to start with battery voltages and generate RF that like turns the propellant into plasma that can be used to propel things into space which is kind of like cool. So again, you can see that this is an application that perhaps doesn't come to mind when you think about power. You guys have any questions? So I guess I have two questions on what you talked about before and then one relating to this. The first was kind of actually kind of answer the first one. It was on like you need special PCB technology in order to design those magnetic components. Like can you just do it with a via or I mean, we use more for the most part we use like the cheapest PCBs that we can get. There are some instances that we need to go to like a fancy or more expensive system. And I actually going to show one at the end but like in general, like I do it and cheap I try to get the chip that we can get. Awesome and also curious on the really printed inductors you mentioned that you had to plate them with the metal is the certain plating methods work better at these megahertz frequencies serve. Yes, the best method, the best plating method to use is the one that you don't use the one that you don't do yourself. So like we realize that plating is like a magic art. Like it's witchcraft. So like we tried to do that we tried to do it ourselves in the lab and like we had like a very low success rate like about like 20% of the attempts that we were trying to like play things where possibly decent, right? But like then we started like, no, this is just not working. So we found a guy and like sound like sketchy but he was as sketchy as it sounds. We just found a guy that like got excited about this project that started doing this for free. So like the guy lives in Pennsylvania for years we've been sending him like scuffles to electroplate and he has never charged anything. Like he's been doing it for free. We don't know how I do it. I don't ask and like, but like I realized that like it's like, I don't recall ever sending a payment for this guy and the guy was like, oh, no, it's fine. I'm like, okay, good. So I would just keep it like that. But plating is hard and that's not our thing. Like I don't, I don't, we try several times and we just like quit on the plating ourselves thing. Someone pointed at me that like perhaps the people that should talk to is that people that make like faucets. So like apparently like cold and those places that like outlets that make faucets, they usually use like plastics and then like electroplating wood, coppering, very, very high quality in very intricate surfaces. So that's something that is in kind of like the list of things that I would eventually do at some point but like we haven't really approached it. Very fascinating things. I guess my final question was how does each visitation work in something like a one-use spacecraft or you don't really have error? This is a one-use. So like what we wanted to show, like this is just a show-off that we have. We're trying to find opportunities to put one into space but like it's hard and it takes years. Cause like we need to qualify and find a lot of tests. But like, and also that's not what we do. Like this is like the group that we're working with in Australia are the ones that are like trying to to put it in a satellite. We're just trying to provide the parts of like what it's kind of like cool about like this is that we're able to make the whole power supply replace one of the sites of the CubeSat. So technically it's not using any space. But like a CubeSat needs a frame like needs like surface, like all the sites and we will just pretty much utilizing the wall better. And we are able to demonstrate that we can like make a plasma rocket and still have enough space inside this one view to actually put more devices. We didn't, we were just trying to make sure trust, trust but like that's, and it's still, this is ongoing. Like we have demonstrated trust but like we still don't have a full CubeSat. Did you envision heat dissipation to be a problem? No, this works in vacuum. This really works in vacuum. So like the, so I mean, we're not trying to make this power supply the smallest we could. Like actually like this trip putting in the whole frame of the CubeSat was a good idea because that manages the heat well. But like in vacuum, it's a thermal considerations are incredibly difficult and important. Like we did so many crazy things on this like we put a battery in a vacuum chamber and we were risking contaminating the vacuum chamber but like we did and it worked and it didn't explode. Very interesting. Thank you very much. No problem. One more question, Juan. And Luke, if you wanna go ahead and unmute and video it was a question about efficiency. So Luke, I don't know if you've got your answer or if you had wanted to elaborate on, but go ahead and it had a mute and video, that would be great. Yeah. Hey professor, sorry I came a little late. I wanted to ask what's the efficiency of the power inverter? Which power, the RF generators? Yes. So the by itself is about like 95%, 94, 95%. The DC to RF, but I'll show some numbers and efficiency in a little bit. Yeah, so in vacuum like the thermal junction coefficient is very low, right? So it's really a problem. Like I mean, so the team in our friends in Australia were the ones doing the thermal design and they, I think they have like mechanisms to cool, to dissipate heat to a certain extent. They put us a budget. I don't remember how much the maximum power dissipation that we can incur, but they put us a budget and like how much we can generate and we were under. Like we operate this for like hours, like it will work. Curious what the temperature is that the case gets to? I have it somewhere, but I didn't put it here. But like it really was not an issue. Yeah, 95%, that's definitely, cool, thank you. No problem. So I just want to show like a completely different topic. And again, I'm going to go a little faster here just mostly because I want to go through the slides so we can see what we're doing. So like as I said before, there hadn't been any power program in Stanford before. So like one of the important things that I needed to do was to like try to work on something that like is able to have an impact in a relatively short amount of time. So like it made no sense to me to like try to make conventional like I don't know like power supplies for data centers or for solar inverter or things like that. Because there is a lot of people working on it and they've been working on it for a long time. So just thinking that like, you know, I can come to Stanford and like get a bunch of people that is brilliant so they can be getting to the point in which we can build up experience and start contributing to the point that we can make a difference. It takes a long time. So I didn't want to do that. So like instead we wanted to see if like the type of work that we were developing could allow us branch out in areas that other people are not working and it turns out there's many, particularly it was in the realm of high voltage power supply. So like in industry, like this is a company called Bicorps, a large company that like they made incredibly dense power converters. Like they take few hundred volts input and delivers like three or four volts for your three or five volts to your power computer and they can achieve externally high efficiencies like 98% or so kilowatts of power in very low volumes. And it's really amazing. But when you look at commercial high voltage power supplies like the ones that you take few volts on one side and you get thousands of volts on the output, the efficiencies suck. Like this is like things in the 50, 60 percentage. They have power densities that instead of being the kilowatts per cubic inch are in the single digit watts per cubic inch. They're like bad in terms of that regard. But like that is what's commercially available. So we thought like, you know, perhaps there is something to the type of work that we're doing and high frequency to perhaps improve up in this metric. And it turns out that there is. So like, but when I started to look sound like a long time ago in 2013, we presented like some work that we built this, when we were starting to work on these frequencies, we built the first like 40 to 500 volts high frequency power supply. And I'm not gonna go into this in the circuits. I don't think there's a point on that here, but I'm just gonna focus on one of the stages of the converter that's called a rectifier that takes the radio frequency signals and turn them back into this. So when you do that, like as I mentioned, as one of the virtues of operating and very high frequencies is that we can reduce the values of the inductors to the capacitors greatly. Great, we can make them tiny. To the point that there are so tiny that we can actually afford to split it in two. So like this capacitor now is so tiny, is why I don't have, I don't mind having two of these capacitors in series. So I connect the second one, a second capacitors in the return path for the current right here. But like the reason that you can do this at any frequency, it just happens that it only makes sense when you do this at high frequency because the values become manageable. But like so at the frequencies that we're operating, the way we're connecting this capacitor. What this allows us to do is to have this right side of the circuit, completely DC isolated from the left. And what that means if you're not into circuits, the only thing with that, you have to think about what this means is that this allows me to have a multitude of these rectifiers and be able to cascade them in series while connecting all the inputs in parallel and hence achieve voltage multiplication. So like if I have this output that gives me say 500 volts and I have 10 of them, now I have 5,000 volts. And it turns out that like this happened to work great for the type of applications that we were targeting. So this is the first product that we came up with. It was still a little ugly, but like you'll see that it gets pretty really quick. So this is a 40 volts into two kilovolts out 27 megahertz converter. And it has 12 of these individual rectifiers at the output. So we all put it together nicely and we demonstrated an efficiency of over 90%. And again, this happens to be a striking good compared to where we could fight commercially. So like, but it's still not in the final nice shape. So like one of the students actually did I tried to put these all things together. And then like, so we bought one of those power supply that I showed high voltage power supply that I showed earlier. So this is a 30 volts into four kilovolts power supply, 250 watts. And the measured efficiency was 78%. So like my student build this high frequency converters at 10 megahertz converter that delivers the same amount of power, the same voltages, but it's the size of a credit card in an area. So it has the printed inductors and the thickness of around four credit cards once you take into consideration everything. This, that's all the heat sink you need. That's all you need. So it works really great. It allows you to have something that is 20 to 30 times the smaller in volume that way you can find commercially. Not only that, it's super fast. So like, this is a comparison of testing with we just like trying to measure how fast we can reach the nominal voltages of the power supply. And we find that these high frequency power supplies are infinitely faster than the commercial than the commercial powers. So by the time this power supply is, it's on its way to try to reach its steady state voltage. Our power supply like turns on super quick in like microseconds. So like we can have a power supply that not only is very compact, but like let you send pulses of kilovoltages in about a microsecond. And you may ask, why do you want voltages that fast? Well, it turns out that there's a really cool process that we were aware of. And it's called pulse electric field posterization. So if you have a pipe, so this is a pipe that you're passing water through it, let's say, and that water has bacteria, right? If you cut this pipe and turn them into electrodes and then applied pulses of around 20 to 50 kilovolts per centimeter across these electrodes, it happens that the electric field that you introduce into the liquid is so high that it just like ruptures the membranes of any bacteria that is living inside. And you can use these to pasteurize liquids. So, but like, remember like, your parents probably told you not to put, to connect electricity when you near water, but there's a reason for it. But it happens that like, if you were to keep this electric field for long, you would just vaporize, like the losses in the liquid would be so high that everything would just vaporize. But if you only apply this pulse for about few microseconds, you don't let the temperature on the liquid to increase so much, but the electric field is high enough that ruptures the membranes of this bacteria that is living inside. And again, I wish I could have come up with this idea. I'll be very rich by now, but like, it's not, so there's a, it's being commercialized by a company, it's a company called Diversify Technologies that has done this, that like, this is the size of the power supply. It's the size of three, it costs about $200,000. And it's pretty much all, this is a power supply. And all the magic, like all this magic happens in this pipe on the side. This is where liquid flows in and out and then they treat it here. They apply pulses that are about 50 microseconds, about 50,000 volts across this little gap. And it's in earnest, you can see. So we wanted to see if we could make the ghetto version of it. So like, we built a tiny hundred watt, like the power supply that I showed you before. That like, it's very small, we put it in the lab. We use like, two pipes separated by a millimeter to pretty much to recreate the same effect. And we use it to see if we could treat water with it. So like we take water from a pond, like we culture it for bacteria, that we check bacteria content and you find that there's bacteria and it's kind of like nasty. And then we treat it with one pulse of this pulse electric fields. And we turns out that like the bacteria, it's kind of like gone. And I still wouldn't drink it because like the color is still green. Like it doesn't change the color, you just render the bacteria in there. It turns to be like an efficient way to like treat liquids. And we turn turns out that that's how they use this equipment. They use it to treat in sewer plants. So like what they use is to like render the liquids in earth and then they go and add chemicals to like try to treat it and do their thing to clean water. So like, so we were kind of excited about this but then we realized that it was a bad business. Particularly if we are like trying to get like a way to get treat water. Turns out there's a lot of competition, right? Like a short trip to REI, if you're into camping, you'll find that you can buy this like tiny UV lights that you can use to sterilize to treat water or just a filter, right? Or also you can add two drops of chlorine in your liquid and like that's probably safer that all the way we do it. So like we realized that like this is perhaps not the best market for like developing something like this. But instead we were looking for a liquid that like you could not treat with UV light, you could not treat with adding chlorine and it also wouldn't work by treating it with a filter. And we realized that there was one and it was milk. So it turns out that like in the rural places about 40% of the milk that is producing rural farms you can't sell it directly, you need to pasteurize. So like the way farmers do is like they collect milk then take it under motorcycle, bicycle or donkey and then go into the city and like give it to a distributor like usually like Nestle or something like that. And like they pasteurize, process the milk and distribute. But because of road conditions, weather, et cetera about 40% of the milk gets bad and they have to like be dumped. And like you can imagine that has an enormous cost to communities. So we thought like, what about if we make this kind of like post-electric filter authorization thing and make a unit that we can sit next to the cow? Like imagine like we can have like a solar system that provides power to this tank and apply it like the farmer is not in his cow. The milk it's treated online directly like next to the animal and it's treated for bacteria to try to reduce the bacteria content to the point that the product can have a longer lifetime. We're not trying to make this to reduce the bacteria enough to make it for human consumption. Cause like I don't want to take that responsibility. I'm not qualified to do that. But we just want to be able to reduce the bacteria content so that like the farmers have longer time to take the product into market without like and hopefully reducing the amount of milk that is spoiled. So we tested that and we turns out that we were very good at late, it certainly works. Like we can reduce the bacteria content and last like longer. It's really difficult. Like we stopped doing this mostly because like you need like people with expertise in areas that I don't. And like for example, how to make testing with bacteria in terms of doing way more difficult than we thought originally. But like also it turns out that like doing this from a circuit perspective was very hard. Particularly because like milk, it's about 300 times more conductive than tap water. So that little ghetto converter that I showed you before which is not for the task. So we need to scale things up dramatically. So we need to increase the power to like the kilowatt level. So like this is a two kilowatt high voltage pulse generator that we built for this project. But that this one worked. So it has power density for around 250 watts per cubic inch. Again, it's now in order to manage it better than anything that we could find commercially. And it's able to deliver pulses in microseconds fast enough to be used in this type of pulse electric field pulsarization. And it turns out to be very efficient. So then when we start computing how we stand in power density versus commercial offering, this is where we found that perhaps we were hitting on something big. We realized that like now the technology that we can do is able to achieve, I mean, this is not incredibly far comparison because like for one part, commercial systems have a lot more protection mechanisms that we are able to like provide in a prototype basis. But nonetheless, we are really far out in power density that will come, anything that we have had commercially. And like it turns out that like it's exciting. So we did something also very exciting. So like one of the students take the same concept and he took it bananas. Like he went like crazy on this, I'll show you why. So like, so he take it to an extreme. Like again, this is the input voltage for this power supply is 45 volts in. And he designed this system to deliver five cubic volts. And in between he put a printed circuit board a transformer that uses the two layers of this PCB board. And he did it in a way that then like he put it and put like fans like the whole thing like super nice. But like what allows you to have is to have another layer of isolation through the use of this transformer to deliver these five kiloballs. And then he start stacking them together to and he's putting together 20 of this stack together to deliver pulses at the level of a hundred kiloballs. And the reason that he wants to do that is again you take 44, 45 volts input to deliver a hundred KV, a hundred kiloballs out at the kilowatt level. And we started like noticing that again, there's a power density. We were really, really out there from anything that we had on commercial. And like you may ask him, why now he wants a hundred kiloballs? Well, they wanted for X-ray. So like there's a lot of applications like in dental, medical, industrial research that require high voltages for producing X-ray. And this is in the voltages and current levels that we perhaps can make really well. And particularly the area that I'm interested in is city scanners. I don't know if any of you have had to be in town, a city scanner. But like usually when you're in a city scanner, you're placing this machine and there is a gantry that rotates at very, very high speeds around the patient. And how fast they can rotate the machine depends on the weight of the power supplies that are in it. So we want to see if it's possible to reduce the size about like 30 times of some of these power supplies so that the gantry can be rotating at a much faster, like higher speeds and then reduce the amount of X-ray dosage that the patient receives. So that's kind of like one of the applications that we have. And lastly, this is something that happens in the past couple of years and the work has been interrupted into COVID for obvious reasons, but like it's pretty cool. And with that I'll finish with you guys, have any questions? Sorry, I think I'm taking too long. So there was a DARPA, like the Department of Defense can move with a call for proposals for making a very small power supplies, high voltage power supply. So this is the specifications that they requested. They wanted something that was a third of a cubic centimeter, only 200 milliwatts and about three kilovolts and you had to weight less than one gram. So we got really excited because we thought like, you know, this is right in our field of expertise. And after three years, they wanted the people that work on this project, they wanted to make this even more challenging goals. So we applied, we submitted a proposal and we didn't get it. And like one of the reasons that we didn't get the project according to one of the reviewers was that they didn't thought, like they really liked the work that we had done on high voltage that I showed before, but they thought that they couldn't see a path for miniaturization to like scale down in power and size the work that we have presented before. And like the student that helped me write this proposal was beyond pale and like upset. So he asked me if he could work on it some free time on it and if I could give him a budget. So I gave him a thousand dollar budget to like try to make this program on his own and he did and he built it. And in his first attempt, he was able to actually meet and exceed the stage two of this DARPA program with a thousand bucks and a month of his time. And he didn't do it by like doing any new science or anything. He just did a very clever design. He used a point of circuit board like a flexible PCB board and made a very conventional high voltage circuit like similar to what we described. And he just made a time. And then he just origami the hell of it to try to meet the, like to meet the specifications for this program. In the process he broke with like, there was like some issues with like some of the, and that's the point of the reason that they turned out to have different like metrics, but it's able to fit and exceed what he was expecting for that program. And that was also very exciting. And in order to show it off and show it to the program managers in DARPA and tell them like, hey guys, you should have given us the money because we did it anyway for free. We wanted to show it in a very DARPA-esque way. So like what we did is like we decided to test this in using an electro-addition drone. An electro-addition drone, an electro-addition, I don't know if it's something you're familiar with, it's pretty much like a patch that is insulated but it has an interdigitated electrode. And when you apply a large voltage across it, it can like stick electrostatic into surfaces. So we decided to put this on a drone. So we bought the chip, the lightest drone we can buy on eBay. They had a camera. And then like we put it together with the power supplies that we had and like the electro-addition patch on top. And what we do this is to, we use it to, so that we are able to purge the drone into the ceiling. I'm still recording. So like instead of like hovering, using a drone to hover to take some video for the extended period of time that takes an enormous amount of energy, instead we can like take it all the way to the ceiling, turn on the power supply electrostatically purge it, turn off the motor and it can stay there indefinitely while still recording and then just turn off the high voltage power supply and then keep supplying. So again, this is kind of like another area that I was completely not expected that we end up working but like we think that has a lot of potential for the type of like that.