Jessica Boles—Miniaturizing power electronics through piezoelectric energy storage
Good, morning. My. Name is Vladimir Ivanovich I'm the director of MIT Nino and I have an absolute pleasure to, be a host of today's. Nano exploration, webinar, today's. Speaker, is Jessica, balls she's, a graduate, student he professor David paroles group and. Has done some remarkable work in miniaturizing, power, electronics, through piezo, electric energy, storage. In. A moment I will ask Jessica, to take over but before, I do I'll just remind you, just. As a way of providing, the bandwidth, to the rest of us turning. Off your video and muting yourself, is one good thing to do now at the beginning and. Then if you have questions at, the very end of the talk we'll, ask, you to submit them via chat. Or. If not via chat to raise your hand and I'll call upon you with. That being said Jessica. Please do take over. Power. Electronics, are our building blocks for processing, electrical, energy here. I'm showing a variety of things that affect our daily lives where we can go who we can talk to how, we can live and, all of these are critically, dependent on power converters, now. One big focus of the power electronics, field today is figuring out how to make the converters inside of these things much much smaller a smaller, converter, has quite, a few advantages that, ultimately boil down to reduce cost and a wider realm of potential applications. Now. For those not familiar with, power electronics, I first want to introduce the concept of how a converter works at its. Root a power converter pulls energy, in from a source, momentarily. Stores, it and then, delivers it to a load in a different form and for, us that usually, means a different voltage now. This process is, cyclical, which means we do this hundreds, of thousands, sometimes millions of times per second. Now. Whenever we start to think about making. Power converters small learn we often refer to power density, which. Is the power out divided. By the volume, of the converter itself in addition. To power density, we also need, high efficiency, which. Is the percentage of the power we bring in from the source that. We can actually do right to the load. Now. High, efficiency, helps, decrease, our heat dissipation, which, is increasingly, too difficult, to, handle as we increase power density. Now. If we take a look at the inside of one of these power converters, we see that a large proportion of the volume, is actually taken up by energy storage I'm going, to highlight that in the macro charger you see here, now. In a traditional converter, we would use inductors, and capacitors for. Energy storage but. There's nothing that says we have to in fact. Electromagnetic. Energy storage scales, down pretty poorly with volume. Now. If we take a look what happens to an inductor as we scale it in all three dimensions, by Epsilon we'll.
Just Call that a scaling, factor like. This, its, volume is going to increase by epsilon, cubes but. The power it can handle is going to increase by epsilon, to the fourth that, means just a small increase, in volume gives us a drastic, power increase, and that's great for high power applications. But. If we look at this in the opposite, way and we see that as we, scale down volume. Power. Decreases, much faster, meaning, our power density is actually going to drop that's, counterproductive for miniaturization. So. To enable major, advances, in power density going forward we're, going to have to look at alternate, ways to store energy such as piezo, electric energy storage, piezo. Is rely on the piezoelectric and inverse piezo electric effects, which, essentially, means that when we apply a voltage to Appiah those terminals that. Bolted induces, mechanical, stress and strain in the material and vice versa, now. Through, that process we can induce mechanical, resonance like, you see here in this animation, in. Which the piezo storing energy and, mechanical compliance. And inertia somewhat, like a spring. With mass. Now. Pieces. Are attractive, from the stand point so they have cleaner form factors they may be more amenable to that fabrication, than magnetics, they, also provide, opportunities, for isolation. With, multi-port resonators, but. Perhaps most importantly, for our purposes, they have very high quality factors. Which is good for efficiency, as. Very, high power density capabilities. Now. Piazza's were used extensively a little more than a decade ago for, lighting fluorescent, lamps and, I've also been used in energy harvesting, applications, like you see this shoe right here but. Overall pieces. Have still seen very, little commercial, use in power conversion and attempts. To use them in research without magnetics, have shown limited, performance that. Being said the, most effective. Ways to use piezas and converters, have yet to be explored much. Less the full realm of potential, waste. So. To evaluate Pisa, based converters, for miniaturization, we're. Going to dig at that, exact question today. How. Do we best utilize the peers in the power converter, so. We'll start by looking at, what. Are all the possible switching begins it's switching sequences, in topologies, we can use will. Then constrain, those for high efficiency so. We can start to compare them to figure out which ones are best. Finally. We'll go over some of our initial experiment results, and we'll wrap up with some unanswered. Questions for Pisa based converters, going forward. To. Start before we can design a converter with a piezo, we need an electrical, model for it but. This is a Butterworth VanDyke model which is commonly, used for resonators, it's, got a physical capacitance. Which. Is C P between the ps2 terminals, and it's. Also got in parallel with that in LC our branch which, models, the mechanical, resonance properties, for the piezo. Now. That we have a model for our piezo we can start to think about what we would like to have a nappy is a base converter if we. Believe piezas may, be a route to high power density we'd like to use them all by themselves inside. Of a converter without any additional magnetic energy storage that, might bring down the converters overall power density. So. We're going to use the source load system shown. On the right it's. Got just the piezo and we haven't specified anything, about switches yet, now. We'd also like, to soft, charge CP, and what we mean by that is we'd like to always charge and discharge CP. Through, resonance so. We can reuse energy as two rather than lose it when we're changing the piezo terminal, connections, in. Addition. To that we'd like to be capable of efficient, voltage regulation, meaning. We'd like to be able to spanned a wide range of input to output voltage, conversion ratios without. Much efficiency, penalty, and. Finally. We, need the the required switch of limitation, for this converter to be practical, as, simple as possible in order for piezo based converters to be marketable. Now. Keeping those goals in mind we enumerate, all the different ways we can use a piezo in. The source of system starting with its connections, now.
We're Not thinking anything about switches, yet just. What are the possible connections, and, we're going to refer to each of these connections as a stage and we'll look at each stage, from the perspective, of VP, which. Is the voltage each stage cuts across the piezo, so. We'll start with what we call connected. Stages, in which we connect the piezo to the sources system, so. That VP, has some combination, of the innernette or V out across it, BP. Could equal the end negative the end be a minus V out there's lots of options here these. Are the energy transfer stages between the source and the load now. We could also short, the PSS, terminals, and. What we call a zero stage in which we force VP 2 equals zero, and. We could also leave, the piezo terminals, open circuited, and this. Stage will, call an open stage allows. The P to change 3 resonance this is where we would do that charging, and discharge charging. We talked about. Now. With string these stages together to create what we call switching. Sequences, we, can think of a switching sequence as the pattern, of stages, that we cycle, through over. And over again to pull energy in from the source momentarily. Store it and deliver. It to a load. Now. One of our requirements, is we always want to soft charge CP, and so we're going to have to alternate. Are. Connected, in zero stages, with open stages in order to achieve that so. We'll look at an example here. In. This sequence let's say our first, stage. Is V, P equals V in minus V out in this, stage if it got a positive il we're, going to transfer energy from the output I'm sorry, from the input to the output, at. Some point we'll turn that off and we'll, start an open stage in which a positive il, discharges. CP all the. Way down until, VP. Equals, zero at which point we'll, start stage 3 at. Some, point the, inductor current will drop to zero and stage 3 and go negative after. That we'll start stage 4 and use, the negative inductor, current to charge back up CP. All the way to V, out at which point we'll begin stage, 5 and. Finally, we'll use the very end of our negative I'll have, to, charge. VP, all the way back up to VN minus V out and start the process over. Now. We're going to refer to these sequences, by the order, of they're connected, stage voltages, so. We would call this sequence VN minus V out 0. V, out assuming. The open stages exist between each one this. Is going to be our example sequence. Throughout this talk. Now. We might start to ask well what are all the possible, switching sequences, we could use in a P as a base converter to. Find that out we could permute all these different stages and figure out all the different combinations and, then, down select to the ones that actually, meet their practical, criteria, we listed before, so. The result of that is, a Six stage sequence, as you see listed here there are five set down and five step-up versions of these all. Of them require only, the PISA for energy storage all, of them always soft charge CP, and all. Of them are capable of efficient voltage regulation on, top. Of that each, of these can be realized with one or more practical. Topologies, which I'm, going to show here, each. Of these topologies, consists, of four, uni-directional, blocking, switches some, of which may, be implemented with diodes, this. Is on par with the switch practicality, of many other common converters, how. Many of these converters, are mirror images of each other some, look like other converter topologies, first, which capacitors, or, resident converters. But. These are the topologies, that are needed to realize those switching sequences. Okay. Now, that we have a set of switching, sequences and topologies, we'd like to constrain them for optimal behavior so we can start to compare them. Efficiency. Is one of our biggest considerations. When choosing a converter implementation. So. We would probably like to constrain, for additional. High efficiency, behaviors. Now. Our switching sequences, have, already been constrained for soft charging, that was step one listed here but. We might also like to constrain for all positive instantaneous.
Power Transfer, and what, that means is at no point in a cycle are we going to be shoving energy back into the source or, pulling it back out of the load. Also. We might also like to constrain, for soft switching or, zero, voltage switching on all of our switches in, reality. Switches, have capacitance, between their dreams and sources that stores, energy when, there's folded across the switch so. Similar to the piezo CP, we. Would like to charge. And discharges these. Capacitances, through residents, to reuse the energy, they. Store, now. To do that we, need to resonate switch. Voltages, to zero volts before we turn that on so we call it zero voltage switching. Now. In order to accomplish any of this we, first need to understand, what's, happening inside the piezo itself and. We're going to look at that with an example switching. Sequence VN minus V out zero V out with. V out a bit, less than half of BN that. Switching, sequence requires the topology you see here now. We're, going to visualize what's, going on during each stage of the switching sequence using, state planes which, you can think of as roadmaps for the piezo state variables, we're. Going to plot one value, of a variable against, another and not worry about time here, now. Each stage has its own. VP. Requirements, and il, direction, so, for stage one for example we're. Gonna say VP equals the N minus V out and, if. Il, is positive, then energy is going to flow from, the source to the piezo to the output like shown in this diagram. That's. Going to result in the following state, plan trajectory, where on the Left state plane we, see VP, fixed, at a constant, V minus V out but. Il Rises, and then. On the right state plane we see il resonate, in an arc with, VC. Now. Without going into too much detail, here we, can craft the state plane stage-by-stage, by. Looking at these constraints, those, constraints, dictate, what, the trajectories, are going to look like and finally. We're back where we started, now. Requiring that the PISA States ends, where they begin means that constraining.
For Periodic steady state that. Means every, single cycle, the PISA state variables, are going to take the same exact, path around the state plane. From. Here we can start to constrain, for the positive, behaviors. We want in terms of efficiency so. The state plane on the Left gives us a good, insight into how. The piezo interacts, with the source load system and. As. You can see here we've constrained, for soft charging meaning, we're always resonating. VP. To its next stage voltage, before, that stage begins and if, you were to look closely you'll also see that we've constrained for all positive instantaneous. Power transfer. Now. The state plan on the right tells, us where we are in the residence cycle at any given time and so, we can actually use the proportion, of the cycle that each stage comprises, to calculate, what our switching, times need to be. Now. In addition to this if we decide that we'd like to constrain for soft switching or zds and all of our switches we, can use the state plane to better understand, how we would need to alter, the switching sequence to achieve that in this. Particular case we needed to alter our sixth stage and so we needed VP to swing all the way up to the N at point six B for. ZB s turn on of s1. So. Now that we've constrained, the state plane it can, tell us a lot of information about what our time. Domain waveforms should look like for. These high efficiency behaviors. Each. Of the stage transition, points are numbered, on the state plane and these, correspond. To the number of dotted lines you see and the time domain waveforms, so. Even, if we can only measure VP, experimentally. We. Can still get a really good idea of, what. VC, and AIA might be doing based. On whether VP is increasing. Or decreasing and, a slope when doing so. Now. If we plot out the efficiencies, of each switching sequence constrain. For these positive behaviors, we. See that there's quite a bit of variability between, the sequences, themselves, now. The highest efficiency, sequences. Tend, to better utilize. The P as resident, cycle for energy transfer, so, they spend less, of the cycle and treat all stages like zero stages, and open stages. These. Switching. Sequences, are the ones we should choose in that case for high efficiency converter, design. Now. That we've how to best utilize the piezo a power converter it's time to actually build a converter. But. Before we do that we need to find a piezo, for. Our first prototype we, sought low frequency, piezo is close 200 kilohertz for, the ease of control, and, we also sought high mechanical quality factor materials, for the efficiency applications, now. Pieces can be difficult, to source but the one that we found to be best for our purposes. Is. A, modified, pzt material. 114. Kilohertz, radial. Resonant frequency, and so we operated, it in a radial, vibration, mode and it. Has a mechanical quality factor of 1500, and that's, actually an order of magnitude higher than a lot of magnetic components. So. We take this piezo, and we, built a pretty simple prototype with it this is just a 2 layer board it's got Gansett Schottky, diodes and everything.
You're About to see here is an, open-loop operation. Now. When we run this prototype in the real world now we're, getting waveforms, that not very closely in form with our expected waveforms. For the switching sequences, we want, and on, this top set of waveforms here are the simulated waveforms the. Bottom set are the experimental, the. Similarities, we're seeing here are great news for being able to use these switching sequences, to, get the high efficiency behaviors, we desire. In. Each, of these waveforms we're seeing VP always, resonate to its next stage voltage meaning, we're, getting soft charging of the piezo we. Also see all of our switch, nodes resonating, to their next stage voltages, meaning, we're getting soft switching and based. On how BP behaves, we can infer. That we're also getting all positive instantaneous, power transfer. So. This particular operating, point has an efficiency of ninety seven point one percent. Now. If we look at our experimental waveforms. As we change voltage conversion ratio we. See that our zero stage increases, in length in this. Top waveform, here we see our. Zero stage is almost non-existent and, that's, because we're at a conversion ratio very close to point five but. As we decrease that. Conversion ratio we, see our zero stage become much more dominant this. Is how a Six stage switching, sequence regulates. Its voltage output. Now. We can plot this prototypes, voltage conversion ranges, as, a function of frequency to, map out the converters, whole operating, range of region now, this is not purely, frequency, regulation each, of these points have been tuned specifically, for this switching sequence. Characteristics. We want but. We see that as this, converter, can span quite a wide range of operating points, while. Still achieving the, high efficiency behavior, as we want. Now. Here we plot efficiency, versus power for, a variety of set down videos below. 0.5, and we, see that efficiency, tends to decrease as gain does here if we look at the general trend, but. Overall we can get a decent. Range of good, performance, across a, wide. Range of power levels now. The downward, spikes, we're seeing here occur at spurious, frequencies, which. Are caused by imperfections, and the piece of manufacturing, or how we mounted it in this first prototype. Now. If we look at how efficiency, varies with, power for, gain greater, than 0.5, we. Find that the gain has no impact on efficiency. At. A fixed VN and we can get an even wider range of, power, levels with good performance, this. Is great news for regulation, purposes, here, we can span a gain range at 2 with. Essentially, no change in efficiency and it's. Worth pointing out that our peak efficiency here. Is above, 99%. That's. Much much higher than, most converters, so. At. This point we've shown that piezo, based converters, can achieve high performance through. Effective utilization, of the piezo but. There's still a long road of development ahead from an answer is a ssin. Some. Unanswered questions include. How, would, we choose a Pisa what are the best geometries, best materials, or. How would we institute closed leaf control to get all of our high efficiency, behaviors, we'll. Take a closer look at that one. The. Behaviors, we want require. Elements. Of all three of the different control schemes you see here modulating, frequency modulating. Pulse width changing. Dead time all of, those together are pretty, difficult to implement. We. Might also ask well what happens when we scale at power or what. Happens when we scale up frequency. The. Resonant frequency, of a piezo depends. On its geometry. Specifically. The length of, the. Piezo, is length in the direction that it vibrates so. Here. I'm showing that for a thickness cyber ational mode and as, the. Thickness decreases our, frequency. Increases. So. This is definitely something we would need to do in order to miniaturize, a converters, we would need to drive up the, piezas frequency. In. Addition. To that we might ask well how are we going to handle high step up or step down ratios, if our efficiency, starts to go down or, how, would we get isolation, it.
Turns Out that both of those needs may be able to be met with, what, we call piezo electric transformers. Piezo. Electric transformers. Are two-port, devices, that. Typically. Have ports along different dimensions, of vibration, but there's also other ways to do it and in. Addition to the advantages, of the resonators, or single port devices they. Can provide isolation. As, well as an inherent voltage conversion range, voltage. Conversion ratio I'm, sorry but, you see down, and the. Electrical model shown at the bottom. So. With that we can conclude that we can design a converter, based on a single piezo, with, high efficiency behaviors. Like soft charging, 0 - switching and all, positive, instantaneous, power flow but, also practical. Characteristics. Like voltage regulation capability, and practical. Switch implementations. But. The big. On point here is that it's the effective, utilization of, the piezo that enables, the competitive, performance we're seeing with no additional, energy storage, components, and I, should emphasize that the. Piezo we use in this prototype is just an off-the-shelf, part, so. We could expect. To get even better performance if, we actually optimize. The piezo itself. Now. In terms of miniaturization, this is one step in a long process of, development that's going to have to happen but. We do believe P as a base converters, are promising, and that they may in fact open, new application, spaces for power electronics. With. That I'd like to acknowledge our sponsors Texas, Instruments, NSF, and the Masdar, Institute a. Cooperative, program thanks. For your support and thanks, for your time. Are. There any specific questions, that the. Audience would like to first we'll bring up if so you. Can do so by either sending. A chats or by. Raising your hand and, I, will call upon you I see John kesakhian has. Raised his hands Oh, John, please uh take. Over and ask a question. Great. Jessica. Great. Presentation. In. Spite of the new glitches. And. And I'd like to talk to you a little bit happier. Freed up, about. The mechanisms. Because I have a challenge. But. My, question, is you. Did not indicate the frequency, at which you were. Operating. Your experimental. System. I assume it's. At the radio frequency. Radial. Resonant, frequency, of the piezo electric resonators. Correct. So. It's not directly, at the, radio frequencies, but it's very close and, in fact it's slightly above, the radial frequency because at that point we're, operating in our inductive region, for the piezo, there's.
A Very, narrow. Inductive, region and so operating. In that region is what enables us to get the charging. In the DBS. And. So I. To quantify that that, that's what you were asking yeah we were operating probably, between about. A hundred eighteen kilohertz to a certain, 25 kilohertz so very low frequency, for now right. Thank. You. Thank. You. So. Ali. There, is a question in the chat would you like to just, unmute yourself and ask it directly. I'm. Repeat. One more time and, having some trouble hearing. So. The. Comparison, between a MEMS based optical, switch and a piezo, switch you're trying to see a value of, the, piezo material, being inserted inside a MEMS type structure. Is. It is there a side-by-side comparison you. Can make those. Immense. Yeah. So in terms of optics. Or anything like that I can't speak for those, specific applications, but in power, conversion at least, piezo. Tend. To give us a really, good confirmation of those power density, and efficiency, and. So, in. Terms of other men's devices, we haven't explored as much just because. Perhaps. But fundamentally, there doesn't appear to be an immediate, advantage there, so. It doesn't mean they're I mean, it. Just, pieces. Or what, has out. So far that. There. Is a another, question just came in first of all by starting, by saying things, were great and clear presentation, and. Then, second the question, itself is. Have. You considered, the dielectric, loss when adjusting, switching, frequency. That's. A very good question a very good point, so. We have not looked into that specifically for this prototype but biologic loss is a major component that, will need to be considered. As we, go to higher frequency, as you go to higher dosages. So. What I was showing in this presentation. Was. Assist to just a mechanical, loss in the model that we showed but for. Everyone else in reality, across the capacitance, there, was also a, leakage, through the dielectric of the capacitance, in which, we're going to lose energy the capacitance, is storing. So. Not. In detail we have not looked into that but we certainly need to. If. I may also add another question that voltages. That you're operating, at you have shown us that as you get to higher voltages. You get even higher performance, responses. Meaning the efficiency, keeps. Improving, which, is wonderful, at. The same time a miniaturized, electronics. Right you would imagine would be using. Portable type, applications, so. Is there a limit to the how. High the voltage, should be from. Perspective of a user of handling this particular miniaturized. System, or, is that simply not the case because really. What you would do is use these in more isolated systems like nano. Sacks or you. Know MRI systems. Yeah. So I think I. Think, there's there's two questions there that I'll attempt to answer so the first is that. Well. Okay so I guess the first, point I would make is that the fuses themselves, like high voltage. Low power and. So that dictates, what. Applications, we would want to put them towards, so. It's likely that's a very low voltage tiny, stuff may. Not perform as well with the piezo, now. In terms of consumer electronics or anything else, the. PISA is not going to necessarily. Dictate, what, voltages were using it's just going to be a matter of disappear, those what, our voltage need is there and. So, it. Certainly. Will, be a trade-off in the low voltage situations. On whether or not we need to use. Something like this or whether a different, solution could, be better but, the, high voltage low power is. Where the giant. Well. Jessica. Again thank, you for a fantastic talk, I very, much look. Forward to seeing all of you next. Week when. The nano explorations, continues Jessica. Again thank you today thank. You so much for coming Christina.