Bayley Wang: right, welcome, welcome. this is another episode of our podcast and with me today we have Austin Brown. Austin, would you like to introduce yourself?
Austin Brown: Sure. my name is Austin Brown. I No Bailey from doing some amount of MIT with you. we both worked in the miters hacker space. A lot of fun stuff came out of there. Anyways, did six years at MIT, Undergraded Masters. Masters had my own company doing various consulting stuff for a little bit. I did a lot okay, back at MIT I did mechanical and electrical stuff. My my my biggest interest in life is the flow of power, whether it be electrical or mechanical, and especially the intersection between those two I find really fascinating. so anyways, motor stuff, we'll get there in a little bit, but at the moment I do motor design at Boston Dynamics. and that's a pretty cool spot, but I also do a lot of stuff outside of work. Like solar power stuff, which is right there, and hydroelectric dam stuff, which we can get
Bayley Wang: Yeah, yeah, all right. Cool cool, yeah. So Austin's on this on this podcast with c podcast with us today because he's one of the rare machine designers that are still based out of the US, like electric machine designers. It's one thing to buy a motor and like write firmware for it, it's another thing to build a motor from scratch and really understand all of the optimizations. And I think this is a really interesting feel to a lot of you guys. Obviously state of the art robots are electric, the old like whatever hydraulic drive robots are sort of fading into memory now. But motors are also really important for things like renewables and electrification. And potentially even, you know, this power generation kind of nonsense. All these hot topics which sort of interplay at the intersection of energy, real estate, geopolitics, and sort of move move the world forward quite literally. So I guess we'll start with you know, machine design. There's a lot of electric machines, there's a lot of motors out there and a lot of opinions about them. Let's talk about you know, various misconceptions in electrification and motor design.
Austin Brown: before going into misconceptions a bit, maybe we should just start with like what is a motor. At
Bayley Wang: Yeah, yeah.
Austin Brown: the highest level, Like what's the point, right? So generally you have either mechanical or electrical energy and you want the other one, right? so what a motor does, the motors that we're all familiar with, they take electrical current, put it through a coil which generates a magnetic field. And magnetic fields are special because they cross the electrical and the mechanical domain. by flowing the current you generate You can generate forces, right? And you can do mechanical work. So that's that basically the energy. If you're trying to use a motor to make mechanical work, the energy flows from electrical, generates a magnetic field, and then the way I think about it, you could generate a magnetic field, and then you extract mechanical work by basically collapsing that magnetic field. You like a a magnetic field stores energy in a volume and And by collapsing that magnetic field, you like by you know physically moving one part of the motor to collapse that magnetic field, you reduce the volume of energy stored and then that is output as mechanical work. And by having multiple phases of your motor, you can you can sort of do that process continuously and generate smooth output power. So that's sort of And then you know, obviously the process is exactly the opposite if you're doing power generation, which is you start with mechanical motion, you can like use permanent magnets or electromagnets to generate a magnetic field, and then it goes the other way. So I think a lot in terms of misconceptions, I think the biggest thing that people really need to think about is a lot of times people say, like, you know, I want a motor with a lot of torque, therefore it has to have like really high amperage, right? But that's not really true. basically a motor is very much like a transformer. the more numbers of turns you get, the higher the voltage, but the lower the current. And you can think of like a one turn motor for example, like one turn around each tooth. that would obviously have extremely high currents because you'd be talking about like fat, you know, copper bars. and very low voltage Voltage output, which you know, if you had a one-volt battery, that might be really, really useful for you. But usually you don't have that, so you do multiple turns. anyways, you know, there's all different ways of making motors. more recently, the sort of state-of-the-art has been I mean, really just getting the efficiency up For a given power output decreasing the volume. I think that's what people care about mostly. you know, in the context of a car, like other than the battery, the motor is probably the heaviest single part. also it's you know, it's obvious it's made of copper, right? So it's very expensive from a material standpoint. and so people basically want to get the motor as small as possible. in robotics, I mean that's that's even more true, right? So, you know, say you have Like a hip actuator, it obviously to fit inside the robot. And if you want to make a robotic hand, there's a huge amount of stuff going on that in that field, which is you know, do you want to put the motors in the fingers, right? Then they've got to be really, really small, right? and you know, also be very efficient to dissipate as little heat as possible. So that's sort of the the lay of the land these days. tons of work being done on that, although it's kind of a weird field because I think as Baylor. said before it's it's rare that people sort of come with the both mechanical and electrical. Like usually like you know you have the electrical people over there and you got the mechanical people over there. And then you know there's not not that many people in the middle. so you know it's it's kind of a niche thing. but it's also weird basically you you almost don't need it ever, right? Like if I'm just making a generic household appliance, I would never design a motor.
Bayley Wang: Yeah, yeah, right. It's a complete waste of time. A very bad idea from a product standpoint to design your own motors with custom laminations, custom magnets. Takes about a year, a year, eighteen months, two years even to get everything running reliably. If
Austin Brown: I mean I don't even know if I'm
Bayley Wang: you want to like shift a product very quickly, like it wouldn't make sense for a toaster. Toasters have continuously rotating parts.
Austin Brown: For almost anything. I mean motors come in a zillion different shapes and sizes and you know you can buy you can buy anything, right? it's just when you want that last whatever percent of performance that but I mean, you know, building your own motors is not unheard of, right? Any b company is building their own motors.
Bayley Wang: I mean we we say that but very often the motor I want is not commercially available, which is really annoying. It seems like the the list of commercially available motors is all the one all the ones except like the two I need for the project.
Austin Brown: Yeah, yeah. Basically any any industrial motor is stupid to design yourself, right? Like that's one hundred percent true. Like you can buy any NEMA frame, p you know, ten thousand pound, whatever, like that. what are the big industries like robotics these days, automotive and industrial Yeah.
Bayley Wang: There aren't a lot of state of the art industrial motors, I would say. The industrial guys are still for the most part running on older designs with lower power density and much worse efficiency curves or heck induction motors plugged straight into the wall. Like the the torque output is whatever it you whatever it wants to be at that time.
Austin Brown: Right, yeah. I mean
Bayley Wang: That is actually a hot topic, right? This this idea of modernizing the electrical machines and factories. moving machines in industrial settings is a very palpable fraction of the world's power consumption right now.
Austin Brown: That being said, large induction motors can be fairly efficient. small induction motors are just like total garbage in terms of but big induction motors are as far as I know in the ninety, ninety five percent.
Bayley Wang: I I do I do remember that it's surprising how good they get without any form of like control. They they've been like carefully designed to pr produce the right number of newton meters at 1750 RPM.
Austin Brown: Yeah, yeah, right. It's it's although they have weird characteristics. we're actually learning this at the dam.
Bayley Wang: give the listeners some context about the dam.
Austin Brown: Sure, okay. So some context about the dam. So I work at Boston Dynamics right now, and what I said in my interview, I was like, I really like motors and I have a l you know, robotics sounds like a lot of fun, but I'm actually more of a power guy than a robotic guy. Those two things are a little different. and I mean, yeah, I'm actually having loads of fun at Boston Dynamics, but that's kind of a separate thing. So one of my biggest sort of side projects that I do for fun is me and two other friends purchased off of Facebook Marketplace a small hydroelectric dam.
Bayley Wang: That's gotta be one of the best, best Facebook marketplace purchases I've ever heard of. Like what did you buy off Marketplace today? A damn.
Austin Brown: A dam, right? So it's been a long side project. so we have gotten it sorta quasi-running, right? Our peak power is about sixty-five kilowatts. The facility was supposed to produce about eighty-five. So we're running a little low. I don't know why. We've sort of been getting it working slowly. It's it's a group project, so you know things never really move that fast. Basically in terms of like physical dimensions and stuff, I think we've got like an eight acre reservoir. we've got I think about eighteen feet of drop and depending on the season, our nominal flow is sixty CFS cubic feet per second. So that's I think that equates to like two cubic meters per second of water. it's pretty cool. Right now it's fall, so We're running a little bit less water. So we we have I think 40 CFS right now. But that's sort of what it is in terms of like what does the dam actually look like? it is what about 50 feet across and it's made of granite blocks. There's a three-foot diameter propeller looking thing, that's the turbine. There's it's actually kind of interesting from a power perspective. So three-foot diameter turbine that goes to About 500 RPM, and then there's a what? I think it's like a 5 to 1, maybe a 5.6 to 1 speed increaser planetary gearbox. And then there's an 1800 RPM induction motor. And the induction motor is actually kind of nice in this capacity because in order to connect to the grid, like it's literally just a massive contactor, and you just bang it, right? The thing that's nice about that is in a lot of larger power generation facilities you have to sync things to the grid. Right? Induction
Bayley Wang: Yeah, yeah, right.
Austin Brown: motor that doesn't exist, right? You just bang. so we
Bayley Wang: That's clever. That's clever.
Austin Brown: pre-spin it up to the proper RPM, but there's no sort of phase synchronization that needs to occur. So I did all the controls for that project. I have this like big like hydroelectric dam controller circuit board. Yeah, that's kinda cool. the the weirdest thing about it, so the induction motor that's on it, like normal induction motor, a normal induction motor, eighteen hundred RPM, right? And they sag by fifty, roughly. Yeah, depending on the load. But this one, because the power's going the other way, boosts by fifty. But but still it's weird. Like people see that and they're like, What? That's not right. So
Bayley Wang: I I g I guess you have to be a really special person to see that and be like, wait, that's not right. Yeah.
Austin Brown: Yeah yeah yeah.
Bayley Wang: Most people don't know the lore, but the yeah the the slip going the wrong way is a really interesting concept here. I yeah, I I think I think this is a good point to talk about the I guess various use cases, right? You've talked about the dam, which is sort of this more traditional like constant RPM power generation use case like similar to industrial stuff. We've briefly touched on E V traction motors, which are more dynamic and more state of the art, and then on the other end you have the robotic stuff that you're working on right now, where basically everything only exists in a start stop condition and And there's like virtually no phases of constant velocity. maybe like I'm wondering whether like have you noticed any sort of large overarching design paradigms as if you worked on these three different types of machine?
Austin Brown: Yeah.
Bayley Wang: I'd say
Austin Brown: Say the biggest thing with the robotics thing is that the hardest thing is deciding what your duty cycle is, right? with a car it's fairly clear, like the hardest thing is, you know, in the summer driving up a hill at eighty miles an hour, right? Like that's that's sort of like n what normal people do on a car. And and you know, nobody really drives up a super steep hill at a hundred miles an hour, right? Like that just
Bayley Wang: Yeah, yeah.
Austin Brown: happen. The thing with robotics is it's like, you know, okay it did a backflip. That's sick, right? can it do a double backflip? Or like, you know, do you really need to do a double backflip to like work in a car You know, so that's that's the thing and and there's a lot of like thermal transient stuff, right? You know, that that stuff is a little bit harder to it's not hard to design for. I think it's more hard to set the requirements for and and either
Bayley Wang: Yeah.
Austin Brown: but I think also the overarching thing is nobody really knows what a good robot is gonna do, right? Like yeah.
Bayley Wang: Very true.
Austin Brown: That's sort of an active question in the industry right now is I mean there's huge loads and loads of different sizes and shapes of robots, right? So yeah. in terms of like overarching design philosophies There's not much. a lot of people have gone in a lot of different ways. I think the only thing that I would kinda say is people have mostly gone towards this like quasi-direct drive situation. you know, I mean, but that's that's kinda old news at this point. It's just you have like
Bayley Wang: Yeah.
Austin Brown: large pancake e motor, and then a relatively low, you know
Bayley Wang: Yeah, it's like practically synonymous with no harmonic drive at this point.
Austin Brown: Yeah, basically, right. So, you know, that's that's basically what people go for. in terms of like more interesting stuff, I probably can't get too much into that because that's more boss and dynamic stuff. but in terms of like where the rest of the industry's going. Yeah, I think I Honestly most of the motors that people are using Hallbach arrays are I think one thing that people are more starting to do.
Bayley Wang: They're they're yeah, yeah, they're they're they're they're a bit fringy, but they do get you more torque. Like they're they're hard to build and no no one no one's no one's shipped one yet, but they give you good numbers on paper, which I think is like if you're a if you're an actuator vendor, right? That number is y the heart and soul of your product. So being able to put out that you're two times better or like whatever, one one and a half times better than the competition. it it does look like the metrics are better on paper, although I have no idea what integrating a Hullbach actuator into a robot looks like.
Austin Brown: Hallbuck arrays, I mean I can tell you this from personal experience are real pain to assemble. I have put together
Bayley Wang: I don't I I can't actually I can't actually think of anyone that's shipped a Hallbach array in like serious volume.
Austin Brown: Yeah,
Bayley Wang: people have tried.
Austin Brown: They seem to be I mean so now we get into the whole what I call the spec dicking war.
Bayley Wang: Yeah, yeah, yeah, right. Figures of merit we call them in
Austin Brown: Right, yeah. So I mean when you want to go down that road, like haulbacker ray it up, right? You know, let's go. But you know, i it's sort of at the same time, you know, the robotics industry right now is very much into investor hype situations.
Bayley Wang: Yeah, yeah. So Ugh Yeah. Tell me about it, right? The the investors come and they ask why the robot is different. You mutter something about Hallbach arrays and show them a nice FEM simulation, they're like, wow, he's smart.
Austin Brown: I mean obviously I think actually Boston Dynamics is doing pretty well or you know, going about things in the right way, I would say, you know, not not only focused on these figures of merit. Although, you know, we're speaking broadly here, right? Like there are many companies out there that are you know, solely focused on this sort of like Weird technical stuff. I think in terms of industry trends though, where people are going, the people that are building their own stuff, basically from a motor perspective, there's kind of two knobs that you can really turn. one is the magnet, right? And the knob that you can turn there is higher grade magnets and rays, right? backing up one step further, fundamentally
Bayley Wang: Let's actually back up two steps further. I have a question for you. in terms of like paper figures of merit, how much better are Hallbach arrays versus a like a traditional steel back motor?
Austin Brown: It really depends. well, I think
Bayley Wang: 'Cause my my understanding is peak torque doesn't go up, right? You're still limited by saturation in the teeth and like the overall volume in the of the field.
Austin Brown: People really like to say that saturation is sort of this like hard limit that you just ram into and in my experience that's basically never the case. you are so limited by thermal stuff, like way before you get into any type of saturation.
Bayley Wang: Introduce.
Austin Brown: Yeah. Basically, yeah. saturation. I mean, one time I think I I wrote a s this is my master's thesis. I wrote a simulation that was, you know, for a seven newton meter motor. And I think I accidentally like put a you know, it was for like a fifty amp motor or something, right? And I think I like had screwed up the number of turns or something, so there was a extra factor of ten, right? And I put in, you know, like I don't know, ten thousand amps I think I did a switch. sweep that ended in like ten thousand amps or something. It was like some huge number, right? And surprisingly, the motor was like more linear than you would expect up to like really, really high currents. I think it's kinda
Bayley Wang: Yeah. What happens when you saturate the steel core, right? The the the beast beast goes up in air. Like the steel becomes air.
Austin Brown: Yes, but I think more interestingly, the best way to think about it is Like, okay, when you have something that's saturating It's like there's all these little magnetic domains that are aligning. And like it turns out in steel that like steel saturates it starts to saturate like one point five, one point six Tesla, but has some sort of saturation left to go. I think all the way up to like one point nine or like two Tesla or something. So it turns out there's quite a bit of like, you know, knee in the curve. Like it's a relatively soft thing. The other thing that's more important to consider is that Think of your D and your Q axis in field area and control, right? The magnet is putting a whole bunch of flux down the D axis, right? And your windings are putting a bunch of flux down the Q axis. you start at zero, right? And you're you're like bringing this like vector of flux, like the total flux, I guess, is kind of the hypotenuse of them. So you start like it's actually to generate that amount of flux with the q axis is like a huge amount of current. on small motors because basically magnets have so much like current capacity or flux generating you know capacity sort of inherent to them that but basically long story short saturation is way more of a soft curve than you would ever expect. I think only if your inverter like shorts out would you ever like really run into saturation because I mean, you know, even if you're putting in peak current for like less than a second or something, right, you still do have to worry about thermal transients. Long story short,
Bayley Wang: Yeah, yeah.
Austin Brown: saturation is I wanna say it's kind of a red herring. For most things, I I seriously
Bayley Wang: Yeah.
Austin Brown: that this is a extremely scale-dependent effect. as the scale goes up, basically like all robotics actuators are like this size or smaller, right?
Bayley Wang: Yeah, yeah, right.
Austin Brown: As you get bigger, the physics work out such that your yeah, saturation just becomes way more likely. Like it's a sur it's basically the surface heritage volume thing, or it's a long story short, the copper conductors become more powerful and more able to drive flux in in comparison to a magnet. Yeah.
Bayley Wang: Inter interesting, interesting. We can go. So at at the at the sort of like robotic scale, which is incredibly fixed, like the the the motor ranges from between seventy and one hundred and fifty millimeters in diameter. It seems to seems to mostly be the case that you're bound by other factors and it's like not actually possible to hit saturation torque out of the actuators.
Austin Brown: even saying hitting saturation is kind of like it's sort of a a fake thing. Like if you're using
Bayley Wang: Yeah.
Austin Brown: an inductor, like a piece of ferrite really does have like a you hit it, right, and then it saturates. Whereas with it you know, with a motor, like I've heard of of torque reductions of like twenty percent at high at high torques for, you know, some types of motors, right? but, you know, that's not like that's not a h hitting a hard edge, right? for all intents and purposes here, like
Bayley Wang: Pretty pretty pretty soft. Like you can you can you can still keep going. That's that's very interesting. I wonder you're
Austin Brown: Right. If you have an inverter which is very finely tuned, like to work great, you know
Bayley Wang: Definitely, yeah, yeah.
Austin Brown: if
Bayley Wang: Those inductances changing really gives you a bad time. And it it's
Austin Brown: Right, yes. That's potential.
Bayley Wang: Yeah, yeah. I've always sort of assumed that the hip actuators and these big robots was like running pretty clear f near physical limits, but it sounds like maybe maybe that's not the case. And it's still sort of a as design limit.
Austin Brown: I would say also these things can obviously be true or not true if your actuator is badly designed.
Bayley Wang: Yeah.
Austin Brown: In BattleBots one time we had basically that was on uppercut, which I'm wearing the shirt. that was the BattleBot. I helped design the power system for that for the weapon on that robot. the previous generation of Uppercut before I had joined the team, they were it basically like big outrunner motors. Those things really saturated. I think they had extremely thin Stator teeth and the controllers would actually freak out. That was the thing that would limit the torque.
Bayley Wang: Yeah, that that that makes sense. You basically have no inductance once you saturate that badly, right? Right. Yeah. The what I I'm actually curious now that I think about it, right, what happens at saturation? Do you just get no longer get more torque per amp? Like where where where does the flux go?
Austin Brown: I mean again, like I I think saying at saturation is is kind of a a false flag. Like there's no there's no like BAM I've run into s you know, it's it's such a slow roll off for the
Bayley Wang: Okay.
Austin Brown: So that's But but you know in terms of what actually happens, yeah, basically I think the best way to think about it, the most accurate, which it took me a long time to kind of reason through, is that saturation only really occurs when so okay, in a normal permanent magnet motor, the flux goes down one tooth, through the permanent magnet array, up back the other adjacent tooth, and then back over through the back iron, right? So there's quite a lot of magnetic reluctance through the air gap and through the permanent magnets, right? Saturation only starts to play when the reluctance of the teeth and the back iron and the other tooth starts to become substantial. compared to the reluctance of the the permanent magnets.
Bayley Wang: That makes a lot of sense, right? Because the the air gap has a mu of one. so it's it's it's whatever, thousands of times effectively longer than this than the stator teeth. So you can actually sacrifice like 10x, 20x, dropping the perm permeability of the teeth before it it starts being comparable. That's really interesting. That's really interesting. That that's that that's pretty cool. So sa saturation really is not the right way to think about it for like machines built out built out of these soft but magnetic materials running at plausible. Currents. I think I think the I guess macroscopic time scales. Like ro robot like time scales. I think that's the real thing, right? A a robot takes like six hundreds of milliseconds to complete an action.
Austin Brown: Correct. And the other thing that I'll say here is that it really is scale dependent. So, I think, yeah, people should be aware that basically like if you take a C core, like a magnetic C core, and you put like some number of turns around it, and you have like some air gap, you'll get and you you know, you energize those clothes with some current density, like one amp per square meter, let's say. You'll get some sort of I mean, this is like a very bad motor, like if we put a magnet over here, like it would go into the gap. But anyways, let's just consider the C. If you have some current density, you have like ten turns, if you then take so you'll then you'll get like, you know, let's say one Tesla in the gap, right? Or you call it half a Tesla for this one. If you double every single dimension, you just double everything. You make the wire go from one square millimeter to four square millimeters, right? you know, it like it gets like into the page, blah blah blah. Everything gets bigger by a factor of two. Every linear dimension gets bigger by a factor of two. Turns out the flux in the air gap will actually double. because you have you can go through the math, the reluctance goes up by a factor of two, but the the current going through the C core has gone up by a factor of four.
Bayley Wang: yeah yeah yeah.
Austin Brown: That's what gives you so anyways, all that is to say, but a sort of a permanent magnet kinda is the same at any scale. So what that means is that as you get to like much bigger motors, saturation becomes way bigger of an effect because all of a sudden now your copper conductors are much more like powerful. Or are you like thermal you can run lower current densities to get the same flux. So that means you're generally Generating the windings, so then you can pump them way harder. and that's when saturation will start to play. Personally I don't have a lot of experience with motors above like ten kilowatts.
Bayley Wang: Yeah, yeah.
Austin Brown: It's never been a problem up to those scale of motors, but I I do know that the physics say that as your scale gets bigger, like I'm sure this a hu
Bayley Wang: I see, I see. So so with like a E V with like a E V size motor, your peak torque might become closer, closer to like a a saturation limit, so to speak.
Austin Brown: Correct. Yep. The other thing that's kinda weird, I actually tore down well, it was one of your motor I think you had purchased one for miters like way back in the day. if you look at like a an EV motor, the like the proportion so this is to illustrate the scaling. If you look at the proportions of one of those E V motors, like the stator teeth are a lot fatter than you'd expect. Like the proportion, like really small motors, you end up with very thin stator teeth because they need a lot of space for copper because you You just can't pump the copper that hard. On the previous motor that you bought, the the like teeth, I was looking at the teeth were, you know, I'd say like on a small motor, maybe they're like 30% like stator tooth by like, you know, if you take the circumference, it's like 30% stator. On an EV motor
Bayley Wang: Yeah.
Austin Brown: of like, you know, a reasonable size, maybe it's more like fifty to seventy percent because the they just don't need as much copper. So
Bayley Wang: Yeah, that makes that that makes sense.
Austin Brown: Yeah.
Bayley Wang: talking about EV motors, that's another very interesting topic, right? They live at a completely different scale, and as I understand, they have a there like a completely different set of desi design criteria for EVs where the big focus really is on day-to-day driving efficiency and not any of the specious metrics like torque per cubic inch, torque per cu watt torque per kilogram, right? Because the the car is very heavy, the drivers are very heavy, and the extra kilogram of motor affects How much you pay for the motor, but probably not the performance of the car. So I think that that that's another one of those interesting misconceptions that I I was really big about back when I was we were doing motor stuff together. That doubling the torque per kilogram of your car's motor basically does nothing because the car weighs much more than the motor. Like the the marginal weight savings is like a like a sub percent. But
Austin Brown: Yep.
Bayley Wang: I I I guess it's different for robots. Talking a little more specific, how much of the so like right if I make the motor quote two times better, but just the raw motor, does the actuator won't become two times better because there's like a lot of stuff which isn't the motor, right? There's the gears, the inverter, the housing, the mechanicals, the the like physical structure of the robot. so it's not quite the case that like really pushing for high motor performance on robots makes them infinitely better or like linearly better.
Austin Brown: Yeah, yeah. I mean I that being said, I would say that the motors probably are one of the heaviest parts of the robot. And also their mass is often located fairly distantly.
Bayley Wang: Yeah, yeah, the the the distal the whole distal actuators. I saw a humanoid at Acre last month where like great care was taken to have no distal actuators on the legs. Like everything was driven by linkages from the hips and it was a really really weird design. I don't want to be the guy who has to unbend the linkages when the robot falls. Right, right.
Austin Brown: Yeah, whereas Boston Dynamics has taken completely opposite approach. You know, you can see in the ankles there's huge actuators, right?
Bayley Wang: Yeah, yeah. The the I I do think it's like the result of it's the difference between like thinking about robots and actually using the robots, right? Like once you start putting the robots in the field and trying to do things with them, stuff like the b the ankle linkages getting bent become really not good. and stuff like cabling as well is a pain in the ass, but that's a that's like literally a topic for a whole other podcast is how to deal with the wires in your advanced robot quickly becoming the biggest problem in the industry. I don't know about you, but our our our our fleet has egregious failure rates. It's like a one one cable failure every hundred hours of our arm operating time or something. And the current industry solution is you send the robot back to the vendor who replaces the cables under warranty. And the vendor is like it's like cheaper for the guys in China to do this than to design competent cable ends. But anyway, that's a top topic for a different different day. Axial flux motors, that's a and I guess the the art of marketing your motor, right? Like I I bet some of our listeners are either VCs or like founders trying to start a startup. And the number of like motor and actuation related pitch decks I've seen that are just completely full of slop, like they measure the wrong thing. And I think increasingly i increasingly strange and over optimized forms of axial flux motor are the pinnacle of this.
Austin Brown: Yeah. So basically from a high level, if you are some sort of robotics person, you can more or less walk into a room and know that any investor is gonna know less about magnetics than you do, right? Right because nobody knows anything about magnetics, right? Like you know, walking down the street, like maybe to somebody doesn't even know what a magnet is, right? So if you just like make up some weird terms about Motors, like usually you can bamboozle people, right? the thing is, you know, I've seen a lot of weird motors, and it turns out I would say 98% of the time the simplest thing wins. I famously like to say a motor is really only made out of four components: it is neodymium, steel, copper, and air. And you will actually want as little air as possible. It's just that the two things, you know, the rotor and the stator can't grind on, right?
Bayley Wang: Get Rob
Austin Brown: So you know, you gotta have a really, really thin air gap. but anyways, like it turns out, you know, if you only are really working with three components it's not that there's like no creativity, but a lot of times it just comes more down to good implementation than some sort of like weird physics, right? So but what a lot of people do is they just say like we're gonna have, you know, the coils going every which way and somehow it's gonna magically generate a war torque. Turns out, I mean, okay, just with axial flux motors specifically, a lot of times, you know, people always ask me, like, you know,
Bayley Wang: Come on.
Austin Brown: I saw this cool thing with axial flux motors. What do you think about them? Are they the future? Absolutely not, right? the I mean axial flux motors make sense about three percent of the time. And s and sometimes that is important, but basically they're they have a few big problems. number one, they're actually a real pain to build because like they want to suck together. like the you know, all of the normal force in a radio flux motor, everything's balanced, right? And you see when you put a like a rotor in a stator, it wants to like clunk off to the side. That's because like ninety eight percent In an axial flux motor, it is one hundred percent.
Bayley Wang: Hundred percent of the stater tries to stick to the rotor.
Austin Brown: And so people try and do weird stuff like, you know, we're gonna have two, right? But then the motor is basically impossible to cool, right? Because or you it becomes water cooling mandatory.
Bayley Wang: Yeah, yeah.
Austin Brown: I believe it's Yassa is actually doing pretty good stuff with the axial flux motors. the one advantage that they really have is that it's oftentimes very easy to wind the coil as like a bobbin and then you sort of assemble the bobbin. That's pretty nice. y you can do that with Radial flux motors too, but it's it's quite a nice form factor with an axial flux. There's two other big problems with axial flux motors. number one, the lamination stack gets weird, right? So you are familiar with ETEC floppy rotor syndrome, right?
Bayley Wang: Yeah, yeah.
Austin Brown: Yeah. So basically like your laminations like in a radio flux motor, they're like stacked, right? and everything just kinda works. But in a axial flux motor, like the lamination is kind of like of like it either needs to be a spiral or it needs to be like concentric rings and you can sort of have it like go in on itself. So yeah, you get like all sorts of weird structural issues where because the you know there's huge forces on these parts, right? And it's kinda hard to like
Bayley Wang: If you've ever if you've ever tried to pull one apart, there's a lot of force on them.
Austin Brown: It's it gets weird really fast. and then kind of the most niche thing that I don't really like about Axial Flux Motors is they're very hard to simulate. they're you but you basically have to simulate them in 3D.
Bayley Wang: Yeah, you can't have a two D cross section.
Austin Brown: Right. A radio flux motor is very, very easy to simulate because it's it can be simulated very accurately in two D. So if you want to simulate a hundred thousand motors, you know, no problem, right? You can just do it in two D and it might take a while, but it's not that bad. Whereas if you want to do the actual flux motor, you basically have to do it in three D. You can do like a couple two D cross sections, but it's just weird. what's the point, right? You're doing all this work, you you've not fundamentally changed the physics. But I mean basic Like the way a motor works is you have a permanent magnet, or the motors were used to, permanent magnet, it's making flux through a coil, right? And the only thing you've done is just turned that whole thing sideways and made it harder to mechanically assemble.
Bayley Wang: Yeah, yeah, right. That that's that's what's sort of my impression that with a radial flux motor, you like can draw the 2D cross-section. Once you have a good simulation, it basically directly translates into like a manufacturable motor. Like that cross-section can be sent out to a vendor to stamp the laminations. With the axial flux machine, it's like a shit show of hiring a technician to try to figure out how to stuff the motor afterwards, and there's like it's like full of tape and glue and potting epoxy and the motor sometimes falls apart.
Austin Brown: You know, any motor If if your motor I like to say if your motor is easy to assemble, that means you weren't really pushing performance hard enough. the best motor is something that like if you really, really shove, like it just barely goes together. you want the whole stator to be entirely steel and copper, right? Like no air, right? Yeah. So, you know, either way, it's gonna be a pain, right? Or it should be a pain if it's a high high performance motor, right?
Bayley Wang: I guess another interesting question is we like to pursue like the endless pursuit of performance for everything we build. Do you think that's going to matter in like the first commercial sort of quasi direct drive robots? Or is the focus going to be more on like can we build this thing at scale? 'Cause I d I have a lot of Chinese actuators and none of them are like dense. They're they're mostly aluminum.
Austin Brown: Yeah. That I I don't know. I mean if there's one thing I'm not, it's not I'm not a production guy. I mean
Bayley Wang: Yeah. I I I can't claim to have ever put put put motors into serial production either, but it is it is interesting that we've like worked so hard to build state of the art actuators and like all that all that work is getting thrown away and being replaced with like a Chinese hobby motor inside some aluminum. Like that that's what that's what quickly what the industry is like I won't say it's just that, like Unitree has done some good advancements on manufacturability.
Austin Brown: I mean, yeah. To some extent, like you can always do something bad, right? And it'll always like kinda work, right? And then I think the real question that the world is trying to figure out is if you have bad hardware, can reinforcement learning just like work, right? and
Bayley Wang: I I still think you need to like have enough torque to stand up and lift the target object.
Austin Brown: I mean I have heard so honestly I don't really watch the news in terms of like humanoid robotics. I'd rather spend the weekend like chilling at the dam than like reading it all some treated robot, right? okay, actually going back to the dam, it's lakefront property, right? That's been the truest awesome thing at about the dam. You just like relax, you like hear the turbine humming in background and you're just like watching
Bayley Wang: Peaceful a peaceful weekend trip to the hydro plant with the turbine worrying in the background.
Austin Brown: I mean this is normal, right? Okay, honestly, the the coolest thing about the dam recently is I got this cycling routine working. So it like cycles until the water goes too low and then it shuts off and like lets the reservoir refill. So if the turbine like takes too much water you know, you you basically can't run continuously, right? So I got this cycling routine programmed. Anyways, I'm out in my tent, I slept right next to the dam. I like wake up, you know, I like poke my head out, and I hear the turbine like I basically hear like the main contactor like open at like which is a big clunk. It like it's like it's a a loud contactor, right? And then I hear the turbine like spooling down because it had run itself out of water and I was like, that's cool, right?
Bayley Wang: achieved a degree of autonomy. That's a it's a robot right there. Yeah.
Austin Brown: yeah, like unit tree stuff and I don't know, you know, people are doing all sorts of weird stuff. Like I've heard yeah, I have heard through the Great Prime that those issue that those robots do have pretty substantial thermal issues and that they can run for some amount of time but not steady state. and
Bayley Wang: I would believe that. I would I would definitely believe that I know for the for the ar for most of like the the whatever standard Chinese arms that the imitation learning industry uses, if you like keep them outstretched with full payload, they overheat in a few seconds. And if you squeeze too hard with the gripper, they literally overheat in like two seconds. it's it's not really a feature, but the idea is everything is only transient. Like li lifting heavy objects continuously is not part of the scope use cases right now.
Austin Brown: Yeah. if I if you told me pick up the you know, the heaviest thing you can pick up and hold that stretched, I could probably
Bayley Wang: Yeah, I I guess that's true, right? Like I think that's their argument. Like whenever you complain about this like my machine is overheating, they're like, Well, if it were a person it would also overheat. But that be
Austin Brown: Being said, you know, I think what's how tall is a unit tree? Like what are the humanoids?
Bayley Wang: The the little humanoids are about four feet tall, it's like one one one point two, one point three meters.
Austin Brown: I believe all of the things get worse as you scale up.
Bayley Wang: I believe so. I believe with longer limbs it quickly becomes worse.
Austin Brown: Like your surface air to volume ratio gets worse. And moving like even if you had like solid aluminum links, like moving the heat around inside the link, I think gets worse. So I I think actually that's sort of a hidden cheat code. Like just make the robot, you know, small. I think that
Bayley Wang: You definitely s you definitely see it. All of the commercially viable humanoids right now are like two-thirds scale people. They're like child-sized. And I g I guess you also work with full-size humanoids. We have a full-size humanoid in the lab, and it is a absolutely terrifying robot. 350 newton meters of hit motor, doesn't doesn't fuck around. Yeah. If the if the controller winds up and that leg is go that that foot is going through something, and hopefully it's not you.
Austin Brown: Yeah. I will not comment on anything related to Boston Dynamics. but our our motors do have plenty of torque, that's for sure.
Bayley Wang: The yeah, I guess we're almost almost out of time. I guess the last question, l last sort of topic of discussion that we briefly mentioned earlier was Dexter's hands. I think this is becoming an increasingly hot field of investment over here in Silicon Valley, but unfortunately investors have never been so bamboozled as much as by a guy building Dexter's hands. you see all kinds of combinations like motors and strings and like none of it is manufacturable, all of it probably overheats.
Austin Brown: Right. that I probably cannot comment much on. Yeah.
Bayley Wang: Alright, alright. I guess any any other intra interesting topics, like there's a lot of lot lot of ways we could go. We could talk about neodymium free motors, which I think are a hot topic these days. gosh.
Austin Brown: honestly, you know, I I don't know what's going on there. I mean, what a pain, right? Like you know, basically the rules are changing every day, right? it takes you so long to redesign the motor. And so if the rules change faster then you can redesign the motor, like you know, what are you supposed to do, right? hard
Bayley Wang: No, d the domestic manufacturing of motors is another another another hot one. 'Cause the right the the FCC recently questionably banned the use of drone motors that aren't made in the US. Which I mean it's o the ban's only gonna stick around for two years, but it's creating at least over here in the valley, there's like a lot of narratives based around doing it in the US. Like do do do you think this is ever commercially feasible?
Austin Brown: I mean again, this comes back to the whole thing of like it's commercially viable until the rules change again and it therefore it's not commercially viable. Like, you know, you could buy a whole bunch of equipment and then the rules change and it's no longer worth doing like
Bayley Wang: I guess that's true. It's a it's it's a it's a terrible model to base your business around.
Austin Brown: Right. Yeah. Like let you know, tell me that you want to pot buy like a whole bunch of capital equipment purely based on some rule that was like arbitrarily ruled out with no checking whatsoever and could just be repealed tomorrow for no
Bayley Wang: Yeah, yeah. That's true. A lot of the like if you had built a business based on the the the tariffs, for example, you would have had a bad time last month. As the the
Austin Brown: Right. Like
Bayley Wang: the tariffs like suddenly drop from fifty percent to ten percent and completely change the completely change the picture. I mean there there is an argument from like a national def security point of view that drones are becoming increasingly critical for war, and in the case of a war we cannot expect to buy drones from China. Which is interesting. And as far as I know, I I I guess Free Fly still builds their motors sort of in the US. I ha haven't checked recently, but I I I think you can get a Free Fly drone where the motors are like US compliant.
Austin Brown: Yeah, I mean again, it's it's not even a motor question. It's more of a supply chain question.
Bayley Wang: it's not really an engineering thing, right? It's like can you yeah can you can you get this working at a anywhere remotely economical here in the US? Funny thing, I think if your drone doesn't use radio frequencies, it's exempt. It's an FCC rolling.
Austin Brown: It's just you know, it's annoying when the rules change every day, right? I what I can tell you a bit about is making motors that don't use magnets. Like how do you do that, right?
Bayley Wang: Yeah, yeah.
Austin Brown: The two popular strategies, at least in the automotive area, are induction motors. So induction motors have like a conducting rotor. And the issue with those I believe is I think they don't scale down very well. I think basically the way you can think of an induction motor is let's say you're driving it at sixty hertz. In the case of a like big induction motor, you have like let's say two hertz of slip, right? That means that, you know, if you're putting in sixty Hertz of electrical energy, you're gonna get out fifty eight Hertz. So that that's like, you know, what, like a three percent loss right off the table. Yeah. I believe that that freak like the slip frequency goes up as the motor size goes down. I know that that's true with like big industrial motors, like like you know, megawatt class. I think they have very low slip frequencies. But as You scale the motor down, the slip frequency will go up, and you can only scale it down so much before your slip frequency. At some point, the motor just becomes small enough that it like it just doesn't work. So I'm not sure if that would work. The other way of building small motor. Is switch reluctance motors. And those are basically just like you replace the magnets on the rotor like steel teeth. And people actually really like those for like missiles. there's no multiple parts on the rotor. So they can spin at like 200,000 RPM, no problem. I think the Wikipedia page on switch reluctance motors says that, you know, in the sidewinder missile. One of those missiles had a generator that was a switch reluctance motor because they could like put the turbine exhaust like right onto the switch reluctance motor or something. Like weird you can do weird stuff like that if the rotor is is just a stack of laminations. It's there's no there's nothing that can explode, so they just, you know, push him really, really fast.
Bayley Wang: Th that that is that is fascinating. That's a that's a really cool feature. Yeah. I never thought about it. And I guess you you have no magnets to like have any currents at high frequencies either. Right. That's a that's a that's a really cool feature.
Austin Brown: Yeah. Yeah.
Bayley Wang: what do the PMs do in a motor anyway? Like they they you can build a magless magnetless motor, so like what what does the permanent magnet do? It's not like making energy, it's not like storing energy.
Austin Brown: Well, it does store energy. There is energy stored in the magnetic field, but the way that you can think about it is it generates a flux which then goes through the conductors, right? From a from a physics standpoint, the flux from the permanent magnet goes through the conductors and that makes when you rotate the rotor you'll get like positive, negative, positive, negative on the coils, right? maybe it's Faraday's law says that by doing that you'll get a a voltage induced by the changing flux. And so if you have voltage induced, then if you happen to put current in at the right time, you will get force generated if your motor is designed well. so all the the only thing that the permanent magnets are doing is they're really just driving this flux through the coil. And with other weird motors like induction motors and squish reluctance motors, that's when things get a little bit more complicated. like an induction motor, like there's a charged magnetic there's some big coil which is charged up which is exciting the fixed stator coils, right? But why people really like permanent magnets is that they always stay on. and so, you know, in an induction motor you kinda need to charge you keep the rotor like charged with current basically, which is a source of loss, right? The rotor will experience I squared R losses. Not much, but you know, when it's a huge motor it can be quite substantial. they're also lighter, you know, an induction motor needs to have like this like honk and rotor, right? which is made of steel and copper, like the heaviest possible things. A lot of times permanent magnets will decrease basically the rotor thickness. so, you know, for a big induction motor, like I I could have like some big donut shaped rotor, whereas if I replaced it with permanent magnets it would be like a much thinner permanent magnet array. And that's actually
Bayley Wang: I see I see
Austin Brown: It's very important for robotics because it takes down your inertia. which
Bayley Wang: I see, in the case that the copper had no resistance, you could replace the PM with like a ring of copper with current going through it, and it would be like sort of equivalent. But the the PM is this magic thing that generates a like a biased flux but somehow doesn't have net current flowing through it, which is weird. It it goes back to the question, what the fuck are magnets?
Austin Brown: I don't know why.
Bayley Wang: Better not think about it. Yeah, I I don't know. Like the the worst part is like the the the electrostatic equivalent of a magnet makes a lot more sense than the like a permanently polarized electric charge is a lot more sensible than a permanently polarized magnetic charge.
Austin Brown: I mean, either way you're sort of like what? You know, I I if you
Bayley Wang: 'Cause there's like a physical unit of electric charge, right? You can you can imagine pushing the positive and negative charges to one end, but like there there isn't a a physical like magneton which is is an electron but for magnets.
Austin Brown: But I think if like the the atoms are like arranged a certain way, like the electrons are like going around or something.
Bayley Wang: Yeah, yeah. They're like spinning spinning around. I I I don't I don't I don't really know either. I don't know how
Austin Brown: Make them work.
Bayley Wang: we can end the podcast with this question, what the fuck are magnets? Which I feel like is a good place to end. and yeah, Austin, thanks thanks for joining us. I've I've certainly learned a lot about motors and I hope the audiences have learned a bit about motors too. I guess my my sort of stance on this thing, especially to the investors in the field, is if some guy comes up and says they built a better motor, you should ask them some very hard questions. people have been doing this for a very long time. And as Austin said There are only four materials we get to work with, and they can only be arranged in so many ways. All right. Thank you, thank you.
Austin Brown: Sure. my name is Austin Brown. I No Bailey from doing some amount of MIT with you. we both worked in the miters hacker space. A lot of fun stuff came out of there. Anyways, did six years at MIT, Undergraded Masters. Masters had my own company doing various consulting stuff for a little bit. I did a lot okay, back at MIT I did mechanical and electrical stuff. My my my biggest interest in life is the flow of power, whether it be electrical or mechanical, and especially the intersection between those two I find really fascinating. so anyways, motor stuff, we'll get there in a little bit, but at the moment I do motor design at Boston Dynamics. and that's a pretty cool spot, but I also do a lot of stuff outside of work. Like solar power stuff, which is right there, and hydroelectric dam stuff, which we can get
Bayley Wang: Yeah, yeah, all right. Cool cool, yeah. So Austin's on this on this podcast with c podcast with us today because he's one of the rare machine designers that are still based out of the US, like electric machine designers. It's one thing to buy a motor and like write firmware for it, it's another thing to build a motor from scratch and really understand all of the optimizations. And I think this is a really interesting feel to a lot of you guys. Obviously state of the art robots are electric, the old like whatever hydraulic drive robots are sort of fading into memory now. But motors are also really important for things like renewables and electrification. And potentially even, you know, this power generation kind of nonsense. All these hot topics which sort of interplay at the intersection of energy, real estate, geopolitics, and sort of move move the world forward quite literally. So I guess we'll start with you know, machine design. There's a lot of electric machines, there's a lot of motors out there and a lot of opinions about them. Let's talk about you know, various misconceptions in electrification and motor design.
Austin Brown: before going into misconceptions a bit, maybe we should just start with like what is a motor. At
Bayley Wang: Yeah, yeah.
Austin Brown: the highest level, Like what's the point, right? So generally you have either mechanical or electrical energy and you want the other one, right? so what a motor does, the motors that we're all familiar with, they take electrical current, put it through a coil which generates a magnetic field. And magnetic fields are special because they cross the electrical and the mechanical domain. by flowing the current you generate You can generate forces, right? And you can do mechanical work. So that's that basically the energy. If you're trying to use a motor to make mechanical work, the energy flows from electrical, generates a magnetic field, and then the way I think about it, you could generate a magnetic field, and then you extract mechanical work by basically collapsing that magnetic field. You like a a magnetic field stores energy in a volume and And by collapsing that magnetic field, you like by you know physically moving one part of the motor to collapse that magnetic field, you reduce the volume of energy stored and then that is output as mechanical work. And by having multiple phases of your motor, you can you can sort of do that process continuously and generate smooth output power. So that's sort of And then you know, obviously the process is exactly the opposite if you're doing power generation, which is you start with mechanical motion, you can like use permanent magnets or electromagnets to generate a magnetic field, and then it goes the other way. So I think a lot in terms of misconceptions, I think the biggest thing that people really need to think about is a lot of times people say, like, you know, I want a motor with a lot of torque, therefore it has to have like really high amperage, right? But that's not really true. basically a motor is very much like a transformer. the more numbers of turns you get, the higher the voltage, but the lower the current. And you can think of like a one turn motor for example, like one turn around each tooth. that would obviously have extremely high currents because you'd be talking about like fat, you know, copper bars. and very low voltage Voltage output, which you know, if you had a one-volt battery, that might be really, really useful for you. But usually you don't have that, so you do multiple turns. anyways, you know, there's all different ways of making motors. more recently, the sort of state-of-the-art has been I mean, really just getting the efficiency up For a given power output decreasing the volume. I think that's what people care about mostly. you know, in the context of a car, like other than the battery, the motor is probably the heaviest single part. also it's you know, it's obvious it's made of copper, right? So it's very expensive from a material standpoint. and so people basically want to get the motor as small as possible. in robotics, I mean that's that's even more true, right? So, you know, say you have Like a hip actuator, it obviously to fit inside the robot. And if you want to make a robotic hand, there's a huge amount of stuff going on that in that field, which is you know, do you want to put the motors in the fingers, right? Then they've got to be really, really small, right? and you know, also be very efficient to dissipate as little heat as possible. So that's sort of the the lay of the land these days. tons of work being done on that, although it's kind of a weird field because I think as Baylor. said before it's it's rare that people sort of come with the both mechanical and electrical. Like usually like you know you have the electrical people over there and you got the mechanical people over there. And then you know there's not not that many people in the middle. so you know it's it's kind of a niche thing. but it's also weird basically you you almost don't need it ever, right? Like if I'm just making a generic household appliance, I would never design a motor.
Bayley Wang: Yeah, yeah, right. It's a complete waste of time. A very bad idea from a product standpoint to design your own motors with custom laminations, custom magnets. Takes about a year, a year, eighteen months, two years even to get everything running reliably. If
Austin Brown: I mean I don't even know if I'm
Bayley Wang: you want to like shift a product very quickly, like it wouldn't make sense for a toaster. Toasters have continuously rotating parts.
Austin Brown: For almost anything. I mean motors come in a zillion different shapes and sizes and you know you can buy you can buy anything, right? it's just when you want that last whatever percent of performance that but I mean, you know, building your own motors is not unheard of, right? Any b company is building their own motors.
Bayley Wang: I mean we we say that but very often the motor I want is not commercially available, which is really annoying. It seems like the the list of commercially available motors is all the one all the ones except like the two I need for the project.
Austin Brown: Yeah, yeah. Basically any any industrial motor is stupid to design yourself, right? Like that's one hundred percent true. Like you can buy any NEMA frame, p you know, ten thousand pound, whatever, like that. what are the big industries like robotics these days, automotive and industrial Yeah.
Bayley Wang: There aren't a lot of state of the art industrial motors, I would say. The industrial guys are still for the most part running on older designs with lower power density and much worse efficiency curves or heck induction motors plugged straight into the wall. Like the the torque output is whatever it you whatever it wants to be at that time.
Austin Brown: Right, yeah. I mean
Bayley Wang: That is actually a hot topic, right? This this idea of modernizing the electrical machines and factories. moving machines in industrial settings is a very palpable fraction of the world's power consumption right now.
Austin Brown: That being said, large induction motors can be fairly efficient. small induction motors are just like total garbage in terms of but big induction motors are as far as I know in the ninety, ninety five percent.
Bayley Wang: I I do I do remember that it's surprising how good they get without any form of like control. They they've been like carefully designed to pr produce the right number of newton meters at 1750 RPM.
Austin Brown: Yeah, yeah, right. It's it's although they have weird characteristics. we're actually learning this at the dam.
Bayley Wang: give the listeners some context about the dam.
Austin Brown: Sure, okay. So some context about the dam. So I work at Boston Dynamics right now, and what I said in my interview, I was like, I really like motors and I have a l you know, robotics sounds like a lot of fun, but I'm actually more of a power guy than a robotic guy. Those two things are a little different. and I mean, yeah, I'm actually having loads of fun at Boston Dynamics, but that's kind of a separate thing. So one of my biggest sort of side projects that I do for fun is me and two other friends purchased off of Facebook Marketplace a small hydroelectric dam.
Bayley Wang: That's gotta be one of the best, best Facebook marketplace purchases I've ever heard of. Like what did you buy off Marketplace today? A damn.
Austin Brown: A dam, right? So it's been a long side project. so we have gotten it sorta quasi-running, right? Our peak power is about sixty-five kilowatts. The facility was supposed to produce about eighty-five. So we're running a little low. I don't know why. We've sort of been getting it working slowly. It's it's a group project, so you know things never really move that fast. Basically in terms of like physical dimensions and stuff, I think we've got like an eight acre reservoir. we've got I think about eighteen feet of drop and depending on the season, our nominal flow is sixty CFS cubic feet per second. So that's I think that equates to like two cubic meters per second of water. it's pretty cool. Right now it's fall, so We're running a little bit less water. So we we have I think 40 CFS right now. But that's sort of what it is in terms of like what does the dam actually look like? it is what about 50 feet across and it's made of granite blocks. There's a three-foot diameter propeller looking thing, that's the turbine. There's it's actually kind of interesting from a power perspective. So three-foot diameter turbine that goes to About 500 RPM, and then there's a what? I think it's like a 5 to 1, maybe a 5.6 to 1 speed increaser planetary gearbox. And then there's an 1800 RPM induction motor. And the induction motor is actually kind of nice in this capacity because in order to connect to the grid, like it's literally just a massive contactor, and you just bang it, right? The thing that's nice about that is in a lot of larger power generation facilities you have to sync things to the grid. Right? Induction
Bayley Wang: Yeah, yeah, right.
Austin Brown: motor that doesn't exist, right? You just bang. so we
Bayley Wang: That's clever. That's clever.
Austin Brown: pre-spin it up to the proper RPM, but there's no sort of phase synchronization that needs to occur. So I did all the controls for that project. I have this like big like hydroelectric dam controller circuit board. Yeah, that's kinda cool. the the weirdest thing about it, so the induction motor that's on it, like normal induction motor, a normal induction motor, eighteen hundred RPM, right? And they sag by fifty, roughly. Yeah, depending on the load. But this one, because the power's going the other way, boosts by fifty. But but still it's weird. Like people see that and they're like, What? That's not right. So
Bayley Wang: I I g I guess you have to be a really special person to see that and be like, wait, that's not right. Yeah.
Austin Brown: Yeah yeah yeah.
Bayley Wang: Most people don't know the lore, but the yeah the the slip going the wrong way is a really interesting concept here. I yeah, I I think I think this is a good point to talk about the I guess various use cases, right? You've talked about the dam, which is sort of this more traditional like constant RPM power generation use case like similar to industrial stuff. We've briefly touched on E V traction motors, which are more dynamic and more state of the art, and then on the other end you have the robotic stuff that you're working on right now, where basically everything only exists in a start stop condition and And there's like virtually no phases of constant velocity. maybe like I'm wondering whether like have you noticed any sort of large overarching design paradigms as if you worked on these three different types of machine?
Austin Brown: Yeah.
Bayley Wang: I'd say
Austin Brown: Say the biggest thing with the robotics thing is that the hardest thing is deciding what your duty cycle is, right? with a car it's fairly clear, like the hardest thing is, you know, in the summer driving up a hill at eighty miles an hour, right? Like that's that's sort of like n what normal people do on a car. And and you know, nobody really drives up a super steep hill at a hundred miles an hour, right? Like that just
Bayley Wang: Yeah, yeah.
Austin Brown: happen. The thing with robotics is it's like, you know, okay it did a backflip. That's sick, right? can it do a double backflip? Or like, you know, do you really need to do a double backflip to like work in a car You know, so that's that's the thing and and there's a lot of like thermal transient stuff, right? You know, that that stuff is a little bit harder to it's not hard to design for. I think it's more hard to set the requirements for and and either
Bayley Wang: Yeah.
Austin Brown: but I think also the overarching thing is nobody really knows what a good robot is gonna do, right? Like yeah.
Bayley Wang: Very true.
Austin Brown: That's sort of an active question in the industry right now is I mean there's huge loads and loads of different sizes and shapes of robots, right? So yeah. in terms of like overarching design philosophies There's not much. a lot of people have gone in a lot of different ways. I think the only thing that I would kinda say is people have mostly gone towards this like quasi-direct drive situation. you know, I mean, but that's that's kinda old news at this point. It's just you have like
Bayley Wang: Yeah.
Austin Brown: large pancake e motor, and then a relatively low, you know
Bayley Wang: Yeah, it's like practically synonymous with no harmonic drive at this point.
Austin Brown: Yeah, basically, right. So, you know, that's that's basically what people go for. in terms of like more interesting stuff, I probably can't get too much into that because that's more boss and dynamic stuff. but in terms of like where the rest of the industry's going. Yeah, I think I Honestly most of the motors that people are using Hallbach arrays are I think one thing that people are more starting to do.
Bayley Wang: They're they're yeah, yeah, they're they're they're they're a bit fringy, but they do get you more torque. Like they're they're hard to build and no no one no one's no one's shipped one yet, but they give you good numbers on paper, which I think is like if you're a if you're an actuator vendor, right? That number is y the heart and soul of your product. So being able to put out that you're two times better or like whatever, one one and a half times better than the competition. it it does look like the metrics are better on paper, although I have no idea what integrating a Hullbach actuator into a robot looks like.
Austin Brown: Hallbuck arrays, I mean I can tell you this from personal experience are real pain to assemble. I have put together
Bayley Wang: I don't I I can't actually I can't actually think of anyone that's shipped a Hallbach array in like serious volume.
Austin Brown: Yeah,
Bayley Wang: people have tried.
Austin Brown: They seem to be I mean so now we get into the whole what I call the spec dicking war.
Bayley Wang: Yeah, yeah, yeah, right. Figures of merit we call them in
Austin Brown: Right, yeah. So I mean when you want to go down that road, like haulbacker ray it up, right? You know, let's go. But you know, i it's sort of at the same time, you know, the robotics industry right now is very much into investor hype situations.
Bayley Wang: Yeah, yeah. So Ugh Yeah. Tell me about it, right? The the investors come and they ask why the robot is different. You mutter something about Hallbach arrays and show them a nice FEM simulation, they're like, wow, he's smart.
Austin Brown: I mean obviously I think actually Boston Dynamics is doing pretty well or you know, going about things in the right way, I would say, you know, not not only focused on these figures of merit. Although, you know, we're speaking broadly here, right? Like there are many companies out there that are you know, solely focused on this sort of like Weird technical stuff. I think in terms of industry trends though, where people are going, the people that are building their own stuff, basically from a motor perspective, there's kind of two knobs that you can really turn. one is the magnet, right? And the knob that you can turn there is higher grade magnets and rays, right? backing up one step further, fundamentally
Bayley Wang: Let's actually back up two steps further. I have a question for you. in terms of like paper figures of merit, how much better are Hallbach arrays versus a like a traditional steel back motor?
Austin Brown: It really depends. well, I think
Bayley Wang: 'Cause my my understanding is peak torque doesn't go up, right? You're still limited by saturation in the teeth and like the overall volume in the of the field.
Austin Brown: People really like to say that saturation is sort of this like hard limit that you just ram into and in my experience that's basically never the case. you are so limited by thermal stuff, like way before you get into any type of saturation.
Bayley Wang: Introduce.
Austin Brown: Yeah. Basically, yeah. saturation. I mean, one time I think I I wrote a s this is my master's thesis. I wrote a simulation that was, you know, for a seven newton meter motor. And I think I accidentally like put a you know, it was for like a fifty amp motor or something, right? And I think I like had screwed up the number of turns or something, so there was a extra factor of ten, right? And I put in, you know, like I don't know, ten thousand amps I think I did a switch. sweep that ended in like ten thousand amps or something. It was like some huge number, right? And surprisingly, the motor was like more linear than you would expect up to like really, really high currents. I think it's kinda
Bayley Wang: Yeah. What happens when you saturate the steel core, right? The the the beast beast goes up in air. Like the steel becomes air.
Austin Brown: Yes, but I think more interestingly, the best way to think about it is Like, okay, when you have something that's saturating It's like there's all these little magnetic domains that are aligning. And like it turns out in steel that like steel saturates it starts to saturate like one point five, one point six Tesla, but has some sort of saturation left to go. I think all the way up to like one point nine or like two Tesla or something. So it turns out there's quite a bit of like, you know, knee in the curve. Like it's a relatively soft thing. The other thing that's more important to consider is that Think of your D and your Q axis in field area and control, right? The magnet is putting a whole bunch of flux down the D axis, right? And your windings are putting a bunch of flux down the Q axis. you start at zero, right? And you're you're like bringing this like vector of flux, like the total flux, I guess, is kind of the hypotenuse of them. So you start like it's actually to generate that amount of flux with the q axis is like a huge amount of current. on small motors because basically magnets have so much like current capacity or flux generating you know capacity sort of inherent to them that but basically long story short saturation is way more of a soft curve than you would ever expect. I think only if your inverter like shorts out would you ever like really run into saturation because I mean, you know, even if you're putting in peak current for like less than a second or something, right, you still do have to worry about thermal transients. Long story short,
Bayley Wang: Yeah, yeah.
Austin Brown: saturation is I wanna say it's kind of a red herring. For most things, I I seriously
Bayley Wang: Yeah.
Austin Brown: that this is a extremely scale-dependent effect. as the scale goes up, basically like all robotics actuators are like this size or smaller, right?
Bayley Wang: Yeah, yeah, right.
Austin Brown: As you get bigger, the physics work out such that your yeah, saturation just becomes way more likely. Like it's a sur it's basically the surface heritage volume thing, or it's a long story short, the copper conductors become more powerful and more able to drive flux in in comparison to a magnet. Yeah.
Bayley Wang: Inter interesting, interesting. We can go. So at at the at the sort of like robotic scale, which is incredibly fixed, like the the the motor ranges from between seventy and one hundred and fifty millimeters in diameter. It seems to seems to mostly be the case that you're bound by other factors and it's like not actually possible to hit saturation torque out of the actuators.
Austin Brown: even saying hitting saturation is kind of like it's sort of a a fake thing. Like if you're using
Bayley Wang: Yeah.
Austin Brown: an inductor, like a piece of ferrite really does have like a you hit it, right, and then it saturates. Whereas with it you know, with a motor, like I've heard of of torque reductions of like twenty percent at high at high torques for, you know, some types of motors, right? but, you know, that's not like that's not a h hitting a hard edge, right? for all intents and purposes here, like
Bayley Wang: Pretty pretty pretty soft. Like you can you can you can still keep going. That's that's very interesting. I wonder you're
Austin Brown: Right. If you have an inverter which is very finely tuned, like to work great, you know
Bayley Wang: Definitely, yeah, yeah.
Austin Brown: if
Bayley Wang: Those inductances changing really gives you a bad time. And it it's
Austin Brown: Right, yes. That's potential.
Bayley Wang: Yeah, yeah. I've always sort of assumed that the hip actuators and these big robots was like running pretty clear f near physical limits, but it sounds like maybe maybe that's not the case. And it's still sort of a as design limit.
Austin Brown: I would say also these things can obviously be true or not true if your actuator is badly designed.
Bayley Wang: Yeah.
Austin Brown: In BattleBots one time we had basically that was on uppercut, which I'm wearing the shirt. that was the BattleBot. I helped design the power system for that for the weapon on that robot. the previous generation of Uppercut before I had joined the team, they were it basically like big outrunner motors. Those things really saturated. I think they had extremely thin Stator teeth and the controllers would actually freak out. That was the thing that would limit the torque.
Bayley Wang: Yeah, that that that makes sense. You basically have no inductance once you saturate that badly, right? Right. Yeah. The what I I'm actually curious now that I think about it, right, what happens at saturation? Do you just get no longer get more torque per amp? Like where where where does the flux go?
Austin Brown: I mean again, like I I think saying at saturation is is kind of a a false flag. Like there's no there's no like BAM I've run into s you know, it's it's such a slow roll off for the
Bayley Wang: Okay.
Austin Brown: So that's But but you know in terms of what actually happens, yeah, basically I think the best way to think about it, the most accurate, which it took me a long time to kind of reason through, is that saturation only really occurs when so okay, in a normal permanent magnet motor, the flux goes down one tooth, through the permanent magnet array, up back the other adjacent tooth, and then back over through the back iron, right? So there's quite a lot of magnetic reluctance through the air gap and through the permanent magnets, right? Saturation only starts to play when the reluctance of the teeth and the back iron and the other tooth starts to become substantial. compared to the reluctance of the the permanent magnets.
Bayley Wang: That makes a lot of sense, right? Because the the air gap has a mu of one. so it's it's it's whatever, thousands of times effectively longer than this than the stator teeth. So you can actually sacrifice like 10x, 20x, dropping the perm permeability of the teeth before it it starts being comparable. That's really interesting. That's really interesting. That that's that that's pretty cool. So sa saturation really is not the right way to think about it for like machines built out built out of these soft but magnetic materials running at plausible. Currents. I think I think the I guess macroscopic time scales. Like ro robot like time scales. I think that's the real thing, right? A a robot takes like six hundreds of milliseconds to complete an action.
Austin Brown: Correct. And the other thing that I'll say here is that it really is scale dependent. So, I think, yeah, people should be aware that basically like if you take a C core, like a magnetic C core, and you put like some number of turns around it, and you have like some air gap, you'll get and you you know, you energize those clothes with some current density, like one amp per square meter, let's say. You'll get some sort of I mean, this is like a very bad motor, like if we put a magnet over here, like it would go into the gap. But anyways, let's just consider the C. If you have some current density, you have like ten turns, if you then take so you'll then you'll get like, you know, let's say one Tesla in the gap, right? Or you call it half a Tesla for this one. If you double every single dimension, you just double everything. You make the wire go from one square millimeter to four square millimeters, right? you know, it like it gets like into the page, blah blah blah. Everything gets bigger by a factor of two. Every linear dimension gets bigger by a factor of two. Turns out the flux in the air gap will actually double. because you have you can go through the math, the reluctance goes up by a factor of two, but the the current going through the C core has gone up by a factor of four.
Bayley Wang: yeah yeah yeah.
Austin Brown: That's what gives you so anyways, all that is to say, but a sort of a permanent magnet kinda is the same at any scale. So what that means is that as you get to like much bigger motors, saturation becomes way bigger of an effect because all of a sudden now your copper conductors are much more like powerful. Or are you like thermal you can run lower current densities to get the same flux. So that means you're generally Generating the windings, so then you can pump them way harder. and that's when saturation will start to play. Personally I don't have a lot of experience with motors above like ten kilowatts.
Bayley Wang: Yeah, yeah.
Austin Brown: It's never been a problem up to those scale of motors, but I I do know that the physics say that as your scale gets bigger, like I'm sure this a hu
Bayley Wang: I see, I see. So so with like a E V with like a E V size motor, your peak torque might become closer, closer to like a a saturation limit, so to speak.
Austin Brown: Correct. Yep. The other thing that's kinda weird, I actually tore down well, it was one of your motor I think you had purchased one for miters like way back in the day. if you look at like a an EV motor, the like the proportion so this is to illustrate the scaling. If you look at the proportions of one of those E V motors, like the stator teeth are a lot fatter than you'd expect. Like the proportion, like really small motors, you end up with very thin stator teeth because they need a lot of space for copper because you You just can't pump the copper that hard. On the previous motor that you bought, the the like teeth, I was looking at the teeth were, you know, I'd say like on a small motor, maybe they're like 30% like stator tooth by like, you know, if you take the circumference, it's like 30% stator. On an EV motor
Bayley Wang: Yeah.
Austin Brown: of like, you know, a reasonable size, maybe it's more like fifty to seventy percent because the they just don't need as much copper. So
Bayley Wang: Yeah, that makes that that makes sense.
Austin Brown: Yeah.
Bayley Wang: talking about EV motors, that's another very interesting topic, right? They live at a completely different scale, and as I understand, they have a there like a completely different set of desi design criteria for EVs where the big focus really is on day-to-day driving efficiency and not any of the specious metrics like torque per cubic inch, torque per cu watt torque per kilogram, right? Because the the car is very heavy, the drivers are very heavy, and the extra kilogram of motor affects How much you pay for the motor, but probably not the performance of the car. So I think that that that's another one of those interesting misconceptions that I I was really big about back when I was we were doing motor stuff together. That doubling the torque per kilogram of your car's motor basically does nothing because the car weighs much more than the motor. Like the the marginal weight savings is like a like a sub percent. But
Austin Brown: Yep.
Bayley Wang: I I I guess it's different for robots. Talking a little more specific, how much of the so like right if I make the motor quote two times better, but just the raw motor, does the actuator won't become two times better because there's like a lot of stuff which isn't the motor, right? There's the gears, the inverter, the housing, the mechanicals, the the like physical structure of the robot. so it's not quite the case that like really pushing for high motor performance on robots makes them infinitely better or like linearly better.
Austin Brown: Yeah, yeah. I mean I that being said, I would say that the motors probably are one of the heaviest parts of the robot. And also their mass is often located fairly distantly.
Bayley Wang: Yeah, yeah, the the the distal the whole distal actuators. I saw a humanoid at Acre last month where like great care was taken to have no distal actuators on the legs. Like everything was driven by linkages from the hips and it was a really really weird design. I don't want to be the guy who has to unbend the linkages when the robot falls. Right, right.
Austin Brown: Yeah, whereas Boston Dynamics has taken completely opposite approach. You know, you can see in the ankles there's huge actuators, right?
Bayley Wang: Yeah, yeah. The the I I do think it's like the result of it's the difference between like thinking about robots and actually using the robots, right? Like once you start putting the robots in the field and trying to do things with them, stuff like the b the ankle linkages getting bent become really not good. and stuff like cabling as well is a pain in the ass, but that's a that's like literally a topic for a whole other podcast is how to deal with the wires in your advanced robot quickly becoming the biggest problem in the industry. I don't know about you, but our our our our fleet has egregious failure rates. It's like a one one cable failure every hundred hours of our arm operating time or something. And the current industry solution is you send the robot back to the vendor who replaces the cables under warranty. And the vendor is like it's like cheaper for the guys in China to do this than to design competent cable ends. But anyway, that's a top topic for a different different day. Axial flux motors, that's a and I guess the the art of marketing your motor, right? Like I I bet some of our listeners are either VCs or like founders trying to start a startup. And the number of like motor and actuation related pitch decks I've seen that are just completely full of slop, like they measure the wrong thing. And I think increasingly i increasingly strange and over optimized forms of axial flux motor are the pinnacle of this.
Austin Brown: Yeah. So basically from a high level, if you are some sort of robotics person, you can more or less walk into a room and know that any investor is gonna know less about magnetics than you do, right? Right because nobody knows anything about magnetics, right? Like you know, walking down the street, like maybe to somebody doesn't even know what a magnet is, right? So if you just like make up some weird terms about Motors, like usually you can bamboozle people, right? the thing is, you know, I've seen a lot of weird motors, and it turns out I would say 98% of the time the simplest thing wins. I famously like to say a motor is really only made out of four components: it is neodymium, steel, copper, and air. And you will actually want as little air as possible. It's just that the two things, you know, the rotor and the stator can't grind on, right?
Bayley Wang: Get Rob
Austin Brown: So you know, you gotta have a really, really thin air gap. but anyways, like it turns out, you know, if you only are really working with three components it's not that there's like no creativity, but a lot of times it just comes more down to good implementation than some sort of like weird physics, right? So but what a lot of people do is they just say like we're gonna have, you know, the coils going every which way and somehow it's gonna magically generate a war torque. Turns out, I mean, okay, just with axial flux motors specifically, a lot of times, you know, people always ask me, like, you know,
Bayley Wang: Come on.
Austin Brown: I saw this cool thing with axial flux motors. What do you think about them? Are they the future? Absolutely not, right? the I mean axial flux motors make sense about three percent of the time. And s and sometimes that is important, but basically they're they have a few big problems. number one, they're actually a real pain to build because like they want to suck together. like the you know, all of the normal force in a radio flux motor, everything's balanced, right? And you see when you put a like a rotor in a stator, it wants to like clunk off to the side. That's because like ninety eight percent In an axial flux motor, it is one hundred percent.
Bayley Wang: Hundred percent of the stater tries to stick to the rotor.
Austin Brown: And so people try and do weird stuff like, you know, we're gonna have two, right? But then the motor is basically impossible to cool, right? Because or you it becomes water cooling mandatory.
Bayley Wang: Yeah, yeah.
Austin Brown: I believe it's Yassa is actually doing pretty good stuff with the axial flux motors. the one advantage that they really have is that it's oftentimes very easy to wind the coil as like a bobbin and then you sort of assemble the bobbin. That's pretty nice. y you can do that with Radial flux motors too, but it's it's quite a nice form factor with an axial flux. There's two other big problems with axial flux motors. number one, the lamination stack gets weird, right? So you are familiar with ETEC floppy rotor syndrome, right?
Bayley Wang: Yeah, yeah.
Austin Brown: Yeah. So basically like your laminations like in a radio flux motor, they're like stacked, right? and everything just kinda works. But in a axial flux motor, like the lamination is kind of like of like it either needs to be a spiral or it needs to be like concentric rings and you can sort of have it like go in on itself. So yeah, you get like all sorts of weird structural issues where because the you know there's huge forces on these parts, right? And it's kinda hard to like
Bayley Wang: If you've ever if you've ever tried to pull one apart, there's a lot of force on them.
Austin Brown: It's it gets weird really fast. and then kind of the most niche thing that I don't really like about Axial Flux Motors is they're very hard to simulate. they're you but you basically have to simulate them in 3D.
Bayley Wang: Yeah, you can't have a two D cross section.
Austin Brown: Right. A radio flux motor is very, very easy to simulate because it's it can be simulated very accurately in two D. So if you want to simulate a hundred thousand motors, you know, no problem, right? You can just do it in two D and it might take a while, but it's not that bad. Whereas if you want to do the actual flux motor, you basically have to do it in three D. You can do like a couple two D cross sections, but it's just weird. what's the point, right? You're doing all this work, you you've not fundamentally changed the physics. But I mean basic Like the way a motor works is you have a permanent magnet, or the motors were used to, permanent magnet, it's making flux through a coil, right? And the only thing you've done is just turned that whole thing sideways and made it harder to mechanically assemble.
Bayley Wang: Yeah, yeah, right. That that's that's what's sort of my impression that with a radial flux motor, you like can draw the 2D cross-section. Once you have a good simulation, it basically directly translates into like a manufacturable motor. Like that cross-section can be sent out to a vendor to stamp the laminations. With the axial flux machine, it's like a shit show of hiring a technician to try to figure out how to stuff the motor afterwards, and there's like it's like full of tape and glue and potting epoxy and the motor sometimes falls apart.
Austin Brown: You know, any motor If if your motor I like to say if your motor is easy to assemble, that means you weren't really pushing performance hard enough. the best motor is something that like if you really, really shove, like it just barely goes together. you want the whole stator to be entirely steel and copper, right? Like no air, right? Yeah. So, you know, either way, it's gonna be a pain, right? Or it should be a pain if it's a high high performance motor, right?
Bayley Wang: I guess another interesting question is we like to pursue like the endless pursuit of performance for everything we build. Do you think that's going to matter in like the first commercial sort of quasi direct drive robots? Or is the focus going to be more on like can we build this thing at scale? 'Cause I d I have a lot of Chinese actuators and none of them are like dense. They're they're mostly aluminum.
Austin Brown: Yeah. That I I don't know. I mean if there's one thing I'm not, it's not I'm not a production guy. I mean
Bayley Wang: Yeah. I I I can't claim to have ever put put put motors into serial production either, but it is it is interesting that we've like worked so hard to build state of the art actuators and like all that all that work is getting thrown away and being replaced with like a Chinese hobby motor inside some aluminum. Like that that's what that's what quickly what the industry is like I won't say it's just that, like Unitree has done some good advancements on manufacturability.
Austin Brown: I mean, yeah. To some extent, like you can always do something bad, right? And it'll always like kinda work, right? And then I think the real question that the world is trying to figure out is if you have bad hardware, can reinforcement learning just like work, right? and
Bayley Wang: I I still think you need to like have enough torque to stand up and lift the target object.
Austin Brown: I mean I have heard so honestly I don't really watch the news in terms of like humanoid robotics. I'd rather spend the weekend like chilling at the dam than like reading it all some treated robot, right? okay, actually going back to the dam, it's lakefront property, right? That's been the truest awesome thing at about the dam. You just like relax, you like hear the turbine humming in background and you're just like watching
Bayley Wang: Peaceful a peaceful weekend trip to the hydro plant with the turbine worrying in the background.
Austin Brown: I mean this is normal, right? Okay, honestly, the the coolest thing about the dam recently is I got this cycling routine working. So it like cycles until the water goes too low and then it shuts off and like lets the reservoir refill. So if the turbine like takes too much water you know, you you basically can't run continuously, right? So I got this cycling routine programmed. Anyways, I'm out in my tent, I slept right next to the dam. I like wake up, you know, I like poke my head out, and I hear the turbine like I basically hear like the main contactor like open at like which is a big clunk. It like it's like it's a a loud contactor, right? And then I hear the turbine like spooling down because it had run itself out of water and I was like, that's cool, right?
Bayley Wang: achieved a degree of autonomy. That's a it's a robot right there. Yeah.
Austin Brown: yeah, like unit tree stuff and I don't know, you know, people are doing all sorts of weird stuff. Like I've heard yeah, I have heard through the Great Prime that those issue that those robots do have pretty substantial thermal issues and that they can run for some amount of time but not steady state. and
Bayley Wang: I would believe that. I would I would definitely believe that I know for the for the ar for most of like the the whatever standard Chinese arms that the imitation learning industry uses, if you like keep them outstretched with full payload, they overheat in a few seconds. And if you squeeze too hard with the gripper, they literally overheat in like two seconds. it's it's not really a feature, but the idea is everything is only transient. Like li lifting heavy objects continuously is not part of the scope use cases right now.
Austin Brown: Yeah. if I if you told me pick up the you know, the heaviest thing you can pick up and hold that stretched, I could probably
Bayley Wang: Yeah, I I guess that's true, right? Like I think that's their argument. Like whenever you complain about this like my machine is overheating, they're like, Well, if it were a person it would also overheat. But that be
Austin Brown: Being said, you know, I think what's how tall is a unit tree? Like what are the humanoids?
Bayley Wang: The the little humanoids are about four feet tall, it's like one one one point two, one point three meters.
Austin Brown: I believe all of the things get worse as you scale up.
Bayley Wang: I believe so. I believe with longer limbs it quickly becomes worse.
Austin Brown: Like your surface air to volume ratio gets worse. And moving like even if you had like solid aluminum links, like moving the heat around inside the link, I think gets worse. So I I think actually that's sort of a hidden cheat code. Like just make the robot, you know, small. I think that
Bayley Wang: You definitely s you definitely see it. All of the commercially viable humanoids right now are like two-thirds scale people. They're like child-sized. And I g I guess you also work with full-size humanoids. We have a full-size humanoid in the lab, and it is a absolutely terrifying robot. 350 newton meters of hit motor, doesn't doesn't fuck around. Yeah. If the if the controller winds up and that leg is go that that foot is going through something, and hopefully it's not you.
Austin Brown: Yeah. I will not comment on anything related to Boston Dynamics. but our our motors do have plenty of torque, that's for sure.
Bayley Wang: The yeah, I guess we're almost almost out of time. I guess the last question, l last sort of topic of discussion that we briefly mentioned earlier was Dexter's hands. I think this is becoming an increasingly hot field of investment over here in Silicon Valley, but unfortunately investors have never been so bamboozled as much as by a guy building Dexter's hands. you see all kinds of combinations like motors and strings and like none of it is manufacturable, all of it probably overheats.
Austin Brown: Right. that I probably cannot comment much on. Yeah.
Bayley Wang: Alright, alright. I guess any any other intra interesting topics, like there's a lot of lot lot of ways we could go. We could talk about neodymium free motors, which I think are a hot topic these days. gosh.
Austin Brown: honestly, you know, I I don't know what's going on there. I mean, what a pain, right? Like you know, basically the rules are changing every day, right? it takes you so long to redesign the motor. And so if the rules change faster then you can redesign the motor, like you know, what are you supposed to do, right? hard
Bayley Wang: No, d the domestic manufacturing of motors is another another another hot one. 'Cause the right the the FCC recently questionably banned the use of drone motors that aren't made in the US. Which I mean it's o the ban's only gonna stick around for two years, but it's creating at least over here in the valley, there's like a lot of narratives based around doing it in the US. Like do do do you think this is ever commercially feasible?
Austin Brown: I mean again, this comes back to the whole thing of like it's commercially viable until the rules change again and it therefore it's not commercially viable. Like, you know, you could buy a whole bunch of equipment and then the rules change and it's no longer worth doing like
Bayley Wang: I guess that's true. It's a it's it's a it's a terrible model to base your business around.
Austin Brown: Right. Yeah. Like let you know, tell me that you want to pot buy like a whole bunch of capital equipment purely based on some rule that was like arbitrarily ruled out with no checking whatsoever and could just be repealed tomorrow for no
Bayley Wang: Yeah, yeah. That's true. A lot of the like if you had built a business based on the the the tariffs, for example, you would have had a bad time last month. As the the
Austin Brown: Right. Like
Bayley Wang: the tariffs like suddenly drop from fifty percent to ten percent and completely change the completely change the picture. I mean there there is an argument from like a national def security point of view that drones are becoming increasingly critical for war, and in the case of a war we cannot expect to buy drones from China. Which is interesting. And as far as I know, I I I guess Free Fly still builds their motors sort of in the US. I ha haven't checked recently, but I I I think you can get a Free Fly drone where the motors are like US compliant.
Austin Brown: Yeah, I mean again, it's it's not even a motor question. It's more of a supply chain question.
Bayley Wang: it's not really an engineering thing, right? It's like can you yeah can you can you get this working at a anywhere remotely economical here in the US? Funny thing, I think if your drone doesn't use radio frequencies, it's exempt. It's an FCC rolling.
Austin Brown: It's just you know, it's annoying when the rules change every day, right? I what I can tell you a bit about is making motors that don't use magnets. Like how do you do that, right?
Bayley Wang: Yeah, yeah.
Austin Brown: The two popular strategies, at least in the automotive area, are induction motors. So induction motors have like a conducting rotor. And the issue with those I believe is I think they don't scale down very well. I think basically the way you can think of an induction motor is let's say you're driving it at sixty hertz. In the case of a like big induction motor, you have like let's say two hertz of slip, right? That means that, you know, if you're putting in sixty Hertz of electrical energy, you're gonna get out fifty eight Hertz. So that that's like, you know, what, like a three percent loss right off the table. Yeah. I believe that that freak like the slip frequency goes up as the motor size goes down. I know that that's true with like big industrial motors, like like you know, megawatt class. I think they have very low slip frequencies. But as You scale the motor down, the slip frequency will go up, and you can only scale it down so much before your slip frequency. At some point, the motor just becomes small enough that it like it just doesn't work. So I'm not sure if that would work. The other way of building small motor. Is switch reluctance motors. And those are basically just like you replace the magnets on the rotor like steel teeth. And people actually really like those for like missiles. there's no multiple parts on the rotor. So they can spin at like 200,000 RPM, no problem. I think the Wikipedia page on switch reluctance motors says that, you know, in the sidewinder missile. One of those missiles had a generator that was a switch reluctance motor because they could like put the turbine exhaust like right onto the switch reluctance motor or something. Like weird you can do weird stuff like that if the rotor is is just a stack of laminations. It's there's no there's nothing that can explode, so they just, you know, push him really, really fast.
Bayley Wang: Th that that is that is fascinating. That's a that's a really cool feature. Yeah. I never thought about it. And I guess you you have no magnets to like have any currents at high frequencies either. Right. That's a that's a that's a really cool feature.
Austin Brown: Yeah. Yeah.
Bayley Wang: what do the PMs do in a motor anyway? Like they they you can build a magless magnetless motor, so like what what does the permanent magnet do? It's not like making energy, it's not like storing energy.
Austin Brown: Well, it does store energy. There is energy stored in the magnetic field, but the way that you can think about it is it generates a flux which then goes through the conductors, right? From a from a physics standpoint, the flux from the permanent magnet goes through the conductors and that makes when you rotate the rotor you'll get like positive, negative, positive, negative on the coils, right? maybe it's Faraday's law says that by doing that you'll get a a voltage induced by the changing flux. And so if you have voltage induced, then if you happen to put current in at the right time, you will get force generated if your motor is designed well. so all the the only thing that the permanent magnets are doing is they're really just driving this flux through the coil. And with other weird motors like induction motors and squish reluctance motors, that's when things get a little bit more complicated. like an induction motor, like there's a charged magnetic there's some big coil which is charged up which is exciting the fixed stator coils, right? But why people really like permanent magnets is that they always stay on. and so, you know, in an induction motor you kinda need to charge you keep the rotor like charged with current basically, which is a source of loss, right? The rotor will experience I squared R losses. Not much, but you know, when it's a huge motor it can be quite substantial. they're also lighter, you know, an induction motor needs to have like this like honk and rotor, right? which is made of steel and copper, like the heaviest possible things. A lot of times permanent magnets will decrease basically the rotor thickness. so, you know, for a big induction motor, like I I could have like some big donut shaped rotor, whereas if I replaced it with permanent magnets it would be like a much thinner permanent magnet array. And that's actually
Bayley Wang: I see I see
Austin Brown: It's very important for robotics because it takes down your inertia. which
Bayley Wang: I see, in the case that the copper had no resistance, you could replace the PM with like a ring of copper with current going through it, and it would be like sort of equivalent. But the the PM is this magic thing that generates a like a biased flux but somehow doesn't have net current flowing through it, which is weird. It it goes back to the question, what the fuck are magnets?
Austin Brown: I don't know why.
Bayley Wang: Better not think about it. Yeah, I I don't know. Like the the worst part is like the the the electrostatic equivalent of a magnet makes a lot more sense than the like a permanently polarized electric charge is a lot more sensible than a permanently polarized magnetic charge.
Austin Brown: I mean, either way you're sort of like what? You know, I I if you
Bayley Wang: 'Cause there's like a physical unit of electric charge, right? You can you can imagine pushing the positive and negative charges to one end, but like there there isn't a a physical like magneton which is is an electron but for magnets.
Austin Brown: But I think if like the the atoms are like arranged a certain way, like the electrons are like going around or something.
Bayley Wang: Yeah, yeah. They're like spinning spinning around. I I I don't I don't I don't really know either. I don't know how
Austin Brown: Make them work.
Bayley Wang: we can end the podcast with this question, what the fuck are magnets? Which I feel like is a good place to end. and yeah, Austin, thanks thanks for joining us. I've I've certainly learned a lot about motors and I hope the audiences have learned a bit about motors too. I guess my my sort of stance on this thing, especially to the investors in the field, is if some guy comes up and says they built a better motor, you should ask them some very hard questions. people have been doing this for a very long time. And as Austin said There are only four materials we get to work with, and they can only be arranged in so many ways. All right. Thank you, thank you.
