#522 – High Current Power Supplies with Fredrik Kensander

Frederick Kinsander: This is The Amp Hour Podcast. Released December 20th, 2020. Episode 522. High Current Power Supplies with Frederick Kinsander.

Chris Gammell: Welcome to the Amp Hour. I'm Chris Gammell of Contextual Electronics. And I'm Frederick Kinsander of Kraft PowerCon. Welcome, Frederick. How are you doing? Yeah, I'm doing fine. Thank you. Nice to be on the show.

Frederick Kinsander: Yeah, yeah, yeah. You actually wrote in about this. I think you're actually one of our first guests from Sweden. Yeah. Maybe. Maybe. Actually, Simone might be from Sweden. I don't remember where she's from originally, but she doesn't live in Sweden. I'm trying to think of other guests we've had maybe, but yeah, maybe our first guest from Sweden.

Chris Gammell: Yeah, maybe. Yeah, I remember some episode with some guy with that Swedish-like name, but I don't think it was from Sweden.

Frederick Kinsander: Yeah. Yeah. So what is the electronics scene like in Sweden in general?

Chris Gammell: The industry is quite high-tech, I would say. My first job was in Linköping and worked at a company that did encryption stuff. It was quite high-end. Yeah. And we have Saab, which does the fighter jets and yeah, like that. So that's quite...

Frederick Kinsander: I think my former company has some headquarters there. S.E. Braun Bavaria or ABB. Yeah. I think that's part Swedish, part Swiss. Yeah, yeah. ABB is a big one. Which is like a weird bedfellows. It's like, oh, okay. I guess those countries both start with S. Yeah. Yeah. Well, and I mean, I mentioned ABB too, because it's like, we did some high-power stuff, but it seems like, I mean, you're also doing high-power stuff. And so tell us about what you're currently working on.

Chris Gammell: So I work at Kraft PowerCon, where we make what we call rectifiers. And yeah, as I mentioned before the show, when I applied for a job here, they said like, okay, we make rectifier. And I was like, okay, that's four diodes. What's next? No. So it's basically a switch mode power supplies for high current and high voltage.

Frederick Kinsander: Yeah. And like, give us a range. Like, what are some of the power envelopes that you work on?

Chris Gammell: Yeah, it's like the 100 to 200 kilowatts range for the big products, I think. We even have some, it goes up to 700 kilowatts that we make in India, those big things. And so we have different sites. We have both in Sweden, we have in India, we have some stuff going on in China, and we have some sales office in the US and yeah, different places.

Frederick Kinsander: Yeah. I mean, it seems like really specialized equipment. So I would imagine that it's like where you need this kind of power, you're just, people are seeking you out and they're like, yeah, I just have this one need to do this sort of thing.

Chris Gammell: Yeah. And there's like different product areas. So the high voltage stuff is used for cleaning exhaust gases from power plants and the like. So you put like 100 kilowatts on metal plates and you attract particles. Yeah. And then you can like have big mechanical hammers that hammers on the plates and the particles fall down and gets collected.

Frederick Kinsander: Yeah. So that's like a, not like, like a industrial process, like a, it's like in a smokestack or something.

Chris Gammell: Yeah. So it's an in smokestack and they can have different, it's called electrical precipitators, those kinds of filters.

Frederick Kinsander: Yeah. Versus like, I guess another way of doing that would be like actually passing, passing the particulates through like a filter medium, but then you'd have to switch that out. This actually allows it to be reused over and over. Yeah.

Chris Gammell: I guess the flow is much better in this case. Yeah. I guess you would restrict the flow very much if you had, and then you can't really maybe collect all the smallest particles that way. Yeah. Oh, right. Right. So because there's a lot of emission controls going on and I actually know a guy that works like that. He goes out and they measure chimneys and ensure they follow the regulations.

Frederick Kinsander: Yep.

Chris Gammell: Yep. Yeah.

Frederick Kinsander: I knew someone who used to do that and he has to go, he had to go and like inspect towers. It was in Ohio. So like the towers weren't that tall, right? You know, there wasn't like a huge thing, but he would get regularly, like he'd be like climbing towers and like all these birds and making nests up there. They'd come and just like swoop on them because they're like, what are you doing to my home? Yeah. Yeah. Yeah. You know? So what do you just need to go check the filters?

Chris Gammell: Yeah. Yeah. And the products we make for that area, they actually live on top of the power plants. So. Oh, wow. They are from like minus 40 degrees Celsius to plus 50 degrees. So there's some harsh environment requirements. Yeah.

Frederick Kinsander: What about like, I mean, I guess they'd have to run power up to that too. So they'd have to have like lightning arresting and similar kind of stuff.

Chris Gammell: Yeah, I guess so. I haven't been up there myself, but. I'm more of a ground-based creature myself. No, but the power, the mains power is like to one rectifier can be 250 amps of free phase. So there are some big cables going up there, I guess.

Frederick Kinsander: Yeah. Wow. That's crazy.

Chris Gammell: Yeah. Huh.

Frederick Kinsander: And so is that the, that's the ones that you're working on or you're working on something else?

Chris Gammell: No, I'm working more on the high current ones. So that's like for electroplating, putting metal and stuff. So surface treatment like that. So you dip stuff into a chemical bath and you put a lot of current on it. Yeah.

Frederick Kinsander: And what kind of current are we talking about here?

Chris Gammell: Yeah. Here we talk about like one kilo amp to 30, 40 kilo amps. Okay. All right.

Frederick Kinsander: That's something I'm seeing. So then at the power envelope, so 30 kilo amps, and that would be what, like the 50 volt range? Like what's the voltage then? Yeah.

Chris Gammell: Yeah. The voltage, voltage is, yeah. For the electroplating stuff, it's 10 volts, maybe. They go up to 12 and we have some cases with 20 volts.

Frederick Kinsander: Okay. And then about the electroplating process too. So like, what about, what about the high current? Like why, I guess I don't know how electroplating works in the first place, but like, so you're passing current through a liquid and then is it like you, is the thing that you're trying to plate that acts as like an anode or a cathode?

Chris Gammell: Yeah, exactly. You dip it down and it's one part is the anode and the other part is the cathode. You put a lot of current in it and I guess the metal that's in the solution that gets put onto the thing that you want to plate. Okay.

Frederick Kinsander: And so it's just the high current is just for higher throughput or higher amounts of plating or whatever? Yeah. Yeah.

Chris Gammell: So the bigger the area, I guess, of the object you're trying to plate, the more current you need and the faster you want it to go, the more current you need. And it's also used for like for a PCB manufacturing, you know? Oh, cool. Because you put copper on the PCB. So the same kind of process.

Frederick Kinsander: So this would be done at the, like the, at the actual board house, or this would be done at a manufacturer that's making like, like a prepreg that has like pre-populated with copper on it or something?

Chris Gammell: Hmm. It's at the board house where they actually make the PCBs. Oh, okay. So, and the simplest one is the DC, but we also make pulse products because if you're going to plate like a through hole in a, in a PCB, like a via, it's hard to get the same amount of copper in the hole as you get on the surface of the board. Yeah. So the high, the, the aspect ratio, then the higher the aspect ratio goes, the harder it gets. So it's like, I was just reading days, they see, they will use a lot of pulse plating for new 5G boards because like the back planes are usually thick boards, maybe four millimeter thick. And if you have a via that's like 0.25 millimeters, the aspect ratio is really high. Yeah. So they use something called reverse pulse plating. So maybe you go first positive pulse and then you go negative a couple of times higher than that. And the, these products goes in the sub millisecond if you want to. So that's quite fascinating also.

Frederick Kinsander: Yeah. That's, that's great. And, and the same kind of power power or current levels, huh?

Chris Gammell: Yeah. And it's, yeah, it's like, I think our product says 100 amps to 2,400. So it can be kilo amps. So maybe you go 100 amp forward and then one kilo amp backwards and you have a pulse that's one millisecond or 0.5 millisecond long.

Frederick Kinsander: Yeah. I feel like I would love to see like, I would love to see like a science educator do like an illustration of like how, how the ions are moving. Like I, I, I, I guess like kind of, I kind of get an idea why a pulse would matter to get it down into like a high aspect ratio via, but I don't understand the actual like physics, like why a pulse versus like a sine wave or, you know, like, you know, like that doesn't make sense to me.

Chris Gammell: Yeah. For some reason, it's like, I read a little bit about it, but it's like the, if you do DC, the, the most, uh, ions will go to the, um, the high current density areas and you, okay. And you have high, higher current density on the top side and bottom side. You don't have so high current density in the hole before you get copper down into it. So, and the pulse helps to get the copper down into the hole.

Frederick Kinsander: Got it. Okay. So it's like, uh, instead of like slopping paint on the outside, you're like painting a car, you wouldn't use like a paintbrush. You use like one of those Mr. You know, like those, uh, a spray gun type of things instead. And like, you'd use a pulse at a time to get it. Okay. All right. Yeah. I guess that makes sense. Like, like intuitively, I guess that makes sense. But like, I just, yeah, it seems like there's physics at play that I just don't understand. No, no. I think the closest I would have is like, so I used to do when I used to do like dry etch, you know, like it was always hard. It was similar, like hard to do really high aspect ratio, like the as are etching, you know, through holes and, you know, we'd have these weird recipes for it, but it would be, it wasn't necessarily pulse so much. You know, there's always like RF stuff happening, you know, weird 60 kilowatt thingies. I really didn't know what I was doing there. Let's be honest. Yeah. Yeah. Yeah. But, uh, yeah, that's interesting. I mean, do you, do you get to go and visit some of the places that the, uh, that you've, you've, uh, your stuff goes into?

Chris Gammell: Uh, yeah, we said one place that actually just a few kilometers from here where they were making, uh, PCBs, uh, Cobra. Cool. It's called. Uh, and, uh, but they have some just, uh, I think it was just the DC products. So not the fancy pulse. Although most of the pulse products is in like China. Okay.

Frederick Kinsander: Yeah. Like super, super advanced type of boards and mobile type things. Yeah. So like that. Yeah. Yeah. I guess I never, I never really thought about that with like, you know, you can get like laser micro, like I just, we talked about last week on like with bunnies, um, bunny Huang's post about the, um, the laser micro vias and on that board that he was making. And it's like, yeah, okay. I kind of get like the making of the board, but then, yeah, you think you gotta, you gotta plate everything too. And that becomes really, really tough.

Chris Gammell: Yeah. Yeah. Yeah.

Frederick Kinsander: What, uh, what, uh, what, uh, what elements outside of PCB plating are, are these used in? So you said, uh, uh, uh, electroplating and the PCBs, are there other, other industries?

Chris Gammell: Yeah. Ballast water treatment is a really big one. So, you know, uh, big cargo ships, they have some, when they have offloaded their cargo, they need ballast water in the tanks to stabilize the ship. Uh-huh. And, uh, if you pick that water up somewhere and then you go across the globe to some other country, you don't want to let it out because you could bring marine life from one area to another one.

Frederick Kinsander: Mm-hmm.

Chris Gammell: And it actually became a legal requirement just some year or something ago. There's been a lot of debate about it, but it finally became legal. So all ships must have it. So now it's a really big industry. And, uh, yeah, uh, well, I was fascinating like a year ago to see our volumes go from go basically go 10 times up. What we made in a year, we suddenly made in a month. Holy moly.

Frederick Kinsander: Yeah. Yeah. Yeah. That's gotta be some interesting requirements. So first off, I, I, I have, I remember like when I was like in school, it was like zebra mussels were like this invasive species in like great lakes where I grew up. And so that was like a big thing. And, you know, these tankers would come in and they'd be dumping zebra mussels. And so like, that was like a, a thing that I learned about. And it was, yeah, it was basically that problem, you know, I would come and do that. But the, the interesting thing I would think is marine isn't don't marine electronics have like 400 Hertz, like don't they kind of like airplane electronics, like weird, like onboard generation capabilities.

Chris Gammell: And they have higher voltage. So I think the normal, uh, free phase voltage is 440 volt. Okay. So, and they can go up to 480. And, uh, then when you make a product for that voltage, you should also go plus 10%. So we tested up to 528 volts. Okay. Yeah. At the same time, we should also work at 400 volts so we can test it here in house. Right. So it's put some requirements on the primary side of the, of the rectifier to be able to cope with that range.

Frederick Kinsander: Yeah. And yeah, it's, that's, that's pretty extreme, huh? It's, I mean, it's extreme even at 400 volts. Let's, let's be honest, this is way outside the realm that I'm usually working at. So, yeah, it's, uh, so let's talk about like safety a little bit then. So how do you keep yourself safe when you're like in the lab and, and working on this sort of stuff?

Chris Gammell: Yeah. Yeah. We have like, first of all, yeah, we have a safety manual that you should know about and follow the world. And, uh, like you, you should know when you're working on a free phase and you should try to put some protection on it so you don't touch it by accident. And also we try to. So that's like covers or. Yeah. Yeah. Put some plastic covers on it. And, and, um, like the products are made relatively safe. Even if you remove like the front panel, so they can be maybe IP 44 with the front panel on. And even if you remove it, maybe it's IP 21 or something. So you cannot accidentally touch things with high voltage.

Frederick Kinsander: Oh, okay. I actually, I don't, I don't know. These are like, uh, I, whenever I hear IP, I always think of like waterproof standards, but these are like.

Chris Gammell: Yeah. Yeah. Yeah. I'm not sure if it's IP 21, but it's basically, you cannot get your finger into it in the dangerous area. But if you open it up so you can get your finger on that part, so you could put some plastic cover on it. And if you're doing some really, really dangerous, you can have one, one of your colleagues standing at the emergency breaker. Yeah. So that's good. Yeah. To pull that one. If something happens. Do you do reflex testing for your, your coworkers? No. I have something.

Frederick Kinsander: Hey Bob, how are you feeling today? You got good, fast reflexes?

Chris Gammell: Yeah. It can always be, uh, when somebody is like about to do something and you make like a high pitch sound that they think there's a spark that we sometimes jump up. If you want to mess with them. Yeah. Yeah.

Frederick Kinsander: I, I try not to mess with people with high voltage around. Yeah. That's, that's.

Chris Gammell: No, no. No, but, uh, so that, that's the one side. And then you have the output when it's like, if you have a one power module with 300 amps out, you put it into a resistor and that resistor usually starts glowing. So yeah, at least it will be orange. And I've been told to stop before it gets white.

Frederick Kinsander: Okay. So like color temperature type stuff. Yeah. Yeah. Yeah.

Chris Gammell: So yeah. It can, it can get hot in the lab.

Frederick Kinsander: Yeah. I was going to say, Sweden is not a, you know, the winter's here. And so, you know, it's helped to heat up the office. I'm sure, you know, like everybody's got space heaters built on their desk.

Chris Gammell: Yeah. The lab is always hot. And I know we had a period when we were developing a new power module and that power module is 20 kilowatts. So you had a 20 kilowatt heater going on in the lab for quite a few hours.

Frederick Kinsander: Yeah. Yeah. Do you have like specialized ventilation then to try and like cool things down or is it just like you just deal with it?

Chris Gammell: No, you just deal with it. But we also, it's a big factory. So the lab is rather small, but if you go out to the big factory and find some open space and you have a couple of meters up to the roof, it gets easier because the heat rises.

Frederick Kinsander: Yeah. Oh, so you do, you do manufacturing on site?

Chris Gammell: Yeah. So we manufacture everything here. Nice. So that's kind of nice because when I work on stuff, I can go out and pick up a PCB fresh from the production line. What kind of capabilities does your production line have? You mean how many is one thing we can make?

Frederick Kinsander: Actually, I meant how small can you go if you need to?

Chris Gammell: Okay. No, no. So we're not making like PCB assemblies. We purchased like the PCBs from our suppliers. Okay. So here we assemble the product. So we have some metal casing. We put in the heat sinks and the boards from like for the primary switching boards, the transformer, the secondary diodes, some output filter board, and everything is assembled on a line. And then it goes to some module testing. If you look at the modular products. Got it. Okay.

Frederick Kinsander: Yeah. So I imagine your test stands are pretty interesting too.

Chris Gammell: Yeah. Yeah. So that one is like a big hood that goes down over the module automatically to like protect you because it's the first time you power it on. Some things can go bad. So it's good to have safety for your workers. So yeah.

Frederick Kinsander: Yeah. Especially in an automated way, right? It's just like, it's just, they walk away and it's just doing its own thing. Yeah.

Chris Gammell: So it's a hood goes down over the power module and then there's some automated testing going on and on, making sure the output is correct and it, uh, uh, following the specs. Yeah. So that's great.

Frederick Kinsander: Yeah. That's great. Is there like calibration that's required in these kinds of devices? Yeah.

Chris Gammell: Some, some, some, some have some calibrations and, uh, because we have a quite big range also. So some of the oldest power modules for the high current side, they, they are analog. So you have some trimmer potentiometers and stuff like that to, to adjust. And, uh, then you have the, the digital power modules and those also have some kind of calibration in, in the menus of the, of the software. Yeah.

Frederick Kinsander: Yeah. Switch it out to DAX instead of, instead of, uh, trimmer pods. Yeah.

Chris Gammell: And it's even like that, like for the latest product we made, we made sure it could self calibrate itself. So, so the output should be really exact from the power module. There should be no question about it, but we still had to put some calibration menu inside because when that one gets to a customer and they have a current, uh, meter, they clamp it on and they say, no, it's not showing the correct value. So maybe the, that meter is not calibrated, but they still want it to show the correct value. So they want to be able to adjust. So, yeah.

Frederick Kinsander: Yeah. I guess, I guess what I was wondering there too, is like how much, how much requirement is there for like, so if you're pumping 2000 amps into a, you know, a vat of copper, Cooper, Cooper core or whatever is actually plating, do they care if it's like 1 amp, 0.1 amp, 0.01 amps, you know, like less than that? I don't know.

Chris Gammell: Hmm. I'm not so sure, but they still want like, uh, they want like 1% accuracy, less than 1% ripple. Yeah. So, so yeah. 1% on 2000 amps is 20 amps. Like, yeah. Yeah. Yeah.

Frederick Kinsander: That's plus or minus 20 amps. Yeah. Yeah. It seems doable, but at the same time, like that's, that's a, a lot of swing. Like that's just a lot to, to get right.

Chris Gammell: Like, you know, it is. Yeah. And then you all have all these chemists that are making these recipes and they specify it should be like this. And you could argue that maybe it's not so important. Like it's exactly the, like if you have a pulse, it doesn't matter if it's exactly a square wave or not, but if the recipe says a square wave, they want a square wave. So yeah. Can it, you know. Deal with it. Yeah. Yeah.

Frederick Kinsander: Yeah. Okay. Well, can you walk us through like a, you know, like an audio block diagram of what one of these systems looks like? It doesn't have to be like exact to the actual thing, but like you'd mentioned, you know, primary, secondary transformers, things like that. Like what's actually happening internally on one of these boxes?

Chris Gammell: So we could start like on the, on the big stack level because we assemble usually like 10 of them into a stack. So on the stack level, the free phase comes in and you usually have an EMC filter to cope with EMC requirements. So, and then it goes, the power goes to a 24 volt supply that will supply the, all the control boards with 24 volt voltage. And then it goes down, the free phase goes down to all the power modules. And on the input of the power module, you have a small board for connecting the free phase and you have some fuses and you have some kind of common mode choke to take care of some EMC requirements as well.

Frederick Kinsander: Yeah.

Chris Gammell: And then after fuses, you go to rectify bridge, then you come on to the, what we call the primary board. And the primary board has the IGBTs that switches the rectified voltage. So you also have some cap, but it's not so big on our products, like a DC link cap.

Frederick Kinsander: So you'd said, you'd said a rectifier, but then also IGBTs to do like active switching. Is that the idea?

Chris Gammell: Yes. It's primary side switching. So you rectify the free phase into like a seven or 800 volt DC. Okay. And then we, in this, like these high current products, we have like a half bridge. So you have one high side and one low side IGBT. Okay.

Frederick Kinsander: Like a push bowl kind of a topology kind of thing through like a, through a transformer.

Chris Gammell: Yeah. Through a transformer. So I think it's called just half bridge and you have, you have some, two DC caps that has a midpoint. So the midpoint of the transformer on the primary side is into two caps. That's between the plus and the minus on the DC links. You have half the voltage or with IGBT. Okay. If you would have a full bridge, you would have the full voltage. But in this case you have half the half voltage. So that's the, that's the benefit of maybe of the half bridge, but instead it has a higher current than the full bridge.

Frederick Kinsander: Yep. Got it.

Chris Gammell: Yeah. So that goes into the transformer. Then on the output of the transformer, you have the secondary diodes because now it's switching. So now it's not DC anymore. It's AC sort of. Yep. So switching at like a 37 kilohertz for that one. And then you have some interesting phenomena. Also when the diodes are like, they are alternating in their conduction. So when one starts to block, you have the reverse recovery. Basically you need to send some coverage, some current backwards before it blocks. And when it blocks the, yeah, the resistance basically goes infinite. So the current must go somewhere. So you have a voltage spike. So you need a snubber. Okay. So, so there's like an RC snubber on the, on the diode taking care of that voltage. Since these are quite big. Yeah.

Frederick Kinsander: I was going to say, I'm like thinking about the, like the power, you know, we're talking about here too. Yeah. It's like big, probably voltage spikes, but then if you're just snubbing it, that means it's just burning up as heat. Right.

Chris Gammell: Yeah. So maybe, but not so much, maybe five or 10 Watts in heat on, on the snubber, but still you must cope with it. Yeah.

Frederick Kinsander: I just mean like, I mean, relative to like, so like people were listening right now and they're like, uh, my product doesn't use five or 10 Watts ever. You know what I mean? Like, it's just like this, these scale differences are so great. It's awesome. Yeah. Yeah. Yeah.

Chris Gammell: Yeah. And, and, and, and the voltage spike is, uh, on some products. So to be like maybe 480 volt on the secondary side. Wow. So there, there are 600 volt diodes there. So. Wow.

Frederick Kinsander: And it's just, is it just a one-to-one transformer as well?

Chris Gammell: No transformer is, uh, maybe for it's 13 primary in one product. It's 13 primary turns and three plus three on the secondary because you have a center tap.

Frederick Kinsander: Okay. Yep.

Chris Gammell: And, um, and, uh, and also because of that topology. So when you block the voltage with the diode, you basically block double the output voltage. So for like a 50 volt, uh, power module, maybe you will have 90 volt out. If you have high mains because you want to have 50 volts, even when you have low mains.

Frederick Kinsander: Oh, right. Yeah. Right. Yeah. And that's what you were talking about at the beginning, right? If it's, if you're in a boat and you have some like higher voltage, you have to just kind of design for that margin all the way through your product. It sounds like.

Chris Gammell: Yeah. So then you're blocking twice 90 volts. So you're blocking 180 volts. And then we have some nasty reverse recoveries depending on the signs and stray inductance and stuff like that.

Frederick Kinsander: Yeah. Yeah. And does that get worse because these are bigger diodes as well? Like the, the recovery and things like that?

Chris Gammell: Yeah. Maybe the recovery is a bit worse because of that, but also because of the, the like design itself. It's, you have one big transformer, you have some copper bars going to the diodes. You have some copper bar going from the diode to the output inductor. So yeah.

Frederick Kinsander: And it's just all copper because it's got to be as low resistance as possible.

Chris Gammell: Yeah. Wow. So it's, it's a lot of copper bars on the. Yeah. Yeah. As a, as a side note, I, and I worked before I worked at PowerCon, I worked like with the encryption and telecommunication stuff. So it was all really low power. Then I come here and my first day at work, I mean the, out in the lab with a big wrench connecting cables and the, like the cables are like 120 square millimeters. And yeah. Yeah. So, yeah. So, so after the, the secondary diodes, you, you go to a big inductance to be, to, because you have a square where it basically coming out from the. Yeah. From the transformer.

Frederick Kinsander: Right. You're not, you're not like softening edges to lower your EMC or anything like that. Right. No.

Chris Gammell: No. So it goes into the inductance, which makes it, makes it more into like a sort of form from the square way. Yeah. And, um, on the way, I think it's after the, you also have a current transfer, current sensor to sense the output current.

Frederick Kinsander: Not a sense resistor at 0.00001 ohms.

Chris Gammell: No, you can have shunts, but you need to cool them. Right. Um, in this case, it's usually LM sensor. They make whole effect sensors. Okay. So we have both closed loops and open loops. Um, the closed loops are more accurate and they're a bit more expensive. Yeah.

Frederick Kinsander: Yeah. Yeah. I, I, I always imagine that with the current sensors. I always think about them like, as like, you know, clamp sensors and things like that. You know, they're the coils going around a, uh, a wire or something like that. Like not, not precision at all. And then it must get harder as, as you're trying to get more precise. Yeah.

Chris Gammell: Yeah, exactly. And, um, yeah, you also have some caps on the output, uh, to take care on some of that ripple.

Frederick Kinsander: And so that, that, uh, you mentioned the, the sawtooth and the output. Mm-hmm. Is that what you expect to see then? So if you were brave enough to stick a scope on the output, would you see like, is it like a, uh,

Chris Gammell: Yeah. You see, uh, uh, like a 20 amp sawtooth before the output capacitors. Okay. So they are taking in that, uh, current. So the caps are handling that, that sawtooth and helping to smooth it out. Yeah. So in the end on the output, yes, you have less than 1%. So if you have a 300 amp output, you have three amps, maybe that's left.

Speaker ?: Oh my gosh.

Frederick Kinsander: So how much of this is like, uh, is any of this on a PCB or is this all just like, uh, like up on pegs or, or some kind of like, uh, individual elements?

Chris Gammell: Yeah. Yeah. Yeah. It's, it's all, it's all on, uh, like copper bars and the connections. Mm-hmm.

Frederick Kinsander: So like standoffs and things like that.

Chris Gammell: Yeah. Yeah. Copper plates and connecting stuff. Wow. Yeah. And, uh, yeah, the primary board has the current going through it, but we are only talking about 14 amps in that board or in this current product, maybe it's 35 amps of primary current. Sure, sure. Yeah. Yeah.

Frederick Kinsander: Only, only 35 amps, right? Right. So low. So then, so then you must have to, so you're then monitoring this stuff using that current sensor. Yeah. Um, and you're doing some kind of control then to then feed that back or what, back around the transformer with isolation or something like that?

Chris Gammell: Yeah, exactly. So we have some isolation going on to, and then, but the current sensor is isolated by the signs. It's sure. Sure. It goes around the bar and it has no contact, so it's isolated, but we also sense the output voltage. So the output voltage and the output current is going back to the control board that sits at the front of the module next to the primary board. Mm-hmm. So it will read that, it will feed it into its PID regulator and it will increase or decrease its pulse width depending on that. Wow. So, and we run, in that case, we run two regulators in parallel, one for voltage and one for current. Okay. And whoever says it wants the smallest wave, PBM wins. Okay. So that's how you can cap on both. If the customer can set, I want 12 volt and 200 amps, and if it gets 200 amps, it will stop at that point, even if the voltage is maybe three volts. Mm-hmm. Or maybe the resistance is different, so it goes up to 10 volts or 12 volts and regulates on that one. So, and it can also be that if, you know, if you remove the, if you're doing the plating and you're putting stuff into the bath and you're removing stuff from the bath, the resistance changes.

Chris Gammell: Right. So you need to limit both the current and the voltage. So you have set points for both.

Frederick Kinsander: Got it. I have to say, like, even though it sounds like a closed loop system because you are sensing on the, you know, the secondary side and feeding it back. And it's like, it is closed loop, but it, it almost feels like it's open loop because you're just like, well, hope this changes is how we need it to. You know, it's like, I mean, like what is the propagation time as well from like, you know, from sensing a change to then getting it back around and actually impacting the difference? Do you have a spec on how fast you'd expect to be able to respond to a change?

Chris Gammell: Yeah. We basically respond in the next pulse. So the, the switching frequency, if the switching frequency is 47 kilohertz, kilohertz, you have like 27 microseconds. Wow. So you sense it and you do some calculations on it and you regulate your PBM.

Frederick Kinsander: Hmm.

Chris Gammell: Okay.

Frederick Kinsander: And, and, and when you said pulse too, you mean that like, so we have the PWM, is it, is it continuous PWM going or is it, is it, are these actually pulses that are happening?

Chris Gammell: No, no, no, no, it's continuous. Yeah. Oh, it is. Okay. Yeah. So, so yeah, we react quite rather instant.

Frederick Kinsander: And, and so when you do, when you do a pulse, like you're talking about, like for the copper bath for like plating those, those tiny, tiny vias. So is that also driven from the primary side and you're doing like a pulse at a time? You're like saying, Hey, we're only going to turn on for the short amount of time and then turn back off.

Speaker ?: Yeah.

Chris Gammell: Yeah. In that case, there's a little bit of extra stuff coming into the picture. Then it's another product. So we base it on the, the high current power module. Sure. Yeah. So the high current power module supplies the, the free hour or free underdamps. And in that case, I think it's 30 volts. And then it goes into a pulse module and the pulse module has a lot of capacitors. So it basically, it stores that energy. I don't know how many are they are, but yeah, there are, there's a lot of them. And, and then there's a H bridge on that pulse module that actually sends that power out. And the, the output of the, of the product is like five volts at maximum. But the, the actual voltage that it's working with is, is 30 volt. And it's because it needs to create those really sharp edges. If you didn't have that high drive voltage, it would, wouldn't be a square wave. It would be a sine wave coming out.

Frederick Kinsander: Mm-hmm. And so it just like snaps off when it gets to the point where it needs like 30 volts. So it climbs super fast, but then you snap it off using the H bridge or whatever.

Chris Gammell: Yeah. So in that, that is a separate regulator. That one needs to be really fast. You don't get the overshoot and things like that. Yeah. Yeah. I was looking into that product just a couple of weeks ago and I didn't, I was not here when they made that one. Maybe it's 10 years old. It's written in assembly and it's like, okay, we're adding some features here. Let's not touch that part and try to, yeah, it works. Let it be and see what else we can add. Yeah. And, uh, and that one was also interesting because it's like an old processor. I think it's, uh, 20 years old or something. And it was like running out of RAM all the time. So it's like, okay, we have three bytes of RAM now. Okay. Okay. Let's do something with that. Oh my gosh. A new feature.

Frederick Kinsander: Yeah. That's great. Yeah. One thing that's really obvious to me is like, so obviously you've been very open talking about this stuff too. And it's like, so like this kind of industry or like when I was working at Keithley too, and it's like, you know, people come in and like, yeah, maybe you can work on this stuff. But like, there's one, there's expertise in people like you, but two, it's just like, it seems like it's so dependent on material sourcing and like, like tribal knowledge internally of like, like what it takes to get like super sharp edges. Like you're talking about, or build out a supply chain that could actually make this thing happen on a regular basis.

Chris Gammell: Yeah. Yeah. Yeah. And, and Kraft PowerCon has been around for a while. It's, it's founded in 1935. So it's, it's 85 years now and people have come and come and gone over the years. So I guess the knowledge stays in the company and we get new people and we get new ideas and we continue making the old ones also trying to support everything. We can, we can get rectifiers from the Swedish railroad here that was produced maybe 30, 40 years that they want serviced. Wow. Like they'll come back and they'll come back from them. They come back and they say, can you fix this one? And I guess it's because maybe they don't want to do a new EMC measurement of the train. So they cannot exchange it to something new without doing a new testing. Like a reevaluation. Yeah. Yeah. Exactly. Yeah. So it's, they try to get it fixed. And there was one fascinating thing also recently. It was one guy who, who purchased a car repair shop and he found this old battery charger. It said Tudor on it. And the Tudor was the name when the company was founded. So he, somehow he found us and reached out and said, do you know anything about this one? And I went into the archives here and looked around and actually found some reference to some metal, front metal plate. And it was from 1937. Holy moly. That's amazing. Okay. On the, on the 4th of January, 1937, somebody took up out the drawings number for that front plate. Yeah. And it's been the same for me when.

Frederick Kinsander: I mean, did you like at least offer to buy it from him and be like, we got to put that in our lobby or something?

Chris Gammell: Yeah. Yeah. Yeah. Yeah. They are trying to buy it now. I think they want to put it in the lobby. Exactly. Oh, that's so cool. Yeah.

Frederick Kinsander: Yeah. That is, that's crazy that there's, I mean, like just that, that kind of longevity in a tech company. You don't, you don't hear about that often.

Chris Gammell: No, no. And it's even the, the, where, where the company was before because it's a mode from one city to another or some, what's it called? Suburbs of the Gothenburg. Oh yeah. Right. Right. Yeah. So there was a fire and in the factory and they moved to a new place. So some of these documents have survived that fire. So this one was a bit, you can see that it has been wet at some point. Yeah. Yeah. Yeah. And the history is fascinating. Like when I made a couple of PCBs here and you go and take a new PCB number. If you scroll back in that paper, you can see PCB numbers that were taken out before you were born. Yeah. Wow. So like, okay, this company has been around for a while.

Frederick Kinsander: And now I'm. I mean, like, was, did the company have access to PCB houses earlier than others maybe too, because they were supplying them with copper plating capabilities? So they just like were tied into that supply chain?

Chris Gammell: Yeah. I'm not sure. I know that we made stuff ourselves here. At least we assembled the PCBs. We had our own assembly line. Yeah. But later we sold that one to a company that became our supplier. So kind of outsourced it. Yeah.

Frederick Kinsander: Yeah. That happens. I mean, like, and especially like if usually that always seems to happen when it's like, well, do we want to update from, you know, we can't get through whole parts like we used to, but, you know, do we want to update to an SMT assembly lines? Like, nah, we'll just push that off. And then, you know, someone wants to buy the old assembly line or whatever too. Yeah. Yeah. Yeah. Yeah. Yeah. That's, yeah, that's crazy. Wow. So, so the boards, I mean, so like, like I said, the supply chain, I mean, it seems like you guys really, you're making a specialized product. You have specialized knowledge about all this stuff. You had emailed me about all this stuff because it was actually when I started talking about Zephyr.

Chris Gammell: Yeah. Yeah. Yeah.

Frederick Kinsander: Yeah. You had sent me an email that said Zephyr without the RTOS part.

Chris Gammell: Yeah. Yeah. Exactly. Because you've been talking about Zephyr on The Amp Hour, so I got inspired by that one. So what do you mean by that though, without the RTOS part? Yeah. Because you were talking about the hardware abstraction and, and right now I'm working as a software team leader here at Kraft PowerCon and trying to make the software part a lot more modern and modular and flexible. Because it's a company that everything was analog and then somebody invented microcontrollers. Yeah. And then the double E's started writing code and everything went downhill from there. Yeah. Yeah. Exactly. The double E's started writing code and they got some kind of ID from the, from the vendor because it was easy. Not a lot of tools to install. That's right. Yeah. And they write some. So they're using pick parts.

Frederick Kinsander: Is that what you're saying? Yeah.

Chris Gammell: And they were writing code and then they discovered they need to make another product. So they copied that code to another product. Yeah. Yeah. And they even, didn't even have like version control. Maybe it's not, it wasn't, wasn't invented yet. I don't know. But at the beginning, all the code was on the, on the server. So you can see actually they had some rule that if you add a line, you will add a comment about which version you added that line in. So it's like all over the code, you see dash, dash version 1.2, dash, dash version 1.4 for every line.

Frederick Kinsander: So instead of like a useful content, yeah, useful comment, it's just, it's just the version stuff. Yeah.

Chris Gammell: There were useful comments, but we're also like the version ones. At least it was a try. Yeah. Yeah. So, so then when I started working here and the first I did some hardware stuff and PCBs, but now I'm working more with software and we're facing like we have 40 different firmwares that we try to maintain. And a lot of them are just copies of copies of each other. So if you find a bug in one, you might need to fix that one in three, four, five other ones. You had to go and put dash, dash 3.2 or dash, dash 3.7, right? Yeah. At least it was in SVN when I started here. So subversion, but now I'm moving things to like Git and Azure DevOps and trying to make a continuous integration pipeline and all that stuff.

Frederick Kinsander: I just, I just imagining people listening right now and they're like 40, 40 straight minutes of talking about like hardware and like, you know, oh, high power. Yeah, this is awesome.

Chris Gammell: And then boom, firmware.

Frederick Kinsander: Gotcha.

Chris Gammell: Yeah. Yeah. Yeah. Yeah. That's great. It's an important aspect. And I think it's a really good part of the amp power that you go into electronics, but you also go into the firmware part.

Frederick Kinsander: Well, yeah. So we, I mean, like it feels like it's unavoidable. I mean, like, so obviously you guys are making analog stuff, right? And that's a very viable product line. But these days, I just imagine that if you wanted to add new features, you would be doing it with a microcontroller or with, you know, some UI element or some other thing on there. I mean, is that, has that been a tough cultural shift internally?

Chris Gammell: No, not so tough to shift to that, but maybe tough to see the value. People are starting to see the value in software now. Maybe you can offer them not a display that displays current and voltage. You can offer them a display that's fit to your process. You can say, I want five micrometers of copper, or I want this amount of hypochlorite for my ballast water treatment. So you can, you can tailor the display and the GUI to the process to provide value that way. I think that's one important. And the other important things that you software brings is the ability to push out fixes. So if some customer has a problem with the regulator, oh, you need to change the P or I or D constants of the regulator on the old one. You had to send a guy out to solder some resistors. Oh, wow. Now they can do it through the menus. And there has been a couple of cases where, okay, we discovered a problem. And then it's like, oh, we need a hardware fix. Can you do something on software in the meantime? And then I come up with a fix that mitigates the problem until we can exchange some boards or something. Just to like always try to help the customer in the fastest way possible.

Frederick Kinsander: Yeah. I mean, I mean, like lying down situations are just so detrimental to production. Yeah. I can imagine that you would get very angry phone calls if something's not working as they needed to.

Chris Gammell: Exactly. So as the product flora, it grows, you get more and more products and it gets harder and harder to maintain them. So that's where I thought about, okay, we should really make this modular. If we have like a canvas implementation, why not have the same in all products?

Frederick Kinsander: And so it's not currently you're saying? No. Right now there's individual implementations? They're individuals.

Chris Gammell: They're like copies of each other and they're also individual implementations. So try to make it more modular, easier to maintain. And when you make stuff modular, you start thinking about the API between them, the interfaces. How can I abstract the interface of this module to work with any other module that I want to connect it to?

Frederick Kinsander: Yeah. The API is just the connector of the software world.

Chris Gammell: Yeah.

Frederick Kinsander: Which pin is which? Yeah, exactly.

Chris Gammell: So recently, I think it was this spring, we hired a new software guy called Zisis and he had worked with Autosar in the automotive industry. So he had some background in like modular design and we were thinking about that. But people start shivering or something like that when you mention Autosar. It's not the nicest experience for some people, I guess. What is it called again? Autosar. Autosar. It's a framework for automotive.

Frederick Kinsander: Okay. All right. Yeah. So I guess I'd have to, I'll look that up.

Chris Gammell: Yeah. Yeah. Yeah. Yeah. Yeah. And, but I have heard Zephyr mentioned a couple of times on the Amp Hour. So I started looking into it. It's like, it looked really good. It has really good hardware abstraction. And we said to each other, well, why don't we make it like that? And I think it was like the day after that, I listened to the Amp Hour and you and Dave were talking about it. And I was like, yeah, yeah, we are trying because we said that if we make our modules in with the Zephyr abstraction, we cannot reuse them if we start to use Zephyr also. So it's like, yeah, we have something to follow and we don't invent it from, take something that works. Instead, if you try to invent it, you have to iterate over it to get it working. Here we have a reference. You can just implement like that reference.

Frederick Kinsander: Right. Yeah. So even, yeah. So you're just re-implementing like an API layer and you're not going to, so you're going to do that. You were saying you're not necessarily going to use the whole framework. You're not going to use everything from it because I think you said it was specifically because of timing requirements. Is that right?

Chris Gammell: Yeah. It's like some of the products are set up in a specific way. So you have like, it's done with interrupts right now. So maybe you have a one kilohertz interrupt, a hundred hertz interrupt and a 10 hertz interrupt, and they do different things. And there's some priority maybe to the faster one. And the slower one, do some communication stuff. And even in the power module, you have the 37 kilohertz interrupt. That's really important. That must always be served. So if you're going to put it in our R2OS, I don't see the immediate benefit for us. Maybe we can get it to work, but if we don't need it, why go for it? Yeah.

Frederick Kinsander: Yeah.

Chris Gammell: At the same time, I'm a bit tempted because it provides a really good hardware layer also.

Frederick Kinsander: Sure. Sure. But yeah, so this seems like this is like a, this is one of those scenarios where like you have to hit your timing, right? Like so that 37 kilohertz over the 27 microsecond. Yeah. Yeah. Like you cannot miss that, that train as it goes by the station. Things will go boom. Yeah. That's, that's bad. Yeah. Yeah.

Chris Gammell: Yeah. So we were just looking into like, now we're implementing like something called a Profibus, which is a field bus used in industry. They have a lot of buses in industry. So that, that one will talk to a specific ASIC, Profibus ASIC on SPI. So as we make this Profibus module, the SPI calls that it will do will be Zephyr compatible. So we make, make our SPI driver that way also. So we could just take that Profibus module into Zephyr and it would work right off the bat.

Frederick Kinsander: Yeah. Yeah. Yeah.

Chris Gammell: Yeah.

Frederick Kinsander: And so like, if, if you push that code upstream, then someone else could pull that down at some point and, and use that to talk to that same chip you're saying.

Chris Gammell: Yeah, exactly. For Zephyr. Yeah. Yeah. Yeah. So that's why I think it is, yeah, it seemed really, a really good choice because we've also been thinking about R2 was made before not the power modules, but on the stack level, you have a controller that's talking to the display. Yeah. Talking out to the PLC. And as I was listening to the embedded, they discussed about where to use an R2 OS and they had some good arguments for like whenever you have communication. I think they said Alicia and Chris. Yeah.

Frederick Kinsander: Yeah. Yeah. Yeah. That's a good guideline. I mean, and, and so how does that work like internally then? So, so do you have, so you have a processor, like you're, you're a real time processor, your thing that's handling that 27 microsecond must hit deadline for the, for the switching. Yeah. Is that like a standalone thing? And then that has an interface to, to the controller, like you mentioned, or what is like, how many processors are in the system?

Chris Gammell: Yeah. There's a, there's one on each power module and that, that one has its 37 kilohertz and the, and it's getting its set points from the, the, the stack controller and it sends it down either analog in some products. So it's sampled by an ADC or it sends it down by a canvas and then, and that value might be stored in like a global variable. So when the 37 kilohertz interrupt fires off, it has the latest value in some global variable that it will just read. So I think a lot of information between the interrupts are exchanged through global variables.

Frederick Kinsander: Got it. Yeah. Yeah. I was just wondering, like, so you mentioned that one use an RTOS because of that. And then I was just thinking like sometimes when there's like something that's so timing critical, you could put a processor just doing that one thing and making sure that it's always hitting that. But then it could have another interface that's going out to like, you know, like, like you're like, like the, it sounds like the architecture, like you're doing and that, that could maybe do an RTOS or whatever you needed. Yeah. Because it's doing communications. Yeah. Yeah.

Chris Gammell: I've been thinking in the terms of that, because you have this TI line of processors where they have some real time units on them. So they have some different cores just to do some really critical tasks. Oh, like the PRUs? Yeah, the PRUs. Yeah. I was thinking about the PRUs and it's like, oh, maybe I could use the PRUs to do the switching of the IGBTs and then I can use run Linux or something on the other one. It's not so critical. Yeah. Yeah.

Frederick Kinsander: Or like when Jay was on the show, he was talking about that, the STM32 MP1, I think it is. And it's got like a Cortex M4 on board that just does the real time stuff.

Chris Gammell: Yeah. Actually, I have a board with MP1 here. Oh, nice. How do you like it? Yeah. Yeah. I haven't played so much with it, but. Okay. TBD, TBD. TBD. Yeah, exactly. You get a lot of stuff that you want to do and you don't always have time for them. But in that case, we thought about it because we have an STM32 F4 on the stack controller that's responsible for talking to the power modules. So if you switch that one to an MP1, you could basically transfer all the code directly into the F4 part of the MP1. And then you can use the rest of the MP1 for the display and the connectivity and other fancy stuff. Yeah. Where you could use like, I believe it's really useful to use Linux on those parts because you have a lot of connectivity for free in this. Yeah.

Frederick Kinsander: Right. Right. And then you don't have to worry about the real time aspect. No, it's handling itself. Yeah. Yeah. So. How do you analyze all this stuff? I mean, how do you, I mean, obviously you're, if it doesn't go boom, it's good. But, but like, how are you watching all the cycles and testing this stuff on the bench and, and like making sure that you're hitting your, your timing?

Chris Gammell: Yeah. You put your oscilloscope on the things and you measure it. Like you, you must measure like the dead time between the switch pulses, because if they're on at the same time, they will go boom. So you, you, you can set, put some dead time into the PVMs blocks of the SDM 32 and then you measure and put out maximum PVM and you, you can have a current sensor on it, but you can also put the oscilloscope directly on the IGBTs with some high voltage probes. And some, uh, isolation transformer on the, on the oscilloscope itself, or you have isolated, isolated probes.

Frederick Kinsander: You, you mean you don't want to like a chassis ground the whole thing? No, no, not really. Through the scope probe? Yeah, I've done that one.

Chris Gammell: It's like I've seen both, you have the isolated probes. That's good. But some, if you don't have that probes, you can put your, like high voltage probes, but you need to make sure you're not touching the BNC connectors on the scope. Yeah. If they're connected to the 700 volt DC line of the power module.

Frederick Kinsander: Yeah. That's a good way to, good way to waste an oscilloscope. Yeah. Yeah.

Chris Gammell: Need to measure that one. And then like there are some overcurrent protections. You need to verify that that one works.

Chris Gammell: Yeah. Create a shortcut.

Frederick Kinsander: Yeah. I guess that's what I'm wondering about is like, like you're setting up guardrails for yourself. It seems like, you know, making sure that A, B, C and D don't happen. Yeah. But then how do you, how do you validate that? It sounds like you have a test set up for that.

Chris Gammell: You test it up and you start on a low voltage. So the voltage, the voltage we have in the lab, you can just turn it down. Maybe you start at a 24 volt AC and you go up as you, as you see it's working. Yeah. But me personally, sometimes when there's some new stuff I must test and I know everything should work. It's faster. Yeah. It's faster. Just go for it. Either it works or it blows up and then you start on the, on the lower voltage. but if you start on the lower voltage and do all the verification when you don't need it

Frederick Kinsander: it takes more time so it depends time you know i got i've got more i've got more stock than i've

Chris Gammell: got time yeah but of course on a new product you start on that one but when you have a working product and you need to change something i sometimes i put on the hearing protection and the protection glasses and and i go for it oh i just imagining like like you have like a

Frederick Kinsander: repair tech out on the floor and like like oh frederick's doing some testing you just like you're just walking out and dropping like smoking boxes on their desk over and over again yeah oh that's great i mean the it seems like the testing is like a really critical part of all this but like yeah do you have any kind of like automation in in that realm i mean you talked about like doing some continuous integration stuff like are you working towards the like the bench testing side of things

Chris Gammell: as well like automation on that side of things yeah so if you if you look into software we're doing looking to doing more and more like unit testing for all the modules because if you make all this hardware abstraction you can unit test very easily all the modules you have a nice abstraction layer and then you go down to the lower levels at some point you need drivers for the hardware and then you we're looking into making more like hardware i do make hardware boards with all the different like i have a deck on one product so i have a module for that deck i could just make a test board with that deck on it so i can download the firmware like nightly builds and test them

Frederick Kinsander: got it and then have that go into some like a dmm or something like that that's like tracking it or

Chris Gammell: yeah yeah but uh yeah since but since we have production we also have a lot of automated tested in production so yeah like for our pcbs we are setting up test gigs that do automated testing or depending on the volume if it's a low volume board it's many manual and the when it goes to high volume it becomes becomes automated that testing is done but by our supplier so we are like developing developing developing a test stand and send it to them so they run it okay so like oh at the cm you

Frederick Kinsander: mean like they'll have that as part of their like acceptance testing yeah so the boards when they come

Chris Gammell: here they should work and we are monitoring that and making sure we're not having too high too low yield like then you need okay then you need to think about should we we should probably add another test because something it's missed by the test so if you can make all the boards that's not working getting stuck at the cm they will take care of their process because they want high yield

Frederick Kinsander: so yep yep and so what are the actual outputs like so like so one of the modules that you might be testing at the cm using a test board that would have like a like a not like a DAC output but it would actually have like maybe like the the pulse output that's driving the IGBT is that kind of the

Chris Gammell: founder yeah yeah so like if you look at the primary board the the cm they they they have like a a mini a mini variant of the of the power module they have a 24 volt AC input to that board and it's switching and then we're measuring and making sure everything works that the pulses are correct and like that so then it gets verified at that part then the board comes here it gets assembled into a power module and then the power module goes into a module tester here that's automated ah okay yeah and and then the power modules gets assembled into a stack and then that stack goes to the final testing which also an automated test but with personal because you need to connect a lot of cables and the cooling water because you have both water cooled and air cooled but the final test should just

Frederick Kinsander: pass without problems because yeah each individual piece is good yeah yeah and we can see them as the customer

Chris Gammell: this is how it should work at the customer so it should be no false at the final testing yeah but you need and we run it for like an hour or so to make sure it heats up and it works like

Frederick Kinsander: i guess you test drive a car when you manufacture it that's right yeah right yeah okay so you've mentioned the stack a couple times too is that just to like get parallel versions of the same board doing like the same amount of current is that the thought

Chris Gammell: there yeah yeah that's the power of the modular design so we put a lot of modules in a stack and the stack can be configured as parallel output or serial output okay and that's how you can yeah yeah for example the marine products they often use a bit higher voltage so they have 50 volt modules but you can connect four outputs in series to get 200 volts and then you can so if each module has two outputs so on its own it can do 100 volts and then you take it in series with another module to get 200 and then you can make five of those groups in a stack as in parallel so you have 200 volts and 100 amps and you do five of those configurations in parallel so you get 500 amps so 200 volts fiber and amps it's 100 kilowatts

Frederick Kinsander: wow and then uh so then is the transformer in each case of those is the transformer on the actual power module itself is that is or is there like some larger transformer no this one in each uh actually two

Chris Gammell: in each since they have two outputs so the power stage so one power module is 10 kilowatt in that case and each power stage is five kilowatt and then you connect a lot of them so uh that's also one thing we identified when you made we made a product recently that we made into 20 kilowatt because it was like okay what are the customer requirements what are the stage what are the steps they would require because if you make the power stage too small you make a lot of measurements you make a lot of transformers a lot of current sensors a lot of voltage sensors which adds cost to the product so the less control logic because the essence of the product is to deliver power so in that case 20 kilowatt was a good fit but yeah so it depends how flexible you want to be regarding or versus the cost of

Frederick Kinsander: the production yeah yeah it's like the the trade-off in modular design is like how atomic do you want the unit to be it's like you could have every you could say like oh well i'm going to use just like breakout boards for every single chip on the board and then everything is super modular and it's like okay that works but then you have in that example you know the added cost of interconnects and extra pcbs and everything else you might get benefit from being able to switch out every chip but then the overhead to do so is like really really high so it's like where do you where do you kind of draw the boxes around

Chris Gammell: around around these systems yeah exactly yeah and and also then then it becomes quite interesting when you have a lot of these power modules connected in a stack and you want to regulate the output current or the output voltage so because if you put them in parallel they will all see the same voltage the common voltage on the output but you still want to regulate it and you want to you want them to be able to supply uh same amount of current because if one module puts out 50 volt and the other module sees that and it's like oh it's 50 volt on output so my pvm is zero then then everything is fine right right

Frederick Kinsander: so they need to like be aware of like where they are in the stack as well then is that kind of a thought

Chris Gammell: yeah they need to be aware where they are to know if they are in serious because whether you need to figure out the configuration so the you configure it in the menu and then you need to like okay if it's serious connection with two modules then every other module is in is a slave to the one above uh things like that so there's a lot of stuff going into that and then going to like current sharing

Frederick Kinsander: when you're in parallel right right yeah you don't want one to be doing you know 100 amps the other one's

Chris Gammell: doing five amps or something right yeah and when when you are in in series instead everybody sees the same uh current so you must make sure that they actually supply the voltage they are supposed to supply because the current will go for the current from the module before we go through the module after it and yeah it's the same current in the loop when you're in series so so then how does that play back

Frederick Kinsander: then in the software realm then so like you're saying you want to modularize everything on the software side does that mean that like you also have to push that stuff all the way back up this the system yeah

Chris Gammell: there's some stuff going on like uh have like a current share master for example that's looking at maybe one module has that role and it's looking at okay when when i'm outputting uh 50 50 volt and the current is one on ramp okay let's tell everybody to run one on ramp so in that case one module is running in voltage control and the other ones are running in current control uh so there's like a pid in that one to make sure they slowly ramp up to the correct current to get 50 volt on the output for example uh and when you're serious there are like different ways you can solve it you can uh one module can see okay i run when i get this current i have this voltage so i can tell the other modules to run this voltage and you must make sure the whole system doesn't oscillate and the things

Frederick Kinsander: like that so oh right right yeah yeah because i imagine some of these loads are very interesting like in capacitive or inductive kind of probably capacitive mostly right if they're big tanks yeah

Chris Gammell: tanks yeah they're quite resistive and there's some problem when the material goes in and out of the bath oh yeah right and then in the in the high voltage just two big metal plates so they are very capacitive yeah and then you have sometimes long cables which adds inductance and uh yeah yeah yeah so that

Frederick Kinsander: plays plays havoc with your pid loops i'm sure yeah yeah yeah for sure well that's uh yeah this is uh this is quite interesting field i i i wouldn't have guessed that like you know like you say like switch mode power supplies and it's like oh okay that i i think i understand that but then like the just the range of like how it's being put to use it's like it's really quite crazy it's big stuff like like if you look at

Chris Gammell: the final tester area there's like big copper bars in the ceiling going to the load it's like i ran one test with like uh two or three stacks of power modules and it was like 12 000 amps so output current so it's like yeah there's a lot of power yeah controlled by software so yeah yeah that's what's

Frederick Kinsander: yeah that's also kind of crazy about it too you know like it's just oh okay one bit one bit's wrong and

Chris Gammell: well like you said it goes boom yes you need to to put a lot of care into that and uh yeah yeah

Frederick Kinsander: well frederick what else did we know about this uh this field i mean it seems very interesting i mean are you guys uh are you hiring are you uh are you looking for more people yeah at least on the i

Chris Gammell: know that on the software side we are looking for more people because we are uh we are free people now and as i said it's like 30 or 40 different firmwares and the people starting to realize the value of good software so yeah i haven't got to go ahead to recruit but we are looking for instant interesting

Frederick Kinsander: candidates if the right uh resume happens to find its way across your desk you're not yeah yeah no no push it up the ladder you know yeah yeah usually when you get a really good candidate you

Chris Gammell: usually can get an exception that's right yeah oh look look a position just opened up who knew how did this happen exactly and uh yeah i would really like to get uh more competence in the area of regulation loops and stuff like that so if anybody's really good in in that area yeah okay all right so like uh like pid control yeah yeah yeah okay and we're starting to doing more like active pfcs and then you have like six igbtes that you're gonna control with uh with software to get that pfc running

Frederick Kinsander: power factor control you mean yeah exactly yeah yeah yeah yeah so like you're not uh so like when you are in a you're not like making your grid go go sideways no idea exactly and and it can actually increase

Chris Gammell: the efficiency of the products you can have higher efficiency yeah yeah well that's great i mean

Frederick Kinsander: that's i think that is the the right way forward i mean like that you can really start to have finer and finer control of like these really big things but it's like you can you can uh you can get a lot

Chris Gammell: more out of it it seems like yeah i would say i listened to episode 520 inductance and stuff and you you started talking about the the compute module four again a little bit oh yeah yeah and that one you also mentioned on on the on the episode that i was listening to about uh zephyr uh i think it was uh yeah episode 514 yeah yeah yeah so um we're looking into that as well to like add a compute module to control our displays and so it was quite fascinating i was listening to you talked about zephyr and then i went to work and we started discussing how should we control this display oh there's something new called a compute module four and then on the way home i listened to the last part of the episode and you guys started talking about it and uh yeah maybe we're just listening to you here yeah it goes both

Frederick Kinsander: ways of that it's the zeitgeist of the electronics scene it seems like yeah so you on the last episode

Chris Gammell: i listened to the 520 you talked about the routing the like the differential pairs for the hdmi and stuff yeah yeah yeah yeah so that was quite interesting and i i listened to rick hartley thanks to you also he has a really good youtube video about what your differential pairs wish you knew

Frederick Kinsander: okay i actually yeah i don't know if i've seen that rick currently is the the signal integrity guy

Chris Gammell: right yeah yeah exactly yeah yeah and uh apparently there were some guys at john dare that had tested this because apparently they do this stuff in farming equipment also so it was quite fascinating to listen to that one and uh yeah i think you can uh on on a hdmi signal you can skew it by three millimeters but one example was like a 100 megahertz signal you can skew it plus minus 55 millimeters and and and it will still work so wow okay so not always that critical also like i think jay carlson

Frederick Kinsander: talked about what jay carlson was talking about yeah right yeah yeah yeah okay that's great yeah that's good to know like even though you know like have a little margin yeah in the proverbial uh you know digital signals not going boom uh no no it's good to know right yeah yeah that's great well uh frederick thank you for being thanks for listening to the show and thank you for writing in and being on the show this is like super interesting and it sounds like you're doing some really cool stuff there so i hope people reach out to you where can where can they find you if they're if they're interested in

Chris Gammell: maybe pushing a resume across your desk yeah then they should find me on linkedin in that case i guess okay and otherwise i'm mostly on facebook so yeah it's there's only one frederick can sender in the world that i know about so yeah okay probably find me okay well thank you frederick it's been great to having you here yeah thank you for having me it's been a great great to be on the show and i'm a big fan of the show so it's really fun thanks all right we'll talk to you soon yeah okay bye

Frederick Kinsander: talking about high power stuff today we realize nothing makes us feel more powerful than having the backing of our audience on these shows join the other patrons keeping and getting current each week via patreon.com slash the amp hour a special thanks today to our corporate sponsor administered