High Power PCB Design: Pushing the Limits with Caleb Buck

Join host Zach Peterson on this episode of the OnTrack Podcast as he speaks with Caleb Buck, PCB Design Electronics Engineer at Nidec Aerospace, about the critical considerations for high power PCB design. From guitar pedal prototypes etched in his kitchen to aerospace systems handling 500+ amps, Caleb shares his journey and expertise in designing life-critical electronics. Learn about voltage isolation requirements, current distribution challenges, thermal management strategies, and the pitfalls that can lead to melting connectors and component failures.

Discover practical approaches to high power design including IPC 2221 vs IPC 2152 standards, material selection for reliability, and how PCB layout dramatically affects current distribution in parallel switching circuits. Whether you're designing power electronics for automotive, aerospace, or industrial applications, this conversation provides valuable insights for pushing the boundaries of what's possible on a PCB.

Watch the Episode:

More Resources:

• Connect with Caleb
• Learn More about PCB West

Transcript:

Zach Peterson 

Hello everyone and welcome to the Altium On Track podcast. I'm your host, Zach Peterson. Today we're talking with Caleb Buck. Caleb is going to be speaking at this year's PCB West. And of course I haven't had the chance to talk to Caleb much in person, but I do know he has some very interesting topics and so I thought it would be great to get him on the podcast to discuss. Caleb, thank you so much for being here today.

Caleb Buck 

Yeah, thanks for having me. Honored to be here.

Zach Peterson 

Yeah, absolutely, absolutely. Yeah, so you're gonna be at this year's PCB West and you have some interesting discussion topics. But before we get into that, if you could just introduce yourself to our audience, tell us about your background and how you got started in electronics.

Caleb Buck 

Sure. So originally, I really didn't know what I wanted to do after high school. I was really into music at the time, really into art. So I originally pursued a degree in art and design. And in that process, had some really fantastic science classes and started kind of getting my gears turning. And while that was going on, I had started kind of modifying my own

music equipment, my guitars and my guitar pedals at that time. And that really is kind of what really got me started. I've got some show and tell that I dug out from the crypt to show you. This is board that predates probably the 2010 timeframe, but actually designed and etched this in my kitchen. This was a prototype for a pedal that I designed that

Zach Peterson 

you

Caleb Buck 

later transferred to an actual more proper layout that I had, you know, manufactured. I went from a one layer design to a two layer design on this guy, which eventually became, go ahead. Yes, yes. And then, you know, ended up with a full commercial product at one point that I was selling these for a while, kind of internationally, really enjoyed that design process. And I was a I was a like play by ear musician, not really

Zach Peterson 

Doubling, doubling your complexity.

Caleb Buck 

I'm a musician that could read music, but I did everything by ear. like this whole R &D process was by ear, swapping out different components, seeing how they sound, seeing how that makes the instrument feel. But I didn't understand the why. So at that point, I decided to kind of switch gears from art and design and focus on electrical engineering to really understand why this capacitor, you know,

produces a different result in that capacity and really just learn electronics in general. So I went to the University of Arkansas and pursued my degree in electrical engineering. One of my TAs you've actually had on your show, Kirsch Mackey was one of my TAs there. But that's kind of my background and what really interested me in this career field. And then so for the past 11 years, I've been in the

defense aviation and aerospace markets, doing a lot of PCB design as well as some system level activities, system design. Primarily, I would say if I had to say what my expertise is in, it's going to be kind of PCB designs for high power or very thermally challenged systems in small spaces and then high reliability designs where specifically around cleanliness. That's another topic that I've

spoken on it, PCB East and PCB West.

Zach Peterson 

Yeah, I think that's a good segue because you're going to be talking about high power designs at PCB West. But you brought up another point there, which is high reliability. And I think high power designs have to be inherently reliable. But then once you take this into the aerospace defense, even in the automotive sphere, there's this additional level of reliability that you have to start caring about with high power designs. Is that correct?

Caleb Buck 

for sure. And high reliability is like one thing, but I've started using the word life critical a lot more recently because you really have a responsibility as a designer to ensure that if this is a product that supports something that keeps people safe, automotive, like you just mentioned, those are life critical products and we have a responsibility to make sure that what we're designing is reliable in that end use application.

Zach Peterson 

.

Caleb Buck 

whatever it is, whether that's like a cell phone that's a communication product for a civilian or if it's a radio that's a communication product for a soldier, in this application, this might not be life critical. In the soldier application, it might be life critical. If I'm in the middle of nowhere in the wilderness and I'm relying on this, this civilian tool to potentially get me out of there if I get in trouble, this then becomes life critical. So you kind of have to...

consider the exact end use case that is in question for that.

Zach Peterson 

Right. You're really, really designed for the reliability that's required in your ideal deployment environment. Cause you know, I kind of think back to like data center hardware, right? I mean, the, the stresses that they experience in data center are nothing compared to the stresses that you're going to encounter in a vehicle or in aircraft, that kind of thing. So, you know, there's a, there's a reliability when it comes to one product and that's not the same type of reliability in another product.

Caleb Buck 

Yes.

Caleb Buck 

right.

Caleb Buck 

Right.

Zach Peterson 

So if you could give us an overview of your presentation. What can attendees expect to learn and expect to hear about?

Caleb Buck 

Sure, so the presentation that I'll be giving and I'm looking at the schedule here it is October 1st from that's the Wednesday of PCB West it is from 4 to 5 p.m. Pacific time and it's the presentation is called high power PCB design pushing the limits and so what we're gonna try to focus on in that one hour which might be kind of difficult to fit into one hour but I'm gonna kind of break it up into three main

Zach Peterson 

you

Caleb Buck 

kind of sections, voltage specific requirements, current specific requirements, and power specific requirements. And then so in the voltage portion of the talk, we're gonna talk about how you design for specific levels of voltage isolation, how to use IPC 2221 formulas to calculate that. And then for those that aren't aware, the newest release of 2221, which is revision C, they've changed the

Type designator, so there's a new one now. So if you are using some of the older calculators that use the revision B formula, the rev C formula is a little different because there is a new device type. So just be aware of that. I just pulled down the latest Saturn PCB calculator today and it is still using the

revision B formulas, so that's just something to be aware of. And then this kind of covers voltage, current, and power, lightning, designs for lightning protection and EMI protection, what do you do to address that? And then current specific requirements for high power designs, how do you determine the current carrying capacity of the traces or the larger copper features?

We'll do a walkthrough of a 100 amp rated PCB design, actually kind of, all right, here's our requirements, and then we're gonna walk through how we approach, or an example of an approach to solving that 100 amp rated circuit board. In my past work, I've seen a lot of stuff in that 150 amp range on the circuit board, and I've seen several continuous current over 500, which is pretty insane.

But and then also how to avoid areas of high current density, which can create hot spots on the board. And then the power requirements portion of the lecture will be mainly like thermal centric issues. thermal performance of the parts and the board itself. You know, do you have parts that are

Caleb Buck 

Do you intend on getting hot? Are there specific areas of the board that you think will get hot? How do you resolve that? Stack up considerations, copper thicknesses, via design, aspect ratio, consequences of very high currents and very high temperatures, determining the temperature rise. What we really want to know is how hot is the board going to get in some cases? And then there will be some specific

considerations for systems that are connected to high-performance batteries or lithium-ion batteries because they're certainly different than just connecting to a regulated power supply, for example.

Zach Peterson 

Interesting. Yeah, so I've got a couple of questions here. yeah, so with you mentioned kind of the 150 amp range for some common designs. And I've recently had to deal with a 150 amp range high power design. And I don't claim to be like the high power expert. I know enough to be dangerous, probably dangerous to myself.

Caleb Buck 

Yeah, go for it.

Zach Peterson 

But you know, I had to deal with one of those designs recently and it was running at 160 amps and in the first spin of the board Man, did they get hot? I mean, you know melting melting wires onto onto pins coming off of connectors kind of hot, know And even though we've got, know connectors that are that are rated for the correct current and you've got wires that are you know rated for that level of current The heat is a huge issue

And I think it becomes even more of a challenge when people are trying to pack more and more power or current into smaller and smaller areas. So the modules are getting smaller. And then of course, the boards are also getting smaller. So if you could, what are some of the consequences of that? And what are some of the strategies for dealing with it? Because at the end of the day, in the application I was dealing with, we just went heavier copper, more planes, throw more copper at it. Is that the best solution?

Caleb Buck 

It depends like that. think that's the best answer to every question. It really does you really have to from a system level understand everything going on if you're if you're a Just the PCB designer and you're isolated from the knowledge of what it is being the application It's being used in you might miss out on some critical information for example that Well, let's just keep it at a you know around 150 amp range discussion

Zach Peterson 

Yeah.

Zach Peterson

Okay.

Caleb Buck 

If the circuit board is connected to a heat sink that has active cooling, you have a significant amount of thermal performance margin compared to something that's just thrown in a sealed box with no active cooling. They're not comparable. So you really need to understand how it's going to get used. Certainly adding more copper helps to a degree to handle the heat and prevent the conductor itself

from developing hot spots, but that heat will start to spread out and eventually if it doesn't have anywhere to go and be released from the system, you will continue to see the heat build. And it may never reach a steady state condition thermally until you've reached the glass transition temperature, you're laminate and now your board's coming apart potentially. adding copper certainly helps. I'm certainly guilty of...

throwing really, really large copper features in there to handle high currents. there is a point of diminishing returns. Does that answer the question?

Zach Peterson 

A little bit. Yeah. And you brought up something important, which is when you're the PCB designer, but you're isolated from really all of the knowledge needed about where the design is going to be deployed. You, you kind of don't always know what the most optimal option is. mean, if you're doing it for a client, they may just tell you, you know, do A, B and C. And, you know, you as the designer, right, you're just hoping they write you a check when you send them the invoice. So, you know, you do what they tell you to do.

And maybe it's not the most optimal thing, but I think, you you brought up heat sinking, right? Is there active cooling or maybe is it going into an enclosure where there's airflow versus an enclosure with stagnant air? Cause with stagnant air could, you know, now the enclosure basically becomes an oven. So yeah, I think that that answers some of the question. I I'm just interested in what, what else can be done at the system level to kind of, you know, help deal with that aside from just throwing as much copper as possible.

Caleb Buck 

Mm-hmm.

Zach Peterson 

Because suddenly you take what probably should be a four-layer board with two or three ounce copper, and it becomes an eight-layer board with four ounce copper or some ridiculous amount of copper in it.

Caleb Buck 

Yeah, I'm personally not a fan of heavy copper. There are people that do it very successfully, but in my experience, and I've tried to do it before, my experience is that it is much more challenging to fabricate, and your average fabricator is going to struggle if...

you design it with extremely heavy copper. Some suppliers are very good at it, but not everybody is. My very first four ounce design, and the only four ounce design I've ever done, we ordered the special material. It was like three months lead time. We were never able to produce a first article with it because the delamination was so bad on that specific design with that specific material. Eventually that became a three ounce and eventually it became two ounce, just adding more layers of copper.

I had, you know, that was pretty early on in my PCB design days. So now I favor, I really don't like to go above two ounce. Two ounce is very common, very affordable. You can have a lot more options without excessive lead time and cost. then just extra layers is my preference to heavy copper. You can certainly embed, you know, heavy features. could...

You can make the high power path actually not on the circuit board and just neighboring and it's bolted to it like a bus bar. Some of the other things you can do to help with heat is other non-layout heat spreading techniques such as thermal potting that you essentially encapsulate your hot parts in it through a thermally conductive material gets the heat out away, not through the board, but through

thermally conductive potting, it needs to be electrically isolated, of course. I haven't gotten to try this myself, but heat pipes is an option. can embed heat pipes through that gap vapor change process. It moves the heat away from a phase change, if you're aware of heat pipes.

Caleb Buck 

The direct heat sinking is what I have seen most often. If you've got a BGA and it's designed to have a heat sink applied directly to it, or maybe that heat sink even touches the case of whatever it is, that's certainly helpful. I want to kind of go back to what you said about the connectors. it the connector body that was melting? Yeah.

Zach Peterson 

Yeah, it was the connector body. Yeah, because it was cheap, cheap plastic. You know, they, they spec'd headers because that's what they had used on a different design. And they're like, well, you know, we want it to be same form factor and you can just, you know, swap them out, plug and play. And so, you know, like I said, when you're, when you're doing something for a client and they're writing the checks, you give them what they want, even though you try to maybe push them into a direction of something that's going to be more reliable.

Caleb Buck 

Yeah, I'm not going to mention any like manufacturer names, but I wonder if you use the same connector I did that failed. It's not specific to the connector necessarily, but would you mind if I just go ahead and share those slides that I prepared? Because it's super relevant to what you just said. All let me go ahead and share this here.

Zach Peterson 

yeah, yeah, go for it.

Caleb Buck 

OK, so let me start this little presentation here. So this is kind of a very highly generic example of something I've seen multiple times. And it's actually a failure mode that I've observed that's very similar to what you just described. So I think it's just kind of relevant right now. So this is a here in the middle.

We have a group of FETs. This is a common source configuration. So this is a bi-directional switch of some kind. We have a 28 volts DC source of some kind over here on the left. And then we have a load over here on the right. We also have multiple points that we can connect it on the source side as well as the load side. So...

In this specific design case, this fixed 5600 watt load, will end up resulting in 200 amps of current draw through the system. We're just going to say this is 28 volts on this side. In a real system design, you would need to accommodate for the minimum input voltage because that's going to be your maximum current condition. in this case, we're just going to call it 28 volts.

So the questions that we want to answer here and then illustrate the consequences of how this challenge can be addressed is how is the load distributed between the components? How is the load distributed between the connection points? So back to your example you just brought up, you had connectors involved. I have a past example where I saw connectors melt and it was somewhat similar-ish to this type of scenario.

And then where are the areas of high density here? So this dark area is an area that's void of copper. The lighter green area is where there is copper pour. yep. So scenario one here, we have the source and the load connected to all four points as they are arranged here. We end up with

Zach Peterson 

I see.

Caleb Buck 

the load evenly distributed across the components, approximately. the total 200-amp load that will be pulled through this FET array will approximately be equal. However, we do need to consider if that level of current is acceptable for this part. Now, this specific part that I dropped in there is a 300-amp rated part, if you just look at the brief description. But that doesn't tell you.

How long of a duration it can sustain 300 amps and under which? Heat sinking situations that is really accounted for and which gate drive voltage that is accounted for and which RDS on value that is for so you have to do a Quite a bit of analysis to make sure that you are driving the device correctly so that you can reduce your RDS on and reduce the overall heating so

I can confidently say this is the part that, you know, I've used some parts in this family before. 50 amps, three to these, they would get pretty warm. Continuous current for 30 minutes, they would probably be pretty high current. Are you still seeing my screen there? Okay, something just dinged at me. I just wanted to make sure. What is notable here is that the load is poorly distributed across...

Zach Peterson 

yeah, yeah, I am.

So, yeah.

Caleb Buck 

the connection points for the source and the connection points for the load. I'm not giving you an exact number here, but what I am saying is source B will carry more than 50 amps and source A will carry less than 50 amps.

Zach Peterson 

Is that because of the resistance of the copper? You basically have a longer path from source B versus source A and then similarly for source C and D. I see.

Caleb Buck 

Exactly.

Exactly. Yeah, so I had a case in the past where I had an array like this. There were five of them in parallel. And my thought was they're all going to get hot. I need to spread them out so that they don't get hot all in the same spot on the board. And because of differences in how the connection was made to the source side,

It resulted in a change in impedance of the connection to the board, which resulted in I had five groups of FETs and I expected one fifth of the load to be distributed across those five groups. But what ended up happening because of how it was connected and differences in the impedance of the connection on the source side, the

The middle three groups, which should have carried three fifths of the load, carried one third of the load. So this was a 150 amp design. So they were supposed to carry 30 amps per group. And that middle group should have carried 90. It actually carried 50. The outer individual groups should have carried 30 each. They each carried 50 individually. So you had...

Zach Peterson 

Interesting.

Caleb Buck 

the outer connections carrying way more load than the inner connections and those connectors melted because of the difference physically of the path from the source side to the load side. In that case, the source was a lithium ion battery, so very high current potential. It's essentially unregulated output.

Zach Peterson 

Right, right, very high discharge current. Yeah, yeah.

Caleb Buck 

Yeah, yeah, short circuit current on those is pretty outrageous. If you've ever seen any short circuit testing, your cables will literally move when you perform the short circuit. The highest one I've ever measured was, I think, around 25,000 amps. And that's like an individual battery within a larger system.

Zach Peterson 

Yeah.

Zach Peterson 

my goodness.

Zach Peterson 

Yeah, yeah, this is really cool. And I encourage anyone that's listening on audio to head over to YouTube and watch this, because this is a really simple example that's actually a really great reminder of how the PCB layout can have such a huge impact on certain systems. I will say, I usually talk about this in terms of high speed and digital and NRF and that kind of stuff, because that's more my domain. But this is actually really informative, especially for power.

know, power electronics is something I've needed to become a little bit more of an expert in myself. So, so in that case, you know, the simple solution, I guess, is to just spread out each of those FETs that make up that switching stage.

Caleb Buck 

That is an option, but let's say you need something else on this board, so these darker regions. Let's say we don't have copper there because it needs to be used for something else. Maybe our gate driver circuitry is over here. For whatever reason, we can't use that. So to address the issues with the poor distribution of current on our connection points, you could spread out the FETs or you could bring the connection points closer together.

Zach Peterson 

Sure, sure.

Caleb Buck 

if that's an option. Sometimes that's not an option. So if you bring the connection points closer together now, from source D to load D is essentially it looks like the same load path versus C versus B versus A. bringing them closer together would help resolve that. And it also helps resolve the high current density areas that we're building up in the corners.

this the the current the the current from source d here can you see my cursor at all

Zach Peterson 

Yes, I can.

Caleb Buck 

Okay, great. When the current has to turn a corner, it builds up a high current density area around that corner. So, you it would be preferable if you have to turn, you know, 90 degrees like this to kind of expand your copper pour so that it doesn't have to make such a dramatic turn. It has a little more of a straight path to where it's going. But there's been some cases where I've actually used that strategy to help.

redirect multiple sources together and steer the current and kind of rebalance it to a degree before it goes through its next component array, whatever that is. So here's an example where we improved, we optimized that arrangement just a little bit. That's not to say this is like the best way to do this. This is just kind of showing if you end up in this scenario, go ahead. No, I keep going back and forth. I'm on.

Zach Peterson 

Have you gone to the next slide? OK, sorry. It looks like it's stuck on my side.

Caleb Buck 

it may be, I'm sorry. So do you see the slide that says scenario two now?

Zach Peterson 

no I don't.

Caleb Buck 

That's a bummer. Let me just describe it as well as I can while we wait. I apologize, I kind of live out in the sticks. So sometimes my internet isn't the fastest. So scenario two, instead of having the source and the load all connected to all the connection points, scenario two connects the source A and load A, but none of the other connection points.

So what this ends up doing is it will, is your screen updated yet? Okay, that's a bummer. You will pull the full 200 amps through the array, but it will be distributed across the components very, very unevenly. Could you by chance pull up that PDF that I sent you?

Zach Peterson 

It hasn't, no.

Zach Peterson 

Yeah, give me just a moment.

Caleb Buck

Yeah, I'll stop sharing because it is, I think it is very helpful to visualize. And for the people just listening, that's not going to be helpful at all. for those, you know.

Zach Peterson 

Yeah, no worries. Okay. So I can go ahead and share my screen. We can take a look here.

Zach Peterson 

Okay, so here's our current distribution and we were on scenario one and then you get the heat generation during scenario one. Now in this optimized arrangement, one thing I wanted to point out is I think this runs counter to some of the conventional wisdom. I think the conventional wisdom would be to spread everything out.

Caleb Buck 

Yep.

Zach Peterson 

and you would get a better result. But it sounds like you're saying you could spread it out or you could bring these load points closer and you'd probably get the same result.

Caleb Buck 

it depends. If you spread things out, any deviation in those individual paths, because as you spread them out, they become, they have the ability to become more unique from each other, any deviation in those paths will be a little bit more significant to the overall result. So that's why that example I explained before, where I chose to spread things out, because there was a factor

contributing the source impedance that I didn't have control of because I spread them out it had a dramatic effect on the the current distribution of that design if I had them you know I'll just say it smashed closer together the effect would have been less significant with them being closer together

It would still be imbalanced or poorly distributed, as I've said in the slides. But it really depends. If you can guarantee that the entire path from the source to the load is very consistent, essentially has the same characteristic impedance for each channel, yeah, I would prefer to spread the heat out, but...

That's not always been the case. Maybe your contact, maybe your contact, if you're using a contactor, maybe your contact resistance degrades over time and you start adding some resistance in one of the channels, which, you know, further causes more low distribution issues. So it always depends.

Zach Peterson 

you

Zach Peterson 

Yeah, and I think spreading things out would give you a chance to maybe have some more thermal mass nearby some of these FETs, right? Because you have material in between them. so that could also help with the heat distribution. Would that be a great way to think about it?

Caleb Buck 

Yeah, it could. And your load profile, think, is also super important. If this is a very short duration load profile, like it's very high power but for a very short duration, you're not really going to have time to fill that thermal mass with heat. If it is a continuous load for 30 minutes, an hour, four hours, longer, you're going to fully

Fully heat the board eventually so the load profile does does play a part. I still think spreading it out makes a lot of sense but in some cases It doesn't so it really depends Yeah, yeah, so here's go ahead

Zach Peterson 

Sure, sure, yeah, that makes sense. So, yeah, so you mentioned scenario two.

Caleb Buck 

Yeah, so scenario two, we only have one connection point. So same design, one connection point. So then advance to the next slide if you don't mind there. What's gonna result here, and like it says down at the bottom screen, these are not real or simulated values. These are just approximations based on things I have seen multiple times now. So this is the same load, same source, but the way that it is connected through the board.

is now causing an imbalance on the components themselves or a poorly distributed current through the components themselves. So I have on multiple occasions kind of debated this idea with people in the past that, if one path, let's call each of these FETs here a path. If one of these paths,

heats up more than the others that's gonna increase the resistance and then the current will end up choosing to go a different way because of that increased resistance. I've not seen that to be true. It makes sense from a scholastic standpoint, but in practice it chooses the shortest route to get to the lobe.

Zach Peterson 

Well, and also, I mean, yeah, sure, heating it up does increase the resistance, but I mean, does it increase the resistance that much to where, you know, current suddenly starts to reroute to the bottom FET? I mean, that's a simple calculation you should be able to do. And to just assume that it will equalize seems very foolish.

Caleb Buck 

No.

Caleb Buck 

Yeah, I mean, I've had that logic in the past. Before I'd seen this phenomena and seen the connectors melting, that felt reasonable to me. But now that I've seen it several times, I certainly feel differently. So be very careful if you have... The symmetry of the arrangement is super important as the takeaway here. So if you're sourcing your load or not coming at your components, assuming you have multiple in parallel,

If they're not coming at them, you know, symmetrically with respect to each other, you can have your load poorly distributed across your devices and you can have overheating on certain devices. So go ahead and advance to the next slide, the scenario three. Scenario three puts the source and load on opposing corners of the design. So go ahead to the next slide here. Let's see what happens here.

In this case, we still have the high current density areas in the corners just because we're turning a corner with that high current. But the path distance from the source to the load averages out to be approximately the same for each parallel path. So you actually get reasonable distribution of the current across the devices in this scenario. So this is the tactic. And a lot of my past work is with batteries.

frequently are not left with a lot of real estate or options for how you can connect directly to a battery. A lot of my work was for a battery OEM in the past. So you're pretty limited on where you can attach the source. So sometimes you do have to get a little creative and design the board so that, you know, if you have to parallel devices like this, you're ensuring that the

load distribution between the devices is even. I've used the PDN analyzer in Altium for this in the past. That's not what it's for. It's actually kind of challenging to get it to do that, but you can make it produce this information for you. You can also use simulation tools like Ansys and other similar things. Next slide, please.

Caleb Buck 

Yes, and then that's not to say that's the end of the story. Like I kind of said before, you need to know the junction temperature of the FETs. Just optimizing the arrangement of the parts to balance the load doesn't guarantee any kind of success. You need to know how much heat your board is generating. Or is the copper and the copper weight and the width sized appropriately for the max load that you expect and the load profile that you're expecting?

And super super critical. We haven't talked about a ton yet is The VIAs that are being used are they sized appropriately? Do they have an aspect ratio that will be reliable for the exact? Thermal stress that you intend on applying to them So there's there's a lot more to consider Does this does this better answer your question that you originally asked?

Zach Peterson 

Yeah, yeah, this is really, really useful because I think it's one of those things that, you know, if you don't do this type of design all the time, you don't really think about. So super valuable information and I guess just a little cross section of what folks can expect if they come and attend your seminar at PCB West.

Zach Peterson 

So sometimes, not all the time, but sometimes, high power designs generally mean high current, I think, in most people's minds. But they also can mean high voltage. And you brought up isolation requirements earlier. Tell us a little bit about how you deal with high voltage and high currents in the same design.

Caleb Buck 

Mm-hmm.

Caleb Buck 

Yeah, so high voltage, most of the high voltage that I've dealt with has been system level requirements that require some sort of high pot testing, dielectric strength test, systems where there are external connections.

I'll just keep referencing a battery situation because that's a lot of my experience. A battery on an aircraft, let's say, the external connections are going to get tested pretty abusively to make sure that the system is reliable. So there's probably going to be some amount of Hi-Pot done, let's say, on

on from a battery positive and battery negative to chassis or you know let's say it's some sort of other system with a 28 volt power input they're probably going to apply high pot somewhere on the design so let's let's just say it's 1500 volts for example that's what they're going to apply to it so you need to

appropriately apply design rules to keep things far enough apart on your design so that you don't have voltage breakdown locally in areas between traces and copper pores or mounting features and other copper features mounting features and parts. So it's really important to as a designer to get this information up front. I have been done with the design and found out

after the fact that, we forgot to tell you, we needed 1500 volt isolation from chassis to basically everything. That's painful if you're already space constrained. 1500 volt isolation, if you are using the old IPC 2221B formula, A6 is going to be the most conservative type for that formula. That is the external component lead termination un-coded.

Zach Peterson 

you

Caleb Buck 

The revision C is, I think that's called A7 if I believe. So if I set the calculator to solve for 1500 volts, it tells me I need 180 mils. Let's see.

Caleb Buck 

Yeah, A6, 1500 volts, 180 mils of isolation from chassis to whatever it is I'm trying to isolate it from. That's really significant spatially. If you're already space constrained, it really adds up. You end up, let's just say it's a quarter inch screw that we have mounting the board to something. You've got a quarter inch screw and then on each side of that you have 180 mils. You've got five eighths of an inch.

of a radius surrounding your screw that you just lost that you can't use for, I would say anything. I wouldn't put anything that close to chassis knowing that it's getting high potted at 1500 volts. So that's a very significant voltage issue. It's pretty easy to not get in the design quick enough or just miss. then you end up, I saw this here recently, needing to cut

Zach Peterson 

for sure.

Caleb Buck 

cut a copper pour on an outer layer because something is arcing when somebody didn't get that rule applied correctly, whether that was me or somebody before me. Did that get the question answered?

Zach Peterson 

Yeah, yeah, yeah. the spacing is really critical. And you brought up the new revision of IPC 2221. So that's something I'm going to have to dig into myself. Now, the other side of the coin when it comes to high power and high current is high heat. And IPC 2221 is one standard that

attempts to provide some guidance for designing conductors to ensure thermal management or to ensure you stay below some temperature limit. But then there's also the newer IPC 2152 standard.

And we've had one of the original architects, Mike Jopy, on the podcast a few years ago at this point. But he had a lot of really interesting things to say about those two standards. So I wanted to know from you, as someone who works on high power designs, what's your opinion on the differences between IPC 2221 and IPC 2152?

Caleb Buck 

Yeah, I've been using the 2152 formulas as long as I've been using Altium, which is over 10 years now. So I really can't speak to what 2221 used to say. I referenced this new revision C and revision C now points to 2152 from what I can tell.

Zach Peterson 

Okay.

Caleb Buck 

for determining the current carrying capacity of conductors. It just says basically go look at 2152. So I'm not sure what it used to say. That being said, I never used the 2221 formula myself. I've always started with 2152. And I think it works well for features that look like a trace. And I've seen the episode that you just referenced from three years back or so. Very interesting conversation.

Zach Peterson 

you

Caleb Buck 

When your conductor starts to look less like a trace, the predicted temperature rise becomes pretty inaccurate. And in my experience, it is very conservative. So I used to just go by the 2152 formula and I found that my traces were, you know, I'm talking, they're not really traces. They're, you know, many inches wide features.

Zach Peterson 

Yeah.

Caleb Buck 

that are connected to things on each end with pretty large heat sinks. So it's not really behaving like a trace in the middle of some sea of other traces. It's a really large, approaching the size of a bus bar with heat sinking on each end to a degree. So your thermal losses start to really change when the features get really large. And I have found that if you follow the 2152 formula,

you will end up with a lot of margin for safety from what I've seen. That 500 amp circuit board that I mentioned, now, well, let's say 150. A lot of those 150 amp circuit boards was just a couple layers of two ounce copper over a couple inches of conductor width. And I like to see temperature rises not exceed 20 C. Sometimes they go a little more, but.

Zach Peterson 

Yeah.

Caleb Buck 

I've thermally characterized boards, you know, just on the bench and compared it to what the calculator says. And in many cases with the very large copper features, they do not heat up as much as the calculator says. So in that situation, I really do think, and if you really need to know how hot the board gets, I really do think a, you know, CFD type simulation really is critical to understanding.

How hot the board is going to get. If you're really pushing how small can we make this, will this actually fit in our budget envelope for space and how much heat are we going to generate? If that's an important question that needs to be answered at the system level to ensure that the system functions correctly, simulation I think is really important. And I've found, you know, I'm just going to say answers because that's what I've had experience with. I built a board that was 14 by 14, carried 180 amps.

I built it, got it on my bench, tested it in a sealed box, measured it with thermocouples. Then I had an ANSYS team help me simulate it, and we got within one or two degrees of the real world results in a simulation. I'm sure there's other simulation tools you can use to get to the same result, but if thermals are that important, I think it's very valuable to simulate something like that.

Zach Peterson 

Yeah, and you brought up the conservativism of some of the calculations, I think, that come out of the older IPC 2221. I mean, the results are pretty crazy when you start to look at larger pieces of copper that, like you say, aren't really traces anymore. They're basically like copper pores that you're using as a current carrying conductor. And some of the numbers that you pull out of that calculator are just.

Pretty unrealistic if you sit there and think about it for a second. So yeah, I agree with the IPC 2152 is the way to go. And it's interesting now that the 2221C standard is pointing to the 2152 standard. So I guess we're finally making some progress in that area.

Caleb Buck 

Yeah, I don't remember what section it is off the top of my head, but I was looking at it the other day and it references 2152 now.

Zach Peterson 

Yeah, yeah. So we're getting up here on time, but I think we got time for just a couple more questions. for high power devices, we've talked a little bit about the stack up, especially in terms of like, you know, copper weight. But

Material types, specifically for the dielectrics, are also important for high power. So if you could give us some guidance on what to do when selecting materials for these designs, especially when you have to handle long-term reliability.

Caleb Buck 

Yeah, so the easiest low-hanging fruit, you know, kind of metric that I like to make sure a high-power design has is at least some kind of elevated glass transition temperature. So the T sub G, I don't want anything that's, you know, 130 or 150 T sub G. The conductor temperatures, you know, can get alarmingly close to that. So definitely don't want something with very low glass transition temperature. You don't want your

board itself to get close to that trend that temperature where you know the the resin starts to flow again. Next the z-axis expansion the z-axis coefficient of thermal expansion is also really important to consider in relation with what your aspect ratio is on your vias and the total board thickness so

the higher your aspect ratio and the thicker your board, the deeper your via is that you're trying to plate, it gets more difficult to get that plating a consistent thickness all the way through. So if you end up with the plating, you know, a little thinner in the center of the via versus the outside where the plating is coming from, that Z axis expansion is gonna, you know, over time, kind of expand, contract, expand, contract.

And that thinner plating wall can become fatigued and break. So a material with a low coefficient of thermal expansion in the z-axis is pretty critical if you expect to be thermally stressing the vias.

Zach Peterson 

That would be like with repeated thermal expansion and contraction, a lot of thermal cycling.

Caleb Buck 

Yeah, exactly. And then I like to also pay attention to the surface resistivity of the material as well. it'll help maintain that voltage isolation level. If you have a material that naturally has a lower surface resistivity, you're going to be slightly more prone to voltage breakdown earlier on. That kind of segues into.

My other area of expertise, is cleanliness, if the board's not clean enough, you can have residues that develop of voltage breakdowns. So that kind of is beyond materials. But that would be my guidance is seeking those low CTEs, the Z-axis CTE, find something with a really low number there. Because it's.

You're aware, but maybe somebody listen is the X and Y axis has the glass weave So it doesn't have the same amount of movement in X and Y But in the Z axis the direction the weave does not go it has quite a bit more movement

Zach Peterson 

Yeah, that's bringing it back, I guess, full circle. You brought up cleanliness. One material that I think some designers might look at to help ensure cleanliness might be a conformal coating. Would that be something that would be recommended in high-power designs or no?

Caleb Buck 

Yeah, I would say that conformal coating doesn't guarantee cleanliness though. The cleaning process is where the board gets cleaned. The conformal coating helps ensure that it kind of stays in that state as long as possible. Conformal coating is not a permanent moisture barrier. It is moisture permeable. So it will help delay the ingress of moisture onto the surface or under parts.

Zach Peterson 

Sure.

Caleb Buck 

very small components will tend to see the conformal coating as insulation if the conformal coating is very thick. So if you have a very small device that you know is going to, you're going to be operating it with pretty high heat generation, conformal coating over it can act as insulation to that device.

It kind of depends on the insulating properties of the exact conformal coating you're using. I've also seen it the other way around. I've seen fuses take longer to blow with very heavy conformal coating. I'm not saying apply heavy conformal coating. There's a lot of failure modes that result from conformal coating being too heavy. But I've seen fuses not blow at the correct time because they were coated and the coating changed their thermal behavior. They didn't get hot enough, fast enough.

Zach Peterson 

Interesting. Right, right. Yeah. Interesting. Yeah. You just added some thermal mass around the fuse and so it's pulling heat out of the fuse and then it affects circuit protection. Interesting. I would have not...

Caleb Buck 

These are like small surface mount fuses.

Caleb Buck 

Yeah.

And the neighboring copper features as well. Don't pour copper under devices that are supposed to behave thermally unless the manufacturer says, do it this way. They've designed it to operate at a certain point. Yeah.

Zach Peterson 

Right.

Zach Peterson 

Yeah, makes total sense. Well, like I said, we're getting up here on time, but this is just a cross section of the really extremely valuable information that folks are going to get if they attend your presentation at PCB West. So I would like to invite all of the folks out there listening to come and see Caleb Buck at PCB West. And of course, we have a lot of other great folks at PCB West. have Rick Hartley, have Susie Webb, Steven Chavez.

We had Kevin Coates on the podcast a few days ago. He'll be talking there as well. So tons of great people. Just a quick wrap up, Caleb, who are you looking forward to seeing at PCBWEST?

Caleb Buck 

I the names you mentioned are you know, some of the some of my favorites, One of my colleagues Troy Hopkins is speaking this year. I haven't seen him speak before I've seen you speak. I saw you speak last year I believe and was A lot your physics knowledge. It was it's a it's a lot to digest for for people that haven't been and it's It is it is career changing every time I go I leave

Zach Peterson 

You

Caleb Buck 

feeling more informed than I was when I came. It's a phenomenal opportunity for a PCP designer.

Zach Peterson

Absolutely, I would totally agree. Yeah, career changing is a great way to describe PCB West and PCB East. Yeah, Caleb, thank you so much for being here today. I'm excited to come see you at PCB West. We're going to be filming the day of the show and maybe we'll catch you on the show floor and we can catch up there.

Caleb Buck 

Yeah, look forward to seeing you.

Zach Peterson 

Absolutely. To everyone that's out there listening today, we've been talking with Caleb Buck. He will be speaking at this year's PCB West Conference. If you're watching on YouTube, make sure to hit the like button, hit the subscribe button. You'll be able to keep up with all of our podcast episodes and tutorials as they come out. And last but not least, don't stop learning, stay on track, and we'll see you next time. Thanks, everybody.