Thermal Management in PCB Today

Clearing up the confusion about the IPC 2152 with Mike Jouppi, the “Thermal Man''. Mike originally sat on one of the IPC task groups working on standards for thermal management. He will help us be enlightened about thermal management on PCB, which will be very useful to apply to your next PCB design project.

Listen to the Podcast:

Download this episode (right click and save)

Watch the video:

Show Highlights:

• Storytime! Mike's background in PCB design and involvement in IPC task groups
  • He worked in the Airforce as Missile Maintenance Specialist and later had an internship at Hughes Aircraft in Tucson, Arizona, where he did a thermal analysis
  • The TQM mindset—total quality management is what kept him moving forward
  • What does it mean when “Power was Negligible”; this led to Mike's first encounter with the IPC document
• How to take a vacuum environment into consideration? This is the question that leads Mike to start research and develop testing methods
  • Mikes received funding from Lockheed to do internal research and perform testing to understand trace heating
  • Mike shares his journey to seek funding for testing and how he ended up setting up his lab in his basement
  • He developed 68 different charts, 13 of those were raw data, and the remaining ones were all analytical creations with the thermal model
• Mike started working with IPC in 1999, IPC 2152 came out in 2009
• Mike wrote a chapter in Happy Holden’s Printed Circuits Handbook
• Clearing up confusion over what IPC 2221 and IPC 2152 represent, Mike emphasizes using the documentation as a baseline
• Another storytime! How did Mike come up with the name “Thermal Man”? The story involved measuring the thermal properties of an egg. Visit thermalman.com to check out Mike’s website
• More research was performed involving flex, heating vias, and microvias
• Discovery of the original data to the charts that are in IPC 2221 documented in the National Bureau of Standards report from 1955ish
• Mike has the data to prove that copper planes and copper pour affects thermal management
• The major players in terms of temperature response of the trace:
  • Influence of the copper planes
  • Mounting configuration
• Before the IPC took over the industry standards in electronic design, the National Bureau of Standards (the government) handled the management and documentations
• Check out more videos and courses about Thermal Management on Altium Academy Youtube Channel
   

Links and Resources:

Connect with Mike Jouppi on LinkedIn
Visit Mike Jouppi’s Electronics Thermal Management LLC website
OnTrack Episode with Isvan Novak: DesignCon 2020’s ‘Engineer of the Year’ Talks Power Integrity, Picosatellites, and Simulation Tools
Read Happy Holden’s Article on Altium Resource Hub
IPC Website
IPC 2221 and IPC 2152

Connect with Zach on LinkedIn
Full OnTrack Podcast Library
Altium Website
Download your Altium Designer Free Trial
Learn More about Altium Nexar

Claim the special offer for Podcast listeners only

 

Transcript:

Zach Peterson:
I was just wondering that. How are you going to mount it in vacuum and expect that it has any reflection at all to how a board will actually be mounted in a real system?

Mike Jouppi:
Yeah, well, that was the problem with all of this testing is that none of it seemed to simulate any real boards in any practice, and it just drove me crazy. It made no sense.

Zach Peterson:
Hello, everyone, and welcome to the OnTrack Podcast. I am Zach Peterson, your host. And today I'm going to be talking with Mike Jouppi. He originally sat on one of the IPC task groups working on standards for thermal management. And we're going to be talking about thermal management and PCVs today. It should be an illuminating this discussion, so let's get started.

Zach Peterson:
Hey Mike, thank you so much for joining me. And I'm really eager to hear what you have to say, especially since you originally sat on one of the IPC task groups for this important topic.

Mike Jouppi:
Thanks, Zach. Good morning.

Zach Peterson:
Good morning. I think before we get started, I always like to ask folks, maybe tell me a little bit how you got involved in PCV design and your background that led you to work with the IPC.

Mike Jouppi:
Yeah. Well, I told you earlier, I got to be careful with story time.

Zach Peterson:
Whatever you feel comfortable sharing. I'll just say that.

Mike Jouppi:
Well, out of high school, I enlisted in the air force and I was a missile maintenance specialist on ICBMs. I met my wife in the air force and we moved to Tucson, Arizona. After the air force, I went to school. I was able to get an internship with Hughes Aircraft after my junior year at the University of Arizona. And they started me doing thermal analysis on electronics. There were two options for me at the time. There were two jobs that they were looking for. One was to go catch TOW missiles in the Yuma desert, and one was to do thermal analysis. And fortunately, it was thermal analysis.

Zach Peterson:
It's funny how that works. My background in lasers, they gave me two options. It was electrochemical stuff or lasers and so I picked the lasers.

Mike Jouppi:
Those forks in the road dictate a lot. And yeah, also, I just liked doing projects. And when things turn up and work needs to be done, I've got a mindset that's TQM, the total quality management mindset. I like the constant improvement and the feedback. And so my wife and I moved to Colorado in 1996 and I was working for Lockheed Martin, and I was working on a spacecraft called Stardust. And I was working on the electronics that was bringing the solar ray power in and charging the batteries on the spacecraft as well as powering the electronics. And it was on that electronics that I was doing the thermal analysis where an electrical engineer brought me power dissipation for the traces. And it was 10 watts of power. Well, no, it was a little over eight watts of power. And we had 24 watts of power just from the components and that additional power from the traces changed the design.

Zach Peterson:
Now this is power dissipated as heat, not just the power being carried.

Mike Jouppi:
Yeah, it's the dual heating in the electrical traces. It's watts of power and the electrical engineer... So this is 1997-ish. And I started doing thermal analysis on that internship in 1982. Now we're in 1997, and this is the first time that anyone had given me power for the traces before. And I had been asked by electrical engineers in previous years when I first started about the power losses in the traces, and I talked to the people that were mentoring me back then, and they said that the power was negligible.

Mike Jouppi:
Well, here we are in 1997 and they were not negligible at all. And it changed the design. So I asked the electrical engineer, his name was Rick Fern, where he got these powers and he showed me this IPC document that was just being converted from a MIL-STD document. I think it was MIL-STD 275. And it was turning into IPCD 275, which later turned into IPC 2221, which most people now are more familiar with. And in that document it showed these nomographs for sizing electrical traces. And he said he picked those and then he calculated the power based on the size of the traces. And in that document, it also said that if you were working in a space environment that you should take it into consideration, but it didn't tell you how.

Mike Jouppi:
So here I am, I'm looking at this and I don't know what to do. So I called IPC and I got in touch with John Perry and John had just started with IPC and I asked him where these charts came from and how to take a vacuum environment into consideration. And he said he didn't know, and he didn't know where the data came from. And so he went off to do some research and he came back to me and he said that he thought the people at IPC thought that the original information came from an IPC technical paper, IPCTM 117. And it was written by a Dr. Jennings at Sandia Labs and it turned out that wasn't the right paper, and it wasn't right because he had done all his testing for 30 sec... He'd power up a trace for 30 seconds, and that's not enough time to reach steady state. And so it just wasn't the right answer.

Mike Jouppi:
And so I was working at Lockheed and I requested some internal research and development money to go off and do some and testing to understand more about the trace heating. And I got funding for it. And I went through a year of testing and we tested in air and in vacuum and I submitted that paper to IPC. And right when I was getting ready to go out for the conference, I found a document that told me how to do the testing. And that's an IPC document also, and their test methods, and it's IPC test document 2.5.4.1a. And in there it showed a totally different board configuration and I was freaking out.

Mike Jouppi:
Here I am going to a conference, talk about my results, and then I find a whole different setup. And so I went to the conference, I presented, and at the time there's a guy by the name of Ralph Hersey who was heading up the task group, the IPC task group for these nomographs. And he was saying that my test results showed that it validated the existing test data and the charts that were in the old MIL-STDs in this IPC document. And I raised my hand and I said, no, I don't think it does. These internal traces are not similar at all. And so I left and I went back and I asked for some more money from our internal research and development group. And they funded me again.

Mike Jouppi:
And I built boards just like this test method. And so we did more testing. We tested in air and in vacuum again, and they were a little bit different. In this test method... Well, I'm going to be doing another presentation and I'll go into more detail about those kind of things. So long story short, I did more testing and finished that up and wrote another paper and presented that, and that turned into... It was much better results that followed the test methods. And again, the internal trace temperatures were very different than what were in the existing standard. And as a result of that, I ended up volunteering to lead this new task group and start doing more work on it. So I went back to Lockheed and I asked for more money And they turned me down.

Zach Peterson:
It's a life of a researcher, isn't it?

Mike Jouppi:
Oh yeah. There was so much more work to do because the testing we were doing, it didn't have any copper planes in the board. And the copper planes I knew would have a significant effect. And we also weren't mounting the... The test method has you suspending the board in air or in vacuum and all of my boards were mounted in these aluminum chassis and with wed locks and I wanted to work all of that in.

Zach Peterson:
Well, and I was just wondering that. How are you going to mount it in vacuum and expect that it has any reflection at all to how a board will actually be mounted in a real system?

Mike Jouppi:
Yeah, well, that was the problem with all of this testing is that none of it seemed to simulate any real boards in any practice. And it just drove me crazy. It made no sense to go that way. And I just kept doing more work. So Lockheed didn't want to continue the work. So I quit. I called the University of Colorado in Denver, talked to the mechanical engineering department. I talked to the head and I told him that I wanted to continue to do research on this whole trace heating thing. And he offered me a position. I created a class and then we teamed up. There were three other professors, and we wrote a national science foundation grant proposal to fund continuing to do the work. So it took us about six months to write that proposal.

Mike Jouppi:
And we even went out to DC, talked to the National Science Foundation after submitting it, they turned it down, they didn't fund it. They said the work had already been done, but obviously they didn't read our proposal because we explained what the state of the art was and they didn't get it. The excuse I heard was it just wasn't cool enough for them to fund it. So they offered me a PhD position to go and continue my education.

Mike Jouppi:
I had two kids and I couldn't afford to live on a stipend and keep doing the research at the university. So I bought all of the equipment that we had set up for the lab. And I set up a lab in my basement. I hired two of the student engineers that I was working with. And I started my own little company and we started collecting all the data. The cool thing was when I went over to the university, Lockheed donated all of the equipment that we had, all the boards, to the university. And there were a couple other companies, another Lockheed facility down in Texas that built some other FR-4 boards for us. And we had all of this equipment ready to do all the testing at the university. So I purchased all of that from the university, set up the lab and continued to do the testing.

Zach Peterson:
So, real quick, the boards that you're testing, I'm assuming that you're going off on your own in terms of the testing method, since you've essentially discovered that the testing method doesn't really validate anything that is happening in a real system, and then varying all the parameters in the board. So, the thickness, how big is the board, where's the copper located, is there any copper? Things like that.

Mike Jouppi:
So the real catch with this kind of testing is that it takes a long time to go through and power up each trace, run it to steady state at each amp setting and collect enough data to be able to put these charts together. And so the process that I developed was to collect real data on these boards, so FR-4 boards and polyimide boards, two different thicknesses of FR-4 and one thickness of polyimide. And what we did was the whole concept was to collect the data, real data with controlled environment and known variables, and then use that to create a thermal model and correlate the thermal model to the data, and then use the thermal model to then start adding variables and change variables.

Mike Jouppi:
So I was then able to start adding copper planes. I never got to the point of adding wedge locks to it, but that was in the game plan. We ended up developing about 68 different charts. 13 of those were raw data, and then the remaining ones were all analytical creations with the thermal models. And it's just really cool to see the results from each one of them because as we varied the board thickness, as you get thinner, the temperatures go up, as make your board smaller, the temperature goes up.

Zach Peterson:
And that's just because when you make the board smaller and you make it thinner, there's just less thermal mass in the substrate to accept all that heat that you're generating from the trace?

Mike Jouppi:
Well, not so much just accepting it, but how it's conducting through it. You're minimizing the heat transfer path, and then you have less surface area for it to... Because suspended in air or suspended in vacuum, suspended in air, you only have natural... Way we test is in still air. So it's just natural convection from the surfaces to the air and as well as radiation losses and for low delta T, the radiation losses are almost negligible, but when you start getting up into a hundred degrees C delta T's, then the radiation is real noticeable. You can see variations there. And so we started collecting all this data and then I reached out to a software company and I was able to talk them into teaming with me. And we created a software tool to use to take all this data in.

Mike Jouppi:
So this is in the timeframe of 2002 to 2004. And so we developed a software program, had it for sale online. We had a version that you could just download and use. And we had a server version where multiple people could use it in a corporation setting. It was really cool. And I liked it a lot, but I just couldn't get it going in terms of a tool. And I had a small window of opportunity to go back to Lockheed to where I could pick up a pension again. And so my window of opportunity as an entrepreneur ended in 2004. We didn't have enough sales with the software, so I had to shut everything down and put it away. And then I just went back to work at Lockheed as a thermal analyst. And I continued to work with the IPC task group to create a new document and try to get some of the data into that document as well as-

Zach Peterson:
So that was the IPC 2152 standard?

Mike Jouppi:
Yeah. So we hadn't created it yet. That came out in 2009. So, I did all this work up until 2004 then I had to archive it all. I couldn't run a business and work at Lockheed as an engineer. You got to do what or the other. And so I had to put it away, but I kept the intellectual property and maintained that, which was a mixed blessing there. Kind of prevented me from sharing a lot of the research that I'd done with the people around me at Lockheed. So that was a bummer, but I continued to work with IPC and keep in mind that all the work with IPC is as a volunteer. So you're just giving your time and giving everything to them. And so we continued on the path with IPC 2152. I started working with IPC in 1999. I sat down with this guy Dieter Bergman, and we created a 10 year plan. And it was about 2004 or 2005. Do you know Happy Holden, Hap Holden, does that ring a bell for you?

Zach Peterson:
Yeah. Yeah, [inaudible 00:18:33].

Mike Jouppi:
So Hap called me up and he said, how about writing a chapter for the Printed Circuits Handbook? Are you familiar with that?

Zach Peterson:
I've heard of it. Yeah. I've never read it.

Mike Jouppi:
Yeah, well...

Zach Peterson:
Well, that's-

Mike Jouppi:
That's why.

Zach Peterson:
... the reason why.

Mike Jouppi:
Yeah. So I wrote a chapter in there and I thought that would be a good way to get a foundation for what we could put into this new document, 2152. And so I wrote that chapter and then I brought that to the task group. And then as a group, which nice way to write, by committee. We just broke it all apart and then assembled it into what became my IPC 2152, which then got published in 2009. And the reason that I contacted you and wanted to have this discussion is because I was going out on the net and I was seeing a lot of confusion over what IPC 2221 and 2152 really represent. People were just confused about it. They just don't really get it.

Zach Peterson:
Well, it confusion over how to use the documentation in those standards or is it confusion over what physically that documentation is supposed to reflect?

Mike Jouppi:
I think it's both.

Zach Peterson:
Or is it both?

Mike Jouppi:
I think it's both.

Zach Peterson:
That's what I thought too.

Mike Jouppi:
Yeah. If you go-

Zach Peterson:
Yeah, well, because sometimes I'll see questions on forums and someone will ask, how do I calculate the maximum current I should put into this trace or into this plane or something like this. And there are some empirical equations and then there's also the standards.

Mike Jouppi:
Yeah. And back to those equations, we talked about this the other day, and I don't want to forget to bring this up, all the equations for the charts in IPC 2152 are in a couple papers that I wrote and I'm going to put those on my website and it's easy to find [inaudible 00:20:48].

Zach Peterson:
We'll link to that in the show notes.

Mike Jouppi:
It's thermalman.com.

Zach Peterson:
Okay. thermalman.com.

Mike Jouppi:
Yeah. And all one word, thermal man, which is a nickname I picked up... I was working in Alabama on the space station. And this is definitely one of those tangents I'm, but it's a good story. I want to share it.

Zach Peterson:
All right.

Mike Jouppi:
I was driving to work one day and I was listening to NPR and it was George Bush, H.W., was president at the time and he asked professionals to go out and start getting professionals to work with students to help them get going and get motivated for school and other things. And so I took that to heart and I went to the local high school and I recruited a dozen students to work with me on a project. And what we were going to do is measure the thermal properties of an egg.

Mike Jouppi:
And I went to the University of Alabama, talked to the mechanical engineering department. They gave me keys to their heat transfer lab and said you can use all the equipment. So we would go in on Friday afternoons after they got out of school and Saturday mornings, we mounted thermal couples in eggs, hard boiled them, collected the time temperature data, created a thermal model of the egg, correlated the model and wrote a technical paper and presented it at a conference in Huntsville.

Mike Jouppi:
The cool thing was, first off, all those students had papers graduating high school that were published. And there was a guy in the audience that had just come back from the Galapagos Islands and they were trying to recover some rare turtle eggs, and the eggs spoiled on way back transporting them. And he asked if we could possibly use this thermal model to help them predict the temperatures and control the environment for these eggs. And I said, well, yeah, this would work cool for that. And so he suggested writing to the National Science Foundation to go and do some research like that, but I never did it. So if anybody ever watches this that knows that if you need some help on turtle eggs, call me.

Mike Jouppi:
All right. So back on topic. So continue to do a bunch of research on the trace heating and in the task group, we had a list of these things that we wanted to characterize. My view toward IPC 2152 is that it's a foundation for building a thermal model. It's all about computer modeling, all the information's there that gives you very detailed parameters on all the variables. The only thing that I didn't include was I included the [inaudible 00:23:51] of the surface, but I didn't include the convective environment, but it's just a natural convective environment, any standard heat transfer book has it. And so it's really easy. All the information's in that document to model. And then from there, you can add things to that model to help you predict what the real temperatures are.

Mike Jouppi:
And that was the path I took. So as a part of the task group, the things that we wanted to look at were not just conductors and rigid boards, but we wanted to look at flex. We wanted to look at the heating and vias and microvias, the neck down areas, we wanted to look at high current. So one of the troubles with the work I was doing was limited to about 25 amps with my power supply. It would've been nice to go higher and bigger conductors... Just to let people know, I got some people to do other testing. There was a classified lab in Minnesota that did some work. They were into real high speed and they were running into issues. They repeated the same testing with the same kind of boards that I was doing. They compared real well. What was interesting is that we could tell a small difference between the testing I was doing, which is at 5,000 feet here in Denver versus roughly sea level in Minnesota, where these, so other guys were doing their testing. And that-

Zach Peterson:
I could imagine that's important if you're in an aircraft and you're at 30,000 feet.

Mike Jouppi:
Yeah. And we got some work done at Nav C Crane also where we did some work in an altitude chamber and we were able to collect data there. And yeah, altitude... So, we collected all of that and it's hard to compress all of that and put it into a single document. And the document-

Zach Peterson:
Well-

Mike Jouppi:
Go ahead.

Zach Peterson:
... And I was going to bring that up because if you actually look at the standard nomograph that's in IPC 2152, what do you see? You see copper weight, you see preferred temperature rise and then conductor cross section. And that's essentially it. But there's so many more variables. And I think at some point it becomes really difficult to compile all of that into one standard because isn't the goal of the standard to just make it really easy for a designer to get the information they need and move on and do what they do best, which is design?

Mike Jouppi:
Yeah. But I think the hard part was that we couldn't just test all these different configurations. And it just took too long. It was just not possible. It just wasn't feasible. And that's where the concept of creating the thermal model and the standard it gives you all the very, very, very good detail about the variables that would impact a very simple model suspended in air. And now it's up to other users to expand on that.

Zach Peterson:
So, I take it that eventually what you find is that the specifications in that standard don't give you something that's actually accurate.

Mike Jouppi:
Well, it's very accurate for what it represents.

Zach Peterson:
Right, for that one case, but if you expand that out and try and extrapolate to a general case for any other board...

Mike Jouppi:
You're going to see a significant-

Zach Peterson:
I can immediately see that the results aren't accurate.

Mike Jouppi:
Well, no, because you're not comparing apples to apples.

Zach Peterson:
Okay. Yeah.

Mike Jouppi:
Yeah. In fact, as a part of all of this work, I ended up finding the original data to the charts that are in IPC 2221. And that was all documented in the National Bureau of Standards report from 1955ish. And those guys, back then, they only had two sided boards and they tested boards of different thickness and they tested different copper weights. Some of the boards had copper planes on the back of them. They had varying board thicknesses, copper planes, trace thickness. Oh, and the materials were phenolic and epoxy.

Mike Jouppi:
And then they put all of those data points on these charts and back then they didn't have things like Excel. So they took a French curve and drew a best line through all those data points. And so the external trace chart and IPC 2221 represents that data. And so when you look at the external data from IPC 2221, and you compare the external data from IPC 2152, you'll see that IPC 2152 is a little more conservative because it's not influenced like copper planes and all over these other data points.

Zach Peterson:
Sure. So when you have the copper plane on the backside of the trace, obviously copper is a heat conductor, it's going to take whatever heat does dissipate into the substrate, spread it around. And then ultimately give you that more even temperature distribution.

Mike Jouppi:
Yep. Yeah. [inaudible 00:29:25]

Zach Peterson:
This is kind of funny because there are certain more experienced designers than myself who are well known, who have said that things like copper planes and copper pour do absolutely nothing for thermal management. And you're saying that they're wrong and you have the data to prove it.

Mike Jouppi:
Big time.

Zach Peterson:
Big time.

Mike Jouppi:
Well, copper is a thousand times more thermally conducted than the dielectric material.

Zach Peterson:
Well, of course.

Mike Jouppi:
How can it not impact the trace temperature [inaudible 00:30:00]?

Zach Peterson:
Well? That's what makes me wonder why somebody would say that whether it's copper pour or copper plane, whatever it may be, that it doesn't do anything for thermal management. I'm wondering if now they're saying, well, it's this one specific type of board or you have to put an asterisk next to that. It's only when the board is small, whereas you're talking more in a plurality of situations and in general it should.

Mike Jouppi:
Yeah. Whenever I was designing part of the design teams for any circuit board design, I would ask the electrical engineers to give me power dissipations for the traces, and I would lay that out on the board. I'd have all the power for my components. I'd lay that out on the board. I do preliminary analysis, just placement by putting surface heat on the boards with the trace heating as well as the component heating.

Mike Jouppi:
And if things are too hot I just add copper planes [inaudible 00:31:03] thicken the copper planes that are near the sources and that's the simplest way to bring the temperatures down and then tie that copper to my main sink. So I'd run it out to my wedge locks where I'm bolted into an aluminum chassis and I'd dump all the energy out there. And so yeah, and then thermal vias, tie those copper planes together and get that... All you're doing is trying to create a heat transfer path around that dielectric material and get the energy out of the board.

Zach Peterson:
Makes perfect sense. And the planes that you're taking advantage of for heat sinking, you're tying them directly into the chassis. So that makes perfect sense.

Mike Jouppi:
Yeah. And we ran into some issues with that, and I've run some studies where a five mill dielectric spacing between a plane and another plane, you almost don't see much of a gradient if you have just a five mill spacing. If you have a quarter inch spacing, it's almost like being completely isolated, but you bring those planes together. So if there was ever a problem with the planes, if you had power ground and you didn't want to dump to chassis, you just kept the dielectric spacing very narrow between that heat transfer path and you're still dumping a lot of heat out. So yeah, run my copper planes out and then thermal vias through the wedge locks and that would be my standard process.

Zach Peterson:
Yeah. And I think now I might be seeing where the copper pour versus a plane could provide a difference in heat transfer characteristics, because if you want to take advantage of say copper pour as a thermal management element, you would then have to tie it back to the plane essentially everywhere with stitching vias.

Mike Jouppi:
Yeah. Well, it depends on what the pores are like. I've seen these 10th inch square patterns, 100 mills by 100 mills little squares cover a whole surface. Those aren't doing a lot for you. It sure is going to make a difference, but it's the filled copper areas that make the big difference. And I've run studies running traces over areas with and without copper and well, you see a big difference where it's close and where it's not and the temperature going up where you don't have the copper and you can see the gradient. They become very apparent and pretty obvious that coppers doing the job for you.

Zach Peterson:
Yeah, I think intuitively that makes sense. And then also you mentioned the dielectric spacing, but then also the dielectric properties, like the resin content should also have an influence.

Mike Jouppi:
Yeah. Well, that's one of the things that when we've started doing this work with the trace testing, I was able to send board samples for all of our boards to a company that measured the thermal conductivity and the X, Y, and Z plane, as well as measure the specific heat and density for us, which are important for the transient side, time temperature response. And what was interesting is that in most of the literature for dielectric materials, they only give you the Z access thermal conductivity, the through plane, which is really the epoxy and in the X, Y plane, where you've got fibers running through, you actually see an impact from the fibers.

Mike Jouppi:
And there's an increase in the thermal conductivity in the X, Y plane that's probably double what it is the Z axis. And that was a big deal for us for correlating the models to the data. We wanted to have the best information on all the variables that we were using in the model to get a correlation. With any kind of modeling, garbage in, garbage out.

Mike Jouppi:
And the only way to get a good correlation was having known properties for everything that we were running in the model. And the only thing that we didn't measure was the [inaudible 00:35:35] of the surface, but I've got a lot of experience with that, with other work I've done. And so it all panned out pretty well.

Zach Peterson:
Yeah, think what I'm really hearing here is that you had used the term conservative to describe the estimates from IPC 2152. So they're essentially telling you to overestimate the amount or the width of copper that you would need to limit a temperature rise. And so for your particular board, you really do have to do quite a bit of testing just to ensure that you are going to be below your temperature limits.

Zach Peterson:
And I'm sure this is all also going to depend on factors like let's say you have a large IC. It's got multiple power rails coming to it. Those power rails could also be heat dissipators. And so there are so many different variables that it's almost impossible to... Not impossible, but it's an intractable problem to really compile all of these variables into one standard.

Mike Jouppi:
Yeah, it is next to impossible. But what I found is some of the major players in terms of the temperature response of the trace were the influence of the copper planes. And then from other work I've done, your mounting configuration, whether it's wedge locks or bolted interfaces, those are huge also. So I designed a lot of six by eight boards with wedge locks. Well, I could have created a standard for my... And 18 to 24 layer boards a two ounce copper. So that wouldn't do anybody [crosstalk 00:37:16]

Zach Peterson:
That's a lot of copper.

Mike Jouppi:
It's a boatload of copper. We had high power and our applications you can't go fix them when they're heading to Jupiter. And so, yeah, do I create a standard just for me? Well, that didn't make any sense. And how do I assume what everybody else is doing? Can't do that either. So I'm not the task group chairman anymore. I left. If you want to go help those guys, I could set you up.

Zach Peterson:
Well, I'm already involved in another organization. And I had to step down from one standards group because I don't have time to do it anymore.

Mike Jouppi:
I know, right.

Zach Peterson:
Yeah. God. Heaven bless those guys, those people I should say, who have the time and the willingness to participate in these groups because I know they're doing a lot for the industry.

Mike Jouppi:
Yeah. And that's why I wanted to come out and talk about these standards a little bit to help clear the air a little bit about what they are and what they represent. You can't expect it to meet the need for everybody, but you don't want to cause someone to fail either. And so-

Zach Peterson:
Yeah. And so the 2152 standard, was that originally developed with the intention of providing a conservative estimate? Or is that just kind of how it ended up?

Mike Jouppi:
Well, it's kind of how it ended up, but it was following what had been done before.

Zach Peterson:
Okay. So no one really thought that it was broken. They just did what was in the test method.

Mike Jouppi:
We followed the test method, which was the right thing to do and I tried to expand that by using 2152. I call 2152 a baseline. Just a baseline. And then from your baseline, you develop from there.

Zach Peterson:
Sure. Yeah.

Mike Jouppi:
And the hardest thing is getting a correlated model because for the same cross sectional area, you get a different delta T if it's half ounce, one ounce, two ounce or three ounce or four ounce copper, same cross section area. Different [crosstalk 00:39:53].

Zach Peterson:
Yeah. No, that makes perfect sense.

Mike Jouppi:
Oh, well that's not obvious to everybody though.

Mike Jouppi:
Well, okay. That's fair. Yeah. Same cross sectional area, but different weight.

Mike Jouppi:
Right. Yeah. And it's all about the perimeter and the exposed surface area of that conductor. As you get thicker, the perimeter gets smaller and you're not dumping as much energy and if [inaudible 00:40:23] flattening it out, you have more surface area, bigger perimeter. And you see that in the data. So that was really cool. We got to see a lot of those things. And that's what became important to replace the charts in 2221, because that was just this hodgepodge of a bunch of data. And then when we went to multi-layer boards, instead of running more tests and collecting internal trace data, all they did was go to halving the current. And so if you look at the equations that describe the charts in 2221, you'll see a 0.5 factor on the beginning of the equation. And it's just half the current from the external conductors.

Zach Peterson:
I think the justification for that is then that the exposed surface on the surface layer, that top surface, you're just regarding it as being a perfect insulator. And so then once you go into the internal trace, it's essentially having double the conductive area to remove heat.

Mike Jouppi:
Well, they were saying-

Zach Peterson:
Is that the thought process behind it?

Mike Jouppi:
They were saying it was going to be hotter internal.

Zach Peterson:
Oh, I see. Okay.

Mike Jouppi:
So they were saying you could have only half the current.

Zach Peterson:
Yes, I see. I see. Okay. They got it backwards.

Mike Jouppi:
And yeah, it doesn't work that way. And it's real apparent when you look at the data that we've collected for 2152. And it makes sense also because the thermal conductivity of the dielectric material is still better than the thermal conductivity of air and a natural conductive environment. Now, if you start blowing air over the traces or you create some other heat transfer path off the external then the story changes.

Zach Peterson:
Sure, sure. That makes sense.

Mike Jouppi:
So the stand-

Zach Peterson:
So what... Oh, sorry, go ahead.

Mike Jouppi:
Go ahead.

Zach Peterson:
Well, I was going to say, given that I think anybody that thinks about this long enough will realize that the data in that standard is not universally applicable. And given that some of those specifications maybe defy intuition, what's the future of these standards? Does the IPC have any plan or motivation to maybe more comprehensively address thermal management problems? Is there going to be a guide line for maybe developing a thermal model from your data? I think that might be a bit more interesting, like an analysis standard.

Mike Jouppi:
Well, that's what I wanted to do. When I created all of this work and I had the software package, I went to IPC and I wanted to try and sell it through IPC, along with the standard, but the economics just didn't work and I couldn't couldn't make that all come together, but I thought it would've been a perfect combo, especially with training. I think that there needs to be more training. But there was just no funding for it, so we couldn't really move forward with any of it. There was a great master plan, but we just couldn't pull it off in a few years that I had available to make that all happen. And something that people need to realize is that IPC really took over industry standards from the government back around 1995.

Mike Jouppi:
And before IPC had these standards, all of this work was done by the National Bureau of Standards. And when I was at Hughes Aircraft Company, I remember the senior guys complaining about how it took forever to get something through the National Bureau of Standards. They wanted to make sure ranges and you'd submit a form and you'd just wait for years before you get results. And so in 1995, there was a thing called the [inaudible 00:44:26] act and all of the military standards were turned over to the industries that used them.

Mike Jouppi:
It was expensive. People were complaining about the National Bureau of Standards. So government said, well, all right, you want to complain? You guys take it over and you manage them. And so IPC is really nothing more than all of us. The industry is responsible for ourselves. We are responsible for ourselves. So is IPC doing anything? You got to go talk to the people that want to make change and bring that to them to make change. I couldn't make it work any more than... I felt good about getting as far as I got and I just couldn't carry it any further. Plus it was all [inaudible 00:45:16] Lockheed wasn't paying my way because it was a conflict of interest because of my IP and I was stuck. And so I eventually had to just stop spending my own money to help the industry and move on to other things.

Zach Peterson:
Well, it sounds like it's time to pass the torch maybe, or pass on the knowledge at least, and allow the next generation to maybe continue in your stead for those who care a lot about thermal management because I know it's extremely important for high reliability designs, but it's interesting because I think you go to a conference like we just had AltiumLive or maybe you go to one of the other PCB conferences or DesignCon or something else. And thermal management is not high on the list of available presentations. It's a lot of signal and power integrity, which are important, maybe it's a lot of HDI and then a lot of manufacturing, but you're lucky if you see anything on thermal management.

Mike Jouppi:
Yeah. Well, there's a lot of people that recognize that thermal management is important though. And there's a guy, his name's Istvan Novak, he's a signal integrity guy.

Zach Peterson:
I believe he's actually been on the podcast as well.

Mike Jouppi:
Oh, awesome. Well, I want to apologize to him. So he asked me if I would submit a proposal to Oxford University to do a class. And I want to do it on IPC 21 52. I sent a proposal to Oxford. They accepted it, but they said, well instead of... They just wanted a one day class, they said instead of a one day class, how about doing a two day class and do a heat transfer, thermal management on circuit boards one day and then do the IPC on the next day? So I submitted of that. They accepted it. I started to put this... One of the questions to them was so do you want death by PowerPoint or should we do some experiments and things like that? And they go, yeah, we don't want death by PowerPoint.

Mike Jouppi:
Let's do some experiments. And so I had a bunch of stuff from when I started things at the University of Colorado in Denver. We had some really cool setups for testing boards and I was going to bring all of that. And I started pulling everything together. And my equipment's old, my laptop that has the data acquisition software was out of date. I couldn't get the software up and I've got a rule of not spending any more money on this stuff. I would've had to start calling people and begging for free copies and it just didn't happen. So I had to put it on hold and I put that training away also. And that was just a few years ago, but I think the key is to get some training and bring attention to it.

Mike Jouppi:
The next generation has to be looking at things a little different, especially in the pre-design phase. After you've got your board built, it's one thing, but then if you find problems associated with the traces, then you have to go do a respin on it. So the cost impacts. And not only that, bu if you have problems... We had a satellite that was on a launch vehicle, poor guys, somebody had shorted one of the lines to the electronics when they were doing some checkout, ran too much current through one of the traces in one of the boards.

Mike Jouppi:
And when they looked at the IPC, the 2221 charts, which is what they had at the time, I was showing 400 degrees C rise and they had most likely shorted it. And so they're going to have to pull this box off the spacecraft and look at it and inspect it, and then you'd have to run through qual all over again before you could put it back on the spacecraft. It would've cost months. It would've missed their launch date, but I was able to quick model, show that the temperature eyes was nothing and we saved just a ton of time and money.

Zach Peterson:
Yep. 400 degrees versus nothing is a pretty big difference.

Mike Jouppi:
Yeah.

Zach Peterson:
Well, I think we're going to have to leave it at, because we're running a little low on time, but this is all extremely interesting. And if you're okay with it, we would like to link to your LinkedIn profile in the show notes so that anyone who's interested could maybe get in contact with you.

Mike Jouppi:
Yeah, sure. I haven't done any updates to it. I just let it run. LinkedIn does it for pretty good job of researching and reminding you to do things. I don't go and do those, but it is there and I welcome people to contact me. That's fine.

Zach Peterson:
Absolutely, absolutely. Well, Mike, thank you so much for joining us. You know, me, I'm a physics guy, this stuff is all extremely interesting to me. And I think it's also very important as you know, high reliability designs are becoming much more commonplace in areas like commercial space, electric vehicles. I think thermal issues are going to come back in Vogue if you will.

Mike Jouppi:
Well, you can't miss out on it. It's a big part of the design process and you can't ignore it.

Zach Peterson:
Absolutely, absolutely. Thank you very much again, Mike. Everybody check out the show notes to learn more about Mike's work. We'll also have some great resources about thermal management and PCB design, and I think that's all for today. Everybody who's listening, don't stop learning, and stay on track.