Materials Science and Manufacturing of Better PCB
Materials Science and how this can level up your PCB manufacturability.
In this episode, our guest Geoffrey Leeds the product manager at Insulectro talks about how material science can help solve the unique manufacturing challenges fabricators are dealing with HDI designs.
Watch through the end and check the additional resources below.
Listen to Podcast:
Watch the Video:
- Geoffrey Leed’s role as a product manager at Insulectro, a material science distributor
- What is material science and how does it relates to PCB manufacturing
- Ultra HDI designs present unique manufacturing challenges to PCB fabricators
- How are your material choices impacting your design performance?
- Geoffrey explains why having lower CTE materials could be a double-edged sword
- Perfect is the enemy of good enough! You must accept some level of tolerance when your product moves into production and goes out into the real world, It can be the material tolerance or the electrical performance
- The PCB industry has been walking in the packaging industry's footsteps for quite some time and the CHIPS act has become one of the drivers of this movement
- Would a set of alphanumerical rules help designers with HDI designs? Geoffrey answered with the phrase “curse of the easy button”
- Geoffrey recognizes IPC’s effort as the governing body in the standardization of PCB design and manufacturability- heterogeneous
Links and Resources:
- Connect with Geoffrey Leeds on LinkedIn
- Follow Insulectro on LinkedIn
- Visit Insulectro’s website here
- Watch Related Episode: Mike Creeden on Empowering PCB Engineers through PCE-A
Are you accounting for that in your S parameters, when you're going and you're punching it in, and you're doing your full analysis, your Monte Carlo analysis to make sure everything's working properly. It can be very different, and I think for a lot of board designers, they don't sit back and say, "Well, wait a second, how are my material choices impacting my design performance?"
Hello everyone, and welcome to the Altium on Track podcast. I am Zach Peterson, your host, and today I'll be talking with Geoffrey Leeds, product manager for flexible materials at Insulectro. We've actually had someone else from Insulectro on the podcast in the past, talking with Judy, and I'm very excited to be talking with Geoffrey today. Geoffrey, thanks so much for joining us and welcome to the podcast.
Hey Zach, thank you so much for having me. Happy to be here.
Absolutely. So previously we had had Mike Creden from Insulectro on the podcast, and he's such a wealth of knowledge. It was pretty cool to have him on here. So anyone that's interested in listening to that, we're going to have a link in the show notes. You can go and listen to that podcast, but Geoffrey, I wanted to give you the opportunity to maybe introduce yourself to the audience, and then state a little bit more about what Insulectro does.
For sure. All right, so introduce, my name's Geoff Leeds, I'm the product manager at Insulectro, and my day-to-day is... It's not quantifiable in two or three statements, like everything in circuit boards, it's all connected. No pun intended there, but everything from how the board is made, how the board is designed, the resin systems, how they interact with one another. So a lot of my job every day is trying to figure out what the problem might be. Maybe someone's called me for support on a manufacturing issue or someone's called me for some kind of knowledge, like they would Mike Creden for a design related question, and then I really have to take a look at it from a material science standpoint and viewpoint and really understand, how can we pick apart this and frame this problem in a different way.:
Insulectro is... At its fundamentals, we are a material science distributor, and that term, if you tried to Google it, doesn't quite exist, and it's for a reason. A lot of what we do is we are a stocking distributor. We have 12 locations around the US where we keep different materials, whether it be rigid flex, consumables for PCD manufacturing, in stock and our warehouses, but the material science fund comes from what I just laid out before, which is there's multiple ways to solve problems, and we have all those different tools to solve them in our warehouse, and the question is, which solution best fits the problem? Is it a cost thing? Is it a performance thing? Is it a manufacturability problem? Different materials perform differently, and I know that might be prophetic, but sometimes having to just sit down and frame it from that viewpoint for whoever's calling us for support really helps them understand what's going on. So we're primarily educators on the material science side, and then we do the demand fulfillment on the distribution warehouse side of the business.
So when you say that you're a stock distributor in that you guys are in the materials business, it sounds like you primarily work with fabrication houses, providing them with the materials that they would want to order, but then also advising them on which material systems are going to be best for particular builds. Is that the right way to frame it?
Yeah, exactly. We'll get a phone call, there'll be some kind of IPC specification on a print, and a lot of times, our day to day, our core competency is around supporting fabricators and helping fabricators build circuits better, and our overarching goal as a company is to help our customers do exactly that. So while fabricators, I would say, are maybe 60, 70% of who our customer base truly is, we do spend a solid maybe 30% of the time talking to designers and just educating them about the material sets and how their material choices impact the manufacturability of a PCB or what those designer ramifications would be as well.
So that was going to be my next question is, do you ever deal with designers? I guess on the design side, this would, I imagine, be designers working probably at larger companies who are going to go into high volume, who need to make those material choices, number one, for the design, but also to ensure that they can procure enough stock for manufacturing with whoever their contract manufacturer is going to be. Is that correct?
Yeah, actually, I was going to say, that's actually an interesting... I would've normally assumed the same thing for most designers, especially those potentially in the consumer marketplace, but a lot of the designers we deal with stateside happen to be working for the military, the DOD side of the business, and a large part of the PCBs that are manufactured today are DOD related material sets. So with that in mind, a lot of designers that we deal with, they might not necessarily be designing millions of PCBs. They're only designing a couple hundred, maybe a couple thousand of them. That seems to be more of the marketplace that at least I interact with on a day to day. I do field calls and I do support those who are looking for the ability to source material in the US and globally, particularly around the Pacific rim region, and while we do help those individuals out, especially if they want to do the prototype here and then mass manufacturing overseas, I would say the majority of designers we help don't necessarily fit that model or that mold.
So in dealing with designers, it doesn't sound like you guys are like a Digi-Key for materials. That would be kind of a cool concept, but you go on a Digi-Key, you order whatever components you want to use, and then they just kind of ship them out and consign them to your assembly house, and voila, you get your boards back, but you guys are taking much more of a hands-on approach. You don't get that from a component distributor at all unless they've got Avnet's field application engineers from Xilinx or something on staff. So you can get that kind of service, but you guys are taking a much more hands on approach with the material sets. Is that because of the design challenges that you are running up against or that your customers are running up against?
Yeah, I'd say that's a big part of it. Today's designs, especially moving into the high speed digital or the ultra HDI designs that we're starting to see, it really presents a unique manufacturing challenge to a lot of the fabricator based in the US. So when you start using different resin systems, those resin systems change, and those implications from those resin systems might change some assumptions you had in the design phase, and then we might not necessarily be taking into consideration some of the manufacturing limits that are there as well.:
A very interesting problem I just recently ran into was we had the material set that would meet the design criteria, but there wouldn't be enough fabricators who could actually assemble this, so build it. So in the sense that it ramps, they wouldn't be able to meet demand, and then today, that's a very acute pain point for a lot of supply chain risk management policies and organizations. So we actually had to tweak the design a little bit in order to make it more manufacturable, and to your point about the Digi-Key, that's exactly it, right? We always have to help not only the designer, but also the fabricators who build these crazy designs, and it's not, as my boss would call it, this easy button you can hit every time, you just design something up and then you hit, go to quote, it gets printed and made and everything works first time.:
There's a myriad of things that can happen, and unlike components where it's pick, place and reflow in a PCB, you can have 250 to over 800 plus process steps, and each one of those process steps could cause something to change in your assumptions. So it's this interesting world where everything is changing quite rapidly at the moment, and we really have to help all the new designers that are coming into the field understand what those limitations are or maybe where we can call back some performance gains.
So you brought up ultra HDI, and there was actually just, I think, this past week an article in 007 about ultra HDI. So I know this is challenging for, I think, a lot of designers in a few ways, because not every designer does HDI to begin with, and now we're pushing the envelope to even smaller line widths and spacings, and then I'm sure the material sets evolve, as well, to be able to support the manufacturing process for those kinds of designs. So how do the materials start to get different once you get beyond the standard process and then in the HDI processing and then down into ultra HDI?
That's a good question, and if I looked at it from a material science standpoint, there's always this... If you open most material science books, the beginning of every chapter says any material has a certain set of properties. You change or process that material, and during that processing, you change the properties, and when you change the properties, you therefore change the performance.:
So when we start moving into how these things differentiate themselves, so when you're looking at a standard digital PCB versus an ultra HDI PCB, if the processing is the same, you're just changing, let's say, the resolution of the imaging and the etching process is identical, a lot of those performance characteristics or expected performance characteristics of your PCB are going to be similar-ish.:
When you start getting down into the micron range or the micron range, it's really interesting. The packaging industry back in the early two thousands actually found limits to classical deformation models and classical modeling techniques. So they had to figure out this new world as we started shrinking, there's a lot more that's going on in that, and when you start moving into super high frequency, you move from, let's say, an initial response, you're working more in the transient world or the transient response of a signal to the frequency response to the initial input of it, and that becomes what you really need to start designing to, and the material sets behave a little bit differently when you start doing that.:
When you start moving up into the ultra gigahertz range, you're starting to move up to 50 gigahertz, the dialectric constants of the material aren't constant. If you looked at a DKDF sweep compared against frequency, they change. So those assumptions about how the material would perform electrically are a little different. Now in your models, you just have to take that into consideration. "Okay, I'm not working at 500 megahertz, I'm working at 50 gigahertz. How are my materials performing?" When you just start going back and looking at those different assumptions you might have had in your regular PCB or regular HDI design, you start moving into ultra dense, ultra HDI designs, you really have to take into consideration the frequency response of these materials.:
So that's one angle of it, and then the next thing that I don't think a lot of people consider is the energy that's been getting pumped into the smaller and smaller spaces, and energy can be all kinds of things. It can be electrical energy, it can be heat dissipation, it can be mechanical energy from the sense of wallets thermally expanding. The board deforms. So there's changes in your spaces. The distances between those traces grow just a little bit, because the resident side of it increases in volume per degrees C, right?:
So those little underlying assumptions, which we might not have had to get too nuanced of in the old digital world, when you start moving into the super high speed world, you really have to get into those characteristics, and how do those change your models? Are you accounting for that in your S parameters when you're going and you're punching it in and you're doing your full analysis, your Monte Carlo analysis to make sure everything's working properly? It can be very different, and I think for a lot of board designers, they don't sit back and say, "Well, wait a second, how are my material choices impacting my design performance?":
So the material sets themselves, theoretically, are capable of going up to those speeds. The question is, what kind of loss are you trying to get? Then secondarily, what's the environment it's going into? There's a lot more out there today than just old school FR4, which was 20 years ago, you had maybe some megtron stuff, you had some products, but as we start moving into these ultra HDI designs, resin content and glass weave style become massive in terms of the overall loss of the conductor and how that affects your design, and those resin systems start to get very pedantic in the sense that it's DK and DFs.:
There's minutia details there. From a macro sense, they might seem really similar, but when you start getting into the weeds, the application really dictates, when do I start moving between one resin system versus. The other big driver, I think that'll be interesting to see how people conquer this, is lead free assembly. In the military world, it's only really starting to hit them versus the commercial world. We all had to deal with Rojas way back in the early thousands, and how did that impact our designs?:
So with that stated, though, how those problems were solved 15, 20 years ago, we had different manufacturing techniques compared to today, and different manufacturing techniques therefore change the material, which changes its performance. So the best way I can succinctly put this is assumptions about how materials behave when you start moving into this really dense world, you need to stop and take a second and really think through, "If this thing grows by half a millimeter distance between my signal and ground plane, what's my loss going to look like?", because you might find more often than not that maybe that's where my loss is coming from. My eye diagram doesn't look like the way it should. It can be a little fickle at times.
You actually just said something very interesting and you are the second person ever to bring this up, which is what happens when the substrate expands or contracts under elevated temperature or reduced temperature, and how that actually changes the distance between the trace and then the reference plane. You're the second person to ever say that, and I was talking with someone who works in the materials industry, I won't say who, but they had mentioned it'd be really nice if we had lower CTE materials for that reason when you're on thin or die electrics and when you're operating in that ultra HDI range, whether it's for boards or for substrate like PCBs or for packaging, it'd be really nice to have that kind of material.
Yeah, it's a double edged sword, and I say double edged for a couple reasons. On one hand, having something that's constant like that has a very low CT and it doesn't grow and move, it's like, "Oh, that would be ideal," because then my trace assumptions, they're constant. On the reverse side, when you start moving into the actual world, everything around us, it's not static, per se, everything's moving. If it's colder, it's a little smaller. If it's hotter, it's a little bigger, but then when you go and you put that into an assembly or a package, those packages might move and contract while your board isn't, so everything's heating up, but then your board is, again, becoming a rivet in this larger assembly, kind of like a microvia, and it's just that you have other kinds of stresses put onto it.:
So in an ideal world, everything would just move and grow and shrink and grow together. That would be fantastic, but sometimes we might want a little bit of movement, and that's the side I could argue for why we wouldn't want a lot of it, but then when you start looking at why we do, it's exactly that, because all my electricals remain constant, and trying to design an analog system that takes that into consideration and doesn't impact my performance, whether it's on the digital or the RF side, it can get a little complicated. I think those are fourth or fifth order differential equations, and it'll be a little bit interesting to see if someone comes up with that, but it's interesting when I hear people talk about low CTE materials, for sure.:
It's needed, and that's exactly what the packaging industry has done, but then when you look at... There was some interesting studies done by HDPug a number of years ago. They looked at packages that had really low CTEs, and as they were getting bigger and bigger, you would actually see parts of the package localize where the heat was coming from. You turn off parts of this model with the guy, and you only have heat getting pumped out of one side of it. Well, the board underneath it's growing, but the top side of it isn't, so you had this sheer stress in the XY plane, as well as in the Z plane, on these packages. It was a really interesting study, but those are the kinds of problems when you start moving to low CTE materials with only one portion without looking at the whole assembly and the whole shebang.:
Yeah, it's very pedantic when you start getting into it, but sometimes you absolutely need to, and people don't take that into consideration until we're sitting at the cage trying to figure out, "Why did this board fail," and it's because there was a material science property we didn't take into consideration in the design phase.
We always talk about CTE mismatch and yeah, the low CTE material is nice, but you're right, it is about matching it across the entire system, and you can never really do that. Maybe with the right material sets and the right package design and all of this, you can get close enough, but it's never going to be perfect.
Yeah, it's that compromise or that age old expression of, perfect is the enemy of good enough, and having to accept that sometimes is difficult, especially when engineers, whether it's a computer science engineer, where if something's not working, it's your code, and then the electrical side, if something's not working, you haven't got your circuit perfectly, but then when you move into actual production and how this performs in the real world, you have to accept some tolerances, and I think that's one thing that a lot of people don't take into consideration, is, what is my material's tolerance, and how is that material's tolerance taken into consideration for my electrical performance? So it's an interesting way to look at it, but it's absolutely needed when you really start looking at those ultra HDI designs.
Sure, sure, and with ultra HDI, the conversation is shifting there, I think, just at the right time, as the industry conversation is starting to shift more towards, a little bit towards substrate-like PCBs, but then also towards advanced packaging, and so now we just have the advanced packaging symposium, and then I know that there was a big panel on advanced packaging at PCB West, and so it seems to me that companies like Insulectro may have an opportunity to start moving into the advanced packaging market with some of those same solutions that they might implement for ultra HDI PCBs. Is that on the roadmap for Insulectro, or for maybe your competitors?
It's actually here today. So kind of like I referenced earlier, and maybe not to this degree, but the PCB industry has been walking in the packaging industry's footsteps for quite some time.
We're beholden to them.
Yes. We are. Whatever they come up with, we have to break out whatever they have the underside of that package is, and what's really interesting is the whole microvia reliability conversation where people are talking about, "Why is the microvia failing? What's going on? When I started in the industry back in 2015, I ended up working for a chemical distributor for a while. We got really into the weeds about this, because people would see a microvia fail at the base and they're like, "Oh, it's the electros, this is the reason why it's failing," and five minutes on Google, and you'll find tons of papers from 2000 to 2003 where the packaging industry was dealing with microvia reliability issues way back then, and it's almost comical how similar our conversations were today as they are being discussed in these papers, but materials behave differently based on how we process them and how the packaging industry processed materials, their solutions are geared towards packaging manufacturing.:
In the PCB world, we have slightly different materials, different processing steps. They're like cousins, kind of in the same family, but still different. It changes our performance. So we had to solve it in a different way, but it was one of these things where people didn't go back and see, "How was this problem solved before, and what learnings can we take from how those problems were solved and apply them today?" So when we start looking at the packaging industry and how they manufacture panels, et cetera, a lot of those same technologies get trickled down to us, and one of those is, for the inner layer processing side of it, is MSAP, it's modified semi additive processing, and it's a drop in right now for any traditional PCB fabricator today to allow them to get down to two mil lines in spaces if their imaging system can handle it.:
If they can go even lower than that, it really comes down to their etcher and their etcher's ability to move solution through it and make sure that it's not over etching or under etching when you start getting smaller and smaller, it changes how a PCB fab needs to take that into consideration, but that is absolutely what we've been doing since HDI design started to really hit the scene, and as we start seeing densities increase on the PCB, as that real estate becomes more and more valuable, we have to be able to break out some of those crazy packages that we're starting to see.:
So that technology from them, we've been using it for several years now, and it's interesting, and in a number of reasons, because it's comically similar to how we build PCBs. It's just a very, very, very small PCB if you wanted to take it from its simplest sense, but it is on the roadmap, and our customers are fielding designs today for interposers right now, and they're trying to figure out how can they build this, what materials can they build with it, especially when they start getting down to these really thin one and two mil dielectrics.:
So it's been something we've been helping our customers figure out for a number of years, but since the CHIPS act, really, it's been in overdrive, and in the last several months, it's been in, I'd have to say hyper drive. It's just accelerating at a pace that I haven't seen since I joined the industry seven years ago. It's cool to see, but it's a novel problem. It's categorically different in the sense that it's an order of magnitude smaller when you start dealing with a 12 micron line in space and the equipment we have to process that changes, but then that changes the material performance. You're changing the process, and that therefore changes performance of the material. So what do those impacts look like?:
A lot of PCB designers today are being tasked with this, where traditionally that might have been a function of somebody else in the packaging side of it, or maybe some pre-silicon ASIC designers are thinking through this problem a little bit. PCB designers are now having to step up to the plate state side to figure out how to solve these problems, and given the manufacturing base we have in the US today, it's been something we've been supporting our customers and our marketplace with for the last, I'd probably say five-ish years, and now the moment's really coming where the designs are here, the need is here, and we need to ramp and scale quickly, and we internally, at Insulectro, have been trying to figure out how we can scale our educational efforts to help support that effort and that onshoring event.
So you brought up the microvia reliability issue, and for me, I was exposed to that kind of late, because I'm only recently fully jumping into PCBs, and then, of course, working with Altium, but when you look at, I guess, the cross section of a typical interposer or a package substrate, and then you compare that to the typical cross section of an HDI PCB, should we really be so surprised that the solutions in one area telegraph onto the other area? Was that surprising for a lot of people?
No, not at all. It's copper. So on one hand, copper behaves. Yeah, yeah, exactly. It's...:
How do I put it? It's just copper. It's a copper rivet internal to the board, and you have to take into consideration what it's riveting together, right? Two signals, two pieces, two traces, or maybe it's going through a ground plane, but then again, kind of like what I said earlier, it's the processing changes, the performance and behavior.:
So when you start looking at how PCB is processed versus how package is processed, yes, they go through an intellectual step, and yes, they both have chemical desmearing to a certain degree beforehand. Yes, they both have lasers that are drilling the microvia themselves, but those aspect ratios are just wildly different, and if you go and you look at design rules for interposers, again, Google Scholar, it's one of the best research tools you have at your fingertips, and it's free, and you can see how they solve those problems, but then when you looked at PCBs, because there's a lot of...:
The way I always describe it is after the dot com bust, a lot of the talent developing PCBs just stopped going into it in North America, because it all picked up and went overseas. We essentially, as an industry, and fabricators had to become research to scientists themselves and figure out, "My factory is a lab, and I am performing experiments on every single panel," and then they have to start figuring out, in each of these different experiments, what's constant, what's not, and it really only reared its ugly head, I would say, for some of the crazier designs we're starting to see very early on, but then as the industry, just as that started to become more standard and more board shops were faced with this problem, the noise around microvia reliability just got louder and louder and louder and louder.:
I always just didn't understand why individuals couldn't just go back and see one, what was done, and then two, can they not perform a DOE themselves to figure out what's causing some of those issues at the shop? What was really interesting was when we start looking at design implications, and those design implications can be material selections and the physical layout of the board itself, and where those microvias are placed. So we're seeing differences in microvia reliability on ground planes versus going through signal planes, and then the question is, "Well, why?", and it's, to me, a fascinating, I would say, rabbit hole that I could jump into and just keep going down, down and understanding what's going on, but designers, they want this easy button saying, "What are my design rules? If I can stay within these rules, am I okay?", and unfortunately, when you start getting into these ultra HDI designs, there's not necessarily that kind of an easy button.:
Gerry Partida has done a lot of great work, and he's been really working with this stack, stack software to figure out that easy button, so to speak, but again, this problem has been solved before, and you can look at how we solved it, and that was solved classically with a tool like Ansys, which a lot of really high frequency designers are familiar with, with the HFSS package, and when you look at it from that standpoint, actually, I've been working on this little side project with some folks at Ozen Engineering, looking at microvia liability from a fracture energy standpoint, again, because we have to be more nuanced when we start getting into these really crazy designs, and it's interesting just to see these learnings, because when you use a tool like Ansys, you get so much more data out of these models, but they're infinitely more complicated in terms of programming and setting up compared to a tool like the stack software, which is relatively straightforward, comparatively speaking.:
But you can't use those tools and understand what these tools are telling you unless you have some kind of grounding in material science and you understand what's going on on a physical level, and what we're starting to see is a lot of electrical engineers having to stop and say, "I need to either go to my mechanical engineering department and get someone to come over here to explain this to me," or, "I need to start using Google and figuring out how to teach myself about what some of these different characterization properties and materials, or what these different material properties are and how they could potentially impact my design," and that's where a company Insulectro comes in, where we have that material science background and we can help designers understand, if you don't have that mechanical engineer to pull into the room, we can help you understand how the material sets interact electrically and interact mechanically with the panel. So we can help essentially scale that learning for individuals and help reduce the time it takes for them to figure out how to design a board and how some of the different material properties will interact, and either help them or hinder them.
Sure, and you brought up going on to Google Scholar trying to figure this stuff out yourself. I think for some folks, they've never had to do that, ever, and I came from academia, I'm familiar with that, and obviously, you having a hard science background as well, you're obviously familiar with that, but I think sometimes the designers need to have... It's not just an easy button, but almost like the alphanumeric rules of thumb that we see so often for stuff like high speed design, those telegraphed onto the ultra HDI regime.:
As this area gets, I guess, more attention, and as more of the, I guess, experienced class of designers starts to move in here and really embrace it and figure out what those rules are, either through experience or simulation or however they figure it out, will there ever be... Maybe not an easy button, but maybe a short list of alphanumeric rules that are going to help designers get a little bit more of their head around how to work in this area, or does that require a lot more work in terms of standardization, in terms of chiplets, in terms of material sets, in terms of processes, on and on and on? How mature does the industry have to be for it to be much easier for designers to get into this area?
So it is a great question, and then there's a number of different ways I can pick this apart and take a look at it, so I'll try to be succinct to a certain degree, but when it comes to getting that easy button, my boss coined a really great phrase, she calls it the curse of the easy button, because if you think about it... Yeah, you're laughing because you understand exactly what it is. To make something easy, it has to stand on hundreds of thousands of hours of effort to try to figure it out to get you this tiny little bit of information, and as the industry starts going towards these directions of much higher density levels in what we need to do, there certainly will be attempts at trying to make it easy. Necessity is the mother of invention, and in this case, we're seeing a lot of people try to figure out how to help that, and Insulectro is playing our part.:
On our website, we have an education section where we're trying to just make it easier for people who are getting into the industry understand basic terms. When you start talking about one or two core, 1068 glass weave versus something else, one resin system versus the other, core buildup versus sequential cap lamination, terms that, if you didn't pick up a book or someone didn't sit there and explain it to you and mentor you for a couple years or months, you'd have no idea, and for us in the material side, we need people to understand those terms, because now we don't have to sit here and spend three hours on a call. We can spend 30 minutes on a call. So we've tried to make short videos, and we are on our website, we have white papers on there, we have terms and definitions, which I just updated and added a bunch for the advanced packaging industry, because they like acronyms like nobody else.:
It was a little difficult to keep up with. So you've got those kinds of efforts that we're trying to take on. You've got companies like Lockheed Martin, who are publishing an IPC, this big test they did where they tried to isolate variables with their microvia testing and the DQ PON tester. The DQ PON tester, again, an invention that was arisen out of necessity, which I believe is a great way of measuring whether or not a panel weak microvia is on the panel.:
So you've got these different pieces trying to come together, but I think IPC really is, as the governing body, I think they're going to have to step up to the challenge, because one of the things that I heard at the packaging symposium when I was there two weeks ago, and I never thought I would hear this, was IBM, Intel, AMD all asking for standards around chiplets, and communicating between them. That really struck me, because I would've thought, as a tier one OBM, they're designing the silicon, they're going to want to control that ecosystem because they'd be able to get the most gains out of it as they possibly can, but they're reaching that point where, from a manufacturability side and a design and cost implementation, when you start getting down to the three nanometer node and you're doing that monolithic dye, the cost just becomes outrageous. So for them, it's easier to say, "Hey, we need the industry to define standards, because we can still go after computational gains with chiplets and switching chiplets and how we pick and place them and stack them onto silicon interposers, but we need an industry to figure out how to do that," and IPC took building a circuit board, which, if you want to think out of it as its most basic sense, is a composite material that just happens to be electrically conductive, and figured out a way to standardize it.:
As we start moving into this much higher density and ultra density designs and essentially now getting into chiplets, and we're talking insane speeds, we're talking traces... Some of these roadmaps, after 10 years, they're going to be going to sub-micron traces on an interposer. How do we standardize that? I think IPC is the right body and organization to kind of figure that out, and that's unfortunately going to take a lot of time, energy, and effort, and money, which will be interesting to see, but at least IPC can help spread that load, and it's done so very well with PCBs. So I think for everyone in our industry, it's a new challenge. The gauntlet's been laid down, it's something to wake up in the morning and say, "Here's something we can get excited about and go after, and use what we've learned and go after this new industry.":
So I think every single piece of the puzzle is slowly starting to congeal, but the best term that I heard someone say at the symposium was that it's the wild west right now, and when you look at some of these designs, packages, designs, how they're using the PCB, how do you integrate everything together in order to reduce that and make it easier to manufacture and to bring products to market for the consumer, standardization is going to be necessary, and I think IPC is really going to be the governing body that can execute on that.
Yeah, standardization, I agree, is extremely important, because heterogeneous integration is not really heterogeneous if you're only doing it with one manufacturer's products. So maybe Intel can standardize their chiplets and IBM can standardize their chiplets, but if you can't take elements from two different companies and put them together, then you're not really standardized.
Correct, and I think AMD actually proposed, they had a really interesting thing in their slides, but they were proposing... It's almost like a universal serial bus, like the USB, but for silicon chiplet interconnections, and that's being, I think, developed by, IEEE, it's still in its kind of implementation or concept phase, but then when it comes to the manufacturability side of it, that's exactly where IPC fits into the puzzle. We are the manufacturing expertise for electronics in the United States, and how, as an industry, can we help support that organic onshoring effort?:
It's a really interesting challenge too, because again, the manufacturing techniques to build the interposers and to get to those density levels, not only we're beholden to them, but we're walking in their footsteps, so we can help figure that out to make it easier for North America and Europe to onshore these and to build them, and realistically, like anyone else who's prescribed to the IPC universe of standards, they don't have to just be in North America and Europe, but those are the two that are trying to help push this effort at the moment, and those are the ones where we're stepping up to the challenge, geographically speaking.
It will be interesting to see how this evolves around the standardization, because I know there's the universal chiplets interconnect express standard, and I'm sure that there are others that are in development that are being looked at to help create this kind of dream of heterogeneous integration, make it a reality, and allow you to create a mature industry where you can just go and grab stuff off the shelf, throw it onto your interposer, throw it into your package, and voila, everybody's making packaging now.
Yep, and it's really interesting is when you start doing that, though, all these other considerations that we might have taken as assumptions in the past suddenly become no longer true. Power and thermal were two things that I wasn't expecting us to talk about to the degree we did at that symposium, but they were right there, and for all of the power integrity engineers and thermal engineers there, they're finally saying, "Yes, someone's finally listening to us." They're not just throwing this design over the fence saying, "You deal with it." We have to take that consideration on the electrical side, and what that means, though, is those implications are huge. Like we were talking about earlier, when everything expands, it changes distances, and that changes assumptions we have in our models. How do we quantify that? How do we standardize the ways in which we talk about ways we can pull energy out of it?:
Physics has done a fantastic job of setting up the units of measurement, but when it comes to manufacturing, like those units of measurement and those terms and the ways in which we figure out how to build it, it's got to get figured out in a language and in a way where we can all come together and work together to enable that, because it's the only way at the moment, it seems, to get those computational gains that we need without having to do complete system architecture redesign from the fundamentals of the transistor and how logic gates work electrically.:
So it's quite exciting to be in that world and to be part of that conversation, because it takes the PCBs, it takes the packaging engineers, it takes the pre-silicon engineers, it takes the thermal guys, everyone in the room to figure this out. So it really needs to have some kind of standardization wrapper put around it. Otherwise, no individual company is going to be able to hold that cost and figure it out. Otherwise, you're in the same position that the semiconductor industry was 10 years ago, where no one wanted to go from the 300 millimeter to the 450 millimeter wafer, because whoever does that transition first loses.:
So when Intel did it whenever it was a three to four inch, all their competitors just hired the Intel engineers to their facility. So there's this interesting game being played, and now we're moving into this world where that industry has basically said, "Okay, how about we standardize this instead of just trying to control everything," which is... Again, it really blew me away, but it's the universe we're headed into, and it's really fascinating to see it come back on shore, and when it comes to designers and how PCB designers need to figure out, "How do we fit into that space?", well, we need to educate them, and we need to figure out how we can talk to them in a way where they aren't having to go to Google Scholar and having to figure out, "Well, what is good research? What is bad research? Where can we go for places of authority to give us good information without having to go down their own rabbit hole?" People like us in Insulectro, we've got our website, we're trying to help make it easy, but IPC's also doing their part, and there's tons of YouTube videos as well, but some of them might not be succinct, and they don't speak specifically to the PCB designers that are here in the States or here in Europe today.
Yeah, yeah, I totally agree with you. Well, as all of this develops, we'd love to have you come on the show again to discuss this, because it's a quickly moving area of technology. It's evolving so quick, and when you said it went into hyperdrive, I couldn't describe it better, because all of a sudden, I'm having to learn as much as I can about chiplets, packaging, interposers, all of it. So yeah, this has been a lot of fun, and we've been talking about it probably not as much as we should at Altium. So I want to thank you very much for coming on here and discussing this with us.
Absolutely. Thank you so much for the invite. I'd be more than happy to come back on. I can tell you six months from now, this conversation is going to be wildly different, so...
I'm sure it is. Well, thank you again. To everyone that's out there listening, we've been talking with Geoffrey Leeds from Insulectro. We're going to have some great resources in the show notes, so I encourage you to go and click on some of those resources and do some reading and gain as much knowledge as you can about this very fast moving area of technology. Make sure to subscribe, hit that like button if you're watching on YouTube, and last not least, don't stop learning, stay on track, and we'll see you next time.