Judy Warner: Hey Lee, it's so good to see you, my friend. We haven't done this in a while, and I'm really excited to have you on today. Lee W Ritchey: Well, I'm excited to do the same thing. It's always been fun to do this because you're pretty good at prying out information that needs to be recorded, and I think that's what we're gonna do with this session. Judy Warner: We are indeed. Get ready, buddy. I'm gonna hold your feet to the fire! Lee W Ritchey: All right. Judy Warner: today we're gonna talk about something we've talked about in the past, but as you know, nothing stays still in this industry. So I wanted to get you on and talk about SKU. But before we do that, for the two people in the industry that might not know who you are, why don't you give us a brief introduction? Lee W Ritchey: All right, my name is Lee Ritchie. I'm president of Speeding Edge, which is a consulting and training company. Been in a high speed business since it started with ECL long ago and have spent pretty much all of my career doing pushing the envelope, going up and up and up in speed. In the early 70s when we were... doing high speed computing, we were delighted that we got our supercomputer to clock speed all the way to 33 gigahertz, megahertz, 33 megahertz. That's unbelievable that it was that slow, but that semiconductors were that slow, if you want to call it that. At any rate, from that experience, I've been, all my career on startups pushing the envelope with the next new product. And for the last, since the mid-90s, it's been internet products primarily, and to give some insight into what we have done since in 1996, we were struggling to make all of our links run air free at 10 megabits per second. Two years ago, I worked on interface to undersea cables to get the bandwidth up in the cables where the data rate was 400 gigabits per second. Gigabits per second. That's Judy Warner: That's Lee W Ritchey: a Judy Warner: crazy. Lee W Ritchey: 40,000 to 1 change in bandwidth in little over 20 years. So that's how rapid our industry has changed and it has everything to do with the classes I teach because... engineers have got to get up to speed at, almost makes your nose bleed. And so, part of, well differential signaling is what has allowed us to do that. And that's going to be what we talk about here. and I'm currently working on new laminates to deal with the next round, whatever the next speed increase is. Who knows where we end up. I thought five years ago we would never get to 100 megabits per second. Judy Warner: Yeah. Lee W Ritchey: Got that one wrong. Judy Warner: Yeah. Lee W Ritchey: That one went by just like that. Judy Warner: Well, to your point, it just blew my mind. I think because we all got in a time warp during COVID before COVID, I had done a podcast with Jason Ellison. And we I think it was Jason, we were talking about doing, you know, getting to 112. And I'm like, no, we're not. And so I just sort of was busy and wasn't paying explicit attention to high speed. And then I come back. Samtech starts talking about 224 and I'm like what? You mean you hit 112 and now you're shooting for 224 like it felt like Rip Van Winkle what just happened? So it's these are crazy times. So let's talk about why don't you give us a a quick for our listeners just give us a I'm sure everyone knows but give us a quick definition of skew. you know, and in the case of diff pairs, what causes it and the problems it causes. Lee W Ritchey: All right. Well. The Internet as we know it and high-speed signaling as we know it is possible because of differential signaling. Basically what we have is two equal and opposite signal waveforms that are launched from a driver down a transmission line, which is two transmission lines really, and arrive at a receiver that looks at the difference voltage. That's where the differential comes from and decides Judy Warner: Mm-hmm. Lee W Ritchey: is this a one or a zero. Well, how successful that is done depends on the two signals arriving at the same time. When we decide, we have a logic state change, is when the two waveforms cross. If they don't cross in the middle of the waveform, in other words, if they're misaligned, the beta error rate goes up and at some point the link will not work because they're so far apart when they arrive that link fails. And that's what we call skew. Judy Warner: Mm-hmm. Lee W Ritchey: Skew is misalignment of the two signals at the receiver. The primary cause of that is... difference in travel time across the printed circuit board. And of course, printed circuit boards are made from a combination of resin and woven cloth, glass cloth, Judy Warner: Mm-hmm. Lee W Ritchey: and those two have very different dielectric constants. So if one member of the transmission line is in the resin, it's got a dielectric constant around three, and the other one's on glass, it's got a dielectric constant around six, the speeds are going to be substantially different. And so Judy Warner: Mm-hmm. Lee W Ritchey: that is basically the only real cause of skew in modern thin circuit boards. We've got everything else that is involved so well made that it's basically the glass required to make a laminate dimensionally stable that is Judy Warner: Right. Lee W Ritchey: a villain. Judy Warner: Right. And so, yeah, and, and usually that's in the form of a grid, right? That was those it's a woven glass Lee W Ritchey: It's woven Judy Warner: and Lee W Ritchey: cloth, Judy Warner: it's right. Lee W Ritchey: just like linen, just like any Judy Warner: Right. Lee W Ritchey: woven cloth. In fact, the materials are woven on the same looms that used to make cloth. If I went through a weaver Judy Warner: Mmm. Lee W Ritchey: in North Carolina and it was in an old textile mill, it was once upon a time making... Calico or something like that. And you could see all around all the stuff from making cloth. Same technology. Judy Warner: Right, ancient technology, but it works, right? So Lee W Ritchey: Mm-hmm. Judy Warner: because of this issue that's well known and is problematic, tell us a little bit about how engineers are handling that today. And then I know this is a problem you've been looking at, solving, working through for a very long time because of your expertise and focus on high speeds. So kind of... Give us a sense of where we are today and how engineers are dealing with this today with high-speed electronics. Lee W Ritchey: Well, basically there are two techniques being used. One of them is to route the traces in the artwork X and Y. And then when we put the artwork down on the panel of the pin circuit board, we tilt it at an angle so that you're not traveling straight down one fiber and between Judy Warner: Right. Lee W Ritchey: two fibers. And that's the dominant. solutions being used. Early on, the number was used with a 15 degree angle, Judy Warner: Mm. Lee W Ritchey: and that turned out not to be good enough. Later on, research was done and we discovered that at 5 degrees, that pretty well takes care of it. So a very large part of the industry is doing it that way. The disadvantage of that is when you, boards are rectangles and so are the fabrication panels. Judy Warner: Yeah. Lee W Ritchey: So if you cock the artwork at an angle, you waste a lot of material or it may not fit. For example, we build a lot of what we call pizza box boards which are one you rack mounted. Judy Warner: Mm-hmm. Lee W Ritchey: And that's the dominant board in this whole industry. Judy Warner: Right. Lee W Ritchey: And that's wide enough that you almost can't do that, caulking on the panel, Judy Warner: Right. Lee W Ritchey: and make it fit. Judy Warner: Yeah. Lee W Ritchey: And so that's not my solution. Judy Warner: Yeah. Lee W Ritchey: My solution is. There are probably a dozen different weave styles that you can choose from for these thin laminates. And the classics always been what's called 106 glass and 1080 glass. And that was the first place that this problem cropped up because 1080 was automatically how you made a four mill core. That's where this problem first popped up. Well, we discovered there was something called spread glass. the fibers in Judy Warner: course. Lee W Ritchey: each glass bundle were spread out so that you had a uniform surface of glass. And that solved the problem. And for quite a long time we were using a material called 3313, that's a glass weave. And we built all kinds of boards with that at a fabricator up in Oregon and had no skew. And so we thought, oh, that's a wonderful material. And so the test work we did for ISOLA, we made from 3313 and they all failed skew. Every one of them failed. Judy Warner: What? Lee W Ritchey: What's going on here? Turned out it was a different weaver. Judy Warner: Oh Lee W Ritchey: And so, Judy Warner: no, so it's where they were sourcing Lee W Ritchey: yes. Judy Warner: the glass. Lee W Ritchey: And so we got to dig it into it. And the reason that the glass is being spread is not single integrity. it was for drilling laser-drilled blind vias. Because Judy Warner: Mm-hmm. Lee W Ritchey: if the glass was not uniformly spread, if the beam intensity was high enough to burn through the glass fibers, Judy Warner: Mm-hmm. Lee W Ritchey: where there were none, you got a huge, huge erosion, and the holes were basically so badly formed that you couldn't plate them. And so that's why glass got spread. And of course, the driver for that was cell phones. cell phones because they have all the buildup technology. Judy Warner: Mm-hmm. Mm-hmm. Lee W Ritchey: Of course, the spreading that was being done in these two variations, the 33, turn 13 I talked about, they were happy with that. The spreading was okay. It was good enough for a laser-drilled binavia, Judy Warner: I see. Lee W Ritchey: but not good enough for SKU. And so we spent, I don't know how many hours trying to get the IPC standard for how glass weave is specified to get a definition of spreading that was good enough so we could trust and get the same thing from everybody. Judy Warner: Mm-hmm. Lee W Ritchey: But it was very much like herding cats. The committee had a member from pretty much every glass weaver on it. They would not share how they did spreading. Judy Warner: And there is the issue. Lee W Ritchey: There is the issue. And so I have a frustration. I finally just quit trying to get that done. So we searched around for a better solution, and this is it. There is a spread glass weave called 1067, which is— intended for the same application. We discovered that if you just use one layer of that above and one below your transmission line, in one direction, you still get skew. So we discovered by building test boards that if I put two flies of 1067 above the transmission line and two below, we got it. Judy Warner: Now did you, did you do those in different directions when you layered them like that Lee? Lee W Ritchey: No, you really can't Judy Warner: Same Lee W Ritchey: do Judy Warner: direction. Lee W Ritchey: that Judy Warner: Okay. Lee W Ritchey: because of the way the weave is made. I mean, that would be one solution, but it... Judy Warner: It'd be bad for the board. Lee W Ritchey: If you look at how laminate is made, you're going to want to throw a whale on material if you do that. Judy Warner: Yeah, okay, I wasn't sure what that would do. So in this 1067 and doing the second, okay, I have two questions. One, when you do spread glass, you mentioned about the different loss of the glass versus the laminate. What does that do to your loss? Lee W Ritchey: Well, so... We were really talking about the loss in the resin versus the loss in the cloth, I Judy Warner: Yes. Lee W Ritchey: think. We didn't see a significant difference in loss when we had SKU. I think that's your question. Do we also get more loss when we have traveling over the glass bundle versus Judy Warner: Yes. Lee W Ritchey: in the dielectric? The answer is yes, but the number is pretty small. The difference is Judy Warner: Okay. Lee W Ritchey: pretty small. The Judy Warner: Okay. Lee W Ritchey: reason for that is the manufacturers of these systems that are aimed at this market have successfully reduced the loss in both materials. So low that any difference you would see. is second or third order. Judy Warner: Okay, all right, that's good to know. And then in regards to this 1067 glass and putting two layers above and below the transmission lines, is this something, like did you publish that? Like is that something known to engineers? Can they tell it from data sheets if they wanted to implement it? Lee W Ritchey: Well, I've taught that technique in my classes for the last seven or eight years. Judy Warner: Mm-hmm. Lee W Ritchey: So it's in the public domain. Judy Warner: Okay. Lee W Ritchey: I suppose that John and I could have patented that, but we've been interfered with by too many patents in the past. And so we're... We don't want to constrain the problem Judy Warner: Right, Lee W Ritchey: that way. Judy Warner: yeah, I understand. Well, Lee W Ritchey: Life's Judy Warner: maybe Lee W Ritchey: hard enough. Judy Warner: we can, yeah, well, maybe we'll go dig those out. I'm just thinking that people listening to this might wanna be, ooh, oh, yeah, I wanna do that. So I wanna send them somewhere else where they could maybe access things that are in the public domain. That's why I was asking. Okay, so spread glass is the best solution and to do it in this way, now. It sounds to me from talking to you and Kela that there's a fundamental limit of laminates and speeds somewhere around 56 gigabits per second. Can you talk about that a bit? And again, how do we deal with that? As you mentioned in the beginning, like the speed's not gonna slow down. Balance are not gonna slow down. So we've sort of... created a stopgap. That's good. And where do we go from here Lee? Lee W Ritchey: The solutions that I was describing and the designs I was describing were 32 gigabits per second. Judy Warner: Okay. Lee W Ritchey: At 56, even that solution doesn't work. Judy Warner: Hmm. Lee W Ritchey: Because there's not much room for error at that data rate, at 56 gigabits per second, a bit period is— 20 picoseconds. Judy Warner: Mm-hmm. Lee W Ritchey: And so, five picoseconds of skew does that in. Well, Judy Warner: Yeah. Lee W Ritchey: we had as much as five picoseconds with that solution I just talked about. At 32 gigabits, that's okay. So, what we're doing right now is fly over. meaning that we don't have the transmission lines in the board. They're in twin acts that's outside the board. Judy Warner: Okay. Lee W Ritchey: And no surprise, that's a very expensive solution. Judy Warner: Right. Lee W Ritchey: And so as it turns out, right now, the only signals that run at that data rate are between the fiber interface, the SFP module, and the first integrated circuit. So Judy Warner: Mm. Lee W Ritchey: you'll see that from the front panel back to the first integrated circuit that received that data rate. That's a flyover connection. Judy Warner: Hmm. Lee W Ritchey: From there on, everything in the board is at 32 gigabits per second. Judy Warner: I see. So they just use it where they need it Lee W Ritchey: Yeah. Judy Warner: and keep the other transmission light on the board. Lee W Ritchey: And you just hinted at earlier about 112. Judy Warner: Yeah. Lee W Ritchey: And so if they, there is 112 in that path I just described. And you Judy Warner: Mm-hmm. Lee W Ritchey: saw at DesignCon this year, trying to do 224. Okay. Judy Warner: Yeah, yeah, my friends at Samtech, I never did get over to their booth at DesignCon. I wanted to, but I was just too busy podcasting. But what did you see over there? And I know it was just a kind of proof of concept. Dreamy land Lee W Ritchey: Well. Judy Warner: at this point, there's, there's a whole ecosystem and a lot of problems have to get solved before then. But what were your Lee W Ritchey: No Judy Warner: observations? Lee W Ritchey: big surprise, all the connections were a twin ax flyover. Judy Warner: I see. Lee W Ritchey: Where I was actually going with my last little bit is if I take that last inch, if you wanna call it, at 224, Judy Warner: Yeah. Lee W Ritchey: down on the board, I'm gonna have to double the data rate on the board, and that's gonna be 56 gigabits per second. Judy Warner: I see. Lee W Ritchey: And now we're stuck. Now we're stuck because there Judy Warner: Okay. Lee W Ritchey: isn't any material right now that will—assuming you don't have a loss problem, there's no material that is good enough with respect to skew Judy Warner: I see. Lee W Ritchey: to go at that data rate. Judy Warner: Got it. So now what Lee? I mean, we got the flyover cables, but now what? Where do you imagine we're gonna go? And we're always talking about, and I really don't know enough about it, but I imagine you do optics or embedding things in the board. Like, is there any solution that... gives us some hope, I Lee W Ritchey: Well, Judy Warner: guess. Lee W Ritchey: optics is more expensive than twin axes Judy Warner: Yeah. Lee W Ritchey: by another order of magnitude. I mean, people Judy Warner: Yeah. Lee W Ritchey: talk about optic connections in the board, and that is not realistic. What we need to do, and I'm working with two suppliers right now, we need to make a material that does not depend on glass. Judy Warner: Hmm, okay. Lee W Ritchey: because that's the villain. And if you reflect way back, the reason there's glass cloths is that produced the mechanical stability Judy Warner: Mm-hmm. Lee W Ritchey: for a two-layer board Judy Warner: Yeah. Lee W Ritchey: or a four-layer board. Well, when we're talking about this kind of a design, it's rarely under 20 layers, Judy Warner: Yeah, exactly. Lee W Ritchey: of which probably half are plain, copper planes. and the copper planes are now producing the dimensional stability. We don't need glass anymore for that. Judy Warner: Oh, interesting. Lee W Ritchey: because Judy Warner: Okay. Lee W Ritchey: the copper sets the temperature coefficient of expansion of the whole thing, not the glass. Judy Warner: Mmm, okay. Lee W Ritchey: And so I'm working with two suppliers to come up with a limit that does not include glass. Judy Warner: Okay. Lee W Ritchey: We're not there yet, but that's what we're doing. Judy Warner: So will you, I assume that as part of that, and I know you can't divulge who those laminate suppliers are because you're under NDA, but what are you gonna make test boards and how far along in the process are you on that? So, I'm gonna go ahead and start with the test boards. So, I'm gonna start with the test Lee W Ritchey: Yes, Judy Warner: boards. Lee W Ritchey: yes, yes. We have already built some test boards. Judy Warner: Okay. Lee W Ritchey: At the moment, they're not satisfying the oil requirement for multiple reflows. Judy Warner: Hmm. Okay. Lee W Ritchey: And so that's the problem is being worked right now. The multiple reflow thing really is for avionics, where you might have your airplanes being sitting on the desert in Arizona, inside the instruments 150 degrees. Judy Warner: Yeah. Lee W Ritchey: About an hour later, you're at 30,000 feet, we're at 60 below, and you Judy Warner: Yeah. Lee W Ritchey: do that over and over and over again, the print Judy Warner: Yeah. Lee W Ritchey: circuit board gets Judy Warner: Yeah. Lee W Ritchey: cycled, and that's the reason for that 20 reflow business. We Judy Warner: I see. Lee W Ritchey: don't need that in a server farm. It doesn't go on and off. So we don't need that, but to get certified, you gotta do that for material. Judy Warner: Oh, all the blessing and curse of standards. Right? And it's the... Lee W Ritchey: There's another irony in that is we built all these test boards in 2013. The company we were doing it for had two laminates that had no skew. We were delighted with those and wanted them to commercialize the material for a reason we don't understand they didn't do it. Now, ten years ago we had material that satisfied the problem, but the manufacturer chose not to commercialize it. I don't Judy Warner: Woe Lee W Ritchey: know why. Judy Warner: is me. Lee W Ritchey: Yeah. Yeah. Judy Warner: Did that back in 2013 involve glass? And would Lee W Ritchey: Yeah. Judy Warner: it meet the 56 gig ceiling you're talking about? Lee W Ritchey: They wouldn't tell me what was in the material, Judy Warner: Okay, makes sense. Lee W Ritchey: but yes, Judy Warner: Patent. Lee W Ritchey: it would do the job at 56 because the loss was low enough. It would have done the job. Judy Warner: Let's go beat him up, Lee. Lee W Ritchey: I wouldn't be surprised that they lost the recipe. Judy Warner: So. Lee W Ritchey: company had some severe reorganizations. Judy Warner: Mm-hmm. Lee W Ritchey: In fact, the guy who was developing that material is one of the people who is developing the material we're testing right now. Judy Warner: Okay. Well, keep going Lee. Keep fighting Lee W Ritchey: Ha. Judy Warner: the good fight. We appreciate you doing it. This is really interesting conversation and since last time we talked about it, you have some interesting additions that I didn't know about. If you were a high-speed engineer today, where would you go look besides some of the things in the public domain that I'll try to hunt down? audience, where can they go to do better? Maybe they're not all the way, you know, at 56, but maybe they are at 32. And where can they go to learn more and maybe learn some techniques to help them with their skew problems? Lee W Ritchey: Well, the dominant place that topics are explored Judy Warner: Mm hmm. Lee W Ritchey: is DesignCon. Judy Warner: Okay. Lee W Ritchey: All of the things we're talking about have been presented as papers at DesignCon Judy Warner: Okay. Lee W Ritchey: and are available online from DesignCon. Judy Warner: Okay. Lee W Ritchey: I'm not sure if... I'm not sure how you get access to those. Judy Warner: Okay, Lee W Ritchey: Be good. Judy Warner: well I have a relationship with the DesignCon folks. So for our audience, I will go try to hunt that down for you. And whatever I find, I promise I will put in the show notes for you. Whatever I can find. And Lee, I know you have just some, you know, great classic books out on high speed. Where would you like our listeners to go to tap into some of your, your hard-won wisdom that... that you've published over the years. Lee W Ritchey: Well, we have two books. They're e-books now. We self-published, and the reason we did that is we wanted color, and none of the technical book publishers want to do that, Judy Warner: Yeah, Lee W Ritchey: so Judy Warner: which is Lee W Ritchey: we Judy Warner: funny. Lee W Ritchey: self-published, and it was not a cheap enterprise to do that. Judy Warner: Mm-hmm. Lee W Ritchey: Funny thing about when you go to a printer, and say I want 3,000 of these books, and say okay, here's the start button on the press, but before I'll push it, I want a check for $50,000. And it Judy Warner: Dear. Lee W Ritchey: kind of sets you back a little bit when you write that check. Say boy, I hope somebody buys these books. Judy Warner: Yeah, right. Lee W Ritchey: They did, but we don't want to do anymore $50,000 checks, so they're e-books and they're on our website. Judy Warner: Okay. Lee W Ritchey: Anyone who wants to can go on there and buy an e-copy for $25. Judy Warner: Which Lee W Ritchey: The first Judy Warner: is a Lee W Ritchey: volume, Judy Warner: heck of a bargain, yeah. Lee W Ritchey: yeah, it is. If you think about what the prices of books you buy from somebody like Prentice Hall. Judy Warner: Yeah. Lee W Ritchey: Anyway, the first volume is what I would call the fundamentals. Judy Warner: Mm-hmm. Lee W Ritchey: The second one is where all this stuff is, we just got talked about, the high-end Judy Warner: Okay, Lee W Ritchey: stuff. Judy Warner: okay. Lee W Ritchey: And volume two is now a textbook at my university, which is kind of nice. Judy Warner: Yeah. Lee W Ritchey: Nice to know you did something that a professor wants to use. Judy Warner: And passing on the baton. Like I said, this is hard worn wisdom. You got the bumps and bruises to account for it. So that's great news. Lee W Ritchey: I have from probably about four or five thousand designs. Judy Warner: Yeah, just Lee W Ritchey: At Judy Warner: a few. Lee W Ritchey: least that many. Judy Warner: Yeah. Lee W Ritchey: and pushing the envelope every day. Judy Warner: Yeah, well, keep carrying on Mr. Lee Richie. Lee W Ritchey: Yeah. Judy Warner: We need you. Just keep doing it. Cause you know, there's a, there's a few out there trying to solve these problems, but I really appreciate you because you have your feet on the ground working with industry and engineers to really solve these problems. So we appreciate you very much. And thanks so much for coming on and talking Lee W Ritchey: Thank Judy Warner: to Lee W Ritchey: you. Judy Warner: us today and getting us up to speed. Lee W Ritchey: Okay, there's one other thing I should point out. Judy Warner: Okay. Lee W Ritchey: We should have written a volume three, but the first two were hard enough work. We chose not to do that again. And on my website is a digital library, which Judy Warner: Mm-hmm. Lee W Ritchey: is all kinds of articles I've written since then that are my volume three. And Judy Warner: Okay. Lee W Ritchey: you can buy that. You can buy that library. Judy Warner: Okay, I'm glad you inserted that. So again for listeners, I'm going to put all those links below for you so you can go tap in and learn much more and much more depth than we've had the opportunity to talk about but Lee, thanks again. I appreciate you and thanks for coming on today and sharing with engineers how to how to navigate this this pervasive problem, particularly in high speed I hope you'll come back and tell us when you make more progress with those laminate manufacturers. Lee W Ritchey: Alright, you can bet we'll make a lot of noise when it's real. Judy Warner: Okay. All right. Well, don't forget us. We'll help you make a lot of noise. Lee W Ritchey: Okay, my pleasure. Thanks, Judy. Judy Warner: Thank you so much for our listeners. Thanks so much for joining. Lee Richie and I to talk about the problem and askew and that ways you can manage it and tap into more resources. We will see you next week. And don't forget, we're gonna be at IMS the second week of June. And so I hope you'll stop by and see me at booth 1855. We'll see you next week. Until then, remember to always stay connected to the ecosystem. Music. We didn't.