Electronics

Signal Integrity | Impedence Modeling | AI In EDA Tools | Expert – Bert Simonovich

Guests DAVID SCHILD | Uploaded : 01/12/2023


The EEcosystem Podcast

Signal Integrity | Impedence Modeling | AI In EDA Tools | Expert – Bert Simonovich

Bert Simonovich shares insights related to hidden factors he’s discovered in regard to inaccurate impedance models. He’s discovered hidden effects that are a result of anisotropic effects in copper-clad laminates, that can result in signal integrity problems. Bert will be giving two upcoming presentations including EDI COM which takes place next week. He will be giving a deeper and more expanded version of this topic in January at DesignCon 2024.

Links & Resources

 

Bert Simonovich Articles in Signal Integrity Journal

DesignCon 2024

Bert’s LinkedIn Profile 

Judy’s LinkedIn Profile

📘 Visit The EEcosystem Website to receive our newsletter and receive a Free Signal Integrity Handbook from Samtec Download and receive a free 90-day subscription to all of Eric Bogatin’s training at The Signal Integrity Academy

Sponsors Resources
  • Sponsor Resource Links:

    💡For SI/PI/EMI News and Technical Resources and to register for a free subscription visit the Signal Integrity Journal today. https://www.signalintegrityjournal.com/

    👨🏽‍💻For free Technical Resources and to Learn more about Keysight Pathwave EDA Software Solutions visit the homepage now. https://www.keysight.com/us/en/produc…

    📡 For Custom RF and MW PCBs visit the Transline Technology Website to learn more. Visit https://translinetech.com/

    ⛰️Visit Summit Interconnect for all your complex PCB manufacturing needs. https://summitinterconnect.com/ 🔗For all of your high-speed and RF connectors visit the Samtec website and access excellent engineering resources while you are there. https://www.samtec.com/

    🎛 Visit the SIGLENT website to learn more about their test and measurement solutions For Every Bench. Every Engineer. Every Day. https://siglentna.com/

    💽For high complexity EDA solutions visit SIEMENS EDA Website https://eda.sw.siemens.com/en-US/

    🔌 Picotest specializes in high fidelity testing and measurement tools, primarily for power-related applications. Visit their website for more product information and excellent training materials from expert Steve Sandler https://www.picotestonline.com/

Transcript

								 Judy Warner (00:00.898)
Hi, Bert, so good to see you, my own friend. It's been a long time and I'm really excited to talk to you today about EDICon.

Bert Simonovich (00:03.768)
Come on.

Bert Simonovich (00:08.192)
Oh, hi, Judy. It's so nice to see you after so many years. I think it was four years ago, or maybe more now, since we saw each other at DesignCon. So, always great to see you and catch up.

Judy Warner (00:16.658)
Been crazy.

Judy Warner (00:22.322)
I know. So last year I was there, but I never, I was podcasting and never got out of my booth. I didn't see anyone. It was horrible. So this year I promise we'll, we'll find some face time, but in the meantime, why don't you share with our listeners who may not know you a little bit about your background and sort of the areas of interest you have.

Bert Simonovich (00:43.32)
Okay, sure, Judy. Well, I'm Bert Simonovitch, and I graduated from Mohawk College, and I'm an electronic engineering technologist. I'm a senior member of IEEE, and I've served on DesignCon program committee and the EDI, or editorial advisory board. I always get that fumbled up.

Judy Warner (01:07.07)
Okay, why don't you repeat that part.

Bert Simonovich (01:10.332)
Okay, I'm a senior member of IEEE and I serve on the DesignCon program committee and the Signaling Integrity Journal editorial advisory board and I'm also a Keysight certified expert.

Judy Warner (01:25.97)
Oh, I actually didn't know that about you. That's very interesting. I'm not surprised though. So.

Bert Simonovich (01:31.62)
Well, it's been a few years. It's a good little thing to have, you know, part of my repertoire.

Judy Warner (01:42.822)
Yeah, for sure. Well, I know you concern yourself with a lot of particularly board level SI issues. So first of all, I'm going to ask you a question that's mostly for me, but maybe for our listeners too. We're going to be talking about here I go anastrophic. Did I say it right?

Bert Simonovich (02:05.937)
No, it's anisotropic properties. Yes.

Judy Warner (02:08.23)
anisotropic properties. So will you please divine for us anisotropic?

Bert Simonovich (02:15.176)
Oh, what anisotropic properties mean is the material property electrical performance is different depending on the axis, x-axis, x, y, or z-axis. And particularly in printed circuit boards, it's related to how the electric fields are orientated toward the PCB laminate.

So printed circuit board laminates are made of fiberglass and resin. And they're measured by the fabricators. They have different test methods, over 13 different test methods they could use. And depending on the test method, the electric fields are either in plane

Bert Simonovich (03:15.856)
fibers, the fiberglass in the dielectric and or out of plane perpendicular to the fibers. So the implication is transmission lines like to have the electric fields perpendicular and the decay measured in that way we call like dkz for it.

Judy Warner (03:42.979)
Mm-hmm.

Bert Simonovich (03:45.272)
Other test methods have the e-fields running in plane with the dielectric material. So we denote that as dkxy typically. So dk is the effective permutivity of the material. That's one of the most important parts to get your impedance modeling right in your

Judy Warner (03:58.679)
Okay.

Bert Simonovich (04:13.208)
2D field solvers, for instance, or any of the EDA tools. Trying to do impedance, dielectric constant is integral part of that.

So it's an important thing to consider when you're doing your modeling, salentegrahy modeling.

Judy Warner (04:34.018)
So I have a question about that. It's my understanding, don't the laminate manufacturer provide those models and what is it that you're noticing where there may be some mismatch?

Bert Simonovich (04:49.892)
Oh, yes. So typically board shops, they work from construction tables provided by copper clad laminate suppliers, CCL suppliers. And for years, they just take the DK and DF values from there. They use that for, they put it into a 2D field solver to predict the impedance when they do their stack ups. And then,

that defines the line width in space, all the things that's involved in the stack up. So because DK is important, if you use the wrong value that's been measured, your impedance will be wrong. Fortunately, most laminate suppliers measure the laminate for in DKZ or out of plane. And for the most part,

know the impedance will come out right, should come out right, the line width is right everything should come out right. But now there are some laminate suppliers, Asian ones, they use what they call split-post dielectric resonator and it's a simple test and by the way these IPC tests are really designed for quality control. They're not designed to provide accurate numbers really for

signal integrity or anything like that. You know, and they state that in their construction tables, but generally they're, they're good enough to do your impedance for it. So, you know, when they,

Judy Warner (06:21.358)
Got it.

Judy Warner (06:24.82)
Okay.

Bert Simonovich (06:38.392)
So when they do the stack up design, the DKZ is there. So some Laminus suppliers, they say they use XY. If you use those numbers, when you do your characteristic impedance modeling, after you build the board, if it was made exactly as the stack up, like no process control, your impedance will end up being higher when you measure it.

Judy Warner (06:41.294)
okay.

Judy Warner (07:06.41)
Right. And so this is what you're gonna be talking about specifically is you've been researching this dynamic, specifically looking at all those variables.

Bert Simonovich (07:08.802)
So.

Bert Simonovich (07:21.512)
right and the reason I started doing it is helping clients have experienced that.

why are my impedances wrong? Because we're using a different laminate. Let's measure next Y. And when you start drilling down into it, then you realize, oh, there's anisotropic effect in there.

Judy Warner (07:48.358)
I see. So you've been researching this Bert from what you told me before our podcast. You've been researching this because you saw it. You have a consultancy and you saw this happening with clients. And so you start following you went down this rabbit hole and started finding these unusual effects and why and so you're going to do a 25 minute talk.

Bert Simonovich (07:49.484)
So.

Bert Simonovich (08:14.881)
Right.

Judy Warner (08:16.778)
You know, what else did you discover along the way of digging into your client's problems with, um, anisotropic effects and the different ways that, you know, in plane, out of plane, you know, whether it's X, Y, or Z access. So you're saying that if you just go with IPC specs in your board manufacturer, but you have a desired outcome for signal integrity purposes, they're not going to line up.

Bert Simonovich (08:26.064)
Yeah.

Bert Simonovich (08:46.356)
Right.

Bert Simonovich (08:51.78)
IPC spec for the TDR measurement is 20 years, let's just say it's 20 years, the last revision. 20 years ago, our line widths were wider, typically five, six mils type of thing. We're using one ounce copper and our dielectric material was pretty lossy. FR4 was pretty much it.

Judy Warner (09:10.613)
Mm-hmm.

Bert Simonovich (09:20.608)
was pretty lossy. When the board shop, when they design the stack up, they use a 2D field solver to do the impedance. A 2D field solver is a lossless calculation. It's not lossy. But when you do a TDR measurement, it's a lossy measurement defined by the IPC. And basically, if you ever looked at a TDR measurement on a screen,

Judy Warner (09:29.358)
Mm-hmm.

Judy Warner (09:43.551)
Mm.

Bert Simonovich (09:49.228)
You see it starts off on the left-hand side and then has a slow gradual rise. And that's caused mainly by the DC loss of the, of the trace, you know, copper loss. Couple that with the, um, very low loss material we're using today. That slope is fairly steep. So today we use very narrow line widths, like three mils, pretty narrow.

Half ounce copper, and as I say, the dielectric material is very, you know, is low loss, very low loss. Well, the low loss also helps to flatten that, or the higher loss helps to flatten that curve. So 20 years ago, you know, even though there was copper loss and everything, the higher loss material tended to flatten it. So even though it was a lossy measurement, where they measured between 30 and 70 percent,

Judy Warner (10:40.622)
Mm-hmm.

I see.

Bert Simonovich (10:49.104)
That's why you had your tolerance of 10%. Everything was pretty close. And you could build boards and not have an issue. But today, as I say, with the low loss material and the thin line widths, that has that rise. So without anisotropy, you could have a problem in the process variation. The real impedance you want to measure is down near the beginning of the TDR plot at the start. That's really.

Judy Warner (10:53.369)
Mm-hmm.

Bert Simonovich (11:18.368)
will match with what you've predicted. If you're measuring up there and it's violating, it's not truly the measurement. But we've been used to that and things generally work okay. But now if you centered your design, say 100 ohms plus and minus 10%, for instance, and you use the DKXY, the wrong DK. But when you measure it now,

it really wants to be DKZ, is the real DK. So you've lost two or three ohms of the positive margin now. So now you really had, you know, realistically you've not centered it at 100, it's like 101 realistically. So now you only have eight ohms of margin left. And now you throw in your process variation, like a little over-atcha bit, and you know the thicknesses aren't right. You know,

Judy Warner (11:51.112)
I see.

Judy Warner (12:03.275)
I see.

Bert Simonovich (12:16.608)
violation line and they say, oh, scrap or they call up and say, what do you want us to do with the board? You know, so generally then we look at it and say, oh no, it's down here so it's okay. But if you use the right decay, then you would center it properly and things would be right. So that's the one problem for SI is the impedance for the traces. But there's another issue now with vias. Vias go the opposite.

they go in the z direction in. So the decay it wants to see when you model is really decay x, y, or in plane. So if you only use one number for decay, one or the other will be wrong. So if you use decay z for the transmission line, that'll be right. But the v will be inaccurate, the model. And vice versa.

Judy Warner (12:55.933)
Hmm.

Judy Warner (13:02.975)
Oh my goodness.

Judy Warner (13:11.278)
So that's really interesting. So Bert, I'm wondering why you notice it now. Like, I'm assuming it's speed or frequency or something that's driving it. Like you said, that when the spec for IPC was made 20 years ago, it sounded more forgiving. So what's driving this and what made you come across this?

Bert Simonovich (13:35.156)
Okay, good point. 20 years ago also was bit rates for one gig serial links.

Judy Warner (13:42.934)
Ha ha

Bert Simonovich (13:44.152)
Today they're 100 gig, right? In order of magnitude difference. And at 100 gig, well, 56 gig even, but 100 gig, everything matters even more. And, you know, that's the problem. So in the past you probably had enough margins in things, but today those margins are eaten up. And...

Judy Warner (13:46.527)
Yeah.

Judy Warner (13:59.755)
Yeah.

Judy Warner (14:09.422)
Mm-hmm.

Bert Simonovich (14:13.12)
It involves accurate modeling. You need a sharp pencil today. Where in the past you could, yeah.

Judy Warner (14:18.302)
Right, that totally makes sense. Everybody's saying it. Like, everybody's saying right now, Bert, that the way people are framing it to me on the podcast anyways is that as these speeds increase or the frequencies increase, we think we're just going to sort of scale or something but the actual rules of thumb don't apply anymore. Like we need new...

Bert Simonovich (14:40.244)
Yeah, not today. You need accurate modeling.

Judy Warner (14:42.858)
Yeah. Accurate modeling, right? Which is, I can't tell you in the last month, probably the last three podcasts I've done, they've talked about this modeling, the problem with models. And now, I mean, we're talking about the models of chips and different package speeds, this and that, and the other thing, but this has to do, you know, it's a board effect, right? Because you have this laminate that's constructed.

in a way and the electrical fields are moving through the laminate. So I could see how it would be overlooked. Okay, so you're gonna go ahead.

Bert Simonovich (15:20.128)
Well, more importantly as well is with the EDA tools that we have today. Not a problem with it. EDA tools are accurate as what you put into it. They're very good. They're EM solvers, everything into it. But now if you think about it, many EDA tools are moving toward this AI stuff, like making it easier, push button things.

Judy Warner (15:27.259)
I see.

Judy Warner (15:34.588)
Yes, yes.

Bert Simonovich (15:48.192)
get results, you know, feed it your layout, it'll extract it, you know, if your stack-up's got numbers in it, it'll give you answers. Well now, right, so now if your tool doesn't allow for anisotropic effect, what decay does which part of that channel to use? It'll just pick the decay from the stack-up that it imports, but that's one decay. You need two.

Judy Warner (15:57.858)
But are the answers right?

Judy Warner (16:17.366)
Right.

Bert Simonovich (16:18.464)
basically, or to know the anisotropic effect of the material so you could put into a tool. And not all the tools have that capability. So this is just one thing to be aware of as well. When you're doing that extraction, to think about that and see if your tool has that capability in it, to be able to do it.

Judy Warner (16:28.686)
I see.

Judy Warner (16:42.702)
Well, this is fascinating. I think this is gonna be really valuable. Now I know you're doing, do you have a date that you're presenting at EDICon?

Bert Simonovich (16:51.816)
Yes, it's October 4th. I think it's in the afternoon after lunch sometime. Not sure the time. They have it on the website. The schedule is there. And yeah, that's the first session of October. And yeah, I'm doing it then.

Judy Warner (16:55.167)
Okay.

Judy Warner (16:58.586)
Okay. Yeah.

Judy Warner (17:11.202)
Well, I know because I've talked to Pat Handel at SIJ and I've talked to Ben Dan and also, they're giving you a whopping 25 minutes, right? Because the online, you know, attention spans, like we're all burned out of webinars and stuff. So you can only get so much in. So as I understand it, you're gonna give a 25 minute talk on the subject, hopefully gives engineers some ways to manage this more effectively.

Bert Simonovich (17:26.412)
Breath.

Judy Warner (17:39.822)
and be aware of it. And then, but you're gonna be doing, like I think of this as a primer for a much larger presentation, you'll be doing at DesignCon. So can you tell us a little bit about how far you are? I mean, maybe you're not done with that presentation yet, but you know, the length of time you're gonna have and, you know, just tease up a bit of DesignCon as well. And we can have another conversation before then.

Bert Simonovich (17:39.884)
Yeah.

Bert Simonovich (17:50.904)
That's right.

Bert Simonovich (18:04.544)
Yeah. Yeah, so for EDI CON, really, it's like you say, it's a primer. But in order to understand an isotope, and I see, I can't say it.

Judy Warner (18:17.194)
See? Now I screwed you up, Bert. Ha ha ha.

Bert Simonovich (18:21.496)
To understand anisotropic effect, you have to really understand how boards are fabricated, but not just from the PCB shop. I'm gonna start off from really, from raw materials, showing like, you know, you get the glass fibers, how you mix it, and really have, you know, take you through that little bit of a journey as a bit of a background to get you on a baseline.

Then when I start talking about the anisotropic properties and what that is, the physics about it, you'll have a better understanding with it. And then I'll show some examples, just quick examples of especially of a via if you use the wrong number based on the anisotropic properties of it, just the impact of the via impedance and what the implications could be.

Judy Warner (19:00.382)
Right.

Bert Simonovich (19:19.996)
In terms of, say, for 112 gig, we have now this COM metric that we can use, you know, channel operating margins, which spits out, you know, dB margins and things. But one of the big new things for 112 gig is there's an effective return loss number that you strive to get. And if you get the wrong impedance, it affects your return loss.

Judy Warner (19:49.54)
Mm.

Bert Simonovich (19:49.88)
coupled with your insertion loss. There's that logarithm that real smart people figured out, and it's in the tool. But I just show just simple and as a topic a few ohms difference, you reduce that margin that you have now. Doesn't mean it's gonna fail, just means you're losing some margin. And if you're on the edge already, then that may push you out. So that's the V effect, and then

Judy Warner (20:14.678)
Yeah.

Bert Simonovich (20:19.1)
I'm going to talk about, show an example for transmission line, what it would be. So it's really a kind of intro type of thing. Give a bit of background. As I say, I'll be doing design con, and I'll be digging deeper into it. And I'll actually, I've come up with an actual heuristic method working from the construction table.

to come up with a number that you can use. So if you know how that dielectric was measured, I have a simple correction factor that can give you another decay that, it's not perfect, that's what heuristics is, but it'll give you a better answer than if you use the wrong number. So I'll be really going into that detail explaining that. And that's really the meat.

potato of design con for it. You know, but yeah.

Judy Warner (21:22.434)
That sounds so great, especially for me as a, you know, ex board fabricator person, you know, I think people, I know I've actually done talks about that when I was teaching RF engineers about boards, just showing them what things look like, you know, before and after processing and being able to give them that visual and give them a sense of what that is really has, like people are like, thank you. You know, they really appreciate it.

So I think you laying that groundwork at DesignCon will be really, really valuable. Well, go ahead.

Bert Simonovich (21:56.884)
You know, to pick up on that one point, it's also important for, I'm fine, a lot of times I mentor some new grad type things, young engineers, and you know, coming out of school, they don't have the PCB background. And the one thing that has helped me all my life was early in my career, I was lucky enough to actually.

Judy Warner (22:09.684)
Mm-hmm.

Judy Warner (22:18.002)
Nobody does. Well.

Bert Simonovich (22:25.652)
learn how to do PCB layout. And then throughout the career, you learn a bit more about PCBs. And the biggest thing that can help any young engineer, regardless of the field, doing layouts or signal integrity or whatever, is understand board fabrication, what everything does. And if you understand those little nuances, it helps you throughout. So, you know, I've given this kind of presentation to people before.

And they found it really valuable. They finally got the light bulb. Yeah, and said, aha, get it, right? So I think it's useful, even though it might be repeated and other people may have seen parts before, others may not have. So I think it's quite a valuable intro for some of it. And then they say, half of the problem with is that what you don't know, you don't know.

Judy Warner (22:58.89)
That aha moment, yeah.

Judy Warner (23:03.35)
Yeah.

Bert Simonovich (23:23.308)
and I'm always that way and have been as well. But once you know you don't know something, then you can go find out more. So hopefully.

Judy Warner (23:23.57)
Yes. Yeah.

Judy Warner (23:30.302)
Right. You don't even know what to look for, which sounds like the thing that you're teaching on is absolutely sort of off the radar of people's awareness, which is really exciting and really glad you did this research. I think it'll be really valuable, Bert. So, well good luck to you at EDI Con for our listeners. Bert, by the way, has won best paper more than once at Design Con. He is an amazing researcher, always

Bert Simonovich (23:56.565)
Yeah.

Judy Warner (24:00.138)
Um, people love to learn from him. So I will put it in the show notes, but he writes articles for signal integrity journal, he'll of course be at design con at EI con, um, Bert, if you let me, I'll share your LinkedIn profile. And he's always putting great articles up and things that I think can really help you in your career. So make sure to plug into Bert and all the, all the great materials he's putting in and, um, and I'll put the links below.

Bert Simonovich (24:14.776)
Sure.

Judy Warner (24:29.458)
also to EDI Con and then I'll arrange to have them back before Design Con. Burt, thank you so much. I'm very excited to learn and now I'm going to go sign up. I don't know that I signed up. I signed up for a few sessions, but I'm definitely to go sign up for yours now. I'm really intrigued by what you've shared.

Bert Simonovich (24:37.569)
Yeah.

Bert Simonovich (24:48.376)
The nice part is, is if you sign up, you can, it's going to be recorded. So you can watch it after.

Judy Warner (24:54.074)
Okay, yes, this is true. And I've mentioned it before to the listeners, but I'm going to mention it again, in case they didn't hear my other podcast is Pat Handel told me. That if you attend in person, I triple E will give you continuing engineering credits for it. But if you watch it, cause that's their rule. If you watch the recording, you can't get the certificate, which

Bert Simonovich (25:11.884)
That's right.

Judy Warner (25:19.586)
Doesn't matter which one you choose, but I want you to be aware that if you are wanting to apply for those continuing ed credits, that you are gonna want to put it on your schedule. Again, Bert, thank you so much. It's so good to see you, and best of luck on your continued research and EDI Con. And hopefully we'll see each other and I won't cram my schedule full.

Bert Simonovich (25:34.176)
Stand here, Judy.

Bert Simonovich (25:39.608)
Thank you.

Judy Warner (25:45.318)
and I'll be able to see your face, my friend. It's been way too long. It's silly.

Bert Simonovich (25:49.992)
I know, same here, Judy. Good to see you.

Judy Warner (25:53.122)
Good to see you. For our listeners, make sure you go check out the show notes. The link that Bert and I have talked about, I'll put them all below so you can go tap in. By the time this is published, you know, we'll be right on the heels of EDI Con, which is held every Wednesday in October for partial days. So it's really manageable. And like Bert said, you can either watch it live or get the recording once you've registered.

And so go over and grab that. We'll see you next week. Thanks for joining us this week. Until next time, remember to always stay connected to the ecosystem. Cue music.

Bert Simonovich (26:28.632)
Thank you, Judy.