Bert Simonovich on Modeling Copper Roughness

Estimated reading time: 6 minutes

I met with one of our contributors, Bert Simonovich of Lamsim Enterprises, at DesignCon 2019. Bert’s paper on interconnect modeling was nominated as a Best Paper finalist, so I asked him to discuss his paper and some of the challenges that engineers and their customers are facing right now.

Andy Shaughnessy: It’s good to see you again, Bert. I know you just finished presenting your paper. Can you tell us about it?

Bert Simonovich: First of all, it’s nice to see you again too, Andy. I presented another paper this year titled “PCB Interconnect Modeling Demystified.” It was nominated for Best Paper Finalist, so that was quite an honor. My paper was a combination of previous papers that I have presented, plus some new things that I’ve learned. This industry is all about never-ending learning; I learn a little bit more every year. It adds to the accuracy of the modeling technique and seems to work very well.

I’ve been working on my own for the last few years, developing interconnect modeling—how to properly model transmission lines from data sheets and show how accurate things can be. Over that time, I’ve focused on my method and technique. I’ve come up with a novel way to model roughness of the foil, and how to correct the dielectric constants due to the roughness. I put it all together in the paper for people who are in the same position as I was a few years ago and help them get through the challenges because on the SI list, I’ve seen people having difficulty with this topic, such as “How do I apply roughness parameters in popular EDA tools?”

Shaughnessy: What are some of the biggest challenges that engineers are facing regarding this topic?

Simonovich: The big thing is modeling the actual copper roughness itself. It’s hard to get the roughness numbers from what it is out there, but it’s there if you dig for it. The roughness of the copper adds to the loss of the channel. With the new standards—including 28 gigs, 56 gigs, and now, people are talking about 112 gigs—it’s becoming more and more of a challenge. We need a more accurate way of modeling things. Another challenge is trying to meet new standards. For instance, PAM4 is brand new and already embedded in; people now have to implement that, so it’s kind of a new paradigm. People have to think more compared to what they were used to.

Shaughnessy: PAM4 was new or almost new. It made a big splash at DesignCon two years ago, and now, it’s part of it.

Simonovich: Yes, and people are designing real products with it now. Then, there are the test equipment people who have to make equipment to test it all. If you’re a smaller company or a consultant like myself, you don’t have access to make test boards and measure things and do things that way. A lot of times, you need “an answer now before a better answer later,” as my friend Eric Bogatin likes to say. That’s exactly true for a lot of cases. If you’re in that mode and you need to know something now, then the modeling technique I’m showing may give you satisfactory results.

Shaughnessy: So, it’s something that someone who isn’t necessarily a hardcore SI engineer would be able to do?

Simonovich: A lot of the EDA tools have these models—and in this particular case, the Huray model for roughness—but it depends on certain perimeters to put in to feed the model; that’s where people get stuck. They start to ask, “How do I do it?” So, they don’t use it, even though it’s very accurate. For my model technique, I’ve worked on what I call the Cannonball Model.

Shaughnessy: I remember you wrote a column on the Cannonball Model.

Simonovich: Yes, I did a few years ago. The Cannonball Model will get parameters that you need to feed into a Huray model from data sheets alone, and that’s the key. If you use that technique to get these parameters, I’m showing an excellent correlation to measure data with it, and that’s what I wanted to share.

Shaughnessy: That’s good. It seems like I hear about people saying that they have these EMC tools or whatever to do all of this wild stuff. But when it comes down to it, they’re already behind on the job, and so they either don’t use it or don’t use it to the full capacity; they just do what they’ve always done.

Simonovich: A lot of tools depend on fitting parameters against some measured data. If they do it, they will get excellent results too, but they need measured data to fit it and then use those numbers in their simulations. If I use my Cannonball Model, I have a technique to get good results using Rz from data sheets alone. As I was saying earlier, I learn more and more every year. This year, I included an analysis of oxide and oxide alternatives. How much of an impact does the rough oxide alternative have versus the new stuff? How much of a difference does it make on insertion loss? I’ve shown some of that stuff too, which is new data that I received over the last year.

Shaughnessy: Are there any other big challenges that you think are going to stick it to us in the near future?

Simonovich: One challenge is PAM4 and getting your channels working right. The specifications are so tight on the PAM4 that signal integrity is being increasingly required now. You can’t just wing it anymore; you have to do analysis and a lot of things have to be right. Everything matters now when you’re doing it.

Shaughnessy: Congratulations on being nominated. How do they choose?

Simonovich: Thank you. At DesignCon, when you get nominated as a finalist, the technical program committee reviews all of the papers. They choose which papers can be best paper finalists, but the audience has a chance to vote on it as well. Regardless if it’s a best paper session or not, audience members get a survey to vote at the end of every session; they rate the presentation, speaker, and everything. Particularly for best paper finalists, they combine the technical program committee scores with attendee feedback to determine the winners. It seems fairer because there are two aspects—your paper and the presentation. One part of it is only based on the paper, but the presentation is just as important. Did they get anything out of it? Having the audience vote makes a big difference.

Shaughnessy: So, you won’t find out until next DesignCon if you won this year?

Simonovich: Well, you actually find out in late March or early April; they compile all of the data and announce who the winners were. At the next DesignCon before one of the keynotes, they will present awards to the previous year’s winners.

Shaughnessy: Well, best of luck. You’ve won before, right?

Simonovich: Yes. Our paper won last year as well.

Shaughnessy: That’s cool. Thanks for speaking with me.

Simonovich: It’s always a pleasure, Andy. Thank you.