Cannonballs, eBooks, and Signal Integrity

Estimated reading time: 12 minutes

Starkey: I did particularly enjoy the analogy that you've presented at a few conferences regarding this pile of cannonballs that you pass on your way to the office in the morning.

Gaudion: Yes (laughs).

Shaughnessy: How does that figure in? Tell us that story.

Gaudion: Well, the Huray method starting off was called the snowball method. The idea was that the dendrites on the copper looked like a messy pile of snowballs, but actually to get the dimensions and areas of the snowballs off an SEM is quite complicated and there's lots of different dimensions and work to do. So what Bert Simonovich said is, “If we made all the dimensions the same and actually piled them up like a stack of cannonballs, we could use one diameter and a certain base and use a cannonball stack method.”

I was actually driving on the way up to the office and there was a lovely big pile of cannonballs, and I took a photo on the way past, so I use that as the opening slide and I have this picture of cannonballs just stacked there, and people are thinking, "What's this about? I came here to learn about signal integrity, not about cannonballs."

Starkey: As you and I have said many times: All models are wrong, but some are useful.

Gaudion: And that's from a statistician called George E. P. Box, and again, some people think models are actually perfect but we all know that base materials are composite of glass fiber and resin systems and when they're pressed together the ratios of glass and resin change, and so you don't have an exact number on anything.

Starkey: And as a PCB fabricator, if you could achieve a good bond between metal and resin with a super smooth copper, it would make the signal integrity calculations much more straightforward.

Gaudion: It would, yes. We wouldn't make as much money on modeling!

Starkey: Yes, but you wouldn't assemble many multilayer boards because they'd all fall apart in the oven. Everything's got to be a realistic compromise, considering an awful lot of factors and parameters that go into making the finished product.

Gaudion: And that's an interesting point. I was saying that I’ve been speaking to automotive suppliers, but also, for the first time, chemistry suppliers are coming in saying, "Polar, we need to speak to you about roughness, because we're developing a new process that will give smoother copper and we'd like to analyze its impact on signal integrity." So we've started to speak to different people that historically Polar wouldn't really be involved with.

Starkey: I come from a process background myself, and I think the chemistry suppliers have responded very positively to the input that you've made.

Gaudion: That's good.

Starkey: Because as you say, you can specify a foil but then what does the fabricator do to that foil in process? When he buys a thin laminate, the foil that comes as part of the laminate has got known characteristics, with a degree of roughening on the side corresponding to the resin. Then subsequently he has got to make the other surface bondable, so he puts it through some sort of chemical roughening process to give a key for his laminating process. In the past it was something that didn't take into account any signal integrity consideration.

But largely as a consequence of the presentations that Martyn and Neil have done, the process guys now recognize the significance of it, and it's precipitated the development of a whole new range of bonding treatments that don't result in a surface that causes any further complication of the already complicated subject of maintaining signal integrity.

Shaughnessy: So if you're working with the chemical companies then, you'll be able to test their chemicals using impedance testing?

Gaudion: Or we'll give them the tools to test, yes. And certainly some of the laminate suppliers are now using our systems and software to qualify their new materials. It's a very interesting time to be in the industry at the moment and it's good that everyone's busy as well.

Shaughnessy: Somebody was telling me that if they have their boards built in China, that they use shops that have Speedstack. Because people who have no English skills whatsoever can still use Speedstack. It's handy across the language barrier.

Gaudion: We've done a lot of work in terms of many of our products being available in traditional Chinese, simplified Chinese, they're available in German, and they’re available in English -  we provide our tools in a number of different languages.

Shaughnessy: Is there anything that you want to talk about that we haven't covered yet? Anything you're working on, or anything you want to talk about in the future?

Gaudion: This year, in terms of signal integrity, we've added the signal integrity tools and incorporated them within Speedstack. We've got a very rich materials library database, about 17 different base material suppliers that are available online within Speedstack, and now you can generate a high-speed stack up and actually specify the insertion loss from within the stack up tool and provide all the graphs for insertion loss, along with the materials you're going to use, the number of layers, and all the information the supplier needs. So that's a big, big improvement.

But I always remind people that when you talk about materials, sometimes designers have a myth of a perfect material library that if they could spec the material perfectly they wouldn't need to speak to the PCB fabricator. But just recalling what Pete said, actually they still do. So it's important that we all keep talking with each other.

Shaughnessy: They’d rather not talk to anybody (laughs).

Gaudion: Yes, that would be my summary.

Shaughnessy: Thank you, Martyn.

Starkey: Yes, thanks very much indeed for your time Martyn. It’s been a very meaningful discussion.

Gaudion: Thank you.

Visit I-007eBooks to download your copies today:

The Printed Circuit Designer’s Guide to…Secrets of High-Speed PCBs, Part 1

The Printed Circuit Designer’s Guide to…Secrets of High-Speed PCBs, Part 2

• < Previous Page