Issue link: https://iconnect007.uberflip.com/i/1210212
18 DESIGN007 MAGAZINE I FEBRUARY 2020 materials much better than anybody else did. We would sample lots of materials for an entire year because to create plus or minus three sig- ma on the dielectric constant and dissipation factor out to 25 gigahertz, even 35 years ago. Ritchey: You were lucky; you had a pretty spe- cial environment. Holden: We had standard materials, and we had to show the engineers that the dielectric constant is not going to be 4.4. Maybe once in its life, it's going to change lot to lot. Here are the numbers, plus or minus three sigma, not over frequency, but also over temperature and over humidity as a design file book. These are the standard materials they had to use. They couldn't go out and pick any old material they wanted unless it was thoroughly charac- terized. Ritchey: You were lucky enough to work with a good engineering company. Holden: Yes. Making test equipment, it has to be 10 times better than what it's measuring. I was trained by the best because the test equip- ment had to be so sensitive, especially the way to deal with noise was sometimes very unique. Ritchey: I was doing that as a college kid at Tektronix. At that time, Tek and HP were the places to work, and it was good, engineering- wise. Holden: All of my classmates went to work for Tek because I came from Oregon. I was the only traitor that went south to Hewlett-Pack- ard. Ritchey: Did you get run out of town? Holden: They kind of disowned me. Shaughnessy: As Dan Beeker says, he doesn't even run signal integrity analysis. He said, "Design the board right." Is it that simple? Ritchey: I agree with that. Holden: With the semiconductor roadmap, they're shipping 7-nanometer geometries, and they're prototyping 4- and 5-nanometer geom- etries; these small transistors turn on and off fast. That rise and fall time is what you have to deal with, and because of the shrinking voltag- es that power these things, they've had to in- crease the number of ground pins. Ritchey: Both ground and power. Holden: The power ground is increasing enor- mously, and you have to distribute it because a critical length is based on that rise and fall time. They don't want to have to pay for trans- mission lines, so they use up the entire criti- cal length, and the only way to do that as your transistor shrinks is to shrink your pitch—es- pecially when you have more pins now to put on the package. How they allocate their pins can be an enormous headache. Once you attach to it, you have an EMI or signal integrity problem, even if you try to do everything right. They've made the job diffi- cult. Other companies understand that some- body who has to use the IC is going to have to wire it on a circuit board, and if they have too many problems to fight, they're not going to buy that component anymore. They'll check to see if somebody else has a component like it or do it the old way and use three devices instead of the new single device. It's all physics. Ritchey: You're going down the right path. In the latest design John Zasio and I did, we had an IC where the operating voltage was 0.9 volts, and the current was 160 amps. Getting that current into that product without having excessive DC voltage drops turned out to be the hardest job we had to do. One-hundred- sixty amps may not mean anything to you, but let me put it into perspective. Your car requires 60 amps to start it. How big is the cable going from your battery? For what John and I do, our biggest problem is power delivery. It has be- come the hardest part of the job, getting the high currents into the part. If you don't have enough pins, no matter what you do, you can't make it work.