Lee Ritchey On the Future of PCB Design
I spoke with long-time PCB design instructor and Right the First Time author Lee Ritchey during the recent AltiumLive event in Germany. Lee and I discussed everything from the advice he offers to young PCB designers to his thoughts on where the industry is going and what’s missing from today’s electrical engineering curriculum.
Barry Matties: Lee, you’re the keynote speaker here at the AltiumLive event in Munich, Germany. It’s also your first keynote. I'm surprised.
Lee Ritchey: I've been thinking about what I presented in this keynote for a long time. Where are we? Where are we going? Dozens of people have asked me those questions, and I've finally pulled it all together in one place. And a fab shop does not want to hear most of what I mentioned. Neither does a laminate supplier, because we're making their market smaller with integrated circuits. Every year we make it smaller, and probably the best example I can think of is the Xbox 360, which is a four-layer board about 10 inches on a side and it is a super computer. Literally a super computer, so that a kid can play a game and I can't even guess at how many MIPS it has; I don't think that you can buy a stand-alone computer that big, and we call it a game. Not unlike your phone.
Matties: You had a chart in your presentation showing the phone in the center and all the items that it's replaced over time around it. It's a video recorder, it's a radio, and it’s so much more than a phone.
Ritchey: It's four radios. You've got Bluetooth, you've got wireless, you've got GPS, and, oh by the way, you also have a phone. It's all there.
Matties: It's interesting…you also said when you looked ahead 20 years ago that you would have never imagined where we are today. I find that kind of surprising.
Ritchey: Let's go back a little bit. No one expected workstations to be what they were. That wasn't what it started out to be. You may not know that, but we had timeshare, if you remember? And all Andy Bechtolsheim was doing was putting a microprocessor in his dumb terminal so that some of the data didn't have to go to mainframe. Well, it wound up eliminating the mainframe. The PC is the same way. You may not know this, but IBM did not set out to make a PC. I was doing IC testing at the time and when we made a tester for an IC, we had a fixed test pattern. When the chips are simple, you can hard code the test pattern, but when they get more complex that's not true. So IBM came out to California to find a processor that could be a programmable tester for their IC factory. No PC intended at all, and once it was there, their club had been putting word processors and that sort of thing on other products and somebody decided, "Well, heck, we got this microprocessor here. Why don't we put a spreadsheet on it?" The rest is history. It was not intended to be a personal computer at all and you can point to all kinds of products.
Matties: They didn't really think there was a large market for the personal computer.
Ritchey: Not at all, not at all.
Matties: I think there is a famous quote that said they saw there was a market of maybe one.
Ritchey: That was the founder of DEC computers. I have a book called “Accidental Empires” [by Robert X. Cringely] that describes this. For example, Bill Gates didn't start out to make Microsoft. He was selling code to IBM to run the testers, and almost everything that made it to market started the same way. The originator did not have that in mind at all, and that's why I can't forecast the future, because nobody has done that right.
Matties: You also mentioned that the numbers of designers, shops and jobs have dropped because of the density of a board and consolidating into such a small package with so much functionality.
Ritchey: Well, it's the integrated circuit. If you tear your phone apart, you'll discover there's one chip in there. It has those four radios and everything else.
Matties: All in that one chip with how many transistors?
Ritchey: I would guess that there's probably eight or 10 billion.
Matties: In my pocket! Most people don’t think in those terms. What you were saying is there was less of a need because we're designing fewer boards, but on the other hand because of the expanded functionality, there’s a lot of new opportunity for products to enter the marketplace. And it seems like everything has electronics now that years ago just wasn't even practical. Down to tennis shoes.
Ritchey: That's true. There are new products that somebody thought of that you and I didn't that catch on to. I guess the place where I'm sort of dumbfounded is cars. In my car, 50% of the cost is electronics. It has about 40 microprocessors in it, but it's just a car. It has a network in it, and you can point to a lot of things like that, that somebody figured out. I was just talking to a lady a little while ago who said that the high-end cars have 10 or 12 radars on top of everything else, like collision-avoidance radars. My car beeps at me every time I get too close at something. I like that, by the way. I don't dent as many fenders.
Matties: To my point though, it just seems like there's a greater demand for more and more boards every time I look. Because when you look at autonomous cars, and as you're mentioning the number of electronics on board, it just seems to me that there is a greater demand for more boards than we've ever had.
Ritchey: I'm not sure that's true. We may be getting back to where we were when we took 10 boards to make a system, but the way I see it is all the boards are smaller and simpler. I don't need 24 layers anymore.
Matties: Right, so maybe we're not building as many panels, but they're so small, we can get 12x or so on a panel today.
Ritchey: Yes, that's right. And of course that's bad news for the laminate manufacturers, right? On top of that, especially all these differential lengths are equipped with all manner of compensation so that you can make something work at a very high data rate on a fairly poor laminate or cheap laminate. Right now, the board I was talking about where we had 28 gigabit links, that's not even on a low-loss laminate. It's the stuff you call FR-4. Because the silicon got so good, now we're going to push that some more and find out what that brings, but as I said, none of us guessed right when we tried to guess where the limits are or what the next product is going to be. If I could, I wouldn't be here. I'd have so much money from investing in my startups that I wouldn't work anymore. No, I would work, because I don't like sitting around, but I've missed almost every one of those ideas that exploded. I've missed it because I didn't believe it would be a product. I was offered stock in Sun. “Workstations aren't going to go anywhere.” I was offered founders stock in Intel. I said, "CMOS are not going to go anywhere." I hate to tell you how many of those I did.
Matties: Well those two are enough, I get the idea (laughs).
Ritchey: Bob Noyce offered me 10,000 shares of founders stock to be his first test engineering manager.
Matties: Why'd you turn it down?
Ritchey: Well, at the time CMOS was very difficult to make work. It was really slow.
Matties: You felt there was an alternative path?
Ritchey: We were building stuff out of ECL. ECL was the hot, fast stuff, but now CMOS is 10 times faster than ECL. Where we could've never made a 15 peak per second rise time with ECL, we have those for CMOS right now.
Matties: Let's move on to designers. People were asking you what advice you would give to young designers. What would you tell them?
Ritchey: I know because of the speed that things are going and how we're going up the speed curve that every board designer is going to have to deal with five or 10 gigabits per second, and you can't do that if you don't understand how fields and waves work. All those guys need to get on top of that, partly so that they don't get misled by people who are not telling them the right thing.
Matties: And there's a lot of that in the industry, isn't there?
Ritchey: Most of it is.
Matties: One of the questions that was posed in your keynote was about the difference between digital and analog and your response was...
Ritchey: It's all analog.
Matties: Why is that a mystery or even a question to people? What's the mindset that would make them think that?
Ritchey: Here's a bit of history. When we got CMOS to run fast enough to do decent computing, the whole industry rewarded computer science majors. Namely, these were the people who write code, because it didn't much matter how you hooked the parts up. So that got only TTL or whatever, and now we have a whole population of engineers who are good computer scientists who now must cope with the fact that there are fields and waves and they don't know what they are. I spend most of my time on that, because when things were slow we could be sloppy. We didn't spend any time on signal integrity. It all was code, and then the hard part of that is now all the senior management are these computer scientists who don't understand any of it and won't support their engineers to get tools or training, and the companies that don't get that are going away.
But we rewarded 30 years of computer science majors at the expense of real electrical engineering. You can get an electrical engineering degree at my university, Cal State University at Sacramento, without ever taking a fields and waves course, which was the most fundamental course in electrical engineering. How is that possible? What happened was, when the focus came to computer science, you can only have so many units per semester, so which course would you toss out? And that one got made an elective. It shouldn't have been, but it got made an elective.
Matties: Is there a shift going on now where we're seeing that coming back as a requirement, as a fundamental?
Ritchey: Yeah, in fact I'm spending probably one day a month at my university getting the professors up to speed. They don't know fields and waves.
Matties: Someone's got to teach them, right? Well, they had no need to know.
Ritchey: No, and so we are having to put that back in the curriculum.
Matties: How is it going teaching your professors? Are they receptive?
Ritchey: Oh, yeah. They want to teach the right stuff. The biggest problem, and it's not just my university, it's all universities: They have no contact with the real world. The students don't have anything practical when they leave.
Matties: Why is that?
Ritchey: Well, because almost all professors go from being a PhD student to an instructor without ever going out in the field.
Matties: It seems to me there is a large population like yourself, who have the knowledge and real-world experience, who could go back in and do more of what you're doing.
Ritchey: That's what I'm doing is taking it back, but the best schools are the ones that have work study programs. Because then the students get to go out in the real world. The other thing about universities is text books and professors are always 10 years behind the industry.
Matties: Aren’t we seeing more major companies creating their own universities, if you will?
Ritchey: Yes, Qualcomm has a huge one.
Matties: Doesn't that make more sense than going through the traditional path?
Ritchey: They're doing that out of necessity. They would rather not, but the universities still have not got it right. I don't think they will. Qualcomm has a university, Cisco has a university, and IBM always had one.
Matties: Are they teaching the right things?
Ritchey: Yes. If you go back to when I graduated from university, companies had programs and the nickname was “kiddie engineers.” For example, at Tektronix, in your first six months you worked on a small project where you did everything all over the company. So you knew what purchasing had to do and all that business, and you were tutored by a senior engineer. That all went away when Silicon Valley exploded, because we did not have time to train anybody. You had to hire somebody who already knew it and so that all went away. Now it's coming back. But Motorola had kiddie engineers and Tektronix and HP did as well.
Matties: There have been a lot of advancements in EDA tools also. You are seeing a lot more thinking embedded in the tools.
Ritchey: Well, actually that's not quite true.
Matties: Really?
Ritchey: I like to use an example. You may remember Amdahl computers? That was the first computer that competed directly with IBM and it was in Silicon Valley. I was there, and we had a press conference. Lots of fanfare and this was a big computer and all that business and here was a question that was asked in the Q&A. This reporter asked, "Mr. Amdahl, aren't you afraid you're making computers so smart they'll replace thinking?" He said, "What you don't understand, ma'am, is what we have here is an exceedingly fast idiot." This is what all the tools are. There's no smarts in the tools. They let you do analysis quicker, but if you don't have the technical skill, you get garbage in, garbage out. The advantage of these tools for me is that I can do a transient analysis faster.
Matties: Right. It's all about speed.
Ritchey: Yeah.
Matties: We were watching the AltiumLive presentation today, and the message was about speed, and also that the controls, the interface, is more intuitive as well.
Ritchey: Let me illustrate this. When we did the Amdahl machine, we had IBM 1130s, which are .1 MIP with 16k of memory. When they did a transient analysis in keypunch cards, I had a deck of those about an inch thick that I put in the machine. I waited about seven or eight hours and out the other end came another stack and a hand-plotted wave form. Well, all these things have done is make that happen faster, but that's what we had to work with. We still did the same thing with computers; it just took a long, long time and before we had computers we built test structures and measured them. That's why you see me with lots of test boards.
Matties: It seems to me that the wisdom of thousands of years of collective experience are embedded into these tools and that experience has to come across in not necessarily replacing the human interpretation or thinking, but certainly it gives you a point to start from that is by far more advanced that we've ever had.
Ritchey: What the tool does for you is error checking very quickly. You have to put the rules in and it's an exceedingly fast idiot that follows your rules really fast.
Matties: Let's change gears a little. You mentioned your course several times in the keynote. Tell me a little bit about your course, what people learn and what they can expect from it.
Ritchey: I mentioned earlier that most engineers are computer science majors and they don't have the fundamentals. That's where we start. We spend half a day on fundamentals, like what are fields and waves? What are transmission lines? What matters in a transmission line and then we go on from there, and when we are done, they've got all the tools and the information they need to go do a high-speed design. Power delivery engineering, which I mentioned is now the toughest part of design. Surprising how fast that happened and how quick we got to that. Two things happened: The day we started doing differential signaling replacing parallel busses the routing problem diminished dramatically, and because we have more transistors the power requirement shot up. Here's how easy differential pairs are. The terminators are inside the receivers, so all you have to do is put two wires on the board and go have a beer.
Matties: Sound like a good plan.
Ritchey: And they're self-timed. The clock is embedded in the data, so you don't even have to length-match two paths to each other. All that stuff that used to be really tough in laying out boards is real simple now. The flip side of that is power delivery has gotten painfully difficult. I think I mentioned that the latest design John Zasio and I did took two days to make all the SI rules, and over a month to get power delivery right. How's that for a flip?
Matties: What was the greatest challenge in that?
Ritchey: Well, you’ve got to make sure that every copper path from these high-current point of loads, 60 to 80 amp point of loads into the chip, has a low enough DC voltage drop that you don't get in trouble. The new one is 160 amps.
Matties: That's quite a load.
Ritchey: We think maybe we're going to go fishing. If we get that, we have a new problem. How do we get rid of the heat?
Matties: Well, that's the big issue now, isn't it? The thermal abatement.
Ritchey: Yeah, and this chip. To give you an idea of the heat density, if you turn an electric range on so that it's glowing red, that's the heat density on the chip. How do I keep that from going above 125°C? We use heat pipes, which is effectively water cooling, because you can't get rid of the heat with air and that's why I like that 3D bit that's in there. That lets me figure out how to fit my heat sinks in. I'm getting close to done.
Matties: So what are you going to do when you retire? I know you were in wines for quite a while and I think you've moved out of that.
Ritchey: I lost enough money in wine.
Matties: Sure, if you want to make a small fortune in wine, start with a large one.
Ritchey: We did exactly that. So John Zasio is my co-author, and he was my partner in the winery, and my son was the wine maker. And if you're an owner, you get the owner's free wine. When we liquidated, we did the math and we figured out our free wine was about $100 a bottle.
Matties: But it was tasty.
Ritchey: Oh, it was very good. So what am I going to do when I retire? I don't think I'm going to retire. I'm just kidding.
Matties: Just keep your teaching going.
Ritchey: Yes. My answer, and I said this to my kids, is that I’ll quit when the phone stops ringing. Both John and I retired once and it was real boring. Really, really boring. Part of the problem is everybody we liked to do things with was at work. And the real truth is, I don't think people really want to quit working. They just want to have a choice of when they work.
Matties: Buy their time back, exactly.
Ritchey: You pick and choose. I'm in the position when I can say, "No." And it doesn't cause me to miss my rent payment, and that's what you want to do, right?
Matties: It's great catching up with you. Is there anything that we haven't talked about that you feel like we should share with the industry?
Ritchey: I'll probably think of that half an hour from now when we're not talking.
Matties: Well, if you do, we'll just pick up where we left off. Otherwise, thank you very much Lee. It's always a pleasure to talk with you.
Ritchey: My pleasure too.