At PCB West recently, Wally Rhines gave one of the keynotes. It was titled Is Past Prologue? The Future of the PCB Design Industry . Wally said that it wasn't really his title, it was given to him by the show organizers who wanted him to do a retrospective on the PCB industry over the years, and a look to the future. Of course, Mentor is a part of Siemens these days (MentorASiemensCompany) so their outlook for PCB is in the context of all the other software and services that Siemens provides. When people want an example of the complexity of modern systems that involve lots of PCBs, interconnect, sensors, actuators and so on, they usually pick autonomous vehicles. Wally had another great example that ties back into IC design, namely an ASML EUV stepper. In the IC world, we worry about the impact of EUV lithography on IC design and manufacture, and in which process node it will be inserted (7nm it seems). But from the perspective of the PCB world, it has: 1200+ PCBs 100+ control units 4800 cable harnesses 5000+ sensors and actuators 10,000+ network and software signals Now that's a system! It makes an autonomous car look simple. Pre-History In the earliest days of PCB design, skilled layout draftsmen (and probably a few women) did that hard work. You too could "enjoy a real job in industry". The next step was to digitize the drawings, and that led to editing on a CRT (that's cathode-ray-tube to you young'uns, what we had before we had the sort of screens we have today). That was the beginning of PCB design automation. In the EDA industry, we like to think that our tools are so compelling that everyone adopts them because it is just a better way—but the reality is that people adopt them when the existing methodologies (doing it by hand, in this case) are no longer viable. PCB manufacturing also got more automated, so required designs to be captured in electronic form. 1980s In the 1980s, CAD transitioned to PCB schematic capture and layout. I don't know what it was like in the PCB world, in that era I was in the IC world. I would say that the old guard who were used to hand-drawing schematics using their plastic stencils never really made the transition to doing schematics on workstations. They retired and the young guard took over. It reminds me of Max Planck's quote that "science advances one funeral at a time." A lot of companies sprang into existence. The earliest like Calma and Racal-Redac, but also the DMV (the M was Wally's Mentor, although in that era he was still at Texas Instruments), Protel, PADS, CADI, Zuken and other names that you've probably forgotten. Several of these companies also had technology for IC design since the front end (schematic capture in particular) was the same, especially for designs using what became known as "ASIC methodology" where the system company did the schematic capture, and then an ASIC company like VLSI Technology or LSI Logic would do all the IC design magic at the back end. 1990s The 1990s were what Wally calls "the PCB design wars", with a huge number of companies. Consolidation started to happen. The big one, that shook the industry, was Cadence's purchase of Valid. But this did not go that well at first. As Joe Costello, Cadence's CEO at the time, said: Valid was clearly the worst acquisition...we inherited a product overlap and the decisions never really got made. Since every customer faced some of the products they used being killed off, they were all forced to re-evaluate what tools they used rather than just take the easy "keep calm and carry on" approach. As I mentioned in my post about Jaswinder , he was involved in a major program to merge the schematic capture systems for PCB and IC...which went exactly nowhere, and the two systems remain separate today, almost 30 years later. Mentor refocused on PCB in the late 1990s. This was not some smart strategic decision, it was, as Wally admits, due to the fiasco of 8.0 (late-point-slow), the Falcon Framework. The 1990s ended with the market split with Mentor, Cadence, and Racal-Redac all with about 20% market share, and the rest split among the smaller players. CCT and Integraph Two IBM alumni, David Chyan and John Cooper, had done two generations of shape-based routing at IBM, and they came to Mentor and created the router for BoardStation. They then left to create their own company, CCT, taking care to use a different programming language to ensure they were clean of Mentor IP. Mentor decided to OEM from them. In 1996, David and John came to Wally to say they were going to go public. But this was around the time that the stock market had figured 8.0 would never see the light of day and the stock price was down. So when David and John said that, since 85% of their revenue came from Mentor, they would forgo the IPO if Mentor would buy them for $60M, what they expected to get in the IPO. Mentor didn't have the money, so they let them go public and continued to OEM the product. At this point, Cadence and Joe Costello make their second major appearance in the story, when they purchase CCT for $480M. Remember that 85% of their revenue was PCB revenue from Mentor. But that OEM agreement expired after a year, and Cadence decided not to renew it. So Wally and Mentor are in a tough spot with no router. By then they had hired Henry Potts (who I worked with at VLSI Technology, and is now at Altium) and they concluded that the best router in the world belonged to Intergraph, who had picked up Daisix after the Cadenetix debacle (Daisy made a hostile takeover of Cadentix, and mismanaged the merger so badly that the company went bankrupt). Medlock, the CEO of Integraph, thought it was a great business but the board pressured him to focus and divest it. So on Tuesday at 6pm they called Wally and Will Herman, the CEO of Viewlogic. But Will was on a business trip. Wally called back that evening, and by Sunday morning they had a deal and cut a check for $20M. You won't be surprised to know that it is a challenge to get a cashier's check for $20M on a Sunday. On Monday morning, Will Herman was back from his business trip and called to say they would be interested. But the deal was already done. Mentor never lost a single customer in the transition and disaster was averted. But then, as Wally put it, "we made the same mistake as Cadence." Having bought Veribest from Intergraph, they owned ViewDraw (renamed DX Designer) which was the front end to much of Cadence'd tool. They decided to "do to Cadence what they did to us." A lot of companies like Dell and Compaq (this was the height of the PC boom) were pretty much completely ViewDraw front-end users, and Cadence back-end. But, as Wally said: We captured exactly zero of that business. 2000s By now the industry has consolidated down into how it looks today (except Siemens hasn't acquired Mentor yet) Whereas in the early days, PCB software was all about design (schematic capture, routing, simulation), the transition to high-speed serial connections means that analysis came to the fore. Instead of doing all the analysis at the end, a lot of the validation was moved earlier in the design cycle to reduce the design time, and to improve product quality by designing the signal integrity, power integrity, thermal, etc in from the beginning. The Landscape Today That brings us to the PCB market today with four players: Mentor, Cadence, Zuken, and Altium. Everything is being driven by the increased complexity of designs, with the line between IC design, package design, complex 3D packaging, and PCB becoming blurred. Design complexity has increased a lot over the last 25 years, as shown in the table below: And the complexity of the design process has increased, as in the table below. Wally asked the obvious question about this table "of the 91% who designed for signal integrity, only 85% did SI analysis...so what did that other 6% do, just hope?" Summary When we build a chip, we do everything in the computer. Manufacturing a prototype costs tens of millions of dollars and takes months, so everything is modeled and verified before tapeout. System design is going to become like that, with a "digital twin" that is created and analyzed completely before any system is constructed. In my recent post on the EE292A course at Stanford (see Experience Teaching Undergraduates EDA ) Patrick had a comparison of how complex a chip design is compared to a plane, and, in effect, how much more advanced the design approach is for chips with fewer designers, fewer months, more parts, and right first time. Designing systems is going to have to get like that because it is too slow and expensive to build systems as a way of testing them. Autonomous vehicles are a great example since testing a car requires billions of miles of driving. Waymo has done over 5M milies of actual driving on roads, but the only way to do billions of miles is virtually. Sign up for Sunday Brunch, the weekly Breakfast Bytes email.![]()
↧