Reading time ( words)
I-Connect007 recently spoke with Lee Ritchey about the subject of continuous improvement with a focus on DFM, specifically looking at the benefits of reducing the number of respins by just one. A longtime instructor and one of the authors of Right the First Time: A Practical Handbook on High-Speed PCB and System Design, Lee has spent decades preaching the value of solid DFM practices, so we asked him to discuss why so many OEMs accept multiple respins with each design project and what designers could do to eliminate just one spin. He also shares some of the lessons in reducing respins that he learned in the early days of Silicon Valley.
Andy Shaughnessy: We’ve been talking about making 2021 the year of continuous improvement, and we came up with an idea. What if we eliminated just one spin from every cycle and kept that as a goal? Of course, your book Right the First Time came to the top of our conversation. Many companies continue to build respins into the project budget, but what about the wasted hours?
Lee Ritchey: You picked a topic that I actually know something about. I have stories. Some of them are horror stories! But if you have one less respin than the competition, you’re at the market first. That has been historically the business model of Intel and IBM. Of course, Intel and AMD have roughly the same kind of processors, with one exception: AMD has never been first to market. Originally, that was because they had to spin the silicon a couple of times. The people at Intel realized the person who’s first to market gets to define the playing field, and they spent huge amounts of money on simulation tools.
One of my favorite stories is the first Pentium processor. It was the first time a processor had been shipped on the first artwork. They got a lot of press on that. During the press conference, somebody asked how they did it, and the project manager said, “We didn’t cut silicon until we got a DOS prompt on the emulator.” They spent the money on upfront design, and that’s sort of defined that field ever since.
There were two where I have really clear experience. One of them was with 3Com, where they would assume they were going to spin the design and rush to get a prototype. The usual reason was so that the engineers had something to work on, and almost never was that something right.
I did a brief piece on this, which wasn’t even a respin; it had to do with how much money we spent to accelerate getting the prototype of a PCB. In the case I was working on there, the difference between a three-day turn and a 10-day turn was only about $5,000, and they wouldn’t spend it.
One evening, we were doing the budget for the next quarter, and it came out to a little bit less than $2 million a week to run the project. I told management, “If I spend $5,000 and take a week out of this budget, doesn’t that make good sense for our return on investment?” Finally, someone had quantified the cost of time for them because no one had done that.
And that’s not even a respin; that’s just being smart about where you spend premiums. Now, that gives you a yardstick for how much a respin costs. If you’re lucky, and a respin only takes six weeks and $2 million, isn’t it worth spending another week getting it right? I think so.
To read this entire interview, which appeared in the December 2020 issue of Design007 Magazine, click here.