Issue link: https://iconnect007.uberflip.com/i/1256432
20 DESIGN007 MAGAZINE I JUNE 2020 Matties: You're making a good case, though. If the app note is good, and it's not over-materi- alized or doesn't require the best of the materi- als, the vendor is likely to sell more parts. Beeker: The engineers don't want to admit that they're wrong. The most humbling moment in my entire career was five minutes into Rick's signal integrity class when I figured out that every design that I had ever done had worked by accident, not on purpose. I thought, "I hope my boss doesn't figure this out." Matties: How do we tell the people at the top that incorrect app notes matter and they need to change their approach? Beeker: A question to ask them is, "How often in your programs do you have to re-spin the circuit board to pass the EMC? And would you be willing to have your teams find solutions for this?" They have to think about the numbers. Some companies don't do new product devel- opment anymore because they can't afford to go through the multiple iterations; others are out of business now because they missed their market window and exceeded their engineering budget while trying to get the boards to pass EMC, but people sit and blindly re-spin the board using the same rules that didn't work the first time to do it again and hope it's going to pass. In the long run, the reliability of the prod- uct in the field is reduced, and that has to be in the hundreds of millions—if not billions of dol- lars—when you look across the entire industry. Matties: And if you look at an in-the-field prod- uct failure, that's a cost of design. Beeker: But they finally lock upon a combina- tion of band-aids and voodoo that allows them to pass EMC, and then a lot of times, that's how some of these guidelines were created. "It worked on this one. We're going to do it again, and it worked because it was only good on that particular bad architecture." Nobody is willing to start over on that second board, even if it blatantly doesn't work. Most of the time, when I do EMC analysis for my customers, I look at the board stack, and I'm done. Until they fix that, there's nothing else I can do to make it better. Hartley: Exactly. Beeker: Sometimes, for my analysis, it's five minutes, and I'm done. Then, it's, "How do I write this to tell them how ugly their baby is and not make them hate me?" Matties: But there's also a cost of lost opportu- nity because now your designer is busy redo- ing what they could've done right the first time instead of working on the new opportunity. Beeker: Right, instead of working on the new program. That's absolutely key. That's why I say it could be billions of dollars of lost reve- nue. It's a cascading impact. Shaughnessy: The app notes probably didn't matter much until the 1990s when the IC speeds got faster, did they? You could be wrong, and the thing would still work. Beeker: There was a time when everything was wrong. Hartley: Before the '90s, all circuits were a lumped length. The only advice the applica- tion engineer had to give was schematic ad- vice. If they got that right, then everybody was a happy camper. Then, along came distribut- ed elements. Once, speeds were fast enough that things were no longer lumped, and all of How often in your programs do you have to re-spin the circuit board to pass the EMC? And would you be willing to have your teams find solutions for this?