The PCB Design Magazine

PCBD-Aug2017

Issue link: http://iconnect007.uberflip.com/i/860275

Contents of this Issue

Navigation

Page 21 of 73

22 The PCB Design Magazine • August 2017 OK, so I've hit the designers and the fab shops, and now it's the engineers' turn in the barrel. Yes, despite what some will tell you, en- gineers don't know everything about PCB de- sign. For example, I once had a simple require- ment for a controlled impedance board that needed 65 ohms for a PCI bus and 50 ohms for everything else. Well, it's not always that easy when you are trying to keep a minimum 5/5 on the design and the board is packed with parts. Since the 65 ohms was only needed on the out- er layers, the internals were easy to hit 50 ohms. Well sort of. I went around and around with the fab shop working on a stackup that would give me those golden impedance numbers, but alas, I just couldn't get 50 ohms. About as close as I could get with this really densely packed board was about 56 ohms routed with an approximately .012" trace. On that board, it was like building a freeway through a residential area without moving a house. I checked with the engineer each time I got new numbers and he said that 58 ohms is fine. But he said he really wanted it as close to 50 ohms as I could get. Yep, I ripped it up and re-routed it with 56 ohms. After the second set of numbers and rip- ping up both internal and external routing, I asked the engineer why he needed the 50 ohms so badly. After all, the board was mostly ana- log and the digital stuff wasn't fast, so what's driving the 50 ohms? As a former engineering technician, I really wanted to understand this design requirement. His answer was, "Well, I just wanted to be sure all of the layers were the same impedance." At that point, I was standing there with a puzzled look on my face and really hoping I wouldn't get the answer I was going to get, but I asked anyway. "If all you want is for all the impedances to be the same, could they all be 65 ohms?" His answer could have got me thrown in jail had I picked up the chair close by, but he said, "Sure, as long as it's the same on all the layers. That's fine." I re-routed for the third and last time, the board was all 65 ohms on all lay- ers and all 5/5 spaces and traces. Case closed. I'm not really sure how to score that one. Now, after reading over these examples, it's easy to see that everyone can play a part in the ENOUGH BLAME TO GO AROUND bit and, you guessed it, the dimension didn't update. Once I got the panel from the shop, I proudly took it up to the checker and exclaimed, "See, this is what happens when you override a dimension!" His answer was, "I just said to re- move them if you could, not that you had to." So, whose fault was that? His for requesting that I remove them, mine for doing it when I knew I shouldn't, or me for not remembering I did it or maybe not providing a compelling reason not to? After that, I had a good reason for leaving the leading zeroes alone, no matter what the di- mensions are. Lesson learned. Many years later and just as many lessons, I found myself doing design, libraries, schematic work and now serving as an incoming inspector for all bare PCBs. Hey, if I designed it, I should have a good idea what the PCB should look like when it hits the door, right? So, I get some boards in that have little perfectly curved cres- cent "mouse bite" shapes in the pads and traces all over the board. This is where IPC's Tech Net really shines and I recommend everyone to sign up for it. I had never seen anything like that be- fore, so I posted a question to the forum. About a day later, I got a reply. It seems that bubbles in the etching tank can stick to the copper if not rinsed or knocked off by agitators in the tanks. After a couple of exchanges of related questions, I get a call from the QA manager of the com- pany that did our boards. Turns out he was the guy on Tech Net who told me about this issue, and when he put my questions to the problems he found in his own shop, a light went on in his head and he knew that this was the culprit. The boards were replaced and never should have left the shop. Designer 1, fab shop 0, but who's keeping score? " Hey, if I designed it, I should have a good idea what the PCB should look like when it hits the door, right? "

Articles in this issue

Archives of this issue

view archives of The PCB Design Magazine - PCBD-Aug2017