Issue link: https://iconnect007.uberflip.com/i/1401769
AUGUST 2021 I PCB007 MAGAZINE 17 validate a board when we ship it out the back door. How do we validate it? We validate with a TDR, a fundamental frequency around two gigahertz, plus or minus." We would tell customers, "You want the im- pedance at 64 gigahertz or 16 gigahertz? Go calculate it. I can only guarantee two giga- hertz." And it's a broadband measurement, not a single frequency test. Every stackup report had a note at the very bottom of the picture, in- dicating what test frequency was used to calcu- late the reported impedance. Of course, some companies have VNA testing, but that's for very low volumes because it's a slow test. But we would just get ourselves out of discussions saying, "Sorry, I can't guarantee any other fre- quency than my TDR." ere are designers that get mad at you be- cause some other fabricator asked more than you did. I had to fill out a corrective action for a customer because two fabricators asked three more questions than my team did. ey want- ed to know "Why didn't you ask these ques- tions?" True story. In Asia, the customer's al- ways right. It's the way they're trained, and so they don't want to get a customer mad by ask- ing too many questions. ey don't tend to challenge hierarchies and the people who pay your bills. Holden: It takes many, many years to really un- derstand PC board materials, and even then it's a moving target. My group oversaw the design manual that included material charac- terization, and we used to get constant com- plaints because it would take us six months to characterize the material because we would take samples from the lab every two weeks. It also included a second chart of temperature variations of the frequency of the signal and the effect of humidity on dielectric constant dissipation factors. Because too many times we had to troubleshoot some HP test gear that we couldn't operate in Louisiana in Au- gust, whereas the rest of the time, it seemed to work perfectly fine. Engineering never consid- ered that higher humidity will affect the loss on their part for laminates. Thompson: To Happy's point, as far as the tem- perature and humidity, at Prototron we had two facilities, one in Tucson, Arizona, and one in Seattle. Well, you can imagine the humidity in Seattle is such, but humidity in Tucson was vastly different, and trying to keep parity be- tween those two shops was crazy. It was diffi- cult for us, not to mention you not only have humidity, but you have individual press pa- rameters from fabricator to fabricator. So, the press thickness of a particular ply of 1080 of P370 may be this at one for the fabricator, and then something else at a different fabricator, and yet something again at another fabricator. Holden: at's not counting each particular in- nerlayer, or how much copper they etched off. Thompson: Exactly. e nesting value based on the layers, the layer configuration. Is it a pure signal? Is it a split plane or is it a full plane? All those have different nesting values of that pre- preg. Once it heats up and fills those voids in a signal layer, it's going to fill the maximum amount. Korf: Just basics like impedance assumes a loss- less line. If the TDR trace comes out flat, that's impedance. When you get down to 2.5 mil wide lines, the second that TDR hits the trace, Well, you can imagine the humidity in Seattle is such, but humidity in Tucson was vastly different, and trying to keep parity between those two shops was crazy.