Issue link: https://iconnect007.uberflip.com/i/1276973
30 DESIGN007 MAGAZINE I AUGUST 2020 the material you were selecting before you did it. Or you didn't bother to test the extremes in your simulation and things like that, which I've always faulted the EDA tools. EDA tools let us put in a dielectric constant, and then those same EDA tools will come up with a thermal analyzer to ensure the differ- ence in temperature on the board. And yet the impedance calculator and field software they have only allows one dielectric constant and no allow-ability to put in a slope between the temperatures that you just told me the boards are going to see anyway. That dielectric con- stant dissipation factor's not going to be the same at those different tem- peratures, as well as frequen- cies. Yes, we understand temperature gradience, but somehow, nobody's ever accepted the fact that there is an electrical performance gradient. They consider mate- rials to be constants or absolutely flat lines; it's not going to change temperatures, and it's not going to change the frequency or anything else. And they do all of their signal integrity and impedance analysis based on this assumption about mate- rial. And with FR-4, that couldn't be further from the truth. You can get some really expen- sive material that's relatively flat, but it's really expensive. I keep asking EDA people that ques- tion, and they just shake their heads and say, "Nobody's asked for it." Thompson: Happy brings up another good point there. We have two facilities: Redmond, Washington, and Tucson, Arizona. And you can imagine that Washington is moisture- laden. It's rainy most of the time and very cool, which is vastly different from Tucson, where it is extremely hot and dry. The press parameters and temperatures are vastly different between our two facilities. Therefore, the effective dielectric constants of the materials we use vary. And that's all over the map. Recently, I've been gearing over to helping with Tucson more. I use Tucson's effective dielectric con- stants on their materials and their dielectric press values, which are different than what we use here in Redmond. Barry Matties: Pretty much every job that comes in has data that's inaccurate or missing. Thompson: Absolutely. Even having gone through sending a customer effective dielectric constants and calculations and dielectric builds, it does not preclude the customer from c h a n g i n g things and submitting data that has vastly changed. They may say, "Five mil lines for 50 ohms." We get the data, and now they're eight mil lines. And the dielectrics are vastly dif- ferent. Not only does that happen, but we also get people asking for dif- ferent colors. Originally, it was calculated as green, but now it's going to be calculated as black—or worse yet, white. In fact, all of the material colors—from green to blue, red, purple, and black—are all an approximation of 3.9 at 1 GHz. The white, for whatever reason, is 6.3. It's virtually dou- ble the dielectric constant. And why is that? Years ago, when people were doing more whiteboards for RF type considerations, they had problems where they were covering what- ever the substrate was, or the surface finish was. Matties: When you get the fab note, though, part of the conversation we've had with you and others is there's no incentive for them to send you a package that's 100% accurate or complete because the fabricators will do the work. And if you don't, the customer will just go and find another fabricator that will.