Issue link: https://iconnect007.uberflip.com/i/1276973
28 DESIGN007 MAGAZINE I AUGUST 2020 a number that you're going to want to have. Additionally, testing—not just electrical—but HIPOT, HAST, and all the other testing vari- ables should be done. If you have a require- ment for HAST testing, it's going to have to be on your drawing. We also need to know the mask, ID color, and type. We have various dielectrics for our colors of masks. Why is that? We can't feasibly set up every possible color with the automatic flood coater. Some of them have to be hand-screened by an operator, and that person doesn't generate the pressure that the machine has. Ultimately, we may call blue or purple or red something closer to two mils of dielectric thickness over the top of the glass and one mil over the top of the metal. Does that make a huge difference on an eight and eight? Not really, but does it make a huge difference on sub-four and four? Absolutely. When you get down to the real small geom- etries, that additional mil of dielectric thick- ness and dielectric constant change for the solder mask makes a huge difference for con- trolled impedance. Additionally, there are things like allow- able X-outs and impedances. If you have any impedances, either call it out on a README or directly on your drawing notes. Typically, it will say something like, "All five-mil lines on layers one, three, six, and eight to be 50 ohms." Now, one thing that I don't advise is chasing the impedances. We have a couple of our customers where if I send calculations, it will go back to the customer, and they will make those modifications down to the gnat's butt. They'll say, "If I say I need to change your 4-mil trace to a 4.25, I'm going to send data at 4.25." That doesn't mean that other fabricators would need that same change. If it's within 10%, don't mess with it at all. Remember that being the goal. As a designer, engineer, or lay- out person, they're just trying to stay within 10%. They give it to a fabricator, and we take it the next 10 miles. Then, there are cut-outs and radiuses. If you have a cut-out that is drawn or depicted as 90-degree intersections, we're going to need to know what the minimum radius is. Is it 0.032", 0.062", 0.093"? What's the minimum radius to be able to fit that device inside that cut-out? Those are just some of those kinds of things that I'm talking about. One thing that can eliminate the confusion between fabrica- tors and people that are generating fab notes is consulting your fabricator. Call us directly and say, "I'm working on some fab notes, and my board is heat-sensitive. It is controlled imped- ance, and I need this. I have a 0.8-millimeter ball grid array, and I need to ingress and egress out of that very tight pitch BGA." Now, at that stage, I have to ask, "Are you going to have long runs of very small lines and spaces for differential pairs that are ingressing and egressing out of that BGA? Or are you only necking down to a very small distance from a much larger differential size outside the BGA?" And the reason I say that is because I have a number of customers that will ask me, "What do I do to control the differential pairs that are inside that BGA?" If they're less than an inch and most components are an inch, two inches, in maximum size, it means noth- ing. I'd say, "Control those larger differential pairs that are running 12 inches on the outside of the BGA and not those that are inside the BGA." Take the hit of having it going inside that BGA. Mark Thompson