Issue link: https://iconnect007.uberflip.com/i/884055
28 The PCB Magazine • October 2017 predicted. They learned that a change had been made in the supplier's manufacturing process at one location which resulted in a slightly differ- ent dielectric thickness. "That was a problem for us, because they were sending prepreg and core materials from two sources. How could we possibly predict im- pedances for a customer when we couldn't even predict the dielectric impedances of the pre- pregs that were being used?" Thompson mentioned a couple of other challenges, such as when a customer makes a change but the revision does not provide for a reference plane—"they've inadvertently placed a bunch of differential pairs over split sections, and you don't have a proper reference plane." He also mentioned that some things, such as lengths cannot be controlled; lengths of less than 0.300 mils are very difficult to control. Getting to the fab side of things, our panel discussed the challenges specifically for manu- facturing including variation in the etching and plating areas, as well as techniques to minimize any signal losses in a Class 3 mil/aero world. Ac- cording to Thompson, blind, buried or epoxy- filled vias may require an additional wrap plate which could affect impedances. At this point, Steinberger expressed concern, referring to a time, 10−15 years ago, when non- functional pads were removed, but today, the dense board geometries mean this can be risky. "I start to wonder more about pad sizes, backdrill precision, and the fact that I can't get big enough anti-pads on my vias to get the impedance I want… From my end of it, I'd like to re- move all the pads. If I could get away with a pad diameter equal to the drill diameter at my rout- ing layer, where I did want to connect to the trace, I'd do it." Sunny also had thoughts on this. "Technology in general has yet to catch up to what we re- quire. The new types of lay-up, back straights, calculations, al- gorithms—are all trying to get back to what you're saying, to remove those non-functional pads and get that alignment to where we need it to be for perfect manufacturing. We think that probably helps with the impedance con- trol side of things, though there is more testing to be done." Yogen added, "But by using the positive-act- ing electrodeposited photoresist, we can consis- tently make landless vias, so you don't need a pad. Also, you mentioned wrap plating, which you wouldn't need. Once you have proper plat- ing in the hole and on the surface, why would you need a wrap? We should concentrate on cap plating." Steinberger then raised a question: "Is there any conversation about eliminating the via stubs for extremely high-frequency signals? For example, are people going to multiple lamina- tions so that they can keep vias confined to sub- laminates, or are designers driving you to get the absolute minimum via stub length on a back-drilled via, stuff like that? Part of the rea- son for my question in general is—at least in the design spaces I do happen to have experi- ence in—it's all about the vias." In response, Yogen mentioned one of his customers, who had asked for back-drilling right to the layer, leaving very minimum stub, with Sunny adding that they still have the sub- drilling customers at the same time, which re- sults in a "mixed bag." Sunny feels that there is no clear direction as to which way is more pop- ular on a design side, but stated that sub-lami- nate drilling is more expensive than back-drill- ing. Thompson agreed that hav- ing to do either blind, buried, and/or back-drilled vias will cre- ate far costlier scenarios. Continuing the subject of vias, Steinberger said, "The oth- er thing that people aren't aware of is that, if you have even a rel- atively small back-drill stub, there is a fringing capacitance at the end of that stub which is quite significant compared with the total capacitance of the stub as transmission line. You real- ly have to back-drill the stub to something smaller than the pad diameter in order to start really EXPERTS DISCUSSION: SIGNAL INTEGRITY AND IMPEDANCE CONTROL Yogen Patel