Issue link: https://iconnect007.uberflip.com/i/1355791
REAL TIME WITH... IPC APEX EXPO 2021 SHOW & TELL MAGAZINE I I-CONNECT007 105 reduction is more comprehensive. It affects every signal layer in the board. Johnson: ere more chances to improve a lit- tle bit when you're optimizing misregistration and then pretty big gains in specific ways with antipads depending upon how your board is implemented. I've got to tell you, at the start I expected misregistration to deliver the more benefit than the antipads. What's the effect achieved by reducing the antipad to reduce crosstalk? Czaplewski: It's similar to the misregistration effect, but it's maybe a little bit harder to visu- alize. You can think of it this way: there is a sig- nal layer with reference layers above and below it. e reference layers have antipads around the PTHs. e bigger that antipad is, the more open space for crosstalk to occur. ere is less ground coverage of the signal layer. Basically, the larger the antipad, the more opportunity for the signal layers to be exposed to each other and therefore have crosstalk. It's the same situation when you have misregistration, layers are shi- ing, exposing the traces to each other. Having more ground coverage between signal lay- ers is beneficial for crosstalk reduction. Happy Holden: Yes, antipads are kind of the Rodney Dan- gerfield of design rules; they get no respect. People ignore it and, in fact, they want to increase it without real- izing the effect that increasing the antipad, and reducing the current return path and power, has on the multilayer. As pitch goes down and you're going to use mechanical drilling and layer-to-layer misregistration, it's tougher and tougher to keep from hitting that power or ground layer. e tendency is to try to make the antipad larger. But that has the potential to change capacitive effects to other parts of the circuit. It's juggling all these balls and that's what I think the whole paper is about: trying to get performance, reliability, and cost with shortening rise times and sensitivity, signal integrity is a tough question. Johnson: Right. And how did this impact the yields? Was there a yield cost to this? Czaplewski: We polled several PCB manufactur- ers about the impact of reduced layer-to-layer misregistration and reduced antipad diameter on yield and cost. Based on collected registra- tion data, we found that yield really varied by supplier. ree PCB manufacturers had little to no yield impact and could meet the reduced misregistration requirement. But then there were three other PCB manufacturers that would have significant yield loss and obviously cost upli with reduced allowable misregistration. We also found that reducing the antipad diam- eter around backdrilled holes too much could also reduce yield due to broken drill bits from drilling through more copper layers. Since we did this investigation, some of the manufacturers have made significant improve- ments in controlling layer-to- layer misregistration. I think that presenting this idea to the industry is going to drive improvements, and the four-mil maximum layer-to- layer misregistration should be attainable. en there may even be a desire to drive to even lower mis- registration allowances. Johnson: It sounds, then, like some of your suppliers learned something and were able to make some improvements. Czaplewski: We saw significant improvements from one supplier, in particular, which ini- tially estimated a 50–70% yield loss at four mil max layer-to-layer misregistration. Two