Issue link: https://iconnect007.uberflip.com/i/884055
26 The PCB Magazine • October 2017 transition to whatever the closest plane is, re- gardless of whether you have a return via. So you don't need ground vias to have this tran- sition of the return current occur, and it turns out that there is a very simple formula for what amounts to an inductance that the return cur- rent goes through when it transitions from one reference plane to another. This is a piece of knowledge that all signal integrity engineers should understand, and basically none of them do…You get it from closed-form equations. "Those are two immediate topics that I think need to be addressed, and the other thing we had better start thinking about is that we may be approaching a brick wall when it comes to throughput on serial channels. I'm going to take another chance here and say that there is going to come a limit, and we may be starting to get close to that limit." Mark Thompson had further thoughts on this: "We've been spending so much time over the last 10 years beating up the material man- ufacturers from the other side saying, 'Look, we need specific dielectric constants. We need spe- cific loss tangents.' Now we're at a time in the existence of electronics and circuit boards that we're beating up the copper foil manufacturers for surface roughness because we're literally at a stage where 25 gig is the key, and at that stage any surface tooth or surface roughness is an is- sue. It's strange that we've gone from a situation where we beat up material manufacturers to the point where we're beating up copper manufac- turers to get surface roughness down to a minimum." According to Steinberger, Rogers has observed that when they measured transmission loss- es for two different kinds of cop- per—rolled annealed and electro- deposited—they found that the roughness models we've been us- ing modeled the electrodepos- ited copper quite well but over- predicted the losses for the rolled annealed copper—by a lot. He added this: "I find this fascinating be- cause I've never been satisfied with the modeling of conductor losses. I see these bewildering models of conductor rough- ness and I ask, 'Where is the underlying phys- ics? I'm not really seeing it.' I learned a theory that says, 'You've got a very lossy dielectric lay- er next to your electrodeposited copper. That's why you're getting the transmission losses. Not specifically because of the actual physical shape of the copper surface itself.' Maybe people will start asking you for rolled copper instead of electrodeposited copper." Thompson noted that some of their cus- tomers have asked for RA copper, as opposed to the electrodeposited. He also commented, "Re- member when people were asking for things like an 11-degree offset on a panel to minimize having to deal with having a structure sitting over the top of a glass knuckle or weave in the material, or even things like a skin effect due to differential pairs that were dipped in deep gold that didn't have any solder mask over the top of them. It's not as prevalent as it was in the last ten years." According to Steinberger, that could be be- cause "people end up routing at different an- gles anyway so you don't have these really long straight lines—or you shouldn't have these long straight lines—that they had on the test boards where they were measuring these effects to be- gin with. I never was happy with that whole dis- cussion of weave effect. One reason was simply the emotional tone that got tied to the whole subject and the other is that people pointed to these weave effects, particularly the skew, which would be induced on a differential pair and talked about how you had to keep that differ- ential skew down. But you can accept a certain amount of skew. You get a little bit of degradation but as long as you keep the skew within, say, an eighth of a wave- length at the maximum frequen- cy of interest, the effects can be minimal." Sunny Patel added his per- spective at this point: "We make boards for differ- ent technologies, and we can control the impedance to nearly EXPERTS DISCUSSION: SIGNAL INTEGRITY AND IMPEDANCE CONTROL Mark Thompson