I am dealing with much higher frequencies than I used to (the highest I dealt with previously was 1.06Ghz, FibreChannel, and now it is anywhere from 2.5 to 10 Ghz).
So far, for routing, we've used sandwich which looks like this: "GND-sig-sig-GND". Two of these put back to back will create a dual GND-plane (sig-GND-GND- sig) in the middle.
The thickness of the board being an issue, I raised a question, whether we can save one GND out of these two due to skin-effect. The calculations showed, that even at 1oz copper, being around 1.5 mils thick, it is much thicker, than the skin- depth, calculated to be 0.05 mils. With 2oz it is even better.
The concern is - will those return currents, created by a strip line on each side of the shared GND plane, crosstalk? I need a general idea - It seems to me, like it could work. The highest risetime so far is 125ps.
Any ideas would be greatly appreciated.
Thanks for your interest in High-Speed Digital Design.
If I understand you correctly, you have two (or more) cavities for signal transmission, each cavity protected on top and bottom with a solid plane layer.
Theoretically, if the planes are completely solid (no holes), they would act as near-perfect isolation boundaries at your frequency of operation. As you noticed, the planes even with 1/2 oz copper are many, many skin depths thick, so they work really, really well. The faster you go the better they work.
Now, about the holes. Every time you hole the planes you create an opportunity for magnetic fields to leak out of one cavity and into the next one. You might think that the use of dual redundant ground planes would somewhat lessen this effect, in the sense that one plane should help to shield you from fields leading through the holes in the adjacent plane. UNFORTUNATELY, the holes probably line up on both ground planes (unless you are using blind vias), so I don't think the dual- shielding approach will help very much.
The crosstalk effect due to the holes is noticeable (so my EMI friends tell me), but it is going to be much, much smaller than the crosstalk generated by the signal VIAS which run through the holes. The via penetrates (and couples) all the cavities in a very direct way. My general advice to you is this: if the via crosstalk doesn't seem to be causing any problems, then the crosstalk due to the holes won't be a problem, either, and you might as well save the layers and just put one ground between cavities.
Dr. Howard Johnson