What are your thoughts on a Signal-Ground-Signal-Ground stackup instead of the two internal ground layers? The sandwiched signal layers being used for sensitive signals.
I haven’t been able to find any information on this particular idea
Thanks!
Edit: Thanks to the recommendations in the comments I found this Robert Feramce video that says to use this stackup for this very purpose and that the usual Sig-GND-Sig-GND only works well for low density boards.
The usual 4 layer stackup has thinner dielectric 1-2 and 3-4; thicker dielectric “pre-preg” 2-3. Unless you vary that construction then 2-3 will not provide so much added crosstalk control.
I’m afraid this forum needs to focus on KiCad-specific questions, rather than general design advice. I’ll direct you towards eevblog as a starting point. If you’re curious, there are simulations out there on youtube that show off different kinds of stackups with various methods of guard rings/vias. One example is here: Stitching multiple copper pours on the power plane to GND planes - Page 1
Just as a quick note in addition, keep in mind the actual distance between layers. Generally (but not always!) there is a much larger distance between layers 2 and 3 in a 4-layer stack up. That can help or hurt, depending on your goals.
I’d be concerned about flatness with a signal-ground plane-signal-ground plane stackup, in addition to getting layer to layer spacing that you want (if you can stomach the price, you can get whatever you want–but I think the lower cost houses have their preferred stackups). If board flatness is not a concern, go for it, although I’ve had some push back at times from uneven copper stackups from vendors over the years. Kinda bugged me to get told that we needed a good ground plane and then I would have two identical ground planes next to each other, but it solved symmetry and made the customer happy.
Somehow signal-ground-signal-signal doesn’t have problems, but if it didn’t work, no one would make 4 layer boards… in the end, sometimes good enough design practice trumps more expensive but perfect design practice (if such a thing actually existed–perhaps I should say “better” instead of perfect).
Do you mean warping as a result of very asymmetrical (amounts of copper in the layers) stackup?
Keep in mind:
By my observation, single sided boards were once common and now I am pretty sure you cannot get that fabricated in USA.
Two copper layers are sort of a cheap minimum for non-critical designs.
I think that four layer are almost a normal-default.
But it is easy and common enough to get 6 or more layers. In my last contract job we had…(not sure) I think it was >16 layers. Doing via-in-pad and all sorts of shenanigans. I think that as the layer count goes up, the stackup becomes more likely to be custom(??)
Yes. I’m still trying to figure it out, but best I can tell, the vendors I work with, given how usually one spec’s +/-10% board thickness tolerance, one can often do a stackup using 3mil/75um core/prepreg and go up in 1mil/25um increments to get thickness and layer count to where they want it. One vendor told me that they could do 2mil/50um core/prepreg but it wasn’t easy (meaning, they’d charge more, I would guess). As you might guess, these boards aren’t cheap. Low quantity runs so I am rather free to do as I want/need.
I recently was bit by a board flatness problem so it’s fresh on my mind.
