Seeing a lot of mixed infomation around involving ground planes on multilayer boards. For a mixed signal mulitlayer PCB which is the preferred method? I know every board is different and it depends what your board is doing but I see alot of people with isolated ground planes yet Henry Otts advice is to do the exact opposite and partition the board in segments. Any advice or comments on this would be great!
It really depends on your design. Henry Ott says it will simplify your design in general.
Off the top of my head there are a few reasons you don't want split planes:
1) It turns your board into a dipole antenna
2) Crossing a split plane with a trace is bad because it increases inductance and complicates the paths for return currents.
There are reasons for splitting planes, in some cases it can reduce common mode noise or noise on the analog side of the ground plane. If your isolating your analog section completely (which I would not recommend, but is sometimes necessary) then you will need split planes.
I'm preparing some blog pages/content about this topic. But I'll summarize here.
If no charge moves between the analog and the digital sections, then you can pick either options.
Given mixed-signal systems tend to NEED to move digital data and control signals back and forth between the Analog and the Digital sections, there are several approaches;
1) use fiber optics to move data; I saw this used on an IR camera head inside a gymbol (track incoming missiles) with noise floor achieved of 17 nanoVolts/rtHz. There were no beathotes in the images.
2) adjust system timing
3) minimize the digital-interface charge injection; I once used +I and -I and Iref/2 (to set threshold) between FractionalN Synthesizer core-algorithm and the PLL charge pump; reference spur was shockingly low at -105dBc. The key is the balanced +I and -I, where lotta charge was moved but nearly nulled.
4) study the system ringing; understand why your silicon parts have ringing and study how to manage that ringing.
Just some thoughts.