[sdiy] split ground plane

[Magnus Danielson]

Daniel Kruszyna wrote:
> Daniel Kruszyna <dan at krue.net> wrote:
>> I'm working on a pcb with separate analog and digital ground planes.
> 
> Well, it's been a while, but I'm now starting to work on this layout again.
> I've done a lot of reading, and most sources recommend to NOT split the ground
> plane. They recommend spliting the power plane instead, and keeping each set
> of signals separate in the layout.

Splitting of the power and ground planes is OK, but you you should 
consider the cut a total cut except for a bridge over which the signals 
go. This way a quiet power/ground plane can be achieved while not 
creating large loops for emission and also additional parasitic induction.

Cutting the ground and power planes and run signals hap-hazzardly over 
the cut is a bad thing. Doing a ground-plane bridge over which all 
signals goes and ground-only on the sides of the bridge ensures low 
emission and thus low parasitic induction.

Also, let the power-plane(s) not go all the way out so that the edges 
has only groundplane the outermost mms, including towards the ground-cut 
  region.

> This board is going into a modular synth, with separate -15v/+15v and +5v
> supplies. I've tied the grounds of the two supplies together at their source.
> Might it cause problems if the grounds are also connected at each module?
> Shouldn't the grounds be connected at only one point?

You can connect the grounds at both points, but since the +5V feed is 
usually used for digital circuits it may be a good thing to keep it as 
the "dirty ground". However, the grounds should in that case be 
connected through a cap so that sharp transitions does not creep outside 
of the board, which happends when a signal goes between dirty and clean 
grounds.

> Additionally, wouldn't a signal trace that crosses a split ground plane be
> simillar to a patchcord in a modular?

Yes, but usually you have a fairly high serial impedance on the patch 
cables where as local signals can be fairly low impedance and thus 
emission/induction problems is more accute. Especially important with 
digital signals for which the quick risetime makes it a more significant 
concern.

Cheers,
Magnus