PSU design?

[jhaible]

> and Haible wrote:
> >One typical error is to make very broad GND connections
> >and connect everything just anywhere. The currents must
> >be *forced* into a certain path, or they will find the path with
> >least resistance, not least ripple.
> 
> What is the theory behind this? I've always been quite enamoured with
huge
> GND tracks; in power supplies and in just about every other circuit. It
was
> my thinking that the broader the track, the lower the impedance and
> therefore the lower the hum. What am I not considering here?

Well, the topology of a circuit sometimes is more important than the
absolute
component values. To simplify things, imagine three nodes, A, B, C.
A is where the garbage comes in, B is where the garbage will be filtered
out,
and C is where you you want to get a clean signal. It's most important that
your signal flow is from A to B to C. What you will avoid is a direct path
from
A to C (bypassing B).
This applied to the GND path of a PSU: It's not so important what absolute
resistance you have between A (transformer/rectifier) and B (capacitor),
or between B and C (output/reference pin of regulator). But you don't want 
a direct path from A to C around B. A broad copper plane for GND will
provide
a lot of "paths" around the capacitor.

JH.