RussellMc wrote:
> Also note that the shorter and/or less area in a current carrying loop
> the less radiating effect it is liable* to have. A capacitor placed
> close to an IC that generates a current spike (source or sink) from
> power supply  to/from ground will typically* more readily act as a
> short circuit for the spike current AND minimise radiating effects.
> You (and Murphy) can produce long paths that have less radiating
> effects than short ones, but as a general* rule the shorter the path
> the better. An ideal ground plane covering the whole PCB will allow
> you to connect to ground at any point BUT as nothing is ideal,
> minimising paths as much as possible is usually* wise.  (On actual PCB
> layout visualise source and sink points for noise and visualise what
> physical paths current will follow. Minimum path length and minimum
> loop area are usually* the aim.

Good points.  Furthermore, you should keep the bypass current off the groun=
d
plane.  Ground planes are funny things in that they work better the less yo=
u
use them.

Think of every IC as a high frequency current source between supply and
ground.  The bypass cap shunts this high frequency current.  As you say, yo=
u
want to keep this loop short, but you also want to keep this current off th=
e
ground plane.  Otherwise the ground plane becomes a patch antenna and the
bypass cap to IC ground lead the feed point.  I try to run a separate groun=
d
and power net for something like a PIC with 1uF bypass caps between these
two nets.  These local power and ground nets have a single connection to th=
e
rest of the boards's power and ground, with a 10uF cap accross the
connection point.  On four layer boards, the local ground net is often a
polygon in layer 2 just below the PIC and its immediate vicinity, like the
crystal, crystal caps, and of course the bypass caps.

And yes, this has been measured to really work.  I once designed the next
rev of a product originally designed by someone that didn't get grounding.
He thought enclosing everything in metal would magically make the emissions
go away.  There are exceptions, particularly in RF, but shields are often
the first refuge of the incompetent.  This guy used 6 layers with the botto=
m
layer ground, and had a metal shield tied to this layer soldered over the
processor that produced the noise.  This thing barely squeaked by the FCC
emissions test.

My version used a open 4 layer board with layer 3 a pervasive ground plane
and a few local ground areas in layer 2 for the nasty digital parts.  It
beat the FCC limit by over 14dB.


********************************************************************
Embed Inc, Littleton Massachusetts, http://www.embedinc.com/products
(978) 742-9014.  Gold level PIC consultants since 2000.
--=20
http://www.piclist.com PIC/SX FAQ & list archive
View/change your membership options at
http://mailman.mit.edu/mailman/listinfo/piclist
.