Re: gEDA-user: pcb, howto partition power planes?

[Steve Meier <smeier@xxxxxxxxxxxxxxxxxxx>]

Nearly impossible? I disagree here it is standard practice.

Analog signals come in and go out the left side of the board and digital
is in the center to the right side.

Top analog is separated from the bottom analog.

Here is the deal. We start with a signal less the one milli volt and we
increase it over 1000 times. Any noise that crosses over from the
digital into the analog gets multiplied up by that gain as well.

Steve Meier


On Tue, 2008-10-28 at 15:02 -0400, Stuart Brorson wrote:
> >> Sometimes it is necessary/recommended to partition (separate) power or
> >> ground planes, i.e. for ADC or DC/DC-Converters, see page 16 and 17 in
> >>
> >> http://focus.ti.com/lit/ug/slwu028c/slwu028c.pdf
> >>
> >> We can do this in pcb program with (adjoining) polygons.
> >> Disadvantage is, that if we change the size of one of the polygons we
> >> have to manually adjust the other sizes. A other method may be so divide
> >> a large polygon by copper clearing traces (with trace width zero).
> >>
> >> This is related to my question from
> >>
> >> http://archives.seul.org/geda/user/Sep-2008/msg00387.html
> >>
> >> but not identical.
> >>
> >> What is the best way to handle this?
> >>
> >
> > I can't speak to that but just one word of caution: In my 20+ years in
> > engineering I have yet to see one case where splitting a ground plane
> > under high-speed ADCs has worked. Regardless of what application notes
> > say. Usually it didn't work, lots of noise. Or it kind of worked but
> > fell apart the instant somebody whipped out a GSM cell phone or BlackBerry.
> >
> > Myself, I never spilt a ground place. OTOH the industry practice of
> > splitting planes is providing part of my income :-)
> >
> > The only time I split is where required for safety, for example patient
> > isolation per 60601 (ECG, ultrasound etc.).
> 
> I gotta agree strongly with Joerg.  ALthough many app notes recommend
> a split ground plane for mixed signal designs, in the real world it's
> next to impossible to split the plane properly. That is, with signals
> (both digital and analog) running all over your board, power, etc.,
> it's almost impossible to gerrymander your AGND and DGND planes to
> follow their respective signal traces and feeds.
> 
> Therefore, the folks at National Semiconductor push the idea that you
> place analog and digital components as far away from one anther as
> possible on your board.  Also, keep analog and digital signal tracks
> as separated as possible.  But use a *single* solid ground plane
> everywhere for the best noise performance.  I've used this
> advice in practice, and have had no problems (at least with radiation
> and noise).
> 
> Here are some app notes discussing this in greater detail:
> 
> http://www.hottconsultants.com/techtips/split-gnd-plane.html
> http://www.national.com/nationaledge/nov04/adc_article.html
> 
> Cheers,
> 
> Stuart
> 
> 
> _______________________________________________
> geda-user mailing list
> geda-user@xxxxxxxxxxxxxx
> http://www.seul.org/cgi-bin/mailman/listinfo/geda-user



_______________________________________________
geda-user mailing list
geda-user@xxxxxxxxxxxxxx
http://www.seul.org/cgi-bin/mailman/listinfo/geda-user