From: John Popelish
Organization: This space not available for advertising.
X-Mailer: Mozilla 4.7 [en] (Win98; U)
Subject: Re: Bypass capacitor position
Date: Fri, 25 Oct 2002 17:31:55 GMT
NNTP-Posting-Date: Fri, 25 Oct 2002 13:31:55 EDT
Dave VanHorn wrote:
> John popelish wrote:
> > You have gotten some very good answers, but I haven't seen one
> > detail. The pair of traces to the capacitor will produce a bit of I*R
> > and L*(di/dt) drop during changes in the chip's current demand. As
> > far as the stability of the power across the chip, it makes no
> > difference how the total trace length is divided between the ground
> > and power sides of the capacitor, so as long as the traces to the
> > bypass capacitor share no other duties, it doesn't matter whether the
> > cap is closer to ground or Vcc.
> It does, if the cap's ground end dosen't have a low Z path directly to the
> same chip's ground pin.
> That's why I put mine at the chip's ground pin. If you think of the tracks
> as small inductors, then the configuration is a tee filter.
I disagree, A tee filter passes a signal past a shunt path to
ground. In this case, there are two inductances in series with a
capacitance. The total impedance of the series string is not much
affected by how the total inductance is divided between the two ends.
> One inductor off
> to the ps, and one off to the chip. You can't reduce the one between the
> cap and the chip, but you can put it where it does the most good.
> If you move the cap up to the vcc pin, then you have an inductor between the
> cap and the ps, and an inductor between the ground end of the cap, and the
> chip ground.
> I also separate my VCC feeds as much as practical, and it's not uncommon for
> me to have three or four VCC tracks meeting at the PS output cap. I also
> watch my grounds, and section "A" that's fed by a discrete VCC track will
> also have a discrete ground return. It's local ground will be planed to the
> degree possible on both layers, but not contact any other section's ground
> except at the PS output cap.
> Two layer boards, plastic enclosures, no ferrites, foils, or sprays, and I
> usually get the "is that thing on?" reaction at part 15.
Do chips powered by one section communicate with chips powered by