On Fri, 17 Dec 2004 09:49:01 -0800 (PST), Bob Blick wrote:
>=A0>=A0>=A0The first pic is plenty capable of driving the second pic, I =
want
>=A0> > to=A0add the resistor to keep EMI as low as possible but not make=
 the=20
>=A0>=A0>=A0resistor so large that it isn't reliable.
>=A0>=A0
>=A0>=A0
>=A0>=A0Which will probably work on the bench, but fail at 5 degrees lowe=
r
>=A0> in=A0temp, or when the humidity rises, or when the PCB it's built o=
n has
>=A0>=A0slightly thinner copper, etc. I personally see such an approach a=
s=20
>=A0>=A0ASKING for gremlins.
>=A0
>=A0I really doubt if 100 ohms is going to cause it to stop working. But=20
>=A0it might reduce EMI or power supply noise. That's why I'm asking if=20
>=A0anyone has done 2 pics off one crystal on a large number of boards,
>=A0and what they did, and if they did any testing. Previously I have
>=A0just used no resistor, but just on a couple of personal projects and
>=A0didn't care about noise.

Herbert brings up the point I was alluding to but didn't spell out.=20

I did a commercial design several years ago with two PICs connected as=20
you suggest. We bought about 25 prototype boards IIRC and I built a=20
handful to test. I spent an afternoon tweaking the resistor value and=20
they worked great. A few weeks later I built the remainder using a=20
fresh set of PICs from Digikey, rather than the Microchip samples I=20
built the first few with. A small number of them didn't start the=20
second PIC reliably. I tweaked the resistor value again and applied the
"new value" to all the boards. A couple of the other boards now=20
wouldn't start/run reliably. So that led me to buffering the clock on a
second prototype run and all worked reliably. We ultimately built about
5000 of these boards and I hate to think of the call I'd have gotten
when the production boards started having "problems".=20

I really doubt you'll see any additional significant EMI by buffering
the clock signal between PICs. You already have TWO PICs running at
~20MHz and the additional gate is not going to stress your supply
-- the PICs are a much worse HF load than the single gate. You also
have a very short clock trace (2") and unless you have very poor layout
it won't be a problem. If you are really paranoid, add a series
resistor on the output of the buffer, but I doubt you'll need it for
this application.=20

I'm of the same mindset as Herbert -- why tempt fate and Murphy's
helpers over a solution that only adds a few cents to the board is
guaranteed to work?

Matt Pobursky
Maximum Performance Systems

_______________________________________________
http://www.piclist.com PIC/SX FAQ & list archive
View/change your membership options at
http://mailman.mit.edu/mailman/listinfo/piclist