[sdiy] PCB layout rules
Ingo Debus
debus at cityweb.de
Mon Dec 25 16:58:50 CET 2006
Am 23.12.2006 um 00:41 schrieb Antti Huovilainen:
> On Fri, 22 Dec 2006, John Luciani wrote:
>
>> The cap is there for stability. It delivers surge currents through
>> a low
>> impedance path to the IC. Without the capacitor you would have the
>> series inductance of a much longer trace which would result in
>> voltage
>> dips and ringing at the Vcc pin of your IC.
>
> Only if you had no cap there. I was speaking of the choice of cap
> between
> V+ cap V- and V+ cap GND cap V-. The first cannot contaminate
> ground, yet
> should guarantee stability (an opamp after all does not actually use a
> ground for anything).
Decoupling an opamp (running from dual supply) with a cap only from
supply+ to supply- and no cap to ground? I think I read somewhere
that this is a no-no, but I can't remember why. Anyone?
Anyway, some thoughts: yes, an opamp has no ground pin, but often the
load is connected to ground (or virtual ground, when driving another
inverting amp). In this case there's either varying current in the
+supply line or in the -supply, but never in both at the same time.
Just one cap across the supply lines would be a bad idea then, no?
A related topic, some people say it's better to route the power like:
supply -> decoupling cap -> IC rather than supply -> IC -> decoupling
cap. Does anyone here follow this rule? It sometimes makes the layout
look funny.
> Of course you'd use tree/star ground then.
Talking about star versus bus routing: In Eagle there's a tool called
length-freq-ri.ulp that calculates the total length of the trace for
each signal on a PCB. It also calculates the maximum frequency for
that signal from that length (just using f=c/l). Are these results
relevant at all? Of course, when doing a star layout, the total
length can get ridiciously long, because this program just adds all
the traces of one signal. Makes me think of the stories in Bob
Pease's book where people blindly trust the numbers they get from a
computer program.
Ingo
More information about the Synth-diy
mailing list