[sdiy] Nice info on decoupling caps

[David G Dixon]

> Absolutely correct.  I've lost the reference, but a long time 
> ago one of the semiconductor manufacturers did an actual 
> experiment on a board with a large analog circuit.  Their 
> conclusion was that you need one pair of bypass caps for 
> every "few" ICs.  I've never seen a problem that could be 
> fixed with more bypass caps, either, as long as sensible 
> bypassing as specifically recommended by the manufacturer for 
> certain chips (555 timer,
> 311 comparator) is used.  In fact some spec sheets even say 
> explicitely that bypassing next to the chip is Not required.  
> Most x-talk problems I have had were due to radiation or 
> improperly laid out supply lines.
> 
> Guess I'm just "uninitiated". :-)

I'm glad you brought that up, as I seem to remember us having this
conversation a couple years ago.  Certainly, anything that makes really
sharp edges should be decoupled (and I do).  The only true crosstalk issue I
ever had (that wasn't due to "broadcasting") was when I laid out a dual VCA
which shared all the dual and quad opamp chips.  When I changed it so that
the two channels were on opposite ends of the board (as opposed to opposite
sides of the board, lengthwise) the crosstalk disappeared.  This educated me
as to how much crosstalk one can actually have at -107dB (especially at high
frequencies).  Live and learn!