> We have recently done boards for two customers, both
> with very tight guidelines. ...
> Both boards work fine. My question is what is the difference
> between these two sets of guidelines, and which do you all
> prefer?
Obviously, the first one is trying to shield the video signal from
coplanar coupling, the second one doesn't but instead controls
crosstalk by spacing aggressor signals a minimum distance away.
Neither one is necessarily right or best. One should try to
understand exactly what benefit you get out of either approach.
I don't know if the guard trace approach really works well at high
frequencies. I have been wary of it. I understand they are effective
for DC/low frequencies on surface layers; but high frequencies?
Unless you tie the guard trace down with lots of vias, it could differ
substantially from ground. Might it actually increase the coupling of
aggressor signals into the victim trace? Might that happen even if
you do liberally tie it to ground??
I once tried limited modeling of a guard trace, and the results were
not particularly encouraging; but not necessarily damning either.
I suspect that if this were a highly critical signal, and the first
customer didn't also control the spacing from the guard trace to other
traces beyond (outside) it, it might not have worked.
> We have a new customer where, to meet Zo, W will be
> 5 mils and H 12 mils. A spacing based on 3W (or more) is
> achieveable. A spacing based on 10 H is not. Also, this
> customer has no room for terminations! What would you all
> recommend?
First, separate the two issues: terminations, and crosstalk.
The issue of terminations is dependent on length versus frequency
content and desired waveform fidelity, and also the source and load
devices. If they can't/won't terminate this video signal, and the
length is substantial, then they can't expect the most ideal response.
Now the spacing issue depends on how much crosstalk they can tolerate,
what the aggressor signals look like (rise time or slew rate), and all
the physical dimensions, etc.
Just because the first two customers' boards worked fine, doesn't mean
the third one has to follow the same rules.
"Video" can mean many things. Plain old NTSC video is only 4.2 MHz,
which is rather slow these days.
Regards,
Andy Ingraham