Routing Video Tracks (Differential Pair)

Yes, the closer they are, the more they will couple. Unless you want to get into simulation and such, your best bet is to provide the greatest amount of separation as you can. Once you have fanned out from the connector, get as much divergance on the pairs as you can.

Tom D.

You bet, it will cause coupling. Really, the whole path of both signal pairs should be evaluated, not the board alone, if you care for good results in the final product.

If the signals are analog, some kind of cancellation - by crossing over the traces - may be necessary. If the signals are digital, the receiver's clock can be shifted, windowed, excluding the crosstalk spikes.

I do not understand the no vias desire. The connector pins (if not surface mounted) act as vias themselves.

In a fanout, you can get a shield between the pairs almost all the way, from Gnd, V+ or V-. Bit unortodox, but works. Spacing 3x of the pair's spacing min.

Can you explain bit orthodox for me please...

If you have any issues with coupling you can try using an image plane with your CCA if you have any clearance. This is a good idea for any designs which pass analog audio or video information.

Is it neccessary for ground via stiching? Somebody suggested this but i do not think it is neccessary for differential pair....

If you run your traces as shown, it's no longer a differential pair, it's more of a "stripline" or "microstrip", with the exception of not having single reference plane. This will greatly affect your differential pair characteristics and impedance.

I'm pretty sure if you did run your traces as shown (and I wouldn't recommend it), you would have a single ground PLANE, not separate traces, above, or above and below, your video traces. If your video was single ended, not diff pair, the solution above might be OK, but if they are true differential pair signals, I don't think that's the solution. If proven wrong, I'll humbly accept that, I am a PCB designer, but not a "veteran", only about 5 years in the field (but many more in RF and electronics in general).

If you DO end up doing as shown, be sure to model that stack-up and routing in some kind of impedance calculator, so you can be sure your video traces are the correct width and spacing.

Good luck,

Tom D.

Thanks,

I also thought the same about this working well only if it was single ended. I have made my layer stack symmetrical (this is a 18 layer board) and kept ground planes below and above my video signal layers.

I have been a PCB Designer for only 2 years and this was a suggestion from a more "experienced" colleague.

Thanks again, I wil stick with traces close to each other(using my calculations) with ground traces seperating each pair and I will only have video tracks in these layers.

My PCB manufacturer is willing to simulate (at a price of coz) the board for me to check SI and impedance matching.

Thanks again for your inputs Gentlemen

A differential pair to work as such well, have to have a strong coupling to each other, and weak to quite weak coupling to a COMMON ground plane (or a Vcc plane serving as such). The stitched drawing fulfills neither. Think about a tightly twisted pair wires in a large cable somewhere, which may or may not have an outside shield way out. That will help visualizing, what it takes.