Routing video (6MHz) on 4 layers

On a sunny day (Wed, 14 Nov 2007 22:52:08 -0600) it happened Ben Jackson wrote in :

I would put the 75 Ohm near the connectors, as the ADC is likely high input impedance. Then you have a volt or so, and no signal current loops. Then wire from the connectors + resistors to analog ground of the ADC. Keep video at top layer.

One issue is: are the video connectors connected to a metal front, and where is that front connected to the PCB? Or are you using isolated mounted connectors? etc. It is really difficult to answer without knowing distances, connector type, etc.. just think logically.

I have used isolated BNC connectors, and differential input (ground + signal) too in the past, with the resistor directly on the connector. But that had >100 meter cable with ground voltage differences relative to the other end...

At 6 MHz, you can get away with some sloppiness in the form of non-ideal routing. I assume you are using 75 ohm traces, and I agree with Jan, that they should remain on the top if possible. Then you can ignore the power plane issues.

If you drop to the back side, the trace widths have to get wider to account for the larger separation. But you also have to chop up the power planes and/or stitch grounds together. All this is not unusual in a video/RF layout, but when you can avoid it, do so. If there are potential interferers along the path, then you may be forced to drop to the back for isolation. But first try to move the interferes to the back, especially if they are not analog signals.

I would put the 75 ohm termination close to the ADC. If you terminate at the connector, the entire routing path becomes high impedance (no current flow) and more noise-sensitive. Of course, at 6 MHz, the wavelength is much longer than your traces, so you can argue that the routing is not really a transmission line anyway. Again, some sloppiness is usually tolerable at these frequencies.

Steve

If the power plane is electrically quiet (well bypassed, and not expecting non-clock noise) then just treat it as another ground plane. If you're really worried about s/n, split the power plane and make it ground pours adjacent to your critical signals, and power elsewhere.

Yup, that's conservative. Lots of vias g-g.

Given the above, no.

Doesn't matter much, as long as the traces are short relative to the video bandwidth. If the signal risetime is fast enough to cause any serious reflections for your trace lengths, terminate at the ADC. I'd terminate near the ADC anyhow... it just feels better.

John

In that case I need bypass caps where the signals switch layers, right?

...which is much easier than lots of bypass caps.

That was my plan, although the point someone made about the terminators at the connectors keeping the current loop out of my routing sounds good. I suppose I could provide pads for both options and try it out.

Power planes should be whisper quiet and from an RF point of view behave like ground planes. If they don't then something needs to be done about it. More bypassing etc.

Ok, I know this sounds like a helicopter instructor telling his student: "If the ground approaches at an uncomfortable rate you need to do something against that."

Sure but if your power plane is "singing" you'll have a lot of other issues at hand.

You could but generally it is better to calculate the correct trace width for 75ohms or whatever your goal is and place the termination at the very end, right before the ADC. Mind the input capacitance of the ADC. Sometimes it helps to place a buffer right before the ADC.

If the power plane is bypassed in the usual manner, there's no need for extra bypasses near crossovers.

Yeah, if the layout allows it, the split plane with grounds opposite the signals is safer.

You'll likely see little difference, but if there is a difference, terminating at the ADC is probably a little better. It keeps from having that open stub flapping around.

John