Routing over split planes

[cocopa]

Hello! I've read in many books that we should not be routing nets over split planes because the return currents will have no path to return. Split planes refer only to the ground plane that the signal is referenced to?Or we should avoid routing over splits in power planes too? And finally about LVDS signals, since each pair has its return path in the adjacent line, do i have to worry about split planes? Thank you.

 

[shaiko]

we should not be routing nets over split planes because the return currents will have no path to return

The reason why routing over split planes is bad practice is because of impedance mismatch.
When a trace adjuctant to one plane travels over another the effects will be similar to connecting 2 transmittion lines of unmatched impedances.
It's true for both ground and power.

 

[loosemoose]

LVDS is a differential signal. That means you send the signal down both lines. Ad you still need to follow all the high speed signaling rules.
Any signal discontinuities will create reflections.

link to document from Altera : www.altera.com/literature/wp/wp_lvdsboard.pdf

 

[enjunear]

Hello! I've read in many books that we should not be routing nets over split planes because the return currents will have no path to return. Split planes refer only to the ground plane that the signal is referenced to?Or we should avoid routing over splits in power planes too? And finally about LVDS signals, since each pair has its return path in the adjacent line, do i have to worry about split planes? Thank you.

The issue is not that the signals will have NO return path, its that the return currents will have to find a way to get from plane B back to plane A, and that can make the loop area very large (this can also be seen as a change in line impedance, causing signal reflections). Ideally, you want to keep the signal return layer (ground) directly below the signals. If you force the return currents to make a long return path to the source there is an increasing possibility that signals in the environment will couple onto the signal lines (the current paths to/from the load act like a loop of wire... the wider the loop, the better the antenna it makes).

Don't use power planes as data signal returns. Putting high frequency switching signal currents (like data) onto your nice, flat, clean DC power is not a good plan. You could end up having the data signal frequencies getting into the power pins of your other devices, which could really mess up their operation.

Most differential signals on a PCB represent data by flipping the relative magnitude between the two signal wires (voltage on wire 1 > wire 2 = "1"... voltage on wire 1 < wire 2 = "0", etc). This way, the signal amplitude relative to "ground" is meaningless, but a circuit is (by definition) not a circuit without a path for the return current to get back to it's source. The voltage on both signal wires (1 & 2) has to find it's way back the source, so a ground/signal return is required.