Author Topic: On differential pairs  (Read 1363 times)

0 Members and 1 Guest are viewing this topic.

Offline Alex EisenhutTopic starter

  • Super Contributor
  • ***
  • Posts: 3425
  • Country: ca
  • Place text here.
On differential pairs
« on: August 21, 2014, 05:07:11 pm »
It annoys me that there isn't a standard pin configuration so that there's no crossed over pairs when going from one chip to the other.

I always suggest that P and N be swapped to keep the layout "clean" and free of vias. Seems to me that signal integrity is more important than some arbitrary convention of what 1 and 0 means, right?

Especially as the board complexity grows, every via just chokes the other layers, and when you add the ground transition vias you end up using 8 vias per swap, 2 signal and 2 ground vias on the source side, and the same at the destination.

So if it's just fast digital stuff like a clock, why couldn't I swap P and N?

If it's a PCIe bus, I believe the standard allows for link training so it figures all this out?

But what about a 1GHz differential (100ohm) analog link, like what's in front of a high speed ADC? (or the output of a fast DAC for that matter.)

I think it shouldn't matter to swap pairs as long as the rest of the people are aware of it.

But I feel I'm missing something crucial here. Is it just convention that prevents swapping?
Hoarder of 8-bit Commodore relics and 1960s Tektronix 500-series stuff. Unconventional interior decorator.
 


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf