Author Topic: Differential pairs!  (Read 1876 times)

0 Members and 1 Guest are viewing this topic.

Offline Simon

  • Global Moderator
  • *****
  • Posts: 12575
  • Country: gb
  • Did that just blow up? No? might work after all !!
    • Simon's Electronics
Differential pairs!
« on: February 02, 2018, 01:16:51 am »
Right. what do I do? the written documentation is not the most enlightening. the video is second to useless and I'm left wondering what to do.

So I finally worked out how to declare two nets as pairs, I sort of worked out how to route differential pairs but it looks like i must set pins up to be on the right side as the differential router won't automatically swap a layer so that the tracks can cross ?
https://www.simonselectronics.co.uk/shop
Varied stock of test instruments and components including EEVblog gear.
Also, if you want to get ripped off: https://www.ebay.co.uk/usr/simons_electronics?_trksid=p2047675.l2559
 

Offline voltsandjolts

  • Supporter
  • ****
  • Posts: 617
  • Country: gb
Re: Differential pairs!
« Reply #1 on: February 02, 2018, 03:05:01 am »
What bandwidth are you hoping to run your diff pair at?
e.g. for USB full speed you would probably OK with via crossover. For high speed, I dunno, but I'm a pessimist.
 

Online blueskull

  • Supporter
  • ****
  • Posts: 10699
  • Country: cn
  • Power Electronics Guy
Re: Differential pairs!
« Reply #2 on: February 02, 2018, 03:26:11 am »
Via crossovers are okay-ish, but you need to know what you are doing.
First of all, NO STUB IS ALLOWED whenever signal speed is faster than a few hundreds MHz, definitely not GHz.
You can cross a signal to the other side of board, and cross it back, that's okay as long as you keep it close to its complementary line and don't do this too much.
However, if you cross from top layer to middle 1, but the via goes all the way to bottom, then the rest part of the via becomes a stub, which is not good.
For USB2.0, this is probably okay, but for any GHz thing this is bad.
And of course, when you do this via crossing thing you need to consider the length contributed by the via and extra length and compensate accordingly.
 

Offline Joel_l

  • Regular Contributor
  • *
  • Posts: 223
  • Country: us
Re: Differential pairs!
« Reply #3 on: February 02, 2018, 07:54:37 am »
I don't think the question was, should you it, as much as it was, how do you get CS to do it.

I played with this a little today, no progress.
 

Offline voltsandjolts

  • Supporter
  • ****
  • Posts: 617
  • Country: gb
Re: Differential pairs!
« Reply #4 on: February 02, 2018, 09:22:54 am »
The question is first of all 'should you' and then maybe 'how you' cross a diff pair. I typically do 4 layer boards that are not controlled impedance and do not have high bandwdth diff pairs. I would guess that Simon is in a similar situation. So, this method works for me:

Once you have decided what pattern you want to use for the crossover (one or two pairs of vias & position), place vias and set their nets appropriately. Use vias in both traces to match impedance as best you can and to provide a via pair that the diff route tool can 'snap' to.

Manually route the (very short) crossover section. Manually matching the trace lengths in this short section is not really a chore and not an issue for a low bandwidth diff pair.

Now use the diff pair routing tool to route the long sections of the pair - they will 'snap' to the via pairs of your crossover because you set the nets for those.
 

Offline voltsandjolts

  • Supporter
  • ****
  • Posts: 617
  • Country: gb
Re: Differential pairs!
« Reply #5 on: February 02, 2018, 08:41:37 pm »
For example, matched length / matched via crossovers with layer change and without layer change:

 
The following users thanked this post: 3db, mars01


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf