Differential Pair Setup of Class to Class

PS: Running latest version of PCB Editor 24.1

It sounds like you haven't set the min line spacing andf tolerance for the physical CSet for the diff pair. (It does explain this in the note you mention). Example, if the primary gap is 0.2mm, set the tolerance + and - to 0.05 then the min line spacing to 0.2 - 0.05 (so 0.15) and see if that helps. Tolerance and min line spacing is highly recommended when defining diff pairs. There is a supplementary notes on diff pairs on the same website that exaplins why.

I did not set the tolerance (as you surmised) just the trace width and primary gap. I thought with the tolerance set to zero it would use the width / gap number I set as hard and fast numbers, no tolerance allowed.

Tom

You need to read the supplemental notes app (under the diff pairs one) on the same website and that gives you all you need to know about why you really should set tolerances and min line spacing.

Hi Steve, Nice app-note, works very well here - cool beans ! Thanks.

Tom on the spacing rule it is fairly ez to setup, steps below. Couple of thoughts first. On a board there are global spacing rules for everything but sometimes it is handy to be able to over-ride these rules by creating a dedicated spacing rule and applying it to an object. In the case of diff-pairs. A dedicated spacing rule pertains to objects outside the diff-pairs. It does not have anything to do with the actual diff-pair creation or spacing gap of the diff pairs etc.

Assuming you have your DIFF-Pairs working here is how.

IN The CM Left-Pane "Spacing Tab" go to "Spacing Constraint Set" (All Layers). In the right pane under  "Objects" Create a new Spacing Set. I used DP-Keepout as the name. Change a couple of things in there such as Line to line spacing & maybe shape to line spacing
so when you test it out it will be ez to see the results. I used 50Mil so it would be obvious on the board.

Next in the left pane click on Net (All Layers). In the right pane you should see your Diff-Pair Objects. In here all you need to do is apply the spacing rule you just created "DP-Keepout" to the DP objects. Click in the cell "Referenced Spacing Cset" and choose the rule
"DP-Keepout"

That's it. As I alluded to above you now have a dedicated rule that is attached to the DP Objects. With this rule if you try to route a trace close to your DP's you will see that it wont come closer than what the rule was setup to do. If you rip up your DP's and re-route them they should route fine again, but because of that new spacing rule you may find you have to clear more objects. It may be easier to just route in the DP's first using the global spacing and then apply the rule to keep any other objects away as you complete the board.

Hopefully that's the EZ button !

Best Regards.

Yes this does work, the diff pair to “other” spacing is now maintained – thank you.  

But……..

 When I set the “line to pin” clearance so that I can’t route within 18 mils of a pin (connector, IC, etc…), I get DRC’s at the connector where the traces connect which makes sense as they no longer conform to the constraint due to the picth of the connector pins. Is there a way to “allow” this?

 I suppose at the end of the day I can waive those DRC’s. Biggest hurdle for me was getting the trace routing to conform.

 Tom

Was trying to insert a screen shot of the DRC's but keep getting errors. 84k file shouldn't be an issue.

Pic just showed three diff pairs connecting to the connector and the DRC's of pin to line at each connection point. 

Oh, and next up I need to constrain groups of diff pairs to be routed to the same length  +/- tolerance. I believe I can do that by making a net group for each group of signals and apply the phase tolerance to that group.

Tom

Oh, and next up I need to constrain groups of diff pairs to be routed to the same length  +/- tolerance. I believe I can do that by making a net group for each group of signals and apply the phase tolerance to that group.

Tom

This is done with relative propagation delay constraint in electrical domain