Differential Signaling SI Issues

[nelsonys]

I was having a hard time discussing with my other colleague on differential signaling deskewing issue.

There are two deskewing method for differential signaling, a big 'U' and a series of small serpentine routing.
For 1Gbps signaling, I suggested not to use 'U' for deskewing although it saves space and time for routing, as it would impose reflection to the signal, but I hardly got any info to back me up on this statement.

They thought jitter plays a more important role than impedance mismatch, since the differential signals are terminated with a 100ohm and that such short distance (within 2mm) impedance mismatch won't cause any serious reflection problem.

So my question here is how are we going to define a standard for these:
1) What is the worst tolerable impedance mismatch?
2) What comes first for deskewing, jitter or impedance mismatch for differential signaling case?

 

[leekc_84]

Hi Nelson,

Deskew is trying to maintain differential pair in the same length. This is to avoid common mode signal and potential EMI issue.

To answer your questions:
1) Normally +/-20% tolerance is acceptable.
2) As I stated in above.

Hope these help.