how much crossstalk_delta affects timing (mathematically)

[Collang2]

I use PRIME TIME for Static timing analysis.

I would like to know "how much crossstalk_delta affects timing." and "how calculate the crosstalk delta"
thanks

 

[dick_freebird]

Couldn't tell you about the tool, but I can tell you I once
had to respin a couple or mask levels to get rid of crosstalk-
induced determinstic jitter on a pair of 60MHz, 90MHz
product clock traces that ran side by side for a couple of
mm.

I "found the range" by simply adding a capacitor to the
extracted netlist starting with the sum of found pcapacitors
between the two, and simulating (Spectre, in those days).
Loop it and when you hit the observed signature, you're
there.

I wouldn't calculate it because there are many unknowns
(line C, cross-C, destination loading, source dV/dt) that
you probably do not want to firt find, then do that arithmetic.
And of course if this is pre-tapeout you may not even know
all that yet. In such a case, you should create a test case
(say, 1mm of co-routed traces carrying two toggling
waveforms, driving two inverters; observe how "victim" timing
may be "beat forward" or "beat back" by the "aggressor"
waveform when their linear transition regions overlap
(in my case, 60 & 90 overlap at both LH and HL transitions,
at the 60MHz rate - a classic 2-band DJ signature).

Map delta-T vs Ccoupling at 1mm, and you have something
you could scale for separation distance (fringing-C or line-C
formula) and co-route distance, as elements of
coupling-C-per-mm. Which I imagine is kinda what your tool
wants (format, I know not).