Why it matters

SI analysis is easy to run and easy to run incorrectly: wrong parasitics, a missing corner, a threshold left from an experiment. A short checklist with evidence for each line is the last defence.

The checklist

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15

items = {
    "SI enabled in all signoff scenarios": True,
    "parasitics final with coupling": True,
    "clock nets analysed": True,
    "filters documented with hidden totals": False,
    "noise run, critical receivers clean": True,
    "exclusivity claims reviewed": True,
    "last ECO verified": True,
}
open_items = [k for k, v in items.items() if not v]
wns_si = 0.004
assert open_items == ["filters documented with hidden totals"]
ready = not open_items and wns_si >= 0
assert ready is False
print("PASS: not ready; open items", open_items, "although WNS with SI is", wns_si, "ns")

With SI on, the worst setup slack is +0.004 ns and everything looks fine, but one item is open: the filter table with hidden totals is missing, so nobody knows how much delay the thresholds hide. A signoff review treats the missing evidence, not the positive WNS, as the deciding fact. The WNS is 4 ps: the hidden total might be larger than that.

What the review looks for

• Every line has a link to the run and the date; a line without evidence is open.

• The checklist is repeated, not copied, after each ECO that changes routing.

• Open items have an owner and an expiry date before signoff is declared.

Related topics

• False Aggressors