Why it matters

Violations are visible in any report. The risk is in what is just above zero: endpoints that pass with 5 ps and would fail with a slightly different derate or extraction. A near-critical histogram shows how much of the design sits close to the edge.

Metrics

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

slacks = [-0.03, -0.01, 0.00, 0.004, 0.012, 0.018, 0.027, 0.033, 0.041, 0.060, 0.085, 0.120, 0.180, 0.250, 0.400]
edges = [0.0, 0.02, 0.05, 0.10]
bins = {"violating": 0, "0-20 ps": 0, "20-50 ps": 0, "50-100 ps": 0, ">100 ps": 0}
for s in slacks:
    if s < 0:      bins["violating"] += 1
    elif s < 0.02: bins["0-20 ps"] += 1
    elif s < 0.05: bins["20-50 ps"] += 1
    elif s < 0.10: bins["50-100 ps"] += 1
    else:          bins[">100 ps"] += 1
assert bins == {"violating": 2, "0-20 ps": 4, "20-50 ps": 3, "50-100 ps": 2, ">100 ps": 4}
assert sum(bins.values()) == len(slacks)
near = bins["0-20 ps"] + bins["violating"]
assert near == 6
print("PASS: histogram", bins, "; violating plus within 20 ps:", near, "of", len(slacks))

Two endpoints violate, but six of fifteen are within 20 ps of zero or below it: fixing the two violations leaves four more that a 20 ps change in margin would flip (a 10 ps change flips the two at 0.000 and 0.004 ns). The histogram turns 'two violations' into 'a design with 40% of its endpoints on the edge'. The bin boundaries are a choice; what matters is that the same bins are used in every run.

What the review looks for

• The same metrics, bins and scenarios in every report so runs can be compared.

• The histogram is shown for the union of scenarios as well as for the worst one.

• Margins and method versions are printed next to the numbers, not kept in a separate document.

Related topics

• Next: Margin Trends