Why it matters

Spacing and shielding are the physical fixes. They cost routing resources, so it is worth knowing what they deliver. A first-order estimate shows both the gain and the price.

Worked example

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

R, Cg, Cc0 = 2000.0, 8.0, 6.0
def d(c):                                         # lumped delay in ps
    return 0.69 * R * c / 1000.0
rows = {}
for label, spacing in (("minimum", 1.0), ("1.5x", 1.5), ("2x", 2.0)):
    cc = Cc0 / spacing                            # first order: coupling falls as spacing grows
    worst, quiet = d(Cg + 2 * cc), d(Cg + cc)
    rows[label] = (round(worst, 2), round(worst - quiet, 2))
assert rows == {"minimum": (27.6, 8.28), "1.5x": (22.08, 5.52), "2x": (19.32, 4.14)}
shield_delay = d(Cg + Cc0)                        # coupling becomes ground capacitance: k = 1 whatever the aggressor does
assert round(shield_delay, 2) == 19.32
print("PASS: (worst delay, delta) by spacing", rows, "; shielded delay", round(shield_delay, 2), "with delta 0")

Doubling the spacing halves the coupling and cuts the worst-case delay from 27.60 to 19.32 ps while keeping a delta of 4.14 ps. Shielding gives the same worst-case delay in this model but with no variation at all: the shielded delay of 19.32 ps equals the quiet nominal of 19.32 ps, so shielding does not raise nominal delay here. What it does is lower the worst case (27.6 to 19.32 ps) and give up the best-case speed-up (11.04 ps). Nominal delay rises only if the shield adds ground capacitance beyond the coupling it replaces, and shielding costs tracks. The inverse proportionality is a rough rule; real coupling depends on layer, height and the other neighbours, and spacing a wire away from one aggressor can move it closer to another.

What the review looks for

• Spacing and shielding changes are checked against routing capacity: an ECO that cannot be routed has no value.

• Shielded nets have their nominal delay re-checked, since the ground capacitance is now always present.

• Spacing changes are re-extracted: the new parasitics, not the estimate, decide the result.

Related topics

• Fix Order

• Next: Victim and Aggressor Changes