SI ECO Fix Order: Theory Deep Dive

Foundational theory

SI ECO Fix Order is central to SI Timing & Noise. SI fixes use spacing, shielding, sizing, and reroute; order matters because a setup fix can worsen hold or noise elsewhere. Senior STA engineers always tie slack movement to corner, mode, constraint set, and parasitic view — not just a single report line.

Core concepts explained

• SI fixes use spacing, shielding, sizing, and reroute; order matters because a setup fix can worsen hold or noise elsewhere.

• Primary metric: post-ECO SI WNS recovery, reroute count, spacing violations cleared

• Primary artifact: ECO change list, before/after SI report, regression matrix

• Owners: PD owner, SI owner, STA lead

• Setup: max delay path; Hold: min delay path

• WNS/TNS summarize tail risk across endpoints

Why this matters at signoff

At signoff, SI ECO Fix Order failures block tapeout or force risky ECO. Coupling makes delay context-dependent — aggressors matter. Wrong analysis setup wastes weeks of PD effort.

Mental model

SI ECO ORDER
1. spacing / shield on worst aggressors
2. victim upsize (if delta delay)
3. reroute victim to quiet layer
4. re-run SI + setup + hold matrix

Worked intuition

1. Name corner, mode, and view for the failing run.

2. Open post-ECO SI WNS recovery, reroute count, spacing violations cleared and identify worst path group.

3. Read critical path: cell vs net, launch/capture clocks.

4. Check constraints on that path (exceptions, generated clocks).

5. Collect ECO change list, before/after SI report, regression matrix and tag database versions.

6. Classify: constraint bug, view mismatch, or real path.

7. Propose minimal ECO and list MMMC+SI regression.

Common misconceptions

• One typical corner is enough for signoff.

• False paths can be applied broadly to green-wash violations.

• Implementation WNS equals signoff WNS without view mapping.

• Setup fix automatically preserves hold.

Visual reinforcement

SI fix priority

SI ECO ORDER
1. spacing / shield on worst aggressors
2. victim upsize (if delta delay)
3. reroute victim to quiet layer
4. re-run SI + setup + hold matrix

Layer responsibilities

STA OWNERSHIP LAYERS — SI ECO Fix Order

layer              owns                         failure mode
----------------   --------------------------   ---------------------
constraints        clocks, IO, exceptions       false violations
implementation     cells, nets, placement       real path delay
extraction         SPEF/RC corners              view mismatch
signoff policy     derate, MMMC matrix          margin disputes
closure           ECO order, regression        fix breaks other corner

STA deep dive

Coupling makes delay a function of neighbors, not just wire RC.

Concept diagram

SI ON VICTIM

victim net delay = base_RC + coupling_from_aggressors

Metric graph

DELTA DELAY ON CRITICAL PATHS

no SI   ████████  1.2ns
with SI ██████████  1.35ns  (+150ps)

Reports and artifacts

• delta delay report

• aggressor list

• noise peaks

• SI WNS delta

Mini case study

Setup fixed by upsize but SI worsened — spacing ECO on bus bits recovered margin.

Debug branches

• Re-run SI after every timing ECO

• Shield wide switching buses

Senior review question

Ask: what corner/mode/view proves this topic is closed or failing?

Key takeaways

• State corner, mode, view, and database tag with every slack claim.

• Run setup and hold plus MMMC regression after every ECO.

Common pitfalls

• Comparing STA runs with different SPEF or SDC tags.

• Broad false_path to green-wash violations.

• Setup-only ECO without hold check.

Theory reinforcement

Coupling makes delay context-dependent — aggressors matter.