I/O Timing & Boundary Contracts: Design Space

Design space exploration

For I/O Timing & Boundary Contracts, closure options trade margin, schedule, and regression risk.

Option A — conservative

• Conservative margin: helps safe signoff

• Risk: over-pessimism, area/power

• Validate with: first silicon

Option B — balanced

• Balanced derate: helps reasonable margin

• Risk: needs approval

• Validate with: mature node

Option C — aggressive ECO

• Aggressive ECO: helps fast WNS gain

• Risk: hold/SI regressions

• Validate with: late schedule

Option D — RTL restructure

• RTL restructure: helps structural fix

• Risk: schedule hit

• Validate with: chronic paths

DESIGN SPACE — I/O Timing & Boundary Contracts
margin <-> performance <-> schedule
pick based on interface setup/hold WNS, input/output delay coverage, port slack on product scenarios

Design pitfalls

• ECO without regression matrix

• Constraint change without audit

Tradeoff curve

BEFORE / AFTER ECO — I/O Timing & Boundary Contracts

WNS (ps)
  0 |                    --- target
-50 |    ● regressed
-80 |         ● before ECO
-20 |              ● after ECO
    +-------------------------------> ECO iteration
Prove setup AND hold moved safely.

STA deep dive

SDC quality determines whether STA measures reality or fiction.

Concept diagram

SDC FLOW

RTL clocks -> create_clock / generated_clock
board      -> set_input/output_delay
exceptions -> false_path / multicycle (audited)
analysis   -> report_timing

Metric graph

CONSTRAINT QUALITY

audited exceptions  ████████████████  good
stale IO delays     ██████            risk
missing gen clock   ████              high escape

Reports and artifacts

• clock report

• uncertainty budget

• IO delay coverage

• exception audit log

Mini case study

1000 new failing paths overnight: SDC merge dropped a generated_clock on divider output. No PD change needed.

Debug branches

• Diff SDC before ECO

• Validate clocks on failing path

• Audit false paths quarterly

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.

Principal STA review addendum

set_input_delay and set_output_delay translate board and pad timing into the block boundary so internal STA matches system timing.

Metric: interface setup/hold WNS, input/output delay coverage, port slack