Sign-off is not triggered by coverage reaching a predefined threshold. It is an engineering decision based on evidence, risk, and the level of confidence achieved across the design.
The central problem in verification is not coverage collection; it is ensuring that the coverage reflects the real design intent. Verification planning operates as a closed-loop system rather than a linear mapping. Requirements are decomposed into verification plans and test cases, which feed regression execution. Coverage data is continuously merged and analysed, providing feedback into the planning process.
The Illusion of the Single Percentage
Coverage is often reported as a percentage. This is insufficient and potentially misleading.
The DesignCon study highlights that coverage is a multi-dimensional space including structural coverage, functional coverage representing behaviour, and assertion coverage representing observability. Acceptable thresholds for each metric must be defined during planning. This is a system-level decision, not a tooling decision.
The Diminishing Returns of Closure
Early in the verification cycle, broad stimuli and basic scenarios quickly close large coverage gaps. As verification progresses, remaining gaps correspond to corner cases, rare conditions, or hard-to-observe behaviours.
These gaps require directed stimulus, constraint refinement, or formal techniques to close. This behaviour introduces diminishing returns, making it necessary to prioritise coverage holes based on risk rather than attempting to close all metrics uniformly.
Without effective planning, teams continue chasing diminishing returns late in the schedule.
Observability: The Missing Link
Verification planning often overemphasises stimulus generation, which addresses controllability. However, correctness depends equally on observability. Observability ensures that incorrect behaviour is detected.
Assertion density is identified as a key metric for measuring observability across the design. Verification plans must actively define the assertion strategy, checker coverage, and observability metrics. Without sufficient observability, bugs remain undetected, coverage appears complete but is misleading, and sign-off confidence is weak.
Structuring the Sign-Off Decision
The Accellera methodology structures the sign-off process through milestone-based planning. This involves progressing through stages such as pre-alpha, alpha, beta, and final verification, where each stage represents increasing verification completeness and a corresponding reduction in risk.
For sign-off planning to be effective, the criteria must be explicitly defined from the outset. When these elements are clearly established and tracked throughout the verification lifecycle, sign-off becomes a defensible outcome grounded in measurable verification results, rather than an assumption driven by schedule or incomplete indicators of progress.
Source Reference Key:
****: Main Article Text (Mike Bartley, "Verification Planning That Actually Works")
****: DesignCon 2007 (Foster and Yeung, "Planning Formal Verification Closure")
****: DVCon Europe 2018 (Kaunds, Bothe, and Johnson, "Case Study of Verification Planning...")
Top comments (0)