Column: Silicon systems design


Verifi cation metrics and the limits of system confi dence


By Mike Bartley, CEO, Alpinum V


erifi cation programmes today generate more data than at any point in the industry’s history. Coverage dashboards are richer, regressions


are larger and assertion counts are higher than they were even a decade ago. Yet, despite this apparent abundance of metrics, late-stage uncertainty in complex programmes has not diminished; in many cases, it has increased. T e gap between what verifi cation metrics report and the confi dence engineers actually have at system sign-off continues to widen. T e disconnect is not caused by weak execution. Most modern verifi cation teams are highly capable, well-tooled and methodical. T e problem lies elsewhere. Metrics that were originally designed to measure local correctness are increasingly being asked to represent global system behaviour. As designs scale and integration complexity grows, that expectation becomes untenable.


What verifi cation metrics were designed to measure Functional coverage, code coverage, assertion pass rates and regression stability metrics all emerged to answer a specifi c question: has this logic behaved as expected under the scenarios we defi ned? Within a bounded verifi cation environment, these measures remain eff ective. T ey provide structure,


12 April 2026 www.electronicsworld.co.uk


The gap between what verifi cation metrics report and the


confi dence engineers actually have at system sign- off continues to widen


traceability and a shared language for progress tracking. At the block and IP level, the


relationship between stimulus, observation and outcome is relatively direct. Coverage


closure in this context is a meaningful indicator that the verifi cation intent has been exercised. T e problem arises when these same metrics are promoted beyond their original scope and used as proxies for system-level confi dence. As integration scales, verifi cation metrics


increasingly describe activity rather than assurance. High coverage indicates that scenarios were exercised, not that the right behaviours emerged when components interacted under real operating conditions. Figure 1 shows that coverage-driven verifi cation fl ows can reach closure without guaranteeing system-level behaviour.


When coverage closure stops correlating with confi dence Modern systems are no longer monolithic. Chiplets, heterogeneous accelerators, fi rmware-driven control paths and dynamic power management introduce behaviours that can’t be decomposed cleanly into isolated verifi cation units. In these environments, correctness is no longer compositional.


Page 1  |  Page 2  |  Page 3  |  Page 4  |  Page 5  |  Page 6  |  Page 7  |  Page 8  |  Page 9  |  Page 10  |  Page 11  |  Page 12  |  Page 13  |  Page 14  |  Page 15  |  Page 16  |  Page 17  |  Page 18  |  Page 19  |  Page 20  |  Page 21  |  Page 22  |  Page 23  |  Page 24  |  Page 25  |  Page 26  |  Page 27  |  Page 28  |  Page 29  |  Page 30  |  Page 31  |  Page 32  |  Page 33  |  Page 34  |  Page 35  |  Page 36  |  Page 37  |  Page 38  |  Page 39  |  Page 40  |  Page 41  |  Page 42  |  Page 43  |  Page 44  |  Page 45  |  Page 46  |  Page 47  |  Page 48