How does temperature inversion complicate the assumption that hold is always worst at the fast/cold corner?
From PDVerse STA Mentor Guide · pdVerse Mentor Guide
Technical Explanation
At low operating voltage, cell delay can actually increase as temperature drops, because the threshold voltage's temperature dependence starts to dominate over the usual mobility-driven speed-up — this is temperature inversion.
Timing-path Reasoning
Drive current scales roughly as mu(T) times (VDD minus Vth(T)) raised to an exponent around 1.2-1.5 for short channel devices. Classically, higher temperature lowers mobility (slower) so hot is assumed worst for setup and cold is assumed worst for hold/fast paths. But in the low-voltage inversion regime, Vth's own temperature dependence can dominate, meaning delay increases as temperature falls rather than the classical decrease. This breaks the folklore of pruning the cold corner because 'hot dominates' setup and cold always gives the fastest, most hold-critical case.
Equation
Drive current ~ mu(T) * (VDD - Vth(T))^alpha, alpha ~ 1.2-1.5
(short-channel)Expected Result
A rigorous corner set must include both the classical hot-slow/cold-fast assumption and the low-voltage inverted regime, verifying which corner is actually worst for each check type rather than assuming.
Failure Symptoms
A design signed off only at the traditional fast/cold corner for hold may miss a worse-case scenario where, due to inversion, a different temperature point produces the true fastest or slowest delay.
Root Cause
Vth's temperature coefficient dominating over mobility's temperature coefficient at low VDD, inverting the classical delay-vs-temperature relationship.
Mentor Note — Debugging Procedure
Do not drop any temperature corner based on the 'high temp = slowest' heuristic; explicitly verify delay behavior across the full temperature range at the operating voltage in question, especially for low-voltage designs. The old 'drop the cold corner, hot dominates' pruning folklore is precisely the escape pattern temperature inversion warns against — at low voltage you cannot assume the worst-setup corner is the high-temperature one.
Continue learning free
Get a practical low-power chapter
Receive the existing “Low Power and Multivoltage Fundamentals” PDF chapter and its download link by email.
Continue practising