Distributed network of remote sensors for real time prediction of hot spot temperature values

This paper presents a practical realization of an on chip sensor network capable of monitoring an integrated circuit temperature in real-time. The main novelty consists in the fact that the hot spot temperature is estimated based on remote sensor measurements employing an inverse algorithm developed expressly to detect fast thermal transients. The operation of the algorithm is demonstrated not only on the simulated data but also on the measurements of a real integrated circuit equipped with distributed networks of temperature sensors and transistor heat sources. The measurement results can be processed either directly on-chip in the digital filter implementing the inverse algorithm or in the external signal processor connected to the analog sensor outputs.