Skip to Main Content
There has been a large body of research on dynamic thermal management (DTM) to manage the die temperature of integrated circuits against their high power density. Control-theoretic DTM is one of the most effective DTM schemes that guarantee stability criteria while meeting several performance requirements such as response time, steady-state error, overshoot, undershoot, phase margin, gain margin, and so forth. Conventional control-theoretic DTM schemes show reasonable stability and performance for general-purpose processors, but they may not fulfill those requirements for vehicle electronics control units (ECUs) primarily because the ambient temperature of an ECU is dependent on the associated unit temperature that often exceeds 100 °C. This results in a high steady-state die temperature and a very narrow temperature headroom. Furthermore, the unit temperature dynamically changes according to the driving condition that acts as a major disturbance to the DTM system. This paper introduces an advanced control-theretic DTM mechanism for high-performance vehicle ECUs. We model such ambient temperature variation as a disturbance, and adopt a disturbance predictor and compensator that effectively mitigates the effects of ambient temperature variations. We demonstrate that the proposed method is superior to the previous control-theoretic DTM in terms of RMS errors, peak temperature, and thermal violation.