Résumé
Lock-in thermography is a well-known technique used in the detection of small thermal sources. For years, it has been applied to the investigation of integrated circuits, in particular for failure investigation. This paper investigates ways to improve the quality and the readability of lock-in thermal maps. Most of the time, the identification of thermal sources relies on human interpretation, thus leading to subjective conclusions. This work proposes an automated signal detection feature based on statistical analysis of thermal measurements. Notably, variance properties of lock-in phase measurements allow applying statistical tests in order to highlight areas of thermal activity. By doing so, the readability and the detection of thermal hot spots are improved.