A Printed Circuit Board assembly (PCBA) testing approach using infrared thermal signatures is presented. The concept of thermal signature for PCBAs is introduced and proved by experience and simulation. Based on this concept, the testing method is able to detect assembly defects such as presence of the component, polarity, value and solder (shorts and opens) and in some cases component health state, it also can classify the components mounted on the PCB into two classes (fault-free, faulty). According to the thermal signature of each component on the PCBA, they can be also classified in the same classes. In this article focus is put on capacitor defects in a DC/DC converter, especially capacitor value defects. Therefore, they will be the main tested components. For a robust detection of multiple defect scenarios, the Principal Component Analysis (PCA) method is used as an outlier detection algorithm.