Tridiagonal Approaches for Network Identification by Deconvolution

The calculation of accurate and detailed structure functions is crucial to perform a reliable thermal transient analysis using tools such as the differential structure function or local thermal resistance diagram. This study investigates six different Foster-to-Cauer transformation methods for Network Identification by Deconvolution (NID) including Codecasa’s approach and Fernando’s approach. Four of the presented algorithms (Khatwani, Sobhy, de Boor and Golub, polynomial long division) are evaluated with respect to numerical precision and computation time as a function of network length. Results reveal that Sobhy’s approach offers significantly more efficient computations, particularly for longer networks. In these cases, the faster Foster-to-Cauer transformation offers a reduction in total computation time of NID by more than a factor of four. Structure functions with more than 1200 points are calculated in approximately 30 s.