How do we know if a structure function is correct?

This paper deals with the verification of thermal transient evaluation implementations. This subject is important because eg. the upcoming standard will describe the thermal transient measurement as a standard method to estimate the junction-to-case thermal resistance, thus anybody can create their own implementation of the evaluation method. We have to have a method to verify these implementations. For this reason we examined the result of the NID method from different angles. For these examinations we defined a multilayer structure as a reference structure and we analytically expressed the unit-step response and the cumulative structure function of this structure. Using the unit-step response as an input data set for the implementation in question we got an approximation of the structure function. Analyzing this and the reference RC network we could define a practical maximum tolerance for the deviation between the analytical and the calculated functions.