摘要
The transient thermal properties of an IC package are typically characterized by the thermal step-function response and/or by the time-constant spectrum. The temperature response is acquired from measurements or simulations while the time-constant spectrum is obtained from this response, using the NID method (Network Identification by deconvolution). The NID method is accurate only if the calculation is based on the exact step-function response. However, practical measurements provide us with responses which are more or less accurate but never absolutely exact. In our paper we present the sources of deviations and a method to eliminate their effect. We demonstrated the process on examples where the level of the corrected errors can be seen at various time-constants.