Optimization-based Network Identification for Thermal Transient Measurements on LEDs

In transient thermal testing, network identification by deconvolution is a frequently used technique to analyze the thermal properties of a device. To obtain reliable results, the method requires measurement data with a high signal-to-noise ratio. In particular, the derivation and deconvolution steps yield inaccurate results when confronted with noisy data. In this work, an alternative evaluation procedure called optimization-based network identification is applied to thermal transient data. The method uses a multidimensional optimization to generate a piecewise uniform structure function, which has a thermal impedance that matches the measurement data well. To judge the accuracy of the optimization-based network identification and to compare it to other methods objectively, accuracy measures are developed. Several variants of network identification a re tested using a reference structure in dependence of the signal-to-noise ratio. Finally, the optimization-based network identification is applied to transient thermal measurements conducted on LEDs. In addition, a transient dual interface test is evaluated comparing the classical approach to the optimization-based network identification.