抄録
Recently the transfer function method, which computes a frequency characteristic from input voltage and response waveforms, has become a subject of attention in insulation level designs and performence tests of transformers. Its computation principle is based on deconvolution operation between the two waveforms and many application softwares are being developed by using FFT algorithms. However, in this process they have some numerical difficulties which are experienced by most of users and developers. First this paper proposes a numerical recipe of a developed program for the method which almost eliminate such difficulties. Furthermore it is necessary to develop a checking method for their accuracies and reliabilities. In this paper analytical test data are generated for full and chopped wave tests accurately by simulating a transformer circuit from a fractional polynorninal model based on its transfer function characteristic. Test data are estimated for signal-to-noise ratio by so-called coherence function. Finally experimental data are also analyzed for transfer and coherence functions and the results are investigated.
