Evaluation of high-voltage impulse waveforms using model-based deconvolution

Accurate measurement of high voltage (HV) impulse waveforms is of critical importance in the testing of HV apparatus and for fundamental insulation studies. Quantifying the measurement system errors can be performed by following procedures recommended by the international standards for high voltage testing. Application of deconvolution techniques to reduce measurement errors by calculating the input waveform using the measured step response have been explored, but they are particularly sensitive to noise in digitized data; small random errors in the measured waveforms can result in large oscillations in the deconvolved waveforms. This paper presents a technique for effectively performing a deconvolution of measured high voltage impulse waveforms using the step response of the measurement system and input waveform model based upon the impedance parameters of the test circuit. Data are presented that demonstrate a significant advantage of this approach in estimating the deconvolved input and its insensitivity to random noise in the measured data. Results for different analytic models for the input waveforms are also included.