MISMATCHING WINDOWS CAN CAUSE FREQUENCY RESPONSE FUNCTION NOTCHES TO APPEAR AS PEAKS

Abstract This paper examines the errors in frequency response function (FRF) estimatorsH1(ο) andH2(ο) caused by improper input and output signal windowing while using random excitation modal tests. These errors are enhanced when the FRF has resonances and anti-resonances that are closely spaced. An experimental analysis is performed on a free–free beam. The beam is excited with broad band random excitation over a frequency range covering the first five natural frequencies. The driving point accelerance FRF is calculated from both estimatorsH1(ο) andH2(ο). The results show that erroneous FRFs can be obtained when the wrong type of window is employed and these errors are due to digital filter leakage. A theoretical model is proposed in order to explain the discrepancies that were found experimentally. Simulated results from 2 dof system are presented that show the influence of different types of windows on the FRFs.