Fig. 2 Simulation results using frequency domain and time domain averaging, respectively. In a) and b) the signal-to-noise ratios of the input and output signals and the contamination noise are shown in dB scale. In c) the true FRF is overlaid by the estimates using frequency domain and time domain averaging, respectively. In d) the plot in c) is zoomed in around the first resonance, where the frequency domain estimate is plotted with rings, and time domain with plus signs overlaid on the true FRF in solid. In e) the ordinary coherence is shown and in f) a the FRF is zoomed in around the antiresonance. The latter plot shows that time domain averaging is as efficient as the H1 estimator to remove the effect of the output noise. is a measure of how much contaminating noise is present. The effect of the time domain averaging can then be estimated by using the following relation to compute an “efficient coherence function” estimate for the time domain averaged process 2 2 ˆ ( ) 1 1 ˆ ( ) a yx K f K f . (10) 302
RkJQdWJsaXNoZXIy MTMzNzEzMQ==