Hello folks! I have done system identification using sinusoidal sweeps on a system I no longer have access to. Now I realize that the output sensor was partially saturated during the trials. By partially I mean that the measurements hit one end of the range of the sensor but not the other. So what it looks like now is (at lower frequencies) the chirp is assymmetric around the time axis: the amplitude is flat on the negative side but varying according to system dynamics on the positive side. This implies that I (among other things) get a too small etfe at lower frequencies. Without taking new measurements, what would be the best way of solving this problem? Right now I have made a rather lousy attempt by scaling each negative "half period" according to the average of the amplitude of the surrounding positive "half periods". Since it's a chirp they are not really half periods of anything, but I hope you know what I mean! How would you guys solve this? Is there anyway of using the DFT on only the positive portion of the data and assume some kind of (pseudo)periodic extension/interpolation? Thanks so much in advance! Edit: What it comes down to: if you would only have the positive values of the output from a chirp input, how to estimate the transfer function?