Shocking, I know.

While exploring data from last night (see LHO aLOG 11026), I saw what looked to be artificially clean straight lines through the performance of the ISIs. In order to confirm, I switched the windowing function in DTT from the default Hann(ing) to "BMH" (which I think is the Blackman-Harris Window), and indeed the reported ASD at those frequencies where the spectra looked artificial changed. Sadly, with this particularly huge dynamic range of a signal, the spectral leakage artifacts hide the real signal right at the frequencies where tilt bothers us (~0.5 to 10 [Hz]). I also tried the Flat Top window, and that seems to be the best for X DOF which has high dynamic range, but in RY, which doesn't, it "over" estimates the ASD at 0.05 [Hz]. Unclear whether this is a good compromise given that depending on the hour, we're interested in one band or the other affecting the performance.

So, I guess I'll have to use asd2.m found here because "[when not computationally limited,] the spectral [leakage] of a single, long-time window is much smaller than the spectral [leakage] of many short, overlapping windows (like you get with pwelch)". Which really means we should just get the live noise budget up and running.

#yakshaving