We had another look at the MICH noise budget tonight. Rana has verified that my techniques / math aren't too ridiculous.
In the first attachment, you'll notice that the MICH noise is waay above the shot noise of 1mW on the beam splitter. We don't know why. One problem is that the modulation depth of the 55MHz is too low by ~a factor of 10. Kiwamu and his magical resonant circuit are working on fixing this. This will not, however, fix the huge discrepancy here. More investigation and meditation is required! For this measurement, the whitening gain of AS55 was set to 42dB for both I and Q.
In the 2nd attachment, the PSL shutter is closed, so all of these are dark measurements of AS55. (The input matrix on the LSC screen is AS55Q * 1 -> MICH_IN1, so they're the same). All we've done is change the whitening gain before the ADC. For 0dB and 9dB, you can see that the low freq noise didn't change - here we're still limited by the ADC noise. With 21dB and 42dB we're clear of the ADC, so either is fine. Unfortunately, the high freq stuff when the loop is on matches up with the high freq part of the dark noise, so that's part of the problem....