Below is a comparison of the CCW error and CW actuation signal noise spectra, both calibrated into optical frequency noise. For the CCW error signal, this was done by dividing by the cavity response in V/Hz, whereas for the CW actuation signal I simply multiplied by the 100 kHz/V gain of the Tektronix VCO. The cavity response is an estimate based on the recent finesse measurements, and may be off by a factor of 2 or so.
I have been speculating that the CW actuation signal (aka the gyro signal) is dominated by excess noise from the CCW loop. This only appears to be true for a certain few frequency bands, while in other places (especially at low frequencies) it seems that there is other noise coupling in. This could come from anything in the optical path between the initial beam split and the cavity.
I have run into some other strange goings-on with the CW signals. For some reason, the ratio between the CW actuation and error signals does not appear to match up with the measured PDH OLTF (more info on this later).
I hadn't been acquiring the CW error signal, so I decided to look at that today. Below is a plot of the CCW error spectrum, along with the CW error spectrum in three different configurations: one with demodulation completely disabled on the VCO (i.e. loop open), one with G = 4, and one with G = 10. The result makes a little bit of sense in that the CW error noise goes down with increasing G, but it should go down by much more, on the order of the CW open-loop gain. Another interesting effect is that the DC level of the CCW error signal is affected by changing the VCO (on the CW loop, of course) from modulation to continuous or vice-versa. This should in principle have no effect on the CCW loop, and this may indicate some problem with the optical isolation.
I will post more information about the recent measurements and do some more tomorrow.