Its amazing if it turns out that the noise goes down that much with just turning off the feedback on the CW path. Is it really repeatable, or is it just that the CW beam was not resonating and therefore not producing a beat signal in the transmisssion?
We need to look at the noise of the PD, the PDH box, and the phase noise of the AOM drivers and then put them into the loop model and plot them with the noise budget. Do we have data on the open loop gain for the CCW and CW?
I also suggest that instead of using DTT, you make a mDV/NB hybrid that can plot each day's noise curve onto our latest SVN NB model. It would be handy.