Yesterday, we replaced the CW REFL PD with the second PDA225. We noticed that turning it on made the output rail so we did some debugging and found that the switch was broken. Koji did some magic and got it working again. We also changed the CCW REFL setup so that we could point the beam at the PD better (before, we didn't have room for a full mount between the faraday isolator and the gyro enclosure, so we had fixed up a lens mount to do the job and had to use the focusing lens and PD height to center the beam).
The InGaAs diodes have a much better response to 1064 nm so---even with the maximum screening from a Y1S-45---we had to reduce the input power to the experiment in order not to cause the loop to oscillate. (This was, of course, after we turned the gain of the PDH box all the way down to "0.0". We should modify the PDH box to have lower gain so that we can turn up the optical power without the loop becoming unstable).
We hooked everything up and got the gyro fully operational again, with both directions locking well. The measured primary loop had a UGF of ~12 kHz with a puny phase margin of ~8 deg. The thought here is that we should swap out the slow OP27s in the PDH box for faster op amps (in the short term) and eventually design our own servo.
We also noticed that there was a significant DC offset in the primary error signal. We guessed that this was from RFAM from the EOM, so we adjusted the orientation of the HWP before it to minimize it.
We hooked the following signals up to CDS:
Today, I came in to find that the gyro was not locked. When I got it locking again, I noticed that the PZT drive was sitting close to the rail and the slow loop was doing nothing about it. I played around with the slow loop filter and got it into a reasonably stable configuration (pole @ DC, zero @ 0.1 Hz to cancel the internal pole of the PZT).
As I changed settings in the slow loop, I noticed that the cavity was having a hard time locking again. I realized that the error signal offset had returned, and a slight rotation of the HWP brought the cavity back into a lockable state. I decided to remove all other variables by sweeping the PZT directly (as opposed to through the PDH box) and looking at the error signal directly (as opposed to through INMON). I then rotated the HWP to minimize the error signal offset, retuned the phase shift to maximize the response, and then adjusted the input power to make it as high as possible without an oscillation.
Things seemed to be working, but then the same error offset issue came back. I am not sure what it is, but it needs to be investigated thoroughly. This is the plan for tomorrow morning.