The high frequency "tail" of DARM from 5-7 kHZ, which is a region limited by frequency noise, seems to improve when the CARM sensor is switched to REFL B only. I discovered this first by accident: when we run the frequency noise injection for the noise budget, we switch CARM over to REFL B and measure the frequency noise using REFL A as an out of loop sensor.
The process by which I switch the sensor is one that Craig described to me, and I have done several times:
- REFL A and B both have 12 dB digital gain
- I step up 1 dB on REFL B and down 1 dB on REFL A until the REFL B gain is 24 dB and the REFL A gain is 0 dB
- I then step REFL A down to -32 dB of gain
- I turn off the 'REFL_SERVO_IN1EN' switch
To switch back to two sensors, I reverse the steps above.
I have attached the high frequency spectrum of DARM from the 64kHz channel, showing the reduction in noise in DARM when CARM is just sensed on REFL B. I believe this is occurring because this is the region where CARM is gain limited. What I don't understand is how switching to one sensor (and correcting for the gain), is causing such a large change.
I was hoping to switch over to REFL A only and see if this holds true for the other PD, but I forgot I was running ADS on at the time and I railed the ADS causing a lockloss.
I see that this is likely a math error on my part. When I made the change I thought "12 dB on each sensor, 24 dB on one sensor", but I need to add the gains, meaning factor 4 gain on two sensors, factor 8 gain on one, or 18 dB (oops!). I have been inadvertently increasing the CARM gain by a factor of two when moving to one sensor. I'm impressed that I was able to do this without causing any problems. If we feel we have room to increase the CARM gain, we can improve the high frequency tail on DARM. I'm not sure what kind of gain margin we have, but apparently 6 dB extra does not cause immediate problems. Happy accidents and/or math errors?
I think alog67214 and alog67584 are the latest measurements of CARM OLG in Feb. The gain margin was ~13dB at that time. It might be better to measure it again with current 76W configuration.
Looking at the CARM OLGs Naoki posted, we were running with a lower UGF than in O3. In O3, we pushed the CARM UGF up to around 25 kHz to suppress the HF noise we saw. You can't go beyond 37 kHz because of the FSR.