Sheila Georgia Craig Koji
IMC board saturation:
We had trouble staying locked for more than 10-20 minutes after powering up today, which turned out to be because of saturations in the IMC common mode board. The attached screenshot shows IMC-REFL_SERVO_SPLITMON (which is a beckhoff channel calibrated into volts right after the IMC boosts) hitting 10Vs, IMC-L hitting it's ADC limit fractions of a second later (which doesn't matter since it's not in loop). Georgia and Craig are working on redistributing the analog gains to avoid this.
BS ST2 isolation glitches causing DRMI locklosses:
We had some difficulty keeping DRMI locked after it acquired tonight. We went through the next several guardian steps, and found that the big transient glitches that have been causing us a lot of grief are from engaging the beamsplitter ISI ST2 isolation. Georgia saw a transient on the optical lever when the only thing that was happening was that the BS ISI was isolating. It should be possible to recreate this without locking anything. For now we have removed the BS ST2 isolation from the main guardian path.
DRMI ASC:
Georgia also found that we had DRMI locklosses when the SRCL mode hopping avoidance offset gets turned off in ZERO_3F offsets. We decided to try engaging the DRMI ASC to avoid this, which was fairly easy. We changed some of the filter states (PRC1 and PRC2) to match what we are using in full lock, changed the SRM sensor to AS A 72Q (it was a combination of A+B), and increased the gains in several of the loops. Once we had all the loops engaged Georgia tuned the POP A offset to increase the POP18 build up to ~63 counts. We have tested the engagement with the gaurdian twice, it seems fine.
Lownoise ASC:
We tried to go through lownoise ASC once, but lost lock due to the IMC saturation, not related to the ASC changes we were making. I've edited the gaurdian to not reduced the DHARD P gain anymore, and to engage the less aggresive DHARD low pass based on the measurement that Craig and Koji made last night (46559), this hasn't been tested yet.
Maybe not super important since we've not been using the POP QPD for full-lock ASC recently, but the change in ASC-POP_A offsets for DRMI ASC engagement was:
| old | new | |
| PIT_OFFSET | -0.52 | -0.9 |
| YAW_OFFSET | -0.03 | 0.4 |
Also attaching a screenshot of the effect of turning on beam splitter ISI stage 2. This happens in ISC_LOCK guardian state number 104, top plot. When the H1, H2, and H3 outputs are engaged (middle plots) there is a big kick which is seen on the beamsplitter oplev (bottom plots), mostly in pitch. In this particular lock we survived the kick to the beamsplitter but this is not always the case.
I adjusted the MC board gains to take 6 dB of gain from IMC_IN1GAIN and IMC_IN2GAIN, and put it on IMC_FASTGAIN.
I also adjusted the ISC_library.IMC_power_adjust() function to take in a some scaler that allows the user to set the IMC_FASTGAIN setting for 2 watts. This scaler is now set to 6 dB, where is formerly was 0 dB by default since the variable didn't exist.
I then made adjustments to ALS_COMM and CARM_TO_ANALOG so that we were able to get through guardian with no overall gain change.
The final gain settings at full lock with 30 W should be:
Channel Slider Value [dB]
---------------------------------------------------
H1:LSC-REFL_SUM_A_IN2GAIN 8
H1:LSC-REFL_SERVO_IN1GAIN 6
H1:LSC-REFL_SERVO_FASTGAIN 16
H1:LSC-MCL_GAIN 16.9 [magnitude]
H1:IMC-REFL_SERVO_IN1GAIN 22
H1:IMC-REFL_SERVO_IN2GAIN -22
H1:LSC-REFL_SERVO_FASTGAIN -18
-----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
Georgia and I checked a couple of locklosses we had from saturday when we were walking around the LVEA. In at least two locklosses from that day, we saw the saturation of the IMC Splitmon (one example in attachment one).
------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
We were able to get through locking and CARM_TO_ANALOG once with the above settings.
During LOWNOISE_ASC we had another saturation in IMC_SPLITMON, even with the 6 dB gain redistribution. The gain redistribution done above is not enough.
Koji and I spent some hours trying to move it an additional 6 dB. Unfortunately, for some very strange reason this was spoiling ALS_COMM acquisition.
Getting around this will be not so trivial, so we moved everything back such that we have only the original 6 dB gain swap so that morning people may lock again, with slightly improved IMC board voltage levels.
I think you want to take off at least 16dB of the IN1GAIN. There is hardly any noise advantage in running this high a gain at the input. The VCO has a range of ±1.3MHz below 1Hz and an input voltage of ±7V. This will be reduced to ±200kHz with a fast gain of -18dB and a split mon voltage of ±14V. A VCO range of ±200kHz corresponds to a laser frequency range of ±400kHz at MC_F (AOM is double passed).
Alternatively, one could feed back frequency below 1Hz to the test masses.
PS. Make sure the limiter in the fast path is off.