Today we made more tests of the effect of differential CO2 annular heating and the effect of thermalization on various parameters. Specifically, we measured the CARM open loop gain, PRCL open loop gain, and DARM spring at various times during power up. In parallel, the squeezing team also attempted to tune sqz during different points of thermalization. We also intended to measure frequency noise and contrast defect, but were unable to due to locklosses. See alog 68971 for details on last night's CO2 tests and the preliminary results of contrast defect and frequency noise.
The TCS setting that we tested was a differential annular CO2 step. CO2X was increased to 4W and CO2Y was left at 1.7W. The annular heating engages at the Power 25W state. The preliminary results showed marked improvement in frequency noise. The ring heaters were set to the nominal 76W settings: ETMX 1.3W, ETMY 1.2W.
CARM OLG
We measured the CARM olg during the first three hours of IFO thermalization. The results are plotted here. The show that the CARM gain drops in similar fashion to PRCL. However, we are not in danger of instability, and the gain levels off into the second hour of lock. We can increase the CARM gain as we thermalize, and increasing it will likely improve our frequency noise above 4 kHz (see 68967). Gabriele is working on implementation. He will add a comment describing the implementation.
PRCL UGF
We returned to changing the PRCL gain by hand. The PRCL gain does drop steadily as we thermalize, but at a different rate than measured by Gabriele. The UGF servo will need to be recommissioned for a few locks to collect data and then we can fit the thermalization trend.
DARM Spring
I made some measurements of the DARM spring through thermalization and also with different SRCL offsets. See the plot here. I measured both the DARM OLGTF and a PCAL to DARM transfer function to create this plot. I think the trend shows that we need less SRCL offset at the start of the lock, and then thermalization increases the SRCL detuning. For example, within the first hour, a SRCL offset of -80 was sufficient. By the third hour, we were moving towards needing something like the usual SRCL offset of -165. Jeff has commissioned low frequency line injections to track this better than I can with these measurements. I look forward to further results on this front.
Other effects
We were not able to make any well-defined measurements of these quantities, but two things were evident by eye during this lock: The frequency noise was improved, and the jitter noise was worse.
Lockloss
At the end of the third hour, we experienced an ASC lockloss. By eye, it seems to be a CSOFT P ring up, but given the amount of cross coupling it is hard to be sure. During a subsequent lock, I made some sensing matrix measurements of the ASC REFL signals at high power. There are some changes in the PRC2 P sensing that could be causing problems. Or, the problem could be something else entirely.
Unfortunately we don't have enough time right now to determine what exactly is the ASC problem and how fix it. Gabriele and I have chosen to revert the CO2X annular change to the previous setting of 1.7W (common with CO2Y). We know this setting is stable. We have commissioned the PRCL gain, various feedforward and more at this particular setting. We would also like to deliver a stable interferometer to the start of ER15 in the morning. We don't think we should give up on this setting though. We would like to use commissioning time in the next week to debug the ASC problem and regain stability with the new setting. We think this is a good change for the frequency noise. We also think that the increased jitter noise present is a clue that there are perhaps ASC offsets that could be causing the instability. We are making the CARM gain change though, as that is something that will occur independent of the CO2 change, and it will improve our frequency noise above 4 kHz.
We added some logic to the THERMALIZATION guardian to increase the CARM gain.
The gain sliders can be changed only in steps of 1 db, so we are implementing the discrete changes in the table below to follow the decrease in CARM UGF we measured over one lock
| Time from LOWNOISE_LENGTH_CONTROL [min] |
Additional REFL SERVO gain [db] |
|---|---|
| 10 | 1 |
| 25 | 2 |
| 45 | 3 |
| 80 | 4 |
| 160 | 5 |
We are increasing both LSC-REFL_SERVO_IN1GAIN and LSC-REFL_SERVO_IN2GAIN at the same time, starting from the gain they had when the THERMALIZATION guardian is asked to go to THERMALIZED
There was a problem with the new REFL servo gain adjustment. The THERMALIZATION guardian red the initial CARM gains at LOWNOISE_LENGTH_CONTROL, but at that state CARM was using only one REFL signal. So during the last lock the THERMALIZATION guardian was also keeping one of the REFL diode gains low.
I fixed the problem by writing explicitly the initial gain of the two REFL signals (6db and 6db) since those numbers are hard-coded in the LASER_NOISE_SUPPRESSION state.
I restarted the THERMALIZATION guardian and set back the CARM gain slider to 6 and 6 db
This gain change is a discrete step in an analog circuit when an output from a lower gain OpAmp is switched to a higher one. This is implemented as a binary ladder with 1dB step size, so depending on the actual gain values some elements may go down in gain while others go up. This will introduce a transient which may or may not be visible at the anti-symmetric port. The transisiton time is sub micro seconds.
Bye eye, we could see that the jitter in DARM increased with the new CO2 setting. I ran a coherence comparison of DARM with the IMC WFS DC channels during the time we were at 4W CO2X versus when we are at 1.7W CO2X. All other TCS settings are the same. Above 100 Hz, the coherence of DARM with the IMC WFS increases almost everywhere when we are on higher CO2X annular power.
The one place where jitter is worse at CO2X = 1.7W is around 50 Hz. DARM is has high coherence with PRCL there. We think that the PRCL gain was mistuned, and that PRCL also sees input jitter. We think we can mitigate that region with better tuning of the PRCL gain.
However, the increased jitter is one argument against moving to differential CO2 annular power.
Note: we are not applying any jitter subtraction at either of these times.
