Sheila, Georgia
I had a look back at a few locks on Friday and Saturday, to compare in which frequency bands we made improvements. On friday we went from roughly 80 Mpc (Friday morning), to roughly 83 Mpc (Friday evening), for reasons we're not really sure of. With Saturday's power up to 30 W (edit: note the lock stretch I looked at the IMC_PWR_IN reads 26 W rather than 30 W) the range increased to ~93 Mpc.
First attachment left plot shows the DARM spectra, and the right plot shows the cross correlation spectra, red is a ~93 Mpc lock, blue is a ~83 Mpc lock, mint is a ~80 Mpc lock. Note the 93 Mpc lock I chose is before we ran A2L and brought the low frequency noise back down.
Shot noise improves with increasing power, but between all 3 locks we also see an improvement in the uncorrelated noise. Going from 80 Mpc to 83 Mpc this improvement is seen from 60-250 Hz; going from 83 to 93 Mpc the improvement is mostly in the 90 - 200 Hz region. I had to hunt around to find long enough lock stretches with no glitches, and I have belatedly realised that the particular 93 Mpc lock I picked out is while Stefan and Peter were increasing the OMC DCPD current, which might contribute to this difference.
I used Sheila's range integrand calculator to compare the different frequency-bands' contributions to the overall range. The top plot of the second attachment shows the DARM spectra again, and the bottom is the integrand. Ignore the sharp dips - between these three locks we had different combinations of calibrations lines turned on and off. The third attachment shows the cumulative range as a function of frequency for these three locks, and the difference in the cumulative ranges going from 80 to 83 Mpc, and from 83 to 93 Mpc. These show most of the improvement going from 80 to 83 MPc was in the 35 - 100 Hz band, while powering up it was in the 80 - 300 Hz region. The improvement in shot noise above 250 Hz doesn't contribute very much to the range.
For reference, the lock times I used started at:
~80 Mpc, t0 = 1225827083
~83 Mpc, t0 = 1225864223
~93 Mpc t0 = 1225937394
There were a couple of problems with the times I chose. There was a glitch included in the 80 Mpc time which contributed to the extra noise at 80-400 Hz. The 93 Mpc time included times where the power on the OMC DCPDs was changed. I've re-run these scripts with new times, new integrands shown in the first and second attachments. The improvement from 80-83 Mpc is now less dramatic, there's little change in the DARM spectrum for the new 93 Mpc time, though looking at the detchar summary page the range here was maybe closer to 90 Mpc. New plots, with new times in the legends, attached.
===
Daniel was concerned about the validity of the calibration going from 22 W to 30 W. We turned the cal lines off on Friday morning and didn't turn them on again until Sunday evening so we cannot compare the PCAL lines ot the spectrum. We looked at the OMC DCPD sum spectra to see if any changes between locks were seen there, see third attachment.
At high frequency the locks are all different due to different CARM gains yielding different frequency noise couplings, with the most severe coupling in the 26 W (90 Mpc) lock. This might explains why we have more sensitivity improvement in the bucket than at high frequency while powering up.
From 80 Hz - 1.5 kHz all three OMC-DCPD_SUM spectra line up, indicating we are limited by some sensing noise. We're not sure why the noise has this shape, if it's shot noise why is it so lumpy?
If the DARM plant changed significantly between these three locks we'd expect to see a change in the DCPD sum output, that we don't see such a change supports the reported increase in range. I've also included a dark noise plot from a time when the power out of the IMC was 0W.
I ran BruCo on 400 seconds of data following each of the GPS times reported above:
https://ldas-jobs.ligo.caltech.edu/~gabriele.vajente/bruco_lho_80Mpc/
https://ldas-jobs.ligo.caltech.edu/~gabriele.vajente/bruco_lho_83Mpc/
https://ldas-jobs.ligo.caltech.edu/~gabriele.vajente/bruco_lho_90Mpc/
From a first look, here's a comparison:
|
Frequency band |
Difference |
|
20-27 Hz [93Mpc is worse than 83Mpc] |
93Mpc data is worse: more coherence with ASC-REFL_A_RF45_I_YAW |
|
29-38 Hz [83Mpc is better than 80Mpc] |
DHARD_Y more coherent in 80Mpc than 83Mpc |
|
Peak at 48 Hz |
no hint |
|
70-100 Hz |
REFL_A_LF_OUT coherence gets smaller fro 80 to 83 to 90 Mpc |
|
>100 Hz |
no hint |
I looked at the frequencies of all of the discernable peaks around the 60Hz line (60.0078 Hz here), during the time of the ~93Mpc lock (starting at 1225937394). As Sheila has mentioned, the 0.4Hz stuff is the biggest, but there are several others that are visible, that we could attack. I've grouped them roughly by peak height:
| Peak height group | Peak frequency [Hz] | Delta from 60.0078Hz [Hz] |
| 1 | 59.5781 | 0.430 |
| 1 | 60.4219 | 0.414 |
| 2 | 59.4688 | 0.539 |
| 2 | 60.5234 | 0.516 |
| 2 | 59.5078 | 0.437 |
| 2 | 60.3902 | 0.383 |
| 2 | 59.6641 | 0.344 |
| 3 | 58.9141 | 1.094 |
| 3 | 61.1016 | 1.094 |
| 4 | 61.0703 | 1.063 |
| 4 | 59.0234 | 0.984 |
| 4 | 59.3906 | 0.617 |
| 4 | 60.5625 | 0.555 |
| 5 | 61.0078 | 1.000 |
| 5 | 60.9766 | 0.969 |
| 5 | 59.125 | 0.883 |
| 5 | 59.2422 | 0.766 |
Solving the 1.1Hz line on HAM3 (which is also getting into SRCL and other DOFs) will definitely help reduce these wings. If we fix that, and the 0.4Hz stuff, then we'll be in much better shape.
