Virgo Logbook

AdV-PSL (Commissioning of SS)

mwas - 15:14 Wednesday 21 October 2020 (49705)

Recurrent increase in PSL noise also present without IMC

During the night of Oct 13 - Oct 14, the PMC was locked but the IMC was unlocked.

To find if the laser noise increase as it does repeatidly when the IMC is locked we cannot look at the EOM correction, as there is no correction sent with the PMC locked only. But we can look at the PMC error signal.

Figure 1 shows the trend data, and one can see that the RMS increase by about a factor of 2 for a few minutes repeadily. The increase is much less visible than when the IMC is locked, but this is just because the PMC signals is much more noisy when the IMC is not locked and the additional loop on the master laser is locked.

Figure 2 shows the PMC signals in 4 conditions

Purple - PMC only locked and noise is low
Brown - PMC only locked and noise is high
Green - PMC and IMC locked and noise is low
Brown - PMC and IMC locked and noise is high

On figure 3 and 4 I have made spectra of the PMC signal (one per minute, with log spaced frequency) and shown a spectrogram normalized by the median over time for the data for the night of Oct 13-14, where the noise elevation can be seen.

Note that this type of analysis needs to be done quickly. The high sampled data (>100kHz) remains on disk only for ~10 days.

/users/mwas/detchar/protoFastUnlock_20201013

Images attached to this report

49705_1603285513_Screenshot from 2020-10-21 14-53-21.png

49705_1603285888_Screenshot from 2020-10-21 15-01-42.png

49705_1603286015_Screenshot from 2020-10-21 14-41-57.png

49705_1603286023_Screenshot from 2020-10-21 14-42-05.png

Comments to this report:

cleva - 15:33 Wednesday 21 October 2020 (49706)

Might be worth to measure the PMC OLTF to check it is still well adjusted together with the mechanical resonnance compensating electronic filter.

mwas - 19:42 Wednesday 21 October 2020 (49709)

Figure 1 shows that during the time when PMC only was locked (IMC unlocked) the increase in noise in the PMC channels corresponded also to an increase in noise in the master laser DC monitor (but not in the frequency monitors).

Figure 2 and 3 if you compare the normalized spectrogram of the PMC error signal and of the master laser DC monitor, the increase in noise happen at the same times. So clearly they have a common origin.

This raises two questions:

What is exactly the sensor for the PSL_ML_DC_MONIT_FS channel, and where is that sensor located in the optical path?
Is there any feedback applied to the master laser in this configuration (PMC locked, IMC locked)?

If there is no feedback applied, then it points to the master laser as the origin of the problem (or whatever that is between the master laser and the ML DC monitor). Or that the electronics that are providing the feedback are creating their own time varying noise, which could be tested by physically unplugging the cables from the master laser.

Images attached to this comment

49709_1603294683_Screenshot from 2020-10-21 17-35-37.png

49709_1603296407_Screenshot from 2020-10-21 18-06-07.png

49709_1603296411_Screenshot from 2020-10-21 18-06-19.png

cleva - 19:50 Wednesday 21 October 2020 (49712)

see 46557

ML_DC_Monit_FS monitors the ML power before the long fiber

direnzo, narnaud - 11:40 Friday 23 October 2020 (49729)

Triggered by Nicolas, I've run NonNA-corre to estimate the Pearson cross-correlation coefficient between a target channel, in this case the max or the rms of INJ_EOM_CORR from the trend frame, and a set of (330+) channels from trend: [ENV_*,PSL_*,INJ_*], either *_rms or *_max. The results are then ranked on the base of (the absolute value of) this coefficient, and listed in at the links below:

trend rms;
trend max.

Mind also the log files, with an extended list of most correlated channels.

Of course most/some correlations are trivial and expected (like the one with INJ_IMC_REFL, which however I didn't exclude from the analysis as a check) but hopefully something could be of interest. Also, feel free to post comments/requests or to suggest me other similar analyses.

mwas - 14:27 Friday 23 October 2020 (49733)

For the data between Oct 13 and 14, looking at INJ_EOM_CORR is not a good witness channel. The input mode cleaner was unlocked (on purpose) and there was no correction sent to the EOM. For that night a good channel to look at instead is PSL_PMC_REFL_I_FS_rms.

direnzo - 23:16 Monday 26 October 2020 (49749)

Thanks Michal for the hint. I repeated the analysis choosing PSL_PMC_REF_I_FS_rms as the target, and correlated it to [ENV_*_rms,INJ_*_rms,PSL_*_rms]. PSL_ML_AC_rms and many magnetometers resulted in high value of the Pearson's cross-correlation coefficient: link to the results.