With apologies that this is so delayed, I have prepared updated jitter cleaning. Since we're already in Observing, and the IFO isn't yet all the way thermalized, I'll push these and see how they do sometime tomorrow.
To do this, I:
- Made sure everything in /ligo/gitcommon/NoiseCleaning_O4/Frontend_NonSENS/lho-online-cleaning/Jitter/ was up-to-date and committed in the git repository.
- Found a nice looking "glitch free" section of low noise time from many hours into our long lock last night. By "glitch free", I just mean there are no major dips down in range during the time I pick +30 mins.
- Set my new start time in /ligo/gitcommon/NoiseCleaning_O4/Frontend_NonSENS/lho-online-cleaning/Jitter/Jitter_TrainSubtraction.py
- Line 44-ish, I added:
params['gps'] = 1395993819
# ER16. OM2 cold. Many hours into a lock.
- Enabled the nonsens conda environment in my terminal:
- In my terminal, in the Jitter folder, ran the training:
python3 Jitter_TrainSubtraction.py
- Checked the resulting cleaning
- In the sub-folder ..../Jitter/trained_models/ I copied the most recent .pickle file to Jitter_best.pickle:
cp Jitter_results_H_1395993819_1800_2024_04_01_16h30m24s.pickle Jitter_best.pickle
- Back in the ..../Jitter/ folder, I ran the _extractCoeffs.py script:
python3 Jitter_extractCoeffs.py
- Note that this just uses whatever is in the ...../Jitter/trained_models/Jitter_best.pickle file, so make sure to have copied forward the pickle file that you liked the cleaning from.
- Check that there is a new coefficients file in ..../Jitter/CoeffFilesToWriteToEPICS/
- This is where I stopped, since we are in Observing. However, tomorrow, the next step is to (after we've been locked and are thermalized) take a reference of H1:GDS-CALIB_STRAIN_CLEAN, and then run that _writeEPICS.py file, and then see how the cleaning is improved. If it is improved, I will accept the SDF diffs in both safe.snap and Observe.snap.
- To push the coefficients, run the following:
python3 Jitter_writeEPICS.py