Kappa_UIM has seen a huge increase in noise starting ~8 days ago (screenshot attached). It's not clear to me what is causing this yet. This is also shown on the summary pages: https://ldas-jobs.ligo-wa.caltech.edu/~detchar/summary/day/20230728/cal/time_varying_factors/ The UIM Kappa line value started to show noisy activity at GPS 1373905226.2881477.
There's a feature just above the corresponding 15.6 Hz calibration line which appears at the identified time. Fig 1 shows this feature in high resolution. It peaks around 15.605 and is present consistently (in Fscan daily data) since the date of the noise change.
I also computed some shorter-duration spectra with gwpy to double check that its appearance corresponds to the gps time Louis posted. I needed about 500s fft length to resolve the relevant features, and I wanted a couple of averages so I ended up looking at 1000s time periods. Apologies for the messy overlay of figures with not-precisely-matching y-axes! I think the shape difference is clear regardless of the scale. Fig 2 shows some samples right before the change. The change occurs near the end of an observing segment but low-noise data remains available right after, so I looked at both the immediate time after the change (fig 3) and the next observing segment (fig 4). Indeed, it looks like a good match.
This is caused by a mistake in the MICHFF filter. It turns out that the filter I retuned on July 20 has a sharp feature at 15.6 Hz that I did not nottice before. This is injecting MICH noise in DARM at 15.6 Hz. This can be fixed by either tweaking the current filter, or retuning the MICHFF again (being more careful with narrow feartures in the fit!)
I modified the current filter to remove the 15.6 Hz feature and saved it. We should reload the MICHFF filters at the first opportunity
This filter was reloaded since we lost lock. We should see an improvement in our next lock.