I processed one hour of no-squeezing data taken in November 2024 (81468) for the purpose of running a correlated noise budget. This data was taken shortly after noise budget injections were run (80596), so I was able to use the measurement of the jitter noise to perform jitter subtraction from 100-1000 Hz, similar to my work in 85899.

I followed the same procedure as I documented in 85899: I plotted the whitened time series and saw two glitches, which I excised from the data by removing the individual segments with the glitches. I then mean-averaged the remaining segments. As a part of the correlated noise data collection in Nov 2024, we took a full calibration measurement, which I was able to use to generate a model to calibrate the DCPD signals. I used the IMC WFS yaw signal as a witness to subtract jitter from 100-1000 Hz. I then calculated the full correlated noise estimate also as described in 85899 (in that log I called it the "full classical noise estimate", which I now realize is confusing).

The comparison of the Nov 2024 time (O4b) and June 2025 time is shown here.

The noise below 100 Hz is lower in June 2025, however the noise above 100 Hz is higher in June 2025. It also appears that the slope of the noise above 100 Hz has changed slightly. The high frequency noise is also different, possibly because the frequency noise coupling to DARM has changed.

I also plotted the ratio of the noise from 40-300 Hz, showing that the noise below 100 Hz is reduced by up to 10% in amplitude, and above 100 Hz increase by up to 10% in amplitude.

I added the GWINC thermal noise trace to the plot above, and took the ratio of the correlated noise to the total thermal noise to that trace to highlight the change in noise. The ratio shows that the overall slope and amplitude of the excess noise, compared to the full thermal noise trace, has changed.

And going back even further, the correlated noise budget was run in O4a in Dec 2023, which we used in the O4 detector paper. We only had 900 seconds of data, so the results are more noisy. Comparing the three traces and comparing their ratios to the GWINC thermal noise trace. I did not do a jitter subtraction on the Dec 2023 data. The nearest valid calibration report is on Oct 27, which is what I used to calibrate the data.