Abby Wang, Athena Baches, Ansel Neunzert

This is a retrospective look at a change that happened during O3. It was motivated by our interest in understanding more fully the impact of past calibration line changes on CW/persistent data quality. The goal of this study was to (a) check whether the O3 lines list captures all notable calibration line non-linearities, or whether some entries were missed, and (b) to test out new tools for examining the behavior of lines over long periods of time.

Background: A large number of O3 calibration line non-linearities are already included in the O3 lines list. We wanted to check whether any other less apparent combs or multi-peak structures might be associated with the calibration line shift besides this family of lines.

Overview of results

• We found that some (previously noted) O3 combs appeared or disappeared on the same date as the calibration line shift. This is not conclusive evidence that they are causally related, especially since this happened on a Tuesday.
• In particular the 21.47236 Hz and 21.491227 Hz combs shifted. The 21.42848 Hz comb also shifted, but at an earlier time than the calibration line change.
• Lines at 68.5 Hz and 33.5 Hz shifted. These peaks are also teeth of the 1.0 Hz comb with 0.5 Hz offset, but the shift does not occur for all comb teeth.

Methods: We ran lalpulsar_spec_avg_long on 2-hour self-gated SFTs from the O3 Hanford data, and used the new persistence feature to track the strength over time (measured in standard deviations above the mean for each SFT) of all the lines from the O3 known line list between 10Hz and 200Hz.  Subsequently we ran these lines through a clustering tool, which we developed over the summer. The main function of this tool is to group together line artifacts that show similar varying patterns over time. We were able to identify a few combs that showed similar time changes as the calibration lines. In particular, the 21.47236 Hz and the 21.491227 Hz combs both significantly shifted up at the time of calibration line change. The tooth of the 1.0 Hz comb with 0.5 Hz offset located at 33.5 Hz also shifted up significantly at that time, while the tooth of the same comb at 68.5 Hz shifted down. We included plots of these lines along with an example calibration line for comparison. When looking through the data we did not find any other clear correlations with other combs.

Attached Plots: In Figure 1 and Figure 2, the blue lines show the strength of the calibration line at 17.6Hz over the course of O3. The red lines are the average strength of all the teeth of the 21.47236Hz comb (Figure 1) and the 21.491227 Hz comb (Figure 2), respectively. The vertical black line marks the date when the calibration line change happened (Apr 16th 2019). Both combs show significant shift at the same time as the calibration event.

In Figure 3, the red line is the average strength of the 21.428485 Hz comb, which shifts up earlier than the calibration change date.

In Figure 4 and 5, the blue lines are the calibration lines at 17.6Hz and 15.1Hz, respectively. The red lines are the lines at 33.5Hz and 68.5Hz, both of which shift at the same time as the calibration lines.

Take-aways

• This study finds possible additional impacts on non-calibration lines between 10Hz and 200Hz due to the the calibration line change in April 2019, but the impact is at most limited to the 21.47236 Hz and 21.491227 Hz combs seen in O3, and two additional lines at 33.5Hz and 68.5Hz.
• Since this change aligned with a Tuesday maintenance, it’s also quite possible that some shifts are due to other factors (not calibration lines). The third ~21 Hz comb displaying different behavior in time lends a bit of extra weight to this.
• In either case, these results provide information about the 21 Hz combs start dates, which may be useful if they reoccur.
• This example shows the promise of the new features in lalpulsar_spec_avg_long for long-duration tracking and grouping of lines and their change points.