‎

25. Detector behavior over time Appendix

In section 11.2 we covered the median charge and energies of clusters in the background and calibration data. Here in the appendix we present a few additional plots. First, in sec. 25.1 are the equivalent figures to fig. 43 for other time periods. The goodness-of-fit tests mentioned in sec. 11.2 for the selection of the best interval length are shown as well. In sec. 25.2 we have the equivalent figures to fig. 41 for the other run periods.

25.1. Choice of gas gain binning time interval

The following figures show the behavior of the different time intervals for the choice of 'ideal' gas gain time binning for all run periods (not in the sense of Run-2 and Run-3, but those split by significant off time). In addition fig. 2 shows the results of applying a range of goodness of fit tests to the cluster data (we use a plot and not a table for easier visual parsing).

Note that the repository of this thesis contains even more figures related to this in the Figs/behavior_over_time directory.

Figure 1: Figure 125: Kernel density estimation of the median energies split by the somewhat distinct run periods and the time intervals used. A KDE instead of a histogram is used as the binning has too large of an impact for the dataset.

Figure 2: Figure 126: Comparison of the different time intervals in each run period using a set of different goodness of fit tests. The \(\SI{45}{min}\) interval seems optimal in the 30/10/2017 period, but worse in others. The \(\SI{90}{min}\) interval is average in most cases.

Figure 3: Figure 127: Equivalent plot to fig. 43 for data from Feb 2018 to Apr 2018. Ridgeline plot of a kernel density estimation (bandwidth based on Silverman's rule of thumb) of the median cluster energies split by the used time intervals. The overlap of the individual ridges is for easier visual comparison and a KDE was selected over a histogram due to strong binning dependence of the resulting histograms.

Figure 4: Figure 128: Equivalent plot to fig. 43 for data from Oct 2018 to Dec 2018. Ridgeline plot of a kernel density estimation (bandwidth based on Silverman's rule of thumb) of the median cluster energies split by the used time intervals. The overlap of the individual ridges is for easier visual comparison and a KDE was selected over a histogram due to strong binning dependence of the resulting histograms.

\clearpage

25.2. Correlation of gas gain and ambient CAST temperature

Let's look at the rest of the data not shown in sec. 11.2.2. First in fig. 5 we see the same plot as fig. 41, but only for Run-2 data from 2017. The anti-correlation is not quite visible here, instead in parts it seems like the expected correlation of temperature and gas gain is visible. Fig. 6 is the data for Feb 2018 to Apr 2018. Here there seems to be some of the anti correlation, but less than in the Run-3 data presented in the main of the discussion. Finally, fig. 7 shows the gas gain of the center chip plotted directly against the ambient temperature at CAST as a scatter plot. Here the anti correlation becomes very visible as a global trend.

correlation_fePixel_all_chips_gasgain_period_2017-10-30.svg — Figure 5: Figure 129: Normalized data for Run-2 (only 2017) of the temperature sensors from the CAST slow control log files compared to the behavior of the mean peak position in the \cefe pixel spectra (black points), the recovered temperature values recorded during each solar tracking (blue points) and the gas gain values computed based on \SI{90}{min} of data for each chip (smaller points using Viridis color scale). The shift log temperatures nicely follow the trend of the general temperatures. In this period no real anti-correlation is visible. Instead in parts it looks like the expected proportionality between temperature and gas gain appears.

correlation_fePixel_all_chips_gasgain_period_2018-02-15.svg — Figure 6: Figure 130: Normalized data for Run-2 (only Feb. to Apr. of 2018) of the temperature sensors from the CAST slow control log files compared to the behavior of the mean peak position in the \cefe pixel spectra (black points), the recovered temperature values recorded during each solar tracking (blue points) and the gas gain values computed based on \SI{90}{min} of data for each chip (smaller points using Viridis color scale). The shift log temperatures nicely follow the trend of the general temperatures. Here the anti correlation seems to be visible in some parts, but also less extreme than in the end of 2018 Run-3 data, presented in the main section.

Figure 7: Figure 131: Gas gains of the center chip (by \(\SI{90}{min}\) time slices) against the ambient temperature at CAST. As a general trend the anti correlation is very visible. Also visible though is that for the 2017 Run-2 data that effect does not really appear.

Click on any heading marked 'extended' to open it