Pointing model run 2021/11¶
For this run we took data at higher elevation where the previous pointing model datasets were lacking. Alone, this dataset is not enough to constrain a good pointing model so the idea was to combine it with data taken on Pointing model run 2021/09.
Observations¶
The data for this run was taken following a similar procedure to that of Observations.
A summary of the observations is shown in Observation journal.
Night | Number of observations |
---|---|
2021/11/07 | 16 |
Data Analysis¶
The data analysis follows a similar procedure to that of Data Analysis.
EFD Data Mining¶
papermill -p year 2021 -p month 11 -p day 04 -p time_window 2 -p data_path /readonly/repo/main reducing_pointing_data.ipynb reducing_pointing_data/20211004_tw002.ipynb
Generating Pointing File¶
papermill -p pointing_data_file data/20211104/AT_point_data_20211104_tw002.pickle build_pointing_data.ipynb build_pointing_data/20211104_tw002.ipynb
Generating Pointing Model¶
As mentioned before, this dataset consist mostly of observations taken at higher elevation to complement previous observations done with a lower maximum elevation limit.
In order the derive a pointing model we then combined these observations with those taken on Pointing model run 2021/09. These are the only dataset taken with the same set of LUTs for the hexapod and therefore, the only ones we can actually combine.
Fitting a simple 7-parameter pointing model to the data yields;
coeff value sigma
1 IA -290.76 2.414
2 IE -246.83 2.276
3 CA +86.71 1.520
4 AN +39.35 0.555
5 AW +68.34 0.548
6 TF -98.54 9.053
7 TX10 +26.91 4.012
Sky RMS = 4.45
Popn SD = 4.65
Figure 6 below shows the diagnostics plots from this model.
As with the other cases we still obtained some small correlated errors on some of the diagnostics plots. A higher order pointing model is cabable or considerbaly reducing these errors and produces a small improvement on sky RMS.
In this case, we obtain the following model:
And the diagnostics plots for this case is shown in Figure 7.
Nevertheless, we still advise on using the simpler model above as the higher order model will likely suffer from overfitting.