Note
This document describes the process to build the pointing model for the Auxiliary Telescope.
The process of building pointing model for the Vera Rubin observatory is common to both the Auxiliary and Main Telescopes. This tech-note documents the process used to build the pointing model for the Auxiliary Telescope at different stages in the project. Each new pointing model run is described in a dedicated session.
The pointing model is part of the pointing component (ATPtg and MTPtg). Although they are separate components, the code base is the same for both telescopes. That means that all the pointing kernel functionality is exactly the same on both cases, the main differences are in the SAL layer of the components. For more information see the pointing component documentation.
1 General Principles¶
The pointing model contains some realistic physical representation of the most common mechanical properties observed in telescopes plus generic polynomial and harmonics terms.
The model input data consists of a series of observations taken on sky where the azimuth and elevation position of the telescope is registered after centering the start in the field of view. The actual data contains the position with and without the offset required to center the target, plus the rotator angle (in case of AltAz mounts).
Registering the data in a format that can be later used to compute pointing models is supported by the pointing components by using the pointAddData command.
The following in an example dataset taken with the Auxiliary Telescope:
LSST Auxiliary Telescope, 2020 Jan 26 UTC 1 47 15
: ALTAZ
: ROTNL
-30 14 40.2
333.0949589 78.6684282 333.0167132 78.7625213 293.6668950
240.4719461 44.8995879 240.3726938 44.9816645 306.1848271
226.7459782 40.0365061 226.6556824 40.1055927 315.6175556
196.2657687 36.7445377 196.1729864 36.8002718 317.5486844
172.6853974 53.4019806 172.5781383 53.4555115 302.4060731
154.9764958 73.0126568 154.8450967 73.0685376 280.5258586
120.8814629 55.6118420 120.8052043 55.6589219 303.2002837
111.9944970 48.5062414 111.9290073 48.5488632 310.8539746
117.3470395 30.9244814 117.2775891 30.9685963 332.6099343
117.0992858 31.6455324 117.0286376 31.6953479 106.7711023
115.5267384 36.6238119 115.4605335 36.6677284 108.7757071
090.7710199 51.0219367 090.7150525 51.0742177 112.0793148
336.8322735 66.9273300 336.7782862 67.0239479 199.2589405
302.6581446 35.2470238 302.5773588 35.3348704 225.1006034
184.1741942 70.6586995 184.0197970 70.7229965 354.5840128
035.2813650 41.9310769 035.2287450 42.0025788 151.8577754
043.0139225 27.5764544 042.9570499 27.6509279 143.9913851
END
One important aspect to keep in mind when obtaining data for a pointing model is the interplay between telescope position and optical collimation. If the telescope optics is not collimated when taking the data, or if collimation changed over time (due to mechanical reasons or else), one may need to recompute the pointing model either with new data or reformatting the data to be compatible.
1.1 Updates with Pointing Component v3¶
As with version 3 of the pointing component, the pointing data is now published to DDS (and, consequentially, stored into the EFD) when users issue the pointAddData command. This makes it much easier to mine the data from the EFD and process them to generate pointing data.
This is especially useful when an acquisition image is taken when “tagging” a particular position. The process to gather the data then becomes:
- query the EFD for instances of the pointing data event,
- for each instance of the event, query; - the closest image taken, - the telescope position, - the rotator position, - (ATAOS) x and y offsets of the hexapod with respect to the look-up-tables,
- for each image, compute the image x and y offset of the brightest source with respect to the boresight,
- convert image x and y offset to telescope elevation and azimuth offsets,
- compute the telescope elevation and azimuth offsets due to offsets in the hexapod,
- Add the offsets to the “measured” axis in the pointing data,
- Save a corrected pointing file.
This process is made available by two parameterized Jupyter notebooks in ts_analysis_notebooks. The idea of using parameterized notebooks is that users can easily run both of them with papermill, considerably improving reproducibility.
2 Building an initial pointing model¶
When we first started on-sky observations with the Auxiliary Telescope, we had a small CMOS camera and an ocular, that we could fit into the camera socket on Nasmyth 1. Because the camera field of view is small (~30 arcsec), we initially relied a lot more on the ocular to find stars and center them in the field than on the camera.
We started by observing only a handful of targets. Although it is not possible to constrain a good pointing model with these limited datasets, we can at least remove the first order terms of the pointing error (e.g. azimuth and elevation zero points, see fitting pointing model furthermore), which then allows us to use the CMOS camera to obtain data more efficiently.
It is also important to emphasize that during this earlier stages we did not have active control of the optics collimation, so the data taken to build this initial pointing model had to be superseded later.
3 Refining Pointing model¶
The main point to consider when refining the pointing model is that a good pointing model requires a good lookup table for the optics. Nevertheless, one requires a good pointing model in order to build a good optics lookup table.
To solve this circular dependency between pointing model and optics lookup table we perform an iterative approach, improving each one of these at each iteration.
Once an initial pointing model was computed using the CMOS camera, effort was dedicated into obtaining a more active control of the optical alignment.
With the initial pointing model we were able to point and track targets more reliably over a good portion of the sky, even with the small field of view of the available camera, which allowed us to build better lookup tables for the optics.
This work was initiated with the CMOS camera and later redone with LATISS.
With the new camera installed and an initial version of pointing model and optics lookup table, we were able to sample more targets in the sky more efficiently.
4 Observing Runs¶
The following is a list of the different pointing models runs. Each run contains the description of the observations as well as the process to obtain a new pointing model.
5 Fitting Pointing Model¶
The pointing model fit is done using tpoint.
To facilitate use of the software we provide it as a docker image, that can be downloaded with:
docker pull ts-dockerhub.lsst.org/tpoint:latest
Because tpoint generates some plots during steps of the fits, the container must be run with some special options to allow the displaying of those graphs.
On a Mac, make sure you have XQuartz installed, and that the Security option Allow connections from network clients, in Preferences is enabled.
If the option is originally disabled, you will have to enable it and restart XQuartz for it to take effect.
Figure 8 XQuartz preference window. In the “Security” tab, make sure “Allow connections from network clients” is enabled.
Once XQuartz is configured to allow network clients, you also have to add local host to the permission table. To do that run the following command:
xhost + 127.0.0.1
To run the container and analyze some data make sure you have the data in a directory and export it to the container. When starting (e.g. running) the container, it is important to also pass the options to allow it to run the graphs, as state above. The command will look something like the following:
docker run -it --rm -e DISPLAY=host.docker.internal:0 -v /tmp/.X11-unix:/tmp/.X11-unix -v ~/data:/home/saluser/data ts-dockerhub.lsst.org/tpoint:latest
In the command above, it is assumed that the data is stored in a local directory data in the user home folder.
Change it appropriately to match the actual path.
The additional options (-e DISPLAY=... and -v /tmp/...) are the ones required to enable container UI interaction.
Once you run the command above your terminal will be inside the container and ready to run tpoint. To do that, simple do:
$ tpoint
+ - - - - - - - - - - - - - - - - - - - - +
| TPOINT |
| Telescope Pointing Analysis System |
| Version 21.8 |
+ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ +
Copyright 2016 P.T.Wallace. All rights reserved.
Licensed to AURA Inc. for use with the 8m Large Synoptic Survey
Telescope and 1.2m auxiliary telescope. Supplied December 3, 2016.
There are 73 standard pointing terms.
Reading procedures from file procs.dat ...
The library now contains 537 lines.
Reading star catalog entries from file stars.dat ...
The catalog contains 210 stars.
TPOINT ready for use: type HELP for assistance, END to quit.
*
Load the data:
* indat data/20200314/AT_point_file_20200315T022713.dat
LSST Auxiliary Telescope, 2020 Mar 15 UTC 2 27 13
: ALTAZ
: ROTNR
-30 14 40.2
154.5139233 46.1577658 154.4096955 46.2199866 314.4605341
357.5678314 17.9506751 357.4900009 18.0435750 288.4585002
357.1805839 31.6160368 357.1093148 31.6954934 301.9815193
358.2118688 56.8783782 358.1454349 56.9733136 327.0756924
....
316.6631312 56.6064178 316.5753298 56.7093257 327.2345760
313.9002173 42.8905503 313.8108478 42.9843875 313.7268993
313.1359793 31.2112474 313.0527900 31.2975443 301.9927343
319.2054673 17.4259367 319.1254117 17.5238048 287.9991321
END
*
The first thing you may want to try is the fauto option, which will perform a full model fit automatically.
* fauto
...
After running this command you will see outputs indicating that the fit is ongoing and plots with different diagnostics. Sit back and wait until the fit is concluded (which may take some time, depending on the dataset).
Figure 9 Diagnostic plots generated by tpoint when running fauto with the 202003 dataset.
Once the fit is done tpoint print the final model and some quick diagnostics:
...
coeff change value sigma
1 IA +0.014 -269.50 2.568
2 IE +0.00
3 & HACA2 +0.020 -11.89 1.982
4 & HASA7 -0.019 -12.28 1.970
5 & HESACE -0.044 -11.55 2.696
6 & HESA2 +0.004 +2.85 1.291
7 & HESA3 -0.002 -2.53 1.116
8 & HESE -0.050 -223.21 17.816
9 & HESE5 +0.238 -239.98 85.831
10 & HESE7 +0.159 -191.78 55.371
11 & HECE7 +0.242 -228.10 108.508
12 & HESE8 -0.136 +136.23 62.538
13 & HECE8 +0.010 -46.32 19.075
14 NPAE -0.031 -50.97 3.614
15 CA +0.00
16 & HSCA5CE8 -0.019 -6.81 2.001
17 & HSSA6CE6 +0.009 -4.86 1.929
18 & HSCE3 -0.039 -7.39 2.839
19 & HSSE8 -0.012 -5.49 1.406
20 & HSCE8 +0.004 -5.82 1.368
21 AN +0.002 +44.26 1.130
22 AW +0.033 +69.89 1.428
23 TF -0.431 +491.27 131.179
24 TX10 +0.170 -126.69 50.740
Sky RMS = 5.46
Popn SD = 7.72
TX10 and TF cannot reliably be distinguished.
Observation #7 is a very weak outlier candidate.
*
Note that tpoint provides the Root mean squares (RMS) and Power Spectra density (PSD) of the fit, to help diagnose the quality of the fit.
It is also highly recommended to run a “manual fit” of the data. Instructions on how to perform this task can be found in tpoint documentation.