Analysis of August SN Pointing Studies

The following three tests where performed during the August 15 engineering night:
  • Test 1: Extra focus/alignment exposure using wait=autocorrect
  • Test 2: Hexapod trim compensation using compensate_trim=true
  • Test 3: Switch to AOS-1 mode using aos_mode=SN

For now these results should still be considered as preliminary

Extra Exposure

We selected a target position (20:30 , -10 degrees) and repeatedly slewed the telescope to this position from about 15 degrees away.
SISPI then measured the offsets using the kentool tim command which gives the difference between mount coordinates (telra, teldec) and the on sky position). It then sent these offsets to the telescope for a position adjustment. A second image was taken at this position (deltaRA = 0.0, deltaDEC = 0.0 in the exposure request).
Offline we analyzed the pointing offsets in this second image using the center command which gives the offsets between target coordinates and the on sky position. These are the offsets of interested for the SN program.

Note After a regular telescope slew, the differences between the mount and target coordinates (or for the experts telra/teldec and ra/dec) are about 1 arcsec or less. However, when the TCS OFFSET command is used, the mount coordinates change but the target coordinates are not updated. All offsets reported in the telemetry database are based on the tim command and the mount position. Therefore these offsets do not reflect the improvements after the pointing correction.

The data set for Test 1 includes exposures 226629 - 226645. Some exposures are missing due to a GCS error.
It should also be noted that for ~four of the initial offsets (those with the largest target-actual, blue in the plots) the TCS had just been rebooted and the zeropoints had not been re-entered. Alistair realized that after the first one but decided to leave it as is since this was an even more stringent test correcting with a larger offset. In practice, the median AOS-1 target-actual differences for the initial slew, over much of the sky, is around 15 arcsec. This was re-confirmed in tests on July 31. For AOS-5 the target-actual differences are larger.
The following histogram demonstrates how the pointing accuracy improves with the additional exposure and the automatic correction.

Very small statistics but in all cases the total offsets defined as square root of the sum of the squares of the ra and dec offsets is now below 10 arcsec (green). The offset distribution before the pointing correction is much wider (blue).

However, this comes at the price of reduced survey efficiency. Even with the automated system it takes an average of 70 seconds + the exposure time to perform this step. Our test exposures were taken with 30 seconds but it might be possible to reduce this to 10 seconds.

We were also able to test this procedure with obstac. Eric has configured obstac to include the request for this pilot exposure every time it a block of SB observations is submitted to SISPI's observing queue. This test was successful.

Bottomline: A new auto-correction pointing mechanism was implemented in SISPI and supported by obstac that uses an extra exposure improve the telescope pointing for the SN survey to meet the 10" accuracy requirement. In terms of survey efficiency this procedure costs us about 80 seconds every time we slew to one of the SN fields.

Hexapod Motion Compensation and AOS-1 Operation

Our analysis of recent engineering data showed that the pointing accuracy is affected by the actual position of the hexapod. The hexapod position is determined by a look up table (as function of telescope position) and corrections from the active optics system (AOS). Since the TCS pointing model is based on the hexapod look up table only, it is conceivable that compensating the AOS corrections could lead to improved accuracy. We performed two tests to explore this hypothesis. For these tests we took 10 exposures as HA -2, -1, 0, 1, 2 and DEC -5, -10 with tracking on. For a baseline we took 10 exposures in the same position without any corrections.

Baseline Exposures: 226696 - 226705
Test 1 Exposures: 226710 - 226719
Test 2 Exposures: 226721 - 226730

Test 2: Correct Target Coordinates based on Hexapod Trim

Using the DECam pixel scale of 0.264 arcsec/pixel and 15 micron/pixel we determine that 1 micron in hexapod lateral motion correspond to 0.0176 arcsec shift.
We read the hexapod trims from the hexapod application and set the new target coordinates to

target_ra = requested_ra + 0.0176/3600 * trim_x
target_dec = requested_dec + 0.0176/3600 * trim_y

The test results are shown in the the plot below. There is no significant improvement but we discovered an error in the implementation. The target position, which is used by the center command as previously discussed was set incorrectly to the new coordinates instead of the requested coordinates. There are also indications that the sign of the hexapod correction should be flipped from "+" to "-". In other words we need to repeat this test before a conclusion can be reached. This method does not cause any survey inefficiencies.

Test 2: Operate in AOS-1 mode

The hexapod can be configured to accept only focus adjustments from the AOS but force all other trim values to 0, ie the hexapod x, y, tip and tilt values are set directly by the look up table. This is equivalent to the AOS-1 mode but the implementation differs slightly to allow to switch modes on an exposure by exposure basis.
The results for this test are shown in the plot below. During previous engineering studies we found improved pointing accuracy in AOS-1 mode but somewhat surprisingly this test showed no significant improvements.

Mount Position vs. On Sky Position (tim command)

Target Position vs. On Sky Position (center command)