Project

General

Profile

Regression Data: June, July (see also the page on the offset/pointing analysis using this data)

On this page we present an initial analysis of the regression data collected during the June and July engineering runs. These are very rich data sets and this analysis just scratches the surface and we hope that additional studies will be performed. Catalogs with the regression data can be found on the catalogs page. The June regression sample consists of exposures 215724 - 215923 and the July regression data is from exposure 222420 - 222631.

Observing Conditions

Observing conditions in June were quite a bit better than in July as is demonstrated by the following set of plots.

Seeing (r50) [arcsec]
June data in blue (left) and July data in red (right)
Whisker [arcsec]
June data in blue (left) and July data in red (right)
Wind Speed [km/h]
June data in blue (left) and July data in red (right)
Upper Truss Temperature [degrees C]
June data in blue (left) and July data in red (right)

Regression Script

The standard regression script Regression_Observations_All.json from /home/sispi/decam/ExposureScripts/DECam was used for June and July. Moon Avoidance was active so some exposures requested by the script might have been vetoed (we did not check the log files). The BCAM calibration logic was changed in June so the BCAM data is believed to be correct. The telescope LUT table was updated before the July run to take out a 900 micron offset in the y direction (900 microns where subtracted from all dy values in the LUT).

The regression script takes 4 consecutive exposures using the z filter in a 30, 30, 30, 90 seconds cadence on a fixed HA/declination grid. Telescope tracking is on.

The June data was taken with the default AOS-5 mode whereas the July data was taken in the AOS-1 configuration.
In the AOS-5 mode all 5 degrees of freedom of the hexapod are controlled by the AOS/Donut system. In AOS-1 mode only focus (z) is adjusted by the system. In this mode the other 4 hexapod coordinates are set based on the values extracted from the telescope LUT alone.

Exposure Cadence and Telescope Position

The script is written to use absolute (HA, dec) coordinates to position the telescope on the grid point for the first exposure. Relative coordinates (RAoffset = DECoffset = 0.0) are used for the 2, 3 and 4 exposures in each block. Since the telescope is tracking this results in a slight change in HA for each of the 4 exposures (RA is kept constant). In hinsight it might have been better to use absolute coordinates for all exposures to keep the telescope in the telescope at a fixed physical position.

id telra ha teldec tel_el az hexapod_x telescope_lut_dx bcam_dx hexapod_y telescope_lut_dy bcam_dy object
222420 339.005983 -44.403458 -75.067998 40.19 166.25 -1140.67 -1140.67 581.887 -444.415 -444.42 3582.832 Regression Test: HA -3 dec -75
222421 339.005954 -44.248792 -75.067609 40.23 166.29 -1100.7 -1100.70 528.046 -408.148 -408.17 3554.484 Regression Test: HA -3 dec -75
222422 339.005954 -44.002542 -75.067609 40.28 166.34 -1087.59 -1087.56 542.976 -396.458 -396.43 3569.16 Regression Test: HA -3 dec -75
222423 339.005983 -43.760417 -75.067637 40.33 166.39 -1075.045 -1075.044 493.232 -385.27 -385.27 3530.39 Regression Test: HA -3 dec -75
222424 325.755733 -29.903875 -75.060498 42.81 169.84 -343.927 -343.92 -3.504 238.474 238.47 3375.81 Regression Test: HA -2 dec -75
222425 325.755733 -29.745167 -75.059942 42.83 169.89 -326.932 -326.92 -22.494 252.725 252.72 3387.304 Regression Test: HA -2 dec -75
222426 325.755779 -29.494625 -75.059942 42.87 169.96 -313.278 -313.27 -34.88 263.833 263.83 3382.333 Regression Test: HA -2 dec -75
222427 325.755733 -29.252458 -75.059942 42.91 170.03 -299.873 -299.87 -53.506 274.747 274.74 3359.672 Regression Test: HA -2 dec -75

The table shows data for the first 8 exposures of the July regression script run. The first block of 4 was taken at -3 hours (HA) and -75 degrees (dec) and the second block was at -2 hours and -75 degrees. The effect of the Earth's rotation/telescope tracking can be seen in the changing HA (and AZ and EL) coordinates. RA stays fixed. Since the look up table is coded for AZ and EL we retrieve slightly different values for the 4 exposures in each block. In x and y the variation is about 40 - 60 microns. Since we are in AOS-1 mode the hexapod values agree with the telescope_lut values [allx, y values in the table are in microns].

BCAMs

It is interesting to study how the BCAM system responds to these very small changes in telescope and hexapod position. Looking at the first exposures in the table above the hexapod x position changes by +65 microns and the BCAM values change by 89 microns (the () is understood). Close but not as close as expected given the BCAM resolution. It has to be noted that not only the hexapod position changed but that the telescope moved by 0.2 degrees in HA. It would certainly be interesting to perform this test with a modified regression script that uses absolute coordinates for all exposures.
For the following study we look at the change in BCAM value between consecutive exposures in each block of 4 exposures and compare this to the change in hexapod position. We call this delta_BCAM and delta_Hexapod. For each block of 4 exposures in the regression sample we get 3 sets of "delta"s. The following set of plots shows the difference between delta_hexapod and delta_bcam for each of the 4 hexapod degrees of freedom (x, y, tip, tilt). The rows show the difference in delta for the 1st/2nd, 2nd/3rd and 3rd/4th exposure in the block. For tip and tilt the units are arcsec, for the other two the units are microns. Our expectation is that we see very narrow distributions centered at 0 - based on the assumption of 1 -2 micron BCAM resolution and that the telescope doesn't flex for this very small move.

delta Hexapod(x) - delta BCAM (x)
Data from 1s and 2nd, 2nd and 3rd, 3rd and 4th exposure in each block.
delta Hexapod(y) - delta BCAM (y)
Data from 1s and 2nd, 2nd and 3rd, 3rd and 4th exposure in each block.
delta Hexapod(tip) - delta BCAM (tip)
Data from 1s and 2nd, 2nd and 3rd, 3rd and 4th exposure in each block.
delta Hexapod(tilt) - delta BCAM (tilt)
Data from 1s and 2nd, 2nd and 3rd, 3rd and 4th exposure in each block.

The plots show approximately 1 arcsec RMS agreement in the hexapod tip (tilt) position and the BCAM tip (tilt) measurement. Note that this is a relative measurement. Looking at the distributions for x and y we see surprisingly large differences. We hope the BCAM experts can explain why the measured RMS width is so much larger than the BCAM intrinsic resolution.

The next set of plots shows the BCAM differences between the 4 exposures of each block without correcting for the hexapod motion. This allows us to check if there is the system behaves differently after the long slew at the start of the sequence.

delta BCAM (x)
Data from 1s and 2nd (red), 2nd and 3rd (green), 3rd and 4th (blue) exposure in each block.
delta BCAM (y)
Data from 1s and 2nd (red), 2nd and 3rd (green), 3rd and 4th (blue) exposure in each block.
delta BCAM (tip)
Data from 1s and 2nd (red), 2nd and 3rd (green), 3rd and 4th (blue) exposure in each block.
delta BCAM (tilt)
Data from 1s and 2nd (red), 2nd and 3rd (green), 3rd and 4th (blue) exposure in each block.

We notice that for all 4 coordinates but most obviously for x, the distribution for the first delta (the change in value between the 1st and 2nd exposure in a block) has a mean value further away from 0 and an RMS about twice as large as the second and third deltas.
Is it possible that we see some residual vibration after the long slow that affects the BCAM measurements done right when the telescope gets in position for the first image?

Pointing Accuracy and Offsets

Having one data set taken in AOS-5 mode and one with AOS-1 allows us to study the effects of the hexapod motion and the AOS adjustments on the pointing accuracy. The pointing model currently in use at the Blanco was created based on an all-sky scan in AOS-1 mode.

A more detailed analysis of the pointing accuracy for the June data can be found here.

In June we saw large offsets not centered at zero for both coordinate directions

June Ra and DEC offsets [arcsec]
RA offsets (left), DEC offsets (right)

In July without the additional hexpod motion due to the AOS correction these distributions become much narrower and centered at zero

July Ra and DEC offsets [arcsec]
RA offsets (left), DEC offsets (right)

A study of the changes in the offsets over the course of the four exposures in each position can be found here

We also checked for trends in the offsets as function of both RA and dec. The first two plots show RA_offset and DEC_offset in arcsec as function of teldec. The color of the dots encodes the hour angle.

June RA_offset as function of teldec (and HA) [arcsec]
June DEC_offset as function of teldec (and HA) [arcsec]

A strong undulation can be observed. The offsets are also not centered at 0 as was already shown by the previous set of lots.

For July the situation looks quite different:

July RA_offset as function of teldec (and HA) [arcsec]
July DEC_offset as function of teldec (and HA) [arcsec]

The declination dependents of both RA and dec offsets has disappeared. The offsets still depend on HA but to a much smaller extent compared to the situation in June.

An alternate veiw of these graphs with traced HA trends can be found here

Image Quality

We used the June and July regression data to study if the different AOS modes cause visible difference in image quality. As we demonstrated on the top the two sample were taken under quite different observing conditions. In July the conditions (seeing) got worse through the course of the regression script. While this makes absolute comparisons impossible (or at least very difficult) we can still study relative changes from the 1st to the 2nd, 3rd and 4th exposure in each block. Here we expect to see better results in the June data sample where the AOS actively controls alignment. Similar to the BCAM and hexapod studies we define quantities called delta_r50 and delta_whisker as the difference in seeing (r50) between the 2nd and 1st, 3rd and 2nd as well as 4th and 3rd image of each block of 4 exposures taken at the same position.

Seeing

The first set of plots shows the changes in seeing (r50) for the June data.
The left histogram is from the 1st and 2nd images, the next is from the 2nd and 3rd images, and the last histogram is from the 3rd and 4th images in each block.
Here is the same set of plots for the July data.

Seeing was obviously much worse in July. There appears to be an improvement from the first difference to the second difference in both plots (rms value). Note that the 4th exposure in each block had a 90 second exposure time.
Also focus was actively maintained for both tests.

Ellipticity

The first set of plots shows the changes in ellipticity (whisker) for the June data.
The left histogram is from the 1st and 2nd images, the next is from the 2nd and 3rd images, and the last histogram is from the 3rd and 4th images in each block.
Here is the same set of plots for the July data.

There appears to be an improvement from the first difference to the second difference in both plots (rms value). This seems to be true for either AOS mode.