Dark frame properties of the acm
Last updated: Mar08,2019

The acm image dark rate is usually around 0.5 adu/sec for exposure with exposure times of 20 seconds or less. In March2019 I developed the acm_BigRed routine and used it to process 36 nights of acm data. iWith the estimated dark rate of 05. adu/sec, I need to use dark frames from nights where good bias frames were taken, othewise small errors in the assumed mean bias level will dominate the dark rate derivations. In the course of my early March2019 work I found the that many of the darks from early 2018 (Feb2018 and earlier) seemed to be contaminated by lscattered light in the PFIP. Hence, I used a smaller subest of available nights. Below are the nights I will use.

 

Basepath used  = /home/sco/AD/HET_work/acm_nights 
Original number of nights to be surveyed  = 36
  Date         Nimages     Nbias    Ndark     Nopen   Nnone
  20180402       145        19        97        29        0
  20180403       133        15        15       103        0
  20180404       112        15        49        48        0
  20190205       771        32         5       463      271
  20190206        97        28        63         6        0
  20190217       585        11        12       552       10
  20190219       113        56        57         0        0

TEMP NOTES: 
  - Need to remove 20180105. 20170906  (some light source was on in PFIP) 
  - Need table_point_selector to make:    table_point_selector.images 


The procedure I use to mak a quick first dtufy using processed images is described below. I use only nights that had bot bias and dark frames. By processed I mean the image headers have been revised and the bias and fixe bias pattern signals have been subtracted.

 
In each date directory of usable nights I want to run:
  image_fullpath_list list.DARK FIXUP list.dset N        

I wrote dlists.bash to do this for a list of nights this script writes the 
fullpath processed image names # to the file:    List.Processed_DARK
# This typically takes less than a minute to run on scohome2.  
  cd /home/sco/ACM_BIAS_STUDY/reds4
  ./dlists.bash list.Date_darks

I move List.Processed_DARK into  ./Dark_study_1/ 
Then I build the table of dark image data
  acm_dark_table List.Processed_DARK N 

This creates the DARKS table file.  I can then use point_selector on the 
DARKS table file to view the dark rates. 

For examples:   point_selector DARKS exp drate1 N
                point_selector DARKS d2000 drate1 N
                Generic_Points ;  xyplotter_auto DARKS q q 20 N



The dark rate data above was gathered as described in the text on Mar8,2019. Dark rates from a total of 298 dark images gathered on 7 nights (20180402 to 20190219 UT) are shown here. The exposure time used is plotted on the X axis, and there is no strong dependence on dark rate with exposure time. The mean dark rate statistics (rates are in adu/second) are:
 
Mean                     = 0.41761
Median                   = 0.39298
Standard deviation       = 0.18605
Minimum                  = -0.06116
Maximum                  = 0.85765
Number of values         = 298
Mean error of then mean  = 0.01080

The adopted mean dark rate of 0.39 adu/sec is shown as the green line. The 1-sigma scatter of this value is 0.18 adu/sec. This rate is reasonably small and will produce a signal of about 2 adu in our average 5-sec acm exposure time. Compared to the average fsky flux in a Gunn g image of 30-40 adu for 5-sec the dark contribution would constitute a 6% systematic error.



Stacks of fark rate images for sets of 5, 10, and 20 second dark frames. In each case we used 80 iamges per stack. The dark rate images agree quite well except for the very corner portion of the image containing the readout amplifier.

I have decided to modify the acm_process.sh script so that it performs the 2-d dark subrtaction using a choice of one of the dark rate images above. To invesigate how well this works I make visual inspections of the final processed image from acm_BigRed using this recipe:

 

% image_fullpath_list list.OPEN FIXUP list.test N  

% head -3 list.OPEN
/home/sco/AD/HET_work/acm_nights/20180404/acm/20180404T054152.4_acm_sci.fits
/home/sco/AD/HET_work/acm_nights/20180404/acm/20180404T054317.8_acm_sci.fits
/home/sco/AD/HET_work/acm_nights/20180404/acm/20180404T055422.9_acm_sci.fits
% head -3 list.test
/home/sco/ACM_BIAS_STUDY/reds4/20180404/local_red/FIXUP/20180404T054152.4_acm_sci.fits
/home/sco/ACM_BIAS_STUDY/reds4/20180404/local_red/FIXUP/20180404T054317.8_acm_sci.fits
/home/sco/ACM_BIAS_STUDY/reds4/20180404/local_red/FIXUP/20180404T055422.9_acm_sci.fits

% bigds9 list.test 16 16 
 
The night 20180404 has only 48 open images, so I use that night because I can look at the images quickly. In the original processing, acm_process.sh does only the bias correcton and I see a trace of the dark signal in the lower-left corder of each processed image. I show an exmaple in the figure below that demonstrated that the revised acm_process.sh is now subtracting the dark signal.



A dark-corrcted image from acm_process.sh is shown in the left panel. The image (20180404T102256.3_acm_sci.fits) with only bias correction is shown in the right panel. The small amplifier signal in the lower-left corner is clearly visible in the right-head panel. Becasue most of our acm images are taken with exposure times of 5sec, I have hard-coded do_acm_night to call the acm_process.sh routine with the dark rate image $critfils/ACM/DR05.fits.




Back to calling page