NOTCam Calibrations

Content of this page:

• Default calibrations taken in tech/service nights
  
  
• Dark images
• Flat field images
• Bad pixel masks
• Non-linearity correction
• Distortion correction for the WF camera
• Illumination correction for the WF camera

Default calibrations taken in tech/service nights

1. Differential twilight flats
   16 flats (8 bright and 8 faint) per filter in JHKs for WF-Cam (minimum)
   16 flats (8 bright and 8 faint) per filter in ZYJHKs for WF-Cam (goal)
   
   This is done using the script notcam.skyflat described under NOTCam scripts.
   
   Raw flats are always stored for download on our ftp site at /pub/service/calib/. Differential master flats are made available in the NOTCam flats archive on a best effort basis.
   
   
2. NIR photometric standard star fields
   One (minimum) or two (goal) NIR standard star fields at 3 airmasses in JHKs (WF-Cam) if the sky is photometric or clear
   
   This is done using the scripts notcam.as**-jhk described under NOTCam scripts.
   
   Raw standard star images are available for download on our ftp site at /pub/service/calib/.
   
   
3. In-situ spectral lamps for each spectroscopy target and wavelength region
   Wavelength calibration lamps (Argon and/or Xenon) are taken on target, for the wavelength regions used, due to instrument flexure.
   3 spectral flats (Halogen lamp) are taken on target.
   
   This is done using the script notcam.easy-calib described under NOTCam scripts. It takes 2 minutes per wavelength region + 1.5 minutes to mirror cover closing and opening.
   
   These calibration frames are provided together with the science frames they are connected to in the data directory of each proposal.
   
   
4. 6 darks in each of the exposure modes used during the night
   If your data was obtained as "frames t N", darks will be taken in the morning as "dframes t N", vice-versa for "exp t" and "mexp t N" the corresponding dark mode will be used. Useful to make hot pixel masks.
   
   This is done at the end of the night using the script notcam.notcam-calibs described under NOTCam scripts. It may take hours.
   
   Darks are stored for download on our ftp site at /pub/service/calib/.
   
   

Note that telluric standards must be selected by the PI with coordinates and finding charts added, along with how frequently a telluric standard should be taken in the case of very long spectra.

Darks

Note that many aspects of the NOTCam darks are not well understood.

	Dark-0-s	Single dark of zero integration time, taken with the command dark 0. The vertical gradient in each of the 4 quadrants (called "bias tilt" or "dc-gradient"), with highest amplitude in the first few readout lines is a known effect with the Hawaii arrays referred to as "the reset anomaly". Darks are taken regularly as part of our Detector Quality Control. Measurements are done using both readout modes (reset-read-read: dark t or mdark t N, and ramp-sampling: dframe t N). Zero second darks can only be taken in reset-read-read mode. This zero second dark is the new one taken with the new clock-boards, i.e. valid from January 2010 and onwards. Compare with the old zero dark
	Dark-42-s	Single dark of 42s integration time, taken with the command dark 42. Note the increased number of hot pixels with respect to the zero second dark. This dark is taken on Jan 17th 2010 after the upgrade of the clock-boards and practically all the previously present shift-register glow has now been removed.
	Dark-42-s	Single dark of 42s integration time, taken with the command dframe 4.2 10. Note the increased number of hot pixels with respect to the zero second dark. This dark is taken on Jan 17th 2010 after the upgrade of the clock-boards and practically all the previously present shift-register glow has now been removed. Compare with the old 42s dark. Note also the difference between the two 42 second darks taken with the two readout modes reset-read-read (dark 42) and ramp-sampling (dframe 4.2 10) where the latter is a linear regression result of the 10 non-destructive readouts, something which strictly speaking makes sense only if the dark current is linear.
	Dark-360-s	Single dark of 360s integration time, taken with the command dframe 60 10. This dark is from April 18th 2012. The structure of this dark is similar to the 42 second dark taken with the dark 42 command. These are both "long" darks in the sense that the time between reads is longer than 30 seconds in both cases. The zero dark and the dframe 4.2 10 dark look similar because in both cases the time between reads is "short", i.e. shorter than 30 seconds.

Flat field images

	Flat-WF-raw	Raw single skyflat taken in the Ks band with the WF-camera on 08-Sep-2009. The flat field structure for the WF camera is similar for all filters. A number of bad pixel groups are seen. The number of stars present shows that this particular blank field (Blank4) is not good for the near-IR.
	Flat-WF-master	Master skyflat for the WF-camera (from 08-Sep-2009). Composed of dithered twilight bright and faint images observed using the sequencer script notcam.skyflat. The differential flats produced by subtracting faint from bright images (taken at the same pointing within a given time interval of 25 min or less), subtract out the thermal emission, as well as the dark. The dithering ensures that stars are filtered away. Bad pixel correction (by interpolation) is made only for the two dead columns. No spatial smoothing is applied to the flat. Check the NOTCam flatfield archive for master flats in fits format.
	Flat-HR-raw	Raw single skyflat taken with the HR-camera on 08-Sep-2009. The flat field pattern looks similar for all filters.
	Flat-HR-master	Master skyflat for the HR-camera (from 08-Sep-2009). Produced as described above for the WF camera master flats. Check the NOTCam flatfield archive for master flats in fits format.

Stability of flats

The below ratio images of master flats is an example.

	Ratio-WF-flats	Ratio images of WF-cam Ks, H and J master skyflats from 7-Jan-2012 and 12-Oct-2011.
	Ratio-WF-flats	Diagonal cut across the ratio images of WF-cam Ks, H and J master skyflats above.

Bad pixel masks

	Badpixel-zero	Badpixel image showing only zero-valued pixels. These comprise mainly the two dead columns (at x=512 and x=1024; NB! a controller problem and not physically dead pixels on the array!) and a number of individual dead pixels, most of which are located in the upper left corner. Masks have value 0 in general (green) and 1 for bad pixels (red). Check the NOTCam badpixel mask archive for badpixel masks in fits format. A typical zero-pixel image for the SWIR3 array is called "bad_zero_sci.fits", and the fraction of zero valued pixels over the whole array is 0.0027.
	Badpixel-cold1	Badpixel image showing "cold" pixels (Dec-07). These pixels have a much lower response than normal pixels (less than 5σ) . Usually they form small groups. Masks have value 0 in general (green) and 1 for bad pixels (red). Check the NOTCam badpixel mask archive for badpixel masks in fits format. The cold pixel mask of SWIR3 upon installation on 13/12-2007 and until the first opening in March 2008 is called "bad_cold1.fits". In addition to the upper left corner, there is one larger bad-pixel feature with "cold" pixels having typically about 50% response compared to the surrounding good pixels and comprise 0.0021 of the array.
	Badpixel-cold2	Badpixel image showing "cold" pixels (Jun-08). The number of bad features in the SWIR3 array increased slightly after the opening during March 2008. The bad pixels typically have 20-40% response compared to the surrounding good pixels. Masks have value 0 in general (green) and 1 for bad pixels (red). Check the NOTCam badpixel mask archive for badpixel masks in fits format. The mask "bad_cold2.fits" is valid from Apr-08 to the 9th of June 2009 and comprise 0.0032 of the array.
	Badpixel-cold3	Badpixel image showing "cold" pixels (Sep-09). Badpixel image showing the bad features of the SWIR3 array after the incident on the 9/6-2009 (shaking telescope). The number of bad pixel groups increased from one hour to the next, believed to be caused by the telescope shaking. The bad pixels have typically 20-40% response compared to the surrounding good pixels. Masks have value 0 in general (green) and 1 for bad pixels (red). Check the NOTCam badpixel mask archive for badpixel masks in fits format. The mask "bad_cold3.fits" is valid from the 9th of June 2009 and until 13th of Dec 2010. The cold pixels in this mask comprise about 0.0041 of the array.
	Badpixel-cold4	Badpixel image showing "cold" pixels (Feb-11). Badpixel image showing the bad features of the SWIR3 array after the opening of NOTCam in January 2011. The bad pixels have typically 40-80% response compared to the surrounding good pixels. Masks have value 0 in general (green) and 1 for bad pixels (red). Check the NOTCam badpixel mask archive for badpixel masks in fits format. The mask "bad_cold4.fits" is valid from the 12th of Feb 2011 until Dec 2015. The cold pixels in this mask comprise about 0.0037 of the array.
	Badpixel-cold5	Badpixel image showing "cold" pixels (Jan-16). Badpixel image showing the bad features of the SWIR3 array after the opening of NOTCam in January 2016. The bad pixels have typically 20-80% response compared to the surrounding good pixels. Masks have value 0 in general (green) and 1 for bad pixels (red). Check the NOTCam badpixel mask archive for badpixel masks in fits format. The mask "bad_cold5.fits" is valid from January 2016. The cold pixels in this mask comprise about 0.0047 of the array.
	Badpixel-hot-05s	Badpixel image showing "hot" pixels in short (~ 5s) exposures. These pixels have a higher response than normal pixels. They are typically located along the edges of the array. Note that the number of "hot" pixels depends on the integration time, and must be properly treated. Masks have value 0 in general (green) and 1 for bad pixels (red). Check the NOTCam badpixel mask archive for badpixel masks in fits format.

Non-linearity correction

The correction must be applied to the raw images as a first step in the data reduction procedure. The raw image (x) is to be corrected as follows using the correction coefficient images ba and ca:

Here we have assumed that the array behaves linearly for low fluxes. From the "tb06" dataset the non-linearity behaves on the average roughly as follows:

Counts       Non-linear
 (adu)          (%)
-------------------------
  5000        0.05
 10000        0.23   
 15000        0.52
 20000        0.91
 25000        1.42
 30000        2.04
 35000        2.76
 40000        3.55
 45000        4.43
 50000        5.37
--------------------------

Testing the non-linearity correction coefficients based on the "tb06" data set on domeflats at later dates, such as tf19, th01, th28, and ti28, gives rather stable results. The plots below show raw data (symbols) and corrected data (lines) for selected areas in each quadrant: q0 or lower left (white), q1 or lower right (red), q2 or upper right (green), and q3 or upper left (blue). The y-axis shows the count level per second and the x-axis shows the exposure time. The non-linearity correction seems to work well up to levels of at least 46000 ADUs in the raw data.

Distortion correction WF-cam

New models for the distortion in the J, H, and Ks bands were found from sharp (fwhm ~ 0.5'') images of a stellar rich field in Baade's Window in June 2009. About 300 2MASS stars with positions accurate to 0.08'' over the Ks magnitude range from 9 to 14 are present in the selected field.

A total of 260 stars (avoiding multiples unresolved by 2MASS), each at 9 positions on the array, gave a grid of 2340 XY positions with reference coordinates per filter. The IRAF task GEOMAP was used with a 4th order Legendre surface polynomial to produce a spatial transformation for each filter. The result is stored in the IRAF database "notcam.db" which can be downloaded from the NOTCam distortion correction archive.

Blinking an uncorrected with a distortion corrected image.

In version 2.4 of the IRAF quick-look reduction package notcam.cl (download and installation info) the tasks reduce and reduce_bs include optional distortion correction. If selected, the correction is applied after flat-fielding, bad-pixel correction and sky subtraction on the individual images using the IRAF task GEOTRAN. Note that the WCS info in the header is not updated and is not in agreement with the distortion correction. Support does not yet exist in the IRAF WCS interface for the higher order distortion corrections that GEOTRAN is capable of performing on the image.

The model is valid for the SWIR3 array and the internal focus value set to 5650, which is the default for imaging with the WF-camera. (Note that for other internal foci the correction will not work.) The correction may work better for some data than other, but no clear dependence has been found with rotator position or altitude. One corner in particular is hard to correct.

H-band image of the field used to make the model. To the left uncorrected, and to the right corrected.
Result of 9 dithered images combined.

Illumination correction for the WF camera

For the WF-camera H-band Janz, Laurikainen, Lisker et al. obtained calibration data during photometric nights in March 2012 and present a model for the surface correction for this filter in their paper Janz et al. 2014, ApJ 786, 105. It is based on measuring one standard star repeatedly scanning it over 25 positions on the array 15 times.

It is unclear how stable this is with time, but the corrections agree with the trends seen in previous analyses of clusters. Similar modelling for the J and Ks bands with the WF-camera is currently on-going.

The fits file is obtained here: FITS image.

Photometric surface obtained for WF camera and H-band during March 2012 by Janz et al. 2013.
The contours go from 1.05 at the top through 1.02, 0.99, 0.96 to 0.94 at the bottom corners.