Loral 3K with Arcon3.5 on 4M R-C Spectrograph

Updated April 1 1997.

The new Loral 3K CCD plus Blue Air Schmidt combination was first tested with the 4.0-m RC spectrograph during an engineering run on 15-16 February 1995. This is a thinned 3K x 1K CCD with 15 micron pixels. The CCD has a two layer AR coating and is UV flooded to maximise its QE over a wide range of wavelengths. It is flat.

All indications are that it is superior in all respects to both the Blue Air Schmidt + Reticon and the Folded Schmidt plus Tek1024. In particular we believe it to be the best choice of CCD for all the CS CCD runs in the present block of observing time.

Gain, readout noise, etc.:

The Loral 3K has two working amplifiers (lower left LL and lower right LR). However only one can be used at a time. We have set up the two video channels to have almost a factor two difference in gain, to allow the user more freedom of choice. Looking at the table below, it can be seen that LL is better optimized given that the full-well capacity is only 78000 e-. (ie LL, gain 2, gives 1.99 e/adu, 7.7 e- RON, and full well will occur at 39000 ADU). UNFORTUNATELY, the high video gain needed for LL has meant that LL suffers from some stability problems (noise bands, bias drifts) and FOR THE MOMENT, we recommend using the LR amp, at gain 4. The only advantage of gains 1,2,3 with LR is readout speed, use these gain settings only if readout time is critical for your program.

  i                                                             ARCON 3.5  /  Loral 3K
  n                                                                                                                      Full CCD
  d DCS   __Read_Noise___    ____1/Gain___   __Read_Noise__ SingleRead
  e (us)                   (ADU)                   (e-/ADU)                      (e-)                Time (s)
  x                      LL    LR                       LL    LR                    LL    LR
                 ------------------------          --------------              ---------------        ----------
  1   5             2.54  1.48                    4.33  7.82               11.0  11.6           88.2  
  2  10            3.88  2.17                   1.99  3.96                 7.7   8.6           120.5 
  3  15            5.68  2.97                   1.39  2.59                 7.5   7.7           152.4  
  4  20            7.08  3.87                   1.03  1.94                 7.3   7.5           184.4

Dark current is extremely low, 0.48 e-/pixel/hour

QE and System Efficiency:

The QE of the CCD (measured at KPNO) is:
Wavelength QE (%)
3200A  78.9
3650A  73.9
4050A  73.0
5000A  86.6
6000A  93.0
7000A  93.9
8000A  73.9
9000A  41.8

There has been some scepticism expressed that the QE figures below 3000A are very optimistic. We do not yet have any really definitive measurements of our own, but figures of around 30-40 % at 3500A may be nearer the truth. The below system efficiency figures assume that the KPNO QE measurements are correct.

The overall system efficiency (fraction of photons striking the primary mirror which are detected by the CCD) was measured using standard stars. Using grating KPGL1 (632 l/mm 4200A blaze) and a wide (10") spectrograph slit the measured efficiency was:

Wavelength Loral3K  Reticon
3000A   2.4%
3500A  10.6%   8.0%
4000A  14.1%   9.6%
5000A  18.6%  10.4%
6000A  14.3%   8.1%

The third column gives values for the Reticon #2 CCD using the same grating.

Image quality:

With a very narrow (50 mu slit which projects to 0.6666 pix), and at best focus, the measured FWHM of comparison lines is 2.3 pix. For a slit width of 150 mu (2 pix, 1.0") the best FWHM grows to ~2.6 pix, while at 225 mu (3 pix, 1.5") it is ~3.0 pix. There is slight curvature of the focal plane which results in some variation of the FWHM with position on the CCD. With the 150 mu slit the images are 3.3 pix FWHM or better over most of the chip (~4 pix in the extreme corners), while with a 225 mu slit the images are 4.5 pix or better over most of the chip (~5 pix worst case). Even the worst case images are quite symmetrical, and do not show the very broad assymmetric wings seen in out of focus images obtained with the Reticon (B A/Sch) and Tek1k (F/Sch CCDs. In general the images obtained with the Loral are much more uniform than with these other CCDs.


The gollowing table lists the coverage and dispersion (A/pix) obtained with the various gratings available for the R-C spectrograph. It is also valid for the Argus multiple object spectrograph.

Grating l/mm Blaze %
250 158 4000 11431 3.75  
400 158 8000 11431 3.75 *
510 300 10000 5999 2.01 *#
181 316 7500 5708 1.91  
kpgl2 316 4400 5708 1.91  
kpgl3 527 5500 3417 1.16  
420 600 8000 2981 1.02 #
kpgl1 632 4200 2872 0.95  
kpglf 632 8200 2872 0.95  
450 632 11000 2872 0.95  
kpgld 790 8500 2290 0.75  
kpglg 860 11000 2101 0.68  
380 1200 8000 1563 0.48 #

% Littrow value: for the actual RC spectrograph configuration the effective
  blaze wavelength is 0.92 of the Littrow value.
* This grating is silver coated and so does not reflect light below ~ 4000A
# This grating is not very efficient when used in second order.


This CCD fringes redward of about 7000A. The maximum fringe amplitude is +/-1% which occurs at a wavelength of ~8500A. The fringe spaceing is ~40 pixels. Note that the fringe amplitude for the Loral is less than that for the Reticon (+/- 3-4%). We do not yet know how well the fringes are corrected by flat fielding. The spectrograph flexes by less than 2.5 pixels or 0.06 of a fringe spacing from the zenith to +/- 5h HA. Thus dome flat fields obtained with the white spot will probably be adequate for fringe correction in many cases. Note that dome flats can be obtained at two telescope/dome positions: North 0H, +20d 40m, dome pa 218; and South 0H, -81d 00m, dome PA 039. It may help to use the flat field position according to the declination of your objects.

Nonethless, until more experience has been obtained, we recommend that users working redward of 7000A and requiring better than 1% flat fielding obtain quartz flats (using the same slit width as for the object) for each object.

Note that it is possible to switch between Ne and Quartz lamps under software control. Set the manual switch on the comparison lamp in the cage to the "Quartz position" and select Ne as the comparison lamp in setspec/instrpars. The Ne lamp will automaticaly be selected for exposures of type comp and the quartz lamp for pflats.

Steve Heathcote (sheathcoteATnoao.edu)
Alistair Walker (awalkerATnoao.edu)