Re: A brief summary of dark current tests performed on 30 January 2001.
R. Blum and N. van der Bliek

Purpose: Verify use of OSIRIS on side mounted port of the SOAR telescope Instrument Selector Box (ISB) at the Naysmith focus in high resolution spectroscopic mode.

Summary: OSIRIS is an uplooking cryogenic instrument cooled by liquid nitrogen (LN2). Use on SOAR will require that it remain stable with respect to internal temperature through full rotations on the Naysmith ISB. An inversion of the instrument results in a loss of most the LN2 capacity, leaving approximately 8 liters in the tank. The following tests were carried out with OSIRIS oriented variously in up-looking, side-looking, and down-looking positions.


Based on the tests, we believe OSIRIS can be used on SOAR with essentially no modification with respect to LN2 capacity for times upto 10 hours after minimum LN2 volume.

Data taken up to 11 hours after minimum LN2 would probably be generally useful, but performance is not optimal. Performance begins to degrade noticably between 11 and 12 hours after minimum LN2, even though the average dark current is still low. The problem is that many warm pixels begin to appear which do not reliably subtract between images taken closely spaced in time. This results in an unacceptable pixel to pixel scatter for faint object spectroscopy.

Filling the dewar midway through the night would provide essentially indistinguishable performance from current operations and would allow normal observing even on long winter nights.

Test Description:

Dark frames (filter dark slide and pinhole in beam) were made with OSIRIS in the front room (kitchen) at the 4m telescope on Cerro Tololo on the 30th of Jan 2001. The ambient temperture was probably about 22 C. This is much warmer than typical nighttime observing temperature, thus the test is conservative with respect to boil off. The instrument was cooled with a full load of LN2 at 1500 hrs on 29 January. It was re-filled 2.5 hours later, and these tests began at 1100 hr on 30 January without further addition of LN2. Thus the instrument was allowed to come to equilibrium for approximately 17.5 hours which is typical for the start of an observing run.

Dark frames were taken in series of 3.24 sec, 30 sec, and 300 sec with the instrument at various positions:

Time elapsed:

0 : series up-looking with remaining LN2 from fill of 29 January

0:42 (zero hours, 42 min) : series side-looking, spilling much of the LN2

1:20 : series down-looking, now at minimum LN2 volume in tank. Subsequent boil off and dewar orientation limits dark current stability and level.

1:20 to 10:55: repeat series up, side, down

10:55 to 15:10: 300 sec darks down looking (LN2 is far from detector, camera, and filter wheel).


A summary plot of the dark current shows negligible impact in the stability or level of the dark current until after the final series has begun at elapsed time >  10:55. This corresponds to a time of 10 hours since minimum LN2. After this time, the dark can be seen to increase slowly until a time near image 406 which corresponds to 13 hours since minimum LN2, where it then increases more rapidly. Our plot uses the running sum of exposure time for the elapsed time so the true dark current is easier to see, but the actual time shown is compressed by several hours since this does not include overheads in moving the dewar, eating lunch and dinner, etc. The labels represent the true elapsed time for the experiment. The spikes are real, but are not "dark current," see the discussion below.

The dark current can be seen to increase over a long time baseline, but this is very gradual. The initial dark current is approximately 0.1 e-/sec and it increases to 1.3 e-/sec at image 406. This may be compared to the expected background levels for R=3000 spectroscopy. Typical average background rates at J, H, and K on the Blanco 4m are 0.1, 0.6, and 0.9 electrons per second per pixel, respectively.  The read noise is 15 electrons. Therefore SOAR spectroscopy will be dark current limited for long exposures (180 to 300 sec) and faint targets (15th to 16th mag at J and H) when the dark reaches about 0.75 to 1.3 e-/sec. For lower dark or shorter exposures the instrument is read noise limited. At K, the background can typically be larger than the dark current except at times well after minimum LN2 volume (> 10 hr). In all cases portions of the spectrum will be background limited where OH lines are present and at the long wavelength end of the K band.

By plotting the dark data by image number each series is given the same weight in "time." From this plot, it is obvious that there is a non zero bias level  which shows up in short exposures (after subtracting the post-bias value of 96 ADU). This is not "dark" but something systematic in the bias level. The magnitude of this effect does not contribute to the dark level shown in the plot above for the 300 sec exposures (factor of 10 less than dark current).

The true test of how well the instrument performs is differencing images taken in closely spaced sequence (this is the standard mode of observing). The table below shows individual images taken during our test as well as differences corresponding to these images. We find the following results from analysis of difference images:

Upto 10 hours after minimum LN2 volume, the operation of OSIRIS is nearly normal; a small increase in dark current is observed, but image subtraction appears stable.
Upto 11 hrs after minimum LN2 volume, the 300 sec exposures are still useful. The pixel to pixel scatter in the difference of two images taken 15 min apart is typically 10-12 ADU about the array (measured in 10x10 pixel boxes), compared to 7-8 for images taken with normal LN2 levels.
Between 11 and 12 hours after minimum LN2, the pixel to pixel scatter begins to increase from warm pixels which are not subtracting as well. While most of the pixels are well behaved, significant regions show higher standard deviations in 10x10 pixel sub groups. We think R=3000 spectroscopy would be compromised for faint objects (14th to 15th mag at J, H) in this time interval.
The upper right corner shows evidence of a light leak. It subtracts well and is present from the beginning. The upper left corner shows a pronounced "warm spot" which may be a source of increased dark. It subtracts well up to 10 hours after minimum LN2, but there are residuals beginning to show up at later times. The patterrn has a different "character" than the presumed light leak, the former is less "radiant" and does not appear to be scattered or reflected thermal light. It is off the illuminated portion of the array under normal operating conditions (the illuminated portion is the central ~577x577 pixels).


image 10, diff 10 - 9 
time 0 since start, 
3.24 sec

image 15, diff 15 - 14
time 0 since start 
30 sec

image 25, diff 25 - 24
time 0:05 since start
300 sec

image 350, diff 350 - 349
time 9.75 hr since min LN2, 
3.24 sec

image 370, diff 370 - 369
time 9.75 hr since min LN2, 
30 sec

image 376, diff 376 - 375 
time 10.25hr since min LN2, 
300 sec

image 384, diff 384 - 381
time 11 hr since min LN2
300 sec

image 396, diff 396 - 393 
time 12 hr since min LN2, 
300 sec

image 420, last in sequence
time 15.5 hr since min LN2, 
300 sec