UIST: Engineering – Dark Current Gradient

UIST: Engineering – Dark Current Gradient

Below is a potted history of the dark current/gradient saga, in chronological order… Example darks are shown at the bottom of the page.

April 2004: Initial Problem

After a warm-up in January, 2004, the UIST array seemed to have developed a dark-current gradient which could potentially affect the sensitivity of the instrument in the long-wavelength (left) half of the array – particularly for the higher-resolution grisms and/or the IFU, where the dark current and readnoise dominate over the sky-background between the sky lines.

With array at 30K and heater at 35%: 100sec dark frame showed a strong left-right gradient with ~500DN at left to ~100DN at right.

  • LIGHT LEAK? No mechanism movement changed this so it appeared not to be a light leak.
  • THERMAL CYCLE? A warm-up (to 100K) and cool down initially brought darks close to normal, though the dark gradient returned within a few hours.
  • HEATER/ARRAY TEMP? In daytime tests, seemed that there was a link with the heater. Brought temp down to 27K and 25% heater (was at 35%). Gradient in 100sec dark were now at ~50 on left and ~10 on right.

See bottom of page for example “good” and “bad” darks (pre/post April 2004).

October 2004: Detailed investigation

Investigated day-time (i) the affects of changing the heater temperature, and (ii) how the dark current/gradient changed with time. Results outlined in JACdocs (under UIST).

  • AFFECT OF REPEAT EXPOSURES: Evidence that dark current decreases with repeat 100sec exposures, though after ~1hr, starts to increase again, eventually leveling off! So varies with readout and frequency of reads.
  • AFFECT OF NOT READING OUT ARRAY: Dark current decreases if we don’t readout array for 1/2 hour.
  • HEATER TEMP: tried running the array at lower and higher temps, but basically got the same behaviour as before; i.e. much higher dark signal at temps of 29 and 30K, with best performance at 27K. We didn’t go below 26K.

The heater temp. did not change during first two tests.

Bottom line: Any initial drop in dark current, caused by repeating the array tests (short and long darks) a few times at the start of the night, wouldn’t help any in the longer term (with e.g. repeat 200sec spectroscopy exposures).

Gradient in 100sec dark were still between 20-50 on left and 5-10 on right in late 2004.

January 2005: Problem less severe…?

With the array operated at a slightly different temperature (29K as opposed to a nominal 30K), tests in January 2005 indicated a reduction in the mean dark current. However, the overall readnoise was found to be a tad higher: with short exposures (<10-20sec) a mild, intermittent chevron pattern increases the RN in the top half of the array; with long exposures (>200sec) the dark-gradient increasingly contributes to the overall noise on the array, and the noise is then higher in the left(long-wavelength)-half of the array.

The nominal readnoise (RN) versus exposure time plot measured when UIST was commissioned in late 2002 is shown on the main UIST web pages. Testing in January 2005 indicates:

  • RN ~ 45e- in 1sec exposures (nominal in 2003 was 38e-),
  • RN ~ 17e- in 100sec exposures (nominal in 2003 was 15e-),
  • RN ~ 21e- in 240sec exposures (nominal in 2003 was 17e-).

Note, however, that scrutiny of the 240sec data show that the noise is of order 25e- in the left half of the array, though only 17e- in the right half.

BOTTOM LINE: longer exposures = stronger dark-gradient = increased noise on array (before addition of sky noise) at longer wavelengths.

February-March 2005

The gradient seems to have “fixed itself”! We have a slightly noisey array (RN in low-mid 40s as opposed to 38e-), especially in the top two quadrants. But no evidense of the high counts nor the gradient in the 100sec array-test darks.

Spring 2005: UIST off telescope and warmed up!

In March 2005, UIST was removed from telescope and warmed up to install new grisms and fix stuck Grism wheel 2. Back on in June 2005. No obvious affect on the darks or dark current gradient.

October 2007

The gradient returns (same counts) after not seeing it for over two years. Coincided with a new cold-head, which is causing the heater to run a little higher than usual. Its possible that this is creating the gradient…


Example darks are shown below (with the same High and Low settings on the display). Note the appearance of the gradient in the 2004 data, though the subsequent overall reduction in the dark current and the absence of the gradient in the Feb 2005 data.

Array at 30K
Jan 2003

(No data available)
Counts: left~0, right~0
Counts: left~5, right~5
Array at ?K
April 2004

(No data available)
Counts: left~0, right~0
Counts: left~20, right~4
Array at 29K
Feb 2005

Counts: left~0, right~0
Counts: left~0, right~-2!