You are here


Content owned by mark.rawlings

Detector Array

The basic properties of NIRI's array are given here, as well as the problems that occasionally affect the quality of the raw data.

Array Aladdin InSb (Hughes SBRC)
Pixel format 1024x1024 27-micron pixels
Spectral Response 1 to 5.5 microns
Dark Current 0.30 e-/s/pix (shallow well) 1.82 e-/s/pix (deep well)
Read Noise (low background mode) 10 e-/pix
Read Noise (medium background mode) 35 e-/pix
Read Noise (high background mode) 70 e-/pix
Gain 12.3 e-/ADU
Well depth (near-IR, -0.6V) 200,000 e-
Well depth (thermal-IR, -0.9V) 280,000 e-
Quantum efficiency about 90%
Flat field uniformity +/-18% (VIEW)
Flat field repeatability +/-0.3%
Bad pixels1 Shallow-Well ~ 1.9% and Deep-Well ~ 2.3%
Residual image retention2 0.5-1% of a bright (saturated) source in the next frame
Centered Sub-array dimensions 768x768, 512x512, 256x256 pixels

The array has nicely uniform response and very low dark current. Various size centered subarrays may be read out instead of the full 1024x1024 array. The bias voltage may be adjusted to modestly increase the well depth for thermal IR (L and M band) observations.

1Additional bad pixels appear when using the 256x256 subarray. These pixels are the last 8 pixels read out in each quadrant, yielding a bad block of 32 pixels (16 pixels wide x 2 pixels high) located exactly in the center of the detector. Because of this feature, perfectly centering targets on the array is not recommended in 256 subarray mode.

2Image persistence is small in this array, but saturated or nearly saturated images will leave a ghost in one or two subsequent frames. The persistence of a very bright source is typically 0.5 to 1% in the next frame and less than 0.1% by the third frame.

The photo below shows the Aladdin array mounted on this focus stage.

[NIRI Detector Mount]

Below is a close-up photo of the Aladdin array.

[NIRI Aladdin Detector]

Read Modes

Three read modes have been defined to optimize use of the array. For high background environments (e.g., in the thermal IR), the array is read once at the beginning and once at the end of the exposure and the difference is recorded. In medium background situations (e.g., f/6 broad band JHK imaging) the same basic mode is used, but the beginning and end reads are digitally averaged 16 times. In low-background observations (e.g., f/32 observations, 1-2.5um narrow band imaging and 1-2.5um faint object spectroscopy), the array is read 16 times at the beginning and the end of the exposure, with the above digital averaging also taking place during each read.


The saturation level in a single coadd and low-noise read pair is about 16,000 ADU (although the exact well depth varies with position on the detector and intensity of the incident radiation). Because of the way the array is operated (reset-read-read), progressively brighter sources will approach saturation and then begin to get FAINTER as the array begins to saturate in the time between the reset and the first read. Saturated stars often show a central hole, sometimes even becoming negative in the core.


The linearity of the array is best (a few percent deviations) at low- to medium-flux levels and with exposures longer than a few seconds. When the exposures are very short (<1s) or when the well reaches approximately >70% of the full well the non-linearity becomes severe. Observations should therefore attempt to stay in this optimal regime. See the Data Format & Reduction > Warnings section on how to deal with non-linearity.

First Frame

When starting exposures after changing the detector configuration (well-depth, read mode, exposure time, etc) the background or dark current level is different, possibly due to image persistence after saturating the array (NIRI has no shutter). This means that the first exposure of each new sequence will show poor background subtraction and will likely need to be rejected. We therefore recommend that sequences include an extra step (either at the beginning, or repeat the first position at the end, or simply include N+1 different offsets). Short exposures that are not background limited (e.g., standard star measurements) are usually fine without an extra exposure. If using multiple coadds you can add a single exposure of the same integration time before the offset iterator. For example, if you are using 15 second exposures and 4 coadds, you could add one or two 15 second exposures to the observation before the offset iterator. Note that pauses in the sequence and changing the exposure time can induce a new first frame, so you may also see the first frame after changing filters.

Quadrant Boundaries

Quadrant boundaries are sometimes visible due to a mismatch in the background level of exposures which are not background dominated. However, background-limited exposures show no discontinuity, and no special precautions are necessary to avoid placing objects on the boundaries.


There is column crosstalk in columns 534 and 535 in the bottom right quadrant (rows 1 to 512), and these pixels will always have very similar values. Stars which fall on these two columns will therefore appear slightly elongated or boxy.

Components | Gemini Observatory


The website encountered an unexpected error. Please try again later.