[Skip to Content]


General Links
Professional Links
Australian Human Resources Institute Diversity Awards 2013 Finalist

IRIS2 Overview

This page gives a quick summary of the current capabilities of IRIS2 which potential users will need to know before preparing an observing proposal. As it is updated regularly, users are recommended to consult it before submitting (or re-submitting) a proposal. Please contact the Instrument Scientist, Paul Dobbie (pdd -@- aao.gov.au) if you require any further information not listed here.


Optical Design

IRIS2 is a straightforward all-transmissive collimator-camera focal reducer (from f/8 to f/2.2). The optical train contains 10 elements (a window + 4 element collimator + 5 element camera). All elements were specified to the manufacturer to be coated to better than 2% reflectivity. In order to maintain simplicity, scale changes are achieved by changing top-ends. To date, IRIS2 has only ever been used in its widest-field mode at f/8, as the smaller field-of-view and over-sampled pixel scale makes it uncompetitive with facilities available elsewhere.

IRIS2 Layout The figure shows the general IRIS2 layout.
It is available in several forms.
Top end Pixel scale Field of view
f/8 0.4486+-0.0002"/pixel (measured) 7.7'x7.7'
f/15 0.24"/pixel (estimated) 4.1'x4.1'
f/36 0.10"/pixel (estimated) 1.7'x1.7'

Cryogenics & Dewars

The instrument is kept cold by two Cryodyne closed-cycle refrigerators. In order to ensure minimal thermal cycling of the fragile (and valuable) detector, the instrument has two dewars. The main dewar contains all the collimator-camera optics, the detector and the pupil, filter and grism wheels. This dewar has a long cycle time, and since commissioning, it has had to be warmed up only occasionally. It also has an auxiliary liquid nitrogen cooling system which is used to accelerate cool down.

The fore-dewar contains the aperture wheel, and room for future expansions to include a multi-slit juke-box, IFU elements, and polarimetry analysers. It can be warmed up and cooled back down on a short (2-3 day) time-scale, so user provided multi-slit masks, filters, etc., go in here.


IRIS2 contains four user-configurable wheels. Click on the wheel name to see the current wheel configuration.


The current filter complement is listed below - see the IRIS2 Filter page table for more details. All filters are 60mm in diameter, which is required for unvignetted operation at f/8. These filters were purchased as part of the Gemini IR Filter Consortium.

The J,H,K,Ks filters are discussed in Alan Tokunaga's 2001 and 2005 specifications. The Z filter in IRIS2 is often referred to elsewhere as "Y" to avoid confusing it with the (somewhat bluer) SDSS z' filter; however, note that the WFCAM Y filter is different again from the MKO Z/Y filter, which is closer to that defined by Hillenbrand et al. 2002.

Filter Name Cut-on
Z 0.996 1.069
J 1.164 1.325
H 1.485 1.781
Ks 1.982 2.306
K 2.028 2.364
Z 0.996 1.069
Jshort spectroscopic 0.95 1.25
Jlong spectroscopic 1.05 1.35
H spectroscopic 1.44 1.82
He I 1.075 1.091
J continuum * 1.198 1.216
(Pa Beta) *# 1.272 1.292
H continuum 1.558 1.582
[Fe II] 1.632 1.656
Methane Offband (CH4_s) 1.53 1.63
Methane Onband (CH4_l) 1.64 1.74
H2 nu=1-0 S(1) 2.106 2.138
Br Gamma 2.150 2.182
H2 nu=2-1 S(1) 2.231 2.265
K continuum 2.253 2.287
CO(2-0) band head 2.278 2.312
   * The J continuum and Pa beta filters have a red leak in the K-band,
   so must be used in combination with the J broadband filter.
   # The Pa beta filter is currently unavailable.

Information on the sky count rates, throughput and filter bandpasses can be found at the IRIS2 Filter Page.

Spectroscopic Formats

The wavelength coverage of IRIS2 is a function of the grism, slit, and blocking filter used. The resolution is ~2400 in all bands.

Grism Slit Start
Hspect SAPPHIRE_316 OFF_150
Jshort SAPPHIRE_240 SLIT_150
Jlong SAPPHIRE_240 OFF_150

For more details on wavelength formats and wavelength calibration, see the IRIS2 Spectroscopy page.

Detector performance

The IRIS2 HAWAII-1 detector has been configured for use in two read modes:

For each read mode we have configured two read speeds. In almost all situations the NORMAL speed in each mode will be the optimal one (the IRIS2 software offers three speed options, but NORMAL and FAST are the same speeds). The numbers below for DRM NORMAL apply to the Mk2 science-grade array (May 2006 ->); the other numbers were measured with the Mk1 science-grade array (March 2002 -> March 2005), but should not be significantly different for the Mk2, as array performance for the most part depends on the controller.

Mode Speed Read
Time (s)
Read Noise/NDR
Typical Read
Noise (e-)
Full Well
(at Full Well)
0.5982 5.5 10.7 15.1 (for DRM) 180 ~2% Mk2 science array
SLOW 0.9925 4.4 8.7 12.3 (for DRM) 70 ~0.25% Mk1 science array
0.7866 4.3 8.6 4.8 (for 61 reads) 75 ~0.3% "
SLOW 1.3109 5.2 8.6 ~4.8 (for 61 reads) 130 ~1.2% "

Dark current

Measured dark current performance is typically <1 e- per second per pixel. There is some structure to the dark frames, most noticeably a "hot spot" of about 100 pixels near the centre of the top-right quadrant, a slight excess of charge in the lowest rows of each quadrant, and some faint horizontal 1/f noise bands. All of these artifacts subtract out quite well, so we recommend that matching dark frames be obtained for each combination of exposure/cycles to be used in imaging mode.

Quantum Efficiency
QE has not been measured directly for our science arrays. The image below shows the 'standard' Rockwell curve. Rockwell have provided a histogram of QE values across the detector, which peaks at 69%. However, as they don't bother to tell us what wavelength they did that measurement at, we have no idea how to normalise the curve below.
    QE curve
Full Well / Linearity
Tests show that non-linearity varies with read speed. In most applications, the non-linearity can be kept below 0.5% by keeping fluxes to less than ~20,000 ADU. The non-linearity can be straightforwardly calibrated out (see the Linearity Technical page for more information) which raises the usable well-depth considerably.
Residual Images / Mode switching dark current
Tests made with the slit in place illuminated by a flat field lamp, and then followed by darks, seem to indicate residual images are present at the 0.01% level. Thus, if you illuminate the slit with light at a rate of about 6667 ADU/s, or 10,000 ADU in a 1.5s exposure, and then follow this with a 30s dark, you will get a residual image at a level of about 15 ADU (or 0.5 ADU/s). Clearing these residual images seems to be a matter of both taking lots of resets and reads of the detector, and just physically waiting a while.

Switching between read modes (i.e. from DRM to MRM, or vice-versa) also seems to induce additional dark currents. So all acquisition exposures for spectroscopy should be acquired in MRM mode.

Residual images / switching images appears as an additional dark current. This is a characteristic of the early generation of PACE arrays. Even though the 0.01% of the HAWAII-1 array is quite good, short exposures on bright sources, followed by long exposures on faint sources may show residual images. We recommend:

  • Acquire your object in a passband with the lowest sky counts possible. This will usually be J.
  • Acquire your object in MRM mode.
  • Be aware that the first spectrum you take after acquisition may still have slightly higher noise, though if you follow these guidelines, the additional dark current is not usually significant.

Cross Talk between Quadrants

HAWAII arrays are known to suffer from cross-talk between their quadrants. If you have a bright object at pixel (660,700), the inter-quadrant cross talk appears as additional flux in all pixels of row 700 and row 700-512=188. Correction of this effect is straightforward. If you collapse your raw data down into a `spectrum' by adding up all the columns, fold this spectrum at row 512, multiply it by 0.0128, and then subtract it from all columns of the detector, the inter-quadrant cross talk is almost completely eliminated. ORAC-DR will apply this correction automatically to all IRIS2 frames.


Is not implemented for IRIS2, as the readout time is already short enough for most applications.

Image Scale and Orientation

The image scale with both the science-grade arrays and the engineering-grade array measured using 2MASS astrometry and a linear astrometric fit to tangent-plane projected and radial distortion corrected images = 0.4486 +/- 0.0002"/pixel. For non-astrometrically corrected images, this plate scale will be correct at the field centre, but up to 1.4% incorrect in the array corners (see the IRIS2 Distortion page for more information on the form of the IRIS2 astrometric distortion).

The detector orientation as it appears on the Skycat display (and in FITS files) is North to the bottom, E to the right, when IRIS2 is used in its default Cassegrain rotator=90 degree orientation. The instrument can be rotated but this requires an AAO staff member be in the Cass cage while the instrument is rotated to watch out for cable and hose fouling. Whenever the rotator angle is changed, the WCS in the FITS header will be updated to reflect this. Note that the WCS information usually written to the FITS headers has a default pixel scale set to 0.446"/pixel, leading to a noticeable change in apparent offset across the field-of-view between IRIS2 images and any 2MASS astrometric overlay.

With the Cassegrain instrument rotator set to 90 deg, the Mk1 science-grade array was aligned such that true N was rotated by 0.1 deg clockwise relative to the array columns. For the engineering-grade array re-installed in 2005, the corresponding offset was such that true N was rotated by 0.6 deg clockwise relative to the array columns. The Mk2 science-array currently in use has true N rotated 1.0 deg clockwise relative to the array columns.

For J and K spectroscopy, wavelength decreases with pixel number (i.e. redder wavelengths are to the left as seen on the Skycat display).

For H spectroscopy, wavelength increases with pixel number (i.e. redder wavelengths are to the right as seen on the Skycat display). The H grism is reversed relative to the J/K grism so that it can be used with the ofset slit to obtain the whole H bandpass in one go.

Slit Orientation and Re-positioning Accuracy

The slit is rotated slightly relative to the array columns in IRIS2. All spectroscopic acquisition with IRIS2 must be carried out using IRIS2 in imaging mode.

Tests of slit re-positioning show that the slit seems to re-position with no backlash to <0.1" when the instrument is nearly upright. When the telescope is slewed off the meridian East-West, the slit re-positions to < 0.1" as long as the wheel is always moved in the same direction (which it always does during spectroscopic acquisition).

While the slit repositions very precisely however, the place it re-positions to does depend on the telescope's East-West orientation, with the slit moving by ~0.5 pixel when the telescope is slewed 3 hours to the East or West from the meridian.

For significant motions East-West from the last spectroscopic acquisition, the acquisition of a new slit image (to re-determine the slit location) is recommended. See the Acquisition procedures for more details.


As the longest period one would spend imaging any one position before dithering is ~1 minute, and the AAT is perfectly capable of tracking unguided for this period, no guiding is performed with IRIS2 when imaging in broad-band filters. Even with narrow-band filters, exposures up to 5 minutes are possible without guiding. Offsets between dither positions are calculated by ORAC-DR by cross-correlating star positions, rather than relying on the accuracy of the telescope's fine encoders.

For spectroscopy, guiding at both nod positions is required to keep the target within the 1 arcsecond slit. Acquisition of the target is carried out in conjunction with acquiring a guide star, which requires the observer to liaise with the Night Assistant. Instructions on how to acquire the target and guide star are provided for single-object spectroscopy, as well as for multi-object spectroscopy. Guiding while taking spectra of telluric standards is optional, as the integration times are usually quite short (~10 seconds).

Imaging sensitivities & S/N calculator

The table below gives an indication of recent (Sep 2006) zero-points and limiting magnitudes for the Mk2 science-grade array, as well as "typical" long-term average sky brightnesses. The sky brightness shortward of 2 microns has no clear seasonal variation, but can vary by up to 30% between (and even within) nights. The K/Ks sky brightness is roughly a factor of 2 greater in summer (S) than winter (W), so the sky brightness and 5-sigma sensitivities are given separately for each.

The median optical seeing of 1.5" at the AAT corresponds to 1.2" at H-band (FWHM ~ lambda-1/5), which is what should be assumed when preparing proposals. It is worth keeping in mind however that since the commissioning of the dome air-conditioning system, up to half of all IRIS2 nights have experienced sub-arcsecond seeing.

Broadband Imaging Figures of Merit
Filter Zeropoint
(1 ADU/s at 1 airmass)
Flux for mag=0
Sky Brightness
5-sigma limiting magnitude
in 1 hr (on sky)
1.2" seeing, 2" aperture.
Z 21.16 2.9e8 17.0 21.9
J 22.58 5.9e9 15.7 22.0
H 22.79 7.2e9 14.1 21.3
Ks 22.37 4.9e9 13.5 (W)
12.6 (S)
20.7 (W)
20.3 (S)
K 22.35 4.8e9 13.3 (W)
12.3 (S)
20.6 (W)
20.1 (S)

Imaging overheads

The major overheads for broad band imaging is the detector read-time and telescope movement. For the purposes of preparing an observing proposal, you should factor in overheads of ~25%.



Spectroscopic sensitivities & S/N calculator

Providing a single S/N estimate for a given magnitude object in a given passband is only crudely possible. Sky brightnesses and atmospheric transmission within an atmospheric window can vary by factors of up to ten. We have prepared an Exposure Time Calculator for IRIS2 which is based on the median sky brightness per pixel, and median object brightness per pixel as measured in practice with IRIS2. This means that in a given spectrum half the pixels will have S/N better than that listed below and half will be worse. However, these estimates should be adequate for most proposals.


Spectroscopic overheads

The major overhead for spectroscopy is acquisition time.


Readout overheads for spectroscopy exposures acquired in MRM are small. There is a slight delay (1-2s) at the end of an exposure as the least-squares fit is re-done a final time, and a 0.9s readout overhead over the length of the whole exposure. So, for example, for K-band spectroscopy where you might be exposing for 300s with 61 non-destructive reads (before nodding the telescope), the overheads are (0.9s+2s)/300s or 1% of the exposure time.


You should assume an overhead of roughly 5s for a typical telescope offset when nodding along the slit. Again, this is usually a small fraction (<1.2%) of typical exposure times of >300s.


Acquisition is the major overhead for spectroscopy. You will acquire your target using short imaging exposures. You should be able to get away with half a dozen or so such images to put your object down the slit. If your target is faint (or hard to identify) this could take up to 15 minutes. The night assistant will also need to acquire a guide star. This is usually done in parallel to getting the target close to the slit, but can add an extra overhead of a few minutes. You should assume acquisition will take 15 min at the start of your run, improving to 5 min as you gain experience.

Acquisition should always be done in MRM mode itself (e.g. 5s exposures with 2 MRM reads). This ensures you get a `clean' detector with no mode-switching dark current in the first of your deep MRM spectroscopic images.

Multi-object Spectroscopy

Up to 3 multi-object masks may be mounted in IRIS2's slit wheel at one time. With slitlet lengths typically 8-10 arcseconds, up to 50 sources can be observed simultaneously. In order to ensure the wavelength coverage of each slitlet is relatively consistent, the maximum displacement of any slitlet should be limited to no more than 2 arcminutes or so off the nominal slit axis, i.e. the accessible field-of-view for any mask is 7.7' x ~4'. Note that a separate mask is required for H or Jlong spectroscopy from the one prepared for K or Jshort spectroscopy, as the nominal slit axis for the SAPPHIRE_316 grism is offset relative to that for the SAPPHIRE_240 grism. Thus, if complete JHK spectroscopic coverage is sought, then two separate masks are required, and only one field can be observed per IRIS2 observing block. If only H and/or Jlong is required, or only K and/or Jshort, then up to 3 fields can be observed per block. Changing masks currently requires at least 4 days, and usually takes place in between IRIS2 observing blocks.

More details on the practicalities of MOS observing with IRIS2 can be found on the IRIS2 MOS page.

Return to the main IRIS2 page.

These pages contain information on the functionality of the IRIS2 Infrared Imager and Spectrograph. Pages maintained by Stuart Ryder (sdr -@- aao.gov.au) and Paul Dobbie (pdd -@- aao.gov.au). Page last modified by Stuart Ryder.