- New IRIS2 Instrument Scientist (23 March 2007)
- IRIS2 Web page and S/N Calculator Updates (1 December 2006)
- Transformations to Other Photometric Systems Derived (October 18, 2006)
- Mark 2 Science-grade Detector Commissioned (June 2, 2006)
- IRIS2 Engineering detector back on-sky (June 20, 2005)
- IRIS2 Detector Failure (June 3, 2005)
- IRIS2 ORAC-DR Improvements (March 1, 2005)
- New IRIS2 Spectroscopy Data Reduction Recipes for ORAC-DR (Jan 30, 2004)
- IRIS2 Officially Dedicated (Feb 21, 2003)
- IRIS2 Wins Prestigious Engineering Awards (Nov 23, 2002)
- IRIS2 Multi-object Spectroscopy - First Results (August 26, 2002)
- Pipeline Reduction of IRIS2 Data with ORAC-DR (August 8, 2002)
- IRIS2 Science Grade Detector is Installed (May 1, 2002)
- IRIS2 Catches the Afterglow of a Gamma Ray Burst! (Nov 23, 2001)
- IRIS2 Gets Improved Optical Mounts (Nov 20, 2001)
- IRIS2 Goes to the AAT (Oct 30, 2001)
- Pre-commissioning News
As a result of recent personnel changes at the AAO, Paul Dobbie has taken over the role of IRIS2 Instrument Scientist from Stuart Ryder. In general, enquiries about IRIS2 and its capabilities should be directed to Paul (pdd -@- aao.gov.au), except for multi-object spectroscopy (MOS) mode (contact Simon Ellis: sce -@- aao.gov.au) or for ORAC-DR queries (contact Stuart Ryder: sdr -@- aao.gov.au).
The IRIS2 Web pages and utilities have undergone a bit of a revamp. The main changes are:
- All web pages have been migrated to the default AAO header, footer, and side-bar format.
- A discussion of imaging strategies has been added to the Imaging Observations with IRIS2 page.
- Information contained within the old "Wavelength Formats and Calibration Details" page has been incorporated within the Spectroscopic Observations with IRIS2 page.
- The old IRIS2 astrometric distortion page has been merged with the Flexure, Focus, and Distortion page.
- The imaging and spectroscopic count rates, zero-points, and long-term average sky brightnesses have been re-determined with the Mk2 science-grade array. The relevant sensitivity tables and Signal-to-Noise Calculators have been updated accordingly. The main change from previous calculations is a reduction in J-band signal-to-noise ratios of ~30%, due mainly to a higher average sky brightness than initially assumed.
- Signal-to-Noise estimates are now provided for IRIS2's increasingly-popular Z-band filter. Note the passband of this filter is closer to the WFCAM Y filter than to the SDSS z' filter.
The magnitudes and colours derived from IRIS2 with its HgCdTe detector and Mauna Kea Observatory (MKO) filter set are nominally a very close match to those produced by other IR cameras around the world which use a similar array and filter set. In practice however there will always be a slight adjustment needed, usually with some dependence on the colour of the object, before data taken at different observatories can be compared. From observations of a number of standard stars on the night of 2 September 2006, covering an extreme range in stellar colour, it has now been possible to derive transforms between the "natural" MKO-like system of IRIS2, and other photometric systems such as 2MASS and the Las Campanas Observatory. This is especially important now that much of the calibration of reduced IRIS2 images is bootstrapped from 2MASS images in the field of view.
Since the catastrophic failure of the IRIS2 science-grade HAWAII1 detector a year ago, the AAO's engineering-grade HAWAII1 detector has been used in IRIS2. The detector manufacturer, Rockwell Scientific, agreed to conduct a special foundry run to replicate this type of device, but with the benefit of improved fabrication techniques that should make this device more resilient to thermal cycles than its predecessor. Following preliminary testing in the AAO's test dewar which verified the array performance, this new detector was installed in IRIS2 in late-April, and on-sky commissioning was carried out on the nights of 12-14 May 2006. A summary of the new array's performance is as follows:
Cosmetics - The Mark 2 detector has much better cosmetics than the Engineering-grade array, and slightly better than the original "Mark 1" science detector. There are ~5000 "dead" pixels (response more than 3-sigma below the mean), and a comparable number of "hot" pixels. The vast majority of hot pixels are on the array perimeter, but overall the border and corner regions of the Mk2 science detector are much cleaner than the Mk1. There is a cluster of ~300 bad pixels just inside the SE quadrant, and a couple of filamentary features with low response possibly due to debris on the face of the detector.
The broad-scale response of the detector is rather more uniform than the Mk1. The most striking feature however are the sharply-defined arcs of lower-response pixels which cross all 4 quadrants (see image at left). These arcs (nicknamed the "skid marks") are likely an artifact of the manufacturing process, and either subtract or flatfield-out quite well. If nothing else, they provide a convenient "watermark" for distinguishing Mk2 data from earlier IRIS2 data!
Gain - 5.5 e-/ADU for DRM NORMAL/FAST, measured in the test dewar.
Readnoise - As with the Mk1 science and the engineering-grade arrays, the intrinsic readnoise for the Mk2 is 15-16 e- in DRM, and 8-9 e- in MRM. Unlike the engineering-grade however, there is no hotspot to inject the 1/f noise bands, so dark frames are much cleaner.
Full Well / Linearity - As the plot at left shows (click on it for an expanded view), the array response is linear to better than 3% all the way up to 40,000 ADU or so. Beyond 50,000 ADU saturation sets in, so aim to stay below 40,000 ADU. A linearity correction has been derived, and a new primitive for use with ORAC-DR that automatically applies the right linearity correction for either the Mk1 or Mk2 science arrays is now available. See the IRIS2 Linearity page for full details.
Image quality & focus - With the matrix mask in the slit wheel, an array translator setting of 400 delivers image FWHM = 1 pixel or better across the field. As with the Mk1 science grade, the focal plane of IRIS2 is curved, and this is seen in the "best focus" surfaces which are also curved. This means best focus for the whole field is not the best focus in the field centre, but rather the best focus in an annular region inscribed in the field. So the 400 number above aims to deliver an "average" best focus over the whole field. Also as with the Mk1 science grade, the detector is tilted slightly with regard to the instrument focal plane. Previously the tilt was primarily up-down on the detector; now it is primarily tilted across the detector.
Since the matrix mask is not necessarily coplanar with the other slits, we still determine "base focus" from the centre slit. Using fits in the X-direction across the whole centre slit, we get a base focus = 340. The AAO_Inc.tcl file has been updated to use 340 as the base focus.
Residual images - Spectra of bright stars in MRM mode will show residual images after nodding at a level of 0.3% in the subsequent image. Similarly, the first spectral exposure taken after an object is acquired will usually show extra noise, due to a residual image of the sky in the acquisition images. As a result, it is recommended that one or two dummy exposures be taken in MRM mode after acquisition, and before executing a spectroscopy sequence. This will also allow the count rate to be checked, with the aim of keeping the counts per exposure below 4000 ADU or so, to keep the residual image at or below the readnoise. If you need more signal-to-noise in your source or telluric standard, increase the number of Cycles or Nods, rather than the exposure time.
Interquadrant Cross-talk - Another characteristic of HAWAII-1 arrays is inter-quadrant cross-talk, such that bright point sources (or any bright structure) in a given row will cause an excess "ghost" signal across the entire row 512 pixels above or below it (see e.g. this description for SOFI/ISAAC). This effect is repeatable and quantifiable, and the level of cross-talk with the Mk2 science-grade has been found to be similar enough to that of the Mk1 that no modification to the existing /star/bin/oracdr/primitives/IRIS2/_REMOVE_ELECTRONIC_GHOSTING_ primitive is required to correct for this.
Following the failure of the IRIS2 science-grade HAWAII1 detector, the AAO's engineering-grade HAWAII1 detector has been installed in IRIS2 and went into operation from June 17. It will be used until the replacement device being acquired is available.
The performance of the engineering array was quite well characterised when the device was first installed for IRIS2's initial commissioning runs in 2001, for the read-out speed files then in use. In particular, the main flaw of the engineering-grade device is a substantial hot pixel in the lower-right (NE) quadrant which made data from that quadrant almost useless, and which injected some 1/f noise in all exposures.
The read-out speed files now in use with IRIS2 are somewhat different, and there is some evidence to indicate the performance of the bad quadrant may be somewhat better. Similarly, the read-noise and gain will be slightly different.
What was found in 2001 was as follows (extracted from /AAO/iris2/oldiris2.html)
Gain - 5.2 e-/adu for the engineering array.
Readnoise - 15 e- for the engineering array, in the absence of the 1/f noise. More like twice that with the 1/f noise present, which (sad to say) is all the time. The 1/f noise produced by the hotspot in the engineering array appears as a "horizontal banding" on every readout. Experiments with spectroscopic data show that if you have a blank area of the array available, you can median 200-300 columns of the detector to produce a "1/f spectrum" for each image. Subtracting this from every column seems to remove this 1/f noise very effectively, and deliver near the 14-15e- performance the engineering device delivers in the absence of the hot spot. Depending on your science application, it may be worth playing with your data to see if this works for you.
Full Well / Linearity - Engineering array tests show that quadrant 1 (the lower right quadrant) really sucks. Its full well is a strong function of x locations and is at best about 10,000 adu at the left edge, and falls to 2000adu or less at the right edge of the quadrant. Forget about trying to get useful data out of this quadrant. Quadrants 2-4 seem to behave well up to 28000 adu per pixel. All three quadrants show the same non-linearity behaviour. Non-linearity is less than 1% over the range 0-28000 adu.
What we have found in 2005 is as follows
Cosmetics - There are ~3000 dead pixels, most on the array perimeter. About 5 dark patches can be seen, each covering 100-500 pixels, apparently due to debris on the detector surface. The northeast quadrant (which had poor, unstable response back in 2001) has a hot column at Y=959. Dark frames show a bright "flare" at the bottom of this column, as well as a hot-spot of ~75 pixels near the border with the northwest quadrant. Both these artefacts seem to subtract out pretty well using a matching dark frame, though obviously data in the bad row and at the hot-spot are suspect. A new bad-pixel mask has been created for use in ORAC-DR imaging reduction on aatlxa.
Image quality & focus - The image quality and flatness across the array is at least as good as for the science-grade device. Spectroscopic images show arc-line widths of < 2 pixels from a 2 pixel slit across the device. A new base camera focus has been determined (270), and put into the default sequence (AAO_Inc.tcl), so that observations using standard AAO sequences will have the correct camera focus with the engineering array. Focus offsets relative to this for various filters were checked, and (not suprisingly) found to be consistent with those found earlier. The pixel scale and array orientation have been re-determined (0.4488+/-0.0002 arcsec/pix, and -179.4 degrees, respectively), and the WCS in the FITS headers updated accordingly.
Gain - IRAF's FINDGAIN gives the following results for DRM_NORMAL/FAST
For quadrants mapped as follows in the standard skycat view when taking data (ie N to the bottom)
indicating that the detector is delivering equivalent gains from each quadrant. For MRM_NORMAL/SLOW the gain for the whole detector 3.1 e-/ADU.
Quad Gain (e-/ADU) RN (e-) 1 4.58 16.8 2 4.44 19.2 3 4.52 15.7 4 4.55 16.7 Average 4.44 16.2
Readnoise - The engineering-grade array is noisier than the science-grade, presumably due to the known 1/f noise component injected into the detector by the hot spot. Whereas the science-grade routinely delivers a readout noise of 14e- in DRM, and 7e- or less in MRM, the engineering-grade gives more like 16-23e- in DRM, and 12e- in MRM.
The 1/f noise exhibits as low-level horizontal bands apparent in the dark frames, with ~3 ADU p-p amplitude and period ~50 rows, which isn't stable from frame to frame. This has almost no bearing on broad-band images, due to generally much higher sky brightnesses, but has the potential to slightly reduce the sensitivity for faint object spectroscopy (though experience in 2001 would suggest that blank regions of the array can be used to quantify and remove the banding for each exposure).
Linearity - As previously found for the science-grade array, DRM count rates are pretty linear up to ~40,000 ADU, but saturation sets in just below 50,000 ADU. In MRM, counts are similarly linear. However, examination of linearity data shows that Quadrant 1 (the NE quadrant with the hot-spot in it) can only be trusted to remain linear at counts below 20000 adu. Observers for whom accurate photometry or high dynamic range is critical are advised to position their primary targets in one of the other 3 quadrants.
Photometric performance - Observations of FS34 in JHK give zero-points ~0.5 mag brighter (i.e. less throughput) than an analysis of standards observed in Sep 2003 with the science-grade array. However, these data were taken on a night with clouds coming and going throughout the night, so we are not convinced this is a fair measure of the array's performance. More importantly, images of FS34 in each of the 4 quadrants showed
consistent zero-points to 2%, so the full array area really can be used for imaging, modulo the linearity and full-well issues discussed above.
In conclusion, the IRIS2 "engineering-grade" array appears to perform better now than it did during IRIS2 commissioning back in late-2001, presumably since the optimum waveforms and speed files developed for the science-grade array were not yet available. Though not as clean cosmetically, and with more artefacts and pattern noise than the former science-grade array, it should still enable IRIS2 to meet the scientific aims of all currently-scheduled programs (including MOS), without the need for longer integrations or special dithering schemes. As the image below shows, all 4 quadrants give excellent response, though having slightly worse cosmetics and read-noise performance compared with the Science-grade array. IRIS2 observing can now proceed pretty much as planned, while we expect to take delivery of a new HAWAII-1 Science-grade array in the near future.
Ks-band image with the IRIS2 Engineering-grade detector of the globular cluster M4, taken on 17 June 2005.
In preparing IRIS2 for its next observing run in late-May 2005, AAT technical staff encountered problems with reading out the array. After ruling out all possible causes external to the instrument, IRIS2's main dewar was warmed up, and the detector package removed. IRIS2's Science-grade array was then found to have suffered a catastrophic failure, with both the HgCdTe layer and the multiplexer layer having separated from the underlying substrate. We are working closely with the manufacturer, Rockwell Scientific, to understand the exact circumstances leading to this failure.
Efforts are now underway to source a replacement Science-grade array, and in the meantime the Engineering-grade array which was used in IRIS2 commissioning back in 2001 has been re-installed. This array has similar performance to the Science-grade array, but with effectively only 3 working quadrants. Programs requiring imaging or spectroscopy of compact sources should be relatively unaffected. Observers should consult with their support astronomer, or the IRIS2 Instrument Scientist Stuart Ryder, on modifying their observing strategy and ORAC-DR data reduction recipes.
IRIS2 observers have been using the ORAC-DR data reduction pipeline to process data to near-publication quality at the telescope for several years now. However, there have been a few minor steps in this processing that have always eluded our ability to get implemented - until recently that is. We are pleased to be able to indicate the following improvements for future IRIS2 observers:
In addition we have made several improvements to the mosaicing of jittered image data sets from IRIS2 - especially with relevance to the most commonly used recipe (JITTER_SELF_FLAT) used to process these images:
- Improved bad-pixel masks : improved standard bad pixel masks have recently been installed. These are different for imaging and spectroscopic observing, and ORAC-DR now automatically chooses the correct one for you.
- Inter-quadrant cross-talk correction : all Rockwell Hawaii-I HgCdTe infrared detectors suffer from this effect. It is most clearly seen in spectroscopic data taken on bright targets, though is also sometimes seen on imaging fields containing a few bright stars. It results in faint horizontal stripes across the detector at the same row (and +/-512 pixels in the quadrant above or below) as the bright star. All IRIS2 processing (imaging and spectroscopy) will now remove this effect.
- Non-linearity correction and thresholding : non-linearity corrections have been implemented for imaging data reductions. In imaging mode IRIS2 is better than 0.5% linear up to ~20,000 ADU, but non-linear by about 0.8% at 40,000 ADU. The imaging recipes now flag data as "bad" above 40,000 ADU, and linearise data below 40,000 ADU.
- Bad pixels : the default behaviour of most of the imaging recipes in ORAC-DR is to interpolate over bad pixels BEFORE making a mosaic from a jittered data set. While this always creates a "pretty" image, it is not the most valid scientific way to interpret an imaging set of data. So new recipes have been created to preserve bad pixels through the mosaicing process. This generally means that bad pixels on the mosaic get "filled in" during the mosaicing process, but that the corners of the array and the cores of bright stars stay "bad" even once the final mosaic is made.
- Astrometric Distortion and the WCS : due to some problems in the underlying Starlink code, mosaics made from images which were corrected for IRIS2's astrometric distortion had (in the past) world coordinate systems (WCS) that could no longer be correctly interpreted. This has been fixed, and undistorted images now have meaningful WCS systems that tools such as GAIA understand. In particular, images converted back from NDF to FITS format and displayed in DS9 are also correctly understood.
- Oversampling : in good seeing conditions, there may be spatial information in a jittered data set at pixel spacings smaller than the 0.4486"/pixel of IRIS2's sampling. So modified versions of several recipes have been created which oversample individual images before being combined into a final mosaic.
The following variants of the JITTER_SELF FLAT recipe are now available to IRIS2 observers at the AAT:
JITTER_SELF_FLAT - interpolates over bad pixels, removes disortion
JITTER_SELF_FLAT_NODIST - interpolates over bad pixels, leaves disortion in (this is the IRIS2 "standard" JITTER_SELF_FLAT currently distributed with ORAC-DR)
JITTER_SELF_FLAT_BY2 - interpolates over bad pixels, removes distortion and oversamples by 2
JITTER_SELF_FLAT_KEEPBAD - keeps bad pixels, removes distortion
JITTER_SELF_FLAT_KEEPBAD_BY2 - keeps bad pixels, removes distortion, and oversamples by 2
Observers can easily use these recipes at their home institution, provided they have the Spring 2004 Starlink release installed. They are available for download from the IRIS2 ORAC-DR Web page .
Since its introduction in mid-2002, the use of ORAC-DR at the AAT for on-line reduction of IRIS2 imaging has proven to be extremely popular and straightforward. With the recent successful commissioning of the new sapphire grisms for spectroscopy, we have at last been able to implement spectroscopy pipeline data reduction in a similar manner. The recipes and primitives draw heavily from those provided by JAC for UKIRT instruments such as CGS4 and UIST, and operate in much the same fashion. Once again, the aim is for minimum observer intervention at each stage. The primary recipes are:
The figure below shows as an example a spectrum of Eta Carinae in the J-long band (courtesy of Nathan Smith at University of Colorado). The top plot is the raw co-added spectra, and the lower plot is after division by the standard star, flux calibration, and smoothing.
- REDUCE_DARK, REDUCE_ARC, REDUCE_FLAT - for filing darks, arcs, and flats in calibration index files. Note that no actual arc-line identification and dispersion solution fitting is done; instead, an "estimated" linear wavelength scale (usually good to 0.002 microns) is applied to each image based on the grism, slit, and blocking filter used. The spectroscopic quartz flats are divided by a 1000 K blackbody spectrum, and normalised to 1.0.
- STANDARD_STAR(_NOFLAT) - for reducing object-sky pair observations of a telluric standard, nodded between two slit positions (usually Apertures A and B in the TCS). The strongest positive and negative peaks in a cross-cut of the co-added image pairs are automatically traced and optimally-extracted. The negative beam is inverted and cross-correlated with the positive beam, so any shift may be removed before co-adding. If the standard star is in the Bright Star Catalogue, its spectral type and magnitude are read from a table; otherwise, a position-based query is sent to Simbad, and the spectral type and magnitude returned. The extracted and co-added spectra of the standard are divided by a model blackbody spectrum for the inferred temperature. The _NOFLAT option may be used if a suitable flatfield has not yet been obtained.
- POINT_SOURCE(_NOFLAT)(_NOSTD) - for reducing object-sky pair observations of a moderately-bright point source, once again nodded along the slit. Assuming this is the brightest object on the slit, the positive and negative beams are traced and optimally-extracted. The individual frames are flatfielded (unless the _NOFLAT recipe is used). The extracted, shifted, and co-added spectra are divided by the rectified standard star spectrum (unless the _NOSTD recipe is used), and a relative flux calibration applied based on the magnitude of the standard star in the J, H, or K-band. The spectrum is lightly smoothed and any remaining negative pixels clipped to assist autoscaling of the spectrum plot.
- FAINT_POINT_SOURCE(_NOFLAT)(_NOSTD) - as for the POINT_SOURCE recipes, except that no spectral profile tracing is performed. Instead, the optimal extraction parameters as determined from the closest STANDARD_STAR are used. This may be necessary if the source has only a weak continuum, or is in a crowded field.
If you would like to try the ORAC-DR spectroscopy piepline at your home institution (and you already have the Summer 2003 release of Starlink installed), then please contact Stuart Ryder, who can supply all the required code.
At a ceremony held at the AAT on February 21 2003, IRIS2 was officially dedicated by the Australian Federal Science Minister, the Hon. Peter McGauran. Further details are available from the official ministerial Press Release, while images of the "unveiling" can be found on Ray Stathakis' page.
At a gala ceremony held at the Sydney Convention Centre on 11 October 2002, IRIS2 took out top honours in the annual Engineering Excellence Awards of the Institution of Engineers Australia (Sydney Division). In being awarded the Bradfield Award for outstanding engineering achievement, IRIS2 emulated the achievement of its predecessor, IRIS, which won the same award in 1993. Not content with that success however, IRIS2 then went on to pick up an Engineering Excellence Award at the Institution of Engineers Australia national awards ceremony held in Canberra a month later! Congratulations to all the engineers and technical staff involved in the design, fabrication, construction, and commissioning of IRIS2, in being recognised by their peers for their innovation and craftsmenship. The photo below shows the Head of the AAO's mechanical division, John Dawson (left), and IRIS2 Commissioning Scientist Stuart Ryder (right) holding the trophies awarded at Parliament House, Canberra.
Multi-object spectroscopy was attempted for the first time with IRIS2 on Director's nights of July 24 & 25, 2002. Successful results have led to an announcment of opportunity for infrared MOS in 2003A. See the IRIS2 Multi-Object Spectroscopy page for more details.
During the most recent IRIS2 Commissioning and Service Observing run (July 24-28), Brad Cavanagh from the Joint Astronomy Centre installed the ORAC-DR data reduction pipeline at the AAT. With relatively trivial modifications, the same recipes which were written originally to process images from UFTI, and spectroscopy from CGS4, can now process data in "real time" from IRIS2. Two examples of the pipeline's output are shown below; the first is a section of a deep observation for Malcolm Bremer (U. Bristol) obtained in 0.9" seeing over 1 hour in Ks with IRIS2. This "blank" field turns out to contain dozens of sources, and these observations were crucial in allowing Malcolm to select out high-z galaxy candidates from low-mass stars for follow-up VLT multi-object spectroscopy just a few days later.
The second example shows NGC 1300 in Ks. This spiral galaxy, which fills most of the field of view of IRIS2, was observed in a 9-point jitter pattern, interspersed with observations of nearby blank sky, and processed without any observer intervention whatsoever by ORAC-DR using the CHOP_SKY_JITTER recipe.
IRIS2's science grade detector was installed in the main dewar prior to the March/April 2002 IRIS2 observing run. The device demonstrates superb performance. It can be read in 0.6s in DCS mode with a read noise of 10e- and a full well of 200,000e-. This is almost three times the Rockwell suggested well depth, and means observations in K and Ks will be significantly more efficient than expected, with exposures as long as 5-10s being permitted instead of the expected 2-3s. The read time is, as far as we are aware, the fastest yet delivered with HAWAII devices. Multiple Read Mode is able to beat the read noise in longer exposures down to ~ 5e-.
On only its second observing run, IRIS2 has helped play a major role in monitoring the fading afterglow of a Gamma Ray Burst (GRB). On what was scheduled to be a Director's commissioning night for IRIS2, the observers (Stuart Ryder and Jeremy Bailey, assisted by Frank Freeman) were contacted by Paul Price from Caltech, on behalf of the Caltech-NRAO-CARA collaboration, who wished to trigger their pre-approved ATAC override program to obtain Ksimaging of a newly-discovered Gamma Ray Burst source, GRB011121. Undeterred by the fact that the source coordinates of (11h 34m, -76o) meant that it would have to be observed below the pole, the observers swung into action, and within a short period of time, had obtained confirmation images of the afterglow. This was further complicated by:
- the source being at a zenith distance of 74 degrees, so that
- it was partly vignetted by the windscreen, and
- a bug in the new telescope control system meant that offsets for dithering had to be fed in by hand!
Nevertheless, we were able to follow it for almost 7 hours (the Director having given over the rest of his night for this important event) as the source got higher in the sky, but noticeably fainter. Unfortunately, the GRB was already too faint in the J and H bands to permit an Italian collaboration to trigger their own IRIS2 spectroscopic override, but the opportunity to provide extended photometric coverage of a still fairly unique class of object shows that IRIS2 is well on the way to making a significant international scientific impact.
GRB011121, as seen in the Ks band with IRIS2 on the night of 22 Nov 2001 UT.
Final integration testing of IRIS2 revealed poor imaging performance over at least half the field of view. As a result only about one quarter of the array was available for imaging during the first (Oct/Nov 2001) commissioning run. Analysis of the images indicated the likely source of the problem was the field flattener mount. So Roger Haynes, Greg Smith, Urs Klauser and Dwight Horiuchi set too between two IRIS2 runs to redesign, re-install and test a new mount. Amazingly, despite an incredibly tight schedule, this all worked to plan (well done guys!). Image quality of 1 pixel can now be achieved at a focus value of 150 over the entire field. Further iterations on detector alignment when the science device is installed for the next (March 2002) run should deliver slight additional improvements.
Other exciting news is that the 'hot pixel' problem we thought we had with the IRIS2 science-grade device turns out to be a recently discovered effect in recently produced HAWAII arrays. No-one (including Rockwell, the manufacturer) knows why, but when you power up the detector as you cool, you get lots of hot pixels. If you don't power-up when cooling, the detector is clean! (In fact, on the last engineering device cycle when we did this the 1/f-noise-injecting super hotspot was not present, so for the Nov/Dec IRIS2 run, three-quarters of the array should be perfectly usable). The science device will be installed for the March 2002 IRIS2 run, along with the last of the sapphire grisms (the S-J) and hopefully the rest of the IRIS2 filters.
On Wednesday October 24, 2001 IRIS2 arrived at the AAT for its first commissioning and scheduled observing time. Thanks to the heroic efforts from both Site staff, and the members of the IRIS2 team, we were able to get the instrument on the telescope, and on sky by about 9pm on Friday October 25. Commissioning observations then began and proceeded in parallel with Shared Risks Service Observing until the first scheduled observer began work on October 30. For more information see October 2001 Commissioning for a visual record of the progress, or read Bridget Dawson's Diary.
Return to 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.
IRIS2's First Pump Down
Have a look in Anthony Dunk's image gallery to see IRIS2's dewar bolted together for the first time and undergoing vacuum tests. It got down to 0.5 mTorr in an hour!
First Light for IRIS2 Optics
This image shows the results of warm tests of the IRIS2 Optics, as provided by Graseby Specac Inc (UK). When tested warm, the optical elements do not lie in their optimal locations, so the results provide more of a "sanity check" than stringent results. But they do indicate that we seem to be sane so far. The image is 350um on a side and each pixel is 0.7um in size. For reference the IRIS2 pixels will be 18.5um in size. There's a radial plot available also (Note that the pixels here are the 0.7um optical detector pixels, not the 18.5um pixels that IRIS2 will eventually use.
Rendering of IRIS2 at the Cassegrain focus