IRCAL is a 1-2.5 µm camera optimised for use with the LLNL Lick adaptive optics system on the Shane 3 m telescope.  Using diamond-turned gold-coated optics, the camera provides high efficiency diffraction limited imaging throughout the near-IR.  IRCAL incorporates optimisations for obtaining high dynamic range images afforded by adaptive optics, coronagraphic masks, and a cross-dispersed silicon grism for high resolution spectroscopy.


The cryogenic and mechanical components of IRCAL were manufactured by IR Laboratories, Inc (Tucson, AZ), based on an liquid nitrogen cooled ND-8 dewar.  By utilising a vapor cooled shield, the heat load on the liquid cryogen is very low, yielding a hold time in excess of 48 hours under normal conditions.  There are three mechanisms inside the dewar: two filter wheels holding eight one inch diameter interference filters each, and a six position focal plane mask wheel.  The mechanisms are driven with external stepper motors via mechanical feedthroughs.

Light from the AO system enters on the right through a 0.85 inch diameter CaF2 window.  The converging F/28.5 beam forms a focal plane at the aperture mask mechanism, is folded to the collimating parabola, and the collimated beam passes through a filter with a 5 degree tilt.  The cold pupil stop is located on the far side of the filters, before the beam is focussed with the f/33 camera parabola, and folded onto the PICNIC focal plane array.  The two optics boxes are shown baffling the pairs of parabola + fold flat, with baffle tubes in between.


The optical design consists of two nearly balanced off axis parabolas with two folding flats.  The camera optics reimage the AO corrected telescope focal plane, via a cold pupil stop (conjugate to the telescope primary mirror), to the IR focal plane array (FPA).  There is a slight magnification from the F/28.5 output beam of the AO system to a camera focal ratio of F/33.  This Nyquist samples the diffraction limit of the 3 m telescope at 2.2 µm.  The measured plate scale is 75.6 ± 0.2 milliarcsec per pixel, giving a 19.4 arcsec field of view across the 256 x 256 pixel detector.  A 5 mm pupil size was chosen as a compromise between instrument size, and the ability to do grism spectroscopy.  With a 5 mm pupil, it is possible to do moderate resolution spectroscopy with a calcium fluoride and resin replica grating grism. R=700 is chosen to fit the full bandpass of the J, H, or K atmospheric windows across 256 pixels with 2 pixel spectral resolution.

The mirrors are kinematically mounted on an Al optical bench that can be removed from the dewar as a single unit.  The optical assembly can thus be aligned outside the dewar.  Reference holes are machined in the optical bench, to which pinholes that define the optical axis are mounted.  Using a laser, the optics can be aligned by adjusting the elements to pass the laser through the pinholes.  As the entire assembly is made from homogeneous material, the alignment does not change as the optical bench cools from room temperature to 77K.

Two filter wheels and the pupil mask are housed in a filter wheel assembly box that mounts directly to the cold plate.  The filters operate in the collimated beam with a 5 degree tilt to avoid optical ghosts.  The wheels are spring loaded onto ruby balls for heat sinking.  The mechanisms are geared with stainless steel bevelled gears, driven by stepper motors external to the dewar.  The entire optical path is carefully baffled with housings over the optical elements and baffle tubes in between.  The interior surfaces of the housings are sandblasted for surface roughness and anodised black. The interior of the baffle tubes are painted black.

The optical elements were manufactured by Janos Technology Inc in diamond turned aluminum and gold coated.  Post polishing was performed by an IR Labs subcontractor.

IRCAL Filters

Name Central 
( µ m) ( µ m) ( percent)
J  1.238 0.271  82
H 1.656 0.296 85
K 2.195 0.411 75
KS 2.150 0.320 (1)
H2 1-0 S(1)  2.125 0.020 80
H I (n=7-4) 2.167 0.020 78
J CH4 1.183 0.040 86 (2)
K CH4 2.356 0.130 80
2.2/0.04 2.192 0.047 79
H cont 1.570 0.020 (3)
[Fe II]  1.644 0.016 (3)
K cont 2.270 0.020 (3)


(1) Specified, not measured.
(2) Unblocked, needs to be crossed with the J filter for blocking.
(3) On order; specified, not measured.


The detector is an AR coated Rockwell 256 x 256 HgCdTe PICNIC array with 40 µm pixels, sensitive from 0.85 - 2.5 µm.


A preamplifier is mounted externally on the dewar, connected to the detector with stainless steel coaxial cable. The detector clocking and analog to digital conversion is controlled by a set of San Diego State University (SDSU) CCD laboratory generation II readout electronics.  Data are transmitted to and from the readout electronics via optical fiber. The fiber transceiver board is hosted in a VME chassis.  A Force CPU-50 UltraSPARC CPU board running Solaris manages the data acquisition through shared VMEbus memory.  This system is capable of acquiring data continuously at the nominal readout rate for the PICNIC array, 3 µs per pixel (57 ms for a full frame read).  At this rate, the readnoise is ~ 30 electrons RMS for a single correlated double sample (CDS).  We implement Fowler sampling to improve readnoise performance, and achieve ~12 electrons RMS readnoise with 16 samples.  The readnoise is slightly dependent on the length of the cables between the preamp and SDSU readout electronics, and the noise environment.  These readnoise figures are measured in a quiet laboratory.  Performance on the telescope is slightly worse, due to noise pickup from nearby equipment.

The SDSU readout electronics also incorporate a 1 µs per pixel readout mode, used with the IR Laboratories IR emission microscopes which interface to the SDSU readout electronics via a SCSI-3 interface.  Since the detector has four quadrants, this readout mode exceeds the 2.5 Mpixel/s speed of the optical fiber, and we implement a two quadrant readout at this speed.  However, the readnoise is typically 80 electrons RMS in this mode and it has rarely been used.

Since the PICNIC detector uses asynchronous resets on the pixel addressing shift registers, as opposed to the initialising scheme of the NICMOS detectors, it is much easier to implement subarray readouts.  It is straightforward for the case that the subarrays are constrained to start at the first pixel in each quadrant.  This constraint substantially simplifies the implementation of the clocking for the subarrays by making the clock wave forms for an n x m subarray different from the full 256 x 256 array only in the count of clock pulses (a corollary is that the readout for a 1024 x1024 HAWAII array is the same as a 1024 x 1024 ``subarray'').  In practice, subarray readouts are preferable for achieving faster frame rates than faster readout modes, because the readnoise is preserved, and in short exposures the detector is almost always readnoise limited.

The minimum full frame, CDS integration time is 57 ms (3 µs readout) or 24 ms (1 µs readout), though only half the detector well capacity is available because the detector integrates during the first readout of the array.  Although we do not provide a separate single read readout mode, it is possible to save the data without the CDS (as a data-cube of raw reads of the detector) if the second read of the CDS is saturated. Because the readnoise is substantially higher (~ 100 electrons RMS) for a single sample, it is generally preferable to use subarrays to speed up the readout when necessary. IRCAL supports subarrays from 2 x 2 up to the full frame.  At present, the shortest exposure time for very small subarrays is set by a ~ 5 ms delay after the detector output amplifier is turned on.  For a standard CDS readout mode (two full frame resets, two full frame reads), there are four passes through the array to complete the exposure, and it is possible to stream full frame data to disk at ~ 4 Hz (57 ms integration per frame).  It is possible to increase this rate (or increase the fraction of integration time) with subarrays, fewer resets, or resetting the detector line by line instead of pixel by pixel (which substantially increases the fixed pattern bias structure because the time between reset and first read differs from pixel to pixel).  We have implemented some of these modes for obtaining high frame rate data for observing occultations. In this mode it is possible to make movies for studing transient phenomena such as occultations.

Frame from an IRCAL speckle movieClick on the movie frame to see an example of a short movie.  The right-hand panel shows the AO-corrected image, the left shows the varying speckle pattern of uncorrected seeing. The AO corrected images correspond to 10 coadded 57 ms exposures; the speckle movie shows individual 57 ms frames.


To facilitate efficient observations, it is necessary to co-ordinate the interaction of many telescope and adaptive optics sub-systems, all operating on different platforms.  To manage the distributed computing in this inhomogeneous environment, we use a simple message passing methodology, based on the Music/Traffic messaging system developed for the Lick and Keck observatories.  Various components of the software are written in C and Tcl/Tk and communicate via Music messages.  We are in the process of migrating this software towards a keyword model, as used at the Keck Observatory.  The bulk of the camera and AO system user interfaces are written in Tcl/Tk, which also offers the advantage that external scripting capabilities are readily available because the core language is designed for scripting.

A unique feature of the data acquisition software is that it provides a facility for reference bias subtraction during image readout and processing.  A store of dark frames taken with various readout configurations is maintained, and if the current readout configuration matches the extant bias frames, an appropriately scaled (for number of coadds) bias frame is subtracted from the image before it is saved to disk.  Although this does not in general substitute for dark subtraction in data reduction, it is highly successful in suppressing the static bias pattern and bad pixels for the purposes of image display and quick-look data reduction.

Facilities are also in place in the user interface for the management of calibrations.  The software maintains a list of all exposures taken during an observing session in the form of a palette of readout configurations.  This encourages the operator to utilise a homogeneous set of readout configurations throughout the night's observations. The user interface can then generate from this set of readout configurations a script to acquire a complete set of calibration darks.

The user interface incorporates an interface to IDL by loading a dynamic Tcl extension to call the IDL libraries.  IDL scripts are used to perform quick-look image manipulation and display.  There are facilities for the observers to incorporate their own IDL scripts into the image display (even replacing the image display with a customised IDL tool).


IRCAL mounts on the AO system optical bench, on a focus slide that is approximately aligned to the optical axis of the AO system and telescope.  Final alignment of IRCAL to the AO system is accomplished by adjusting two steering mirrors.  Pointing and centering are slightly coupled between the two mirrors, so the alignment is accomplished iteratively.  A point source at the input of the AO system is moved out of focus to observe the shadow of the two intervening pupils (a pupil mask on the DM, and IRCAL's cold pupil). From the shape of the shadow it is possible to determine the pupil registration and adjust the beam centering accordingly.  An in focus image is then used to adjust the beam pointing.  Once this alignment is accomplished, it is stable.  A particular advantage of the recent redesign of the Lick AO system is that the field steering of the wavefront sensor is independent of the field that IRCAL sees.  Since IRCAL looks through the telescope along a stable optical path, we achieve very good stability of the telescope contribution to the thermal background.

The AO system at Lick observatory is a facility instrument, available to the general user community.  Although the calibration and alignment of the AO system in general requires a specialist, the routine observing operations of acquiring objects, dithering exposures (by offsetting the telescope and counter-steering the wavefront sensor to track the guide star) and monitoring the status of the AO loops, do not.

As both the AO system user interface and camera user interface have been written in Tcl/Tk, we have implemented a method for Tcl remote procedure calls (rpc) using the Music messaging system.  This allows buttons in one user interface to invoke actions in the other in a straightforward way.  As the two systems have been developed largely independently, we have found this to be a useful way to present a user with a single screen to operate both the AO system and IR camera.  We have also used this facility to construct quick, simple distributed scripts.  For example, we use this method of rpc to enable remote operation of a serial port for communication with motor controllers.


We have characterised the optical performance of IRCAL with images of pinholes in a cold focal plane mask.  Using a grid of laser cut pinholes, we have determined that the image quality is very nearly diffraction limited across the detector field of view.  Although raytraces of the optical design show negligible aberration across the field of view, there is a slight astigmatism in the corners of the field of view.  We are in the process of characterising the optical performance with phase diversity measurements, which we can also use to cancel camera and static non-common path aberrations at the DM.


Band Camera 
efficiency (1)
throughput (2)
Background Point source 
(mag arcsec-2) SNR = 5   t=300 s
J (1.25 µm)
 0.08 ... 21.8
H (1.65 µm) 0.67  0.16 14.4 20.5
K (2.2 µm) 0.67  0.13 9.3 17.8
KS (2.1 µm) ... 0.14 10.3 18.3
(1) Measured from the front of the dewar window to detected photoelectrons, including filters.
(2) From top of the atmosphere for a 3 m clear aperture telescope, to  detected photoelectrons.

The throughput of IRCAL is excellent, but due to the large number of surfaces in the AO System, and the telescope itself, the total system throughput is low, and the emissivity high, hence the high K-band background.  We expect that the throughput will improve this year, as the 3 m primary has been realuminised and some of the AO system mirrors have been replaced.

K-band image of the Herbig Ae/Be star LkH alpha 234. The image is log stretched.  Total exposure time is 50 s.  The 1.5 arcsec companion is barely visible in the seeing limited image, but the AO system improves the contrast substantially.  Also a companion (1.2 mags fainter) is clearly detected 0.3 arcsec from the primary with adaptive optics.  Note that each point source is surrounded by a partial Airy ring.

Last updated 18 April 2000

James R. Graham -- Email:



(510) 642-8283
University of California
Berkeley, CA 94720-3411

Click here to return to the IRCAL home page.