IRIS Users' Guide

AAO User Manual 30a

Version 2.5

David Allen, December 1993.

Click here to return to the AAO Instrument Manuals page.

The version of this document dating from December 1993 was converted to HTML in April 1996 by Chris Tinney.

A postscript version of this document suitable for direct printout can be found here.


  1. IRIS
    1. Firing up the software
    2. Using the software
    3. Setup requirements
    1. The default display
    2. Changing the display
    1. 5.1 Darks, flats, biases etc.
    2. 5.2 Photometric standard stars
    1. Configurations
    2. Calibrations
    3. Source acquisition
    1. Configurations
    2. Setup
    3. Software
    4. Linear polarization observations
    5. Circular polarization observations
    1. Focusing
    2. Light loss between pixels
    3. Saturation effects
    4. Remnance
    5. Stray reflections
    6. Dust on the Fabry lens
  13. Appendix - STANDARD STARS
  15. Appendix - Filter profiles
  16. Appendix - Configuration Summary


This IRIS manuals has been divided into two parts, this Users' Guide (UM 30a - this document) and a Reduction Guide (UM 30b).

The newcomer needn't peruse this manual in its entirety. Section 1, the first portion of section 2, and all of section 3 and section 4 are probably required reading. It is as well to skim through section 5 as a reminder of the calibrations you will need.

There is also a file IRIS.SUMMARY for those planning telescope applications.

Most observations will require images to be taken at a number of positions of the telescope. The sky is so intense, and the response of the system so structured, that you must subtract a sky frame from your data. This can be done only by observation of several different fields. You can obviously have the night assistant offset the telescope to different positions, but it is more efficient to use the telescope offsetting procedure described in section 6.

Section 7 is relevant only if you intend to take some spectra, and section 8 only if you will attempt the spectral mapping option. Section 9 covers polarimetry.

Section 10 contains a few important matters that don't belong anywhere else, such as the focusing requirements, and Section 11, the close down, is short enough that you don't need to consult it till dawn begins to pale the sky. Section 12 lists the sensitivities.

The list of standard stars, in Appendix 1, is useful to have on hand when observing.

Appendix 2 provides some scripts to use for quick on-line processing of IRIS data. Currently the scripts provided are only useful on the VAXes. These will be updated with UNIX scripts soon.

Appendix 3 provides some filter profiles.

Appendix 4 summarises the most commonly used configurations with IRIS.

We have developed a few command procedures that simplify data assessment at the telescope. These work on the Figaro files generated when data have been transferred to the VAX, and are given in Appendix 2 whence they can be edited into command files.

This guide should be used with a Web browser. Searching for key words usually finds the information you most need quite quickly. A postscript version of this manual will also be made available here, however it will generally be updated less frequently than the on-line version.


The advent of two-dimensional infrared detectors has allowed AAO to construct a camera and low-resolution spectrograph for the wavelength range 1.0-2.5 microns. IRIS was designed and constructed at AAO and commissioned on the telescope in December 1990 and February 1991. If you want to refer to a published description of IRIS the most complete is Allen et al., 1993. Proc. Astr. Soc. Australia, 10, 298.

The detector is a hybrid array of 128x128 pixel format with 60 micron pixels. The IR-sensitive material, mercury-cadmium telluride (HgCdTe), is bonded to a silicon multiplexer by individual indium columns. The array was manufactured by the Rockwell International Science Center of Thousand Oaks, California, and it would be appreciated if this fact were mentioned in all publications based on IRIS data. The Thousand Oaks branch of Rockwell is not a commercial enterprise, but an R&D lab. Few detectors are made for external use, and each is constructed on an individual basis. For all its idiosyncracies, the array acquired by AAO is very much state of the art and one of the best that has been produced.

The detector is housed within an enormous home-grown dewar and operated at a temperature of 82 K. It mounts at the Cassegrain focus and can be fed either at f/15 or with the f/36 chopping secondary. Swapping between these two f/ratios requires a top-end change. The chopping secondary is not driven, and is merely used to provide a convenient f/ratio. It should be used in the axis position.

IRIS has been integrated into AAO's instruments under CCD software control. A single terminal attached to the VAX 4000 serves both to configure the instrument and to operate the detector. Operation includes control of several wheels within the dewar which insert filters and different optical trains. Included in the latter are cross-dispersed transmission echelles for spectroscopy. Collimated beams are provided for the spectroscopic option, and used also for direct imaging.

The following filters are currently provided; [L] = lower wheel, [U] = upper:

BEWARE that the f/15 wide configuration puts an f/1.6 beam through the filters. The narrow-band filters are therefore not always so effective. In particular for [Fe II] you must use the 1.65 filter.

The K' filter was defined by Wainscoat & Cowie (1992 Astron J. 103, 332). The sky background is lowered by a factor of about 2.5 and the star signal is typically 1.1 times that at K. The sensitivity is thus ~0.6 mag better than K. Kn has almost as high sensitivity and greatly reduces the effect of water vapour at the short end of the K' filter. The sky is slightly higher, giving a gain over K of ~0.5 mag. It will probably become the 2-micron filter of choice for moist observatories such as Siding Spring. There is some fringing with the [Fe II] filters, but it appears to cancel when the sky is removed.

A schematic layout of the IRIS instrument. Note there is no shutter.

And here is a more detailed technical diagram. A better resolution version can be found here


The chip is slow to read out. To read and reset every pixel requires nearly 1.5 sec; to read and not reset requires 0.9. The readout proceeds row by row up the array (downwards on the xmem display). An integration starts the moment the relevant pixel has been read, so that the first rows are exposed earlier than the last rows.

There are three ways of reading out the array, and they have different uses. In parlance they tend to be called modes, but from the software's point of view they are methods, and will be called that here. They are selected by METH 1, METH 2 or METH 4. When selected they are announced by the message IRIS_OBS1 (etc.) in the top right portion of the screen.

We recommend:

On a first reading you can safely skip the remainder of this section.

In method 1 the incoming charge accumulates for the exposure time, after which it is read and each well is reset to the bias level. This is the end read, Er. Method 1 should be used only in high background cases, because the readout noise is high (~120 e-). However, it is the most efficient readout because there is scarcely any dead time between reading a pixel and readying it to accept more photons (there are other dead times in the system, at the start and end of exposures when processing and writing to disk occur). Typically, prefer method 1 for broad-band imaging. Be aware that method 1 shows off the whole, ugly bias to perfection, so your raw frames will have odd columns of about 15000 and even columns of 24000 ADU. Be aware also that THE FIRST CYCLE IS NOT USED, so if you ask for a small number of cycles with long integration times you will be working at kow efficiency. If low count rates force this to overcome readout noise, consider using method 4.

For method 2 we use the so-called double-double correlated sample, in which an exposure starts with a reset, Sr, and ends with another, Er. The noise is reduced, to about 80 e-, and the differencing against an Sr gets rid of most of the bias structure. A residual remains, and is actually negative (around -200), for reasons that the chip alone understands. There is a time penalty here: one extra complete read of the array, costing 1.5 seconds, has to be performed for each exposure. Method 2 has gone out of favour (it doesn't flat field well) but is useful for idle mode because it gets rid of most of the bias.

Method 4 is even more costly in time, because every exposure begins with an Sr and ends with an Er, neither of which is used. Between these two is a series of non-destructive reads (NDRs). At the end of the exposure, the entire data set has to be linearised (see IRIS reduction guide) in readiness for a least-squares linear regression fitted through the NDRs to define the incident photon rate.

Method 4 has a number of advantages. The bias is very small, so that in real time you see useful data without needing a previous dark exposure. Long exposures are punctuated by frequent displays so that one can see how the exposure is progressing. And pixels that saturate (as on bright stars) are still correctly measured by using only the span of reads up to saturation.

At the start of the integration there is an additional bias component that decays away over a few seconds. For this reason the first read is thrown away, and for accurate work it is best to take reads no more frequently than every 5 seconds. Nonetheless, each method 4 read has a noise of only 40 e-, and the noise on the fit falls below this figure for large numbers of reads. Thus method 4 is essential for spectroscopy where extremely long exposures are practicable and often desirable.

Do not mix method 4 data with method 1 or 2: there are subtle but important differences in the gain from pixel to pixel so that flat fielding and calibrations will leave some of the alternate column striping.


3.1 Firing up the software

IRIS runs under the CCD software. The startup procedure is as follows.

Log on to the OBSERVER account on AAT40A. When the prompt comes up, type:

(or CCD2, depending on the controller in use)

The second of these commands will create Figaro files on the VAX in .SDF format instead of the default .DST. The difference is noted below, in the reduction guide. .SDF files are preferred because they reduce dead time between integrations.

The third command selects the CCD controller. There should be a note to tell you which controller has been hooked up, on the control-room white board.

Your support astronomer may baffle you with variants on this set, such as VERSION 2OCT
injected before the CCD1 command. The white board will tell you if this is the case.

A different command is required if running with telescope control (see section 6).

After about 40 seconds the screen will be redrawn with an upper boxed area where status is displayed. Below this a string of messages scrolls up the screen while a fixed message requests you to wait for the "Iris: " prompt. One minute later a display will appear at top left of the XMEM screen. A few seconds after that the prompt will appear and you are ready to take data. If the VAX is heavily loaded these times will be optimisitic.

The only exception to this (other than in the event of a failure) will be the appearance of a prompt "Show: ", which occurs if one of the voltages or currents in the electronics, or the detector temperature, are out of range. Alert your support astronomer to the condition, which is usually not serious. You can return to the Iris prompt by typing END.

3.2 Using the software

There are now four sections of screen in use. From top to bottom these are status, commands, messages, and the active typing space.

Most of the messages will pass you by. However, it is as well to get into the habit of watching for the one that tells you a run has been recorded (Data for run xxx recorded in DISK$INST:[CCD_n.yyyyyy]zzzzz0xxx.sdf;1). When using ICL (for mosaics) this message appears instead on the ICL terminal. Also, when you summon a dark integration (with the USE command), watch for the message telling you it has been installed. Scroll back through the message area with the Prev Screen button (or Ctrl P); scroll forwards again with Next Screen (or Ctrl N).

For a relatively uncluttered life, use only the commands listed below. They can be abbreviated to the shortest unambiguous command. In most cases the options will be listed (sometimes in response to a ?). Asterisks identify commands enlarged upon below.

Aborts the run in progess and does not update the run number. No data are recorded to VAX.
Changes look-up table in XMEM (see display section, below).
Appends a comment to the current or previous run.
Sets number of repeat exposures that will be taken and averaged.
Performs a dark integration and loads into XMEM.
See XMEM display section, below.
Takes a trial observation but doesn't keep it.
Prepares to log off.
Sets field flattener wheel. (Double TT at end of FLATT is essential!)
Takes trial integration, but doesn't record it.
Sets appropriate position of grism wheel: OPEN for imaging, IJ or HK for the relevant echelle.
Chooses method 1 or 2 for idle.
Selects idle exposure time. Exceeding 5 seconds is rash but allowed.
Records a GLANCE exposure.
Sets lens wheel.
Sets lower filter (OPEN, J,H,K,KP [= K' filter], KN [=Kn] or narrow-band chosen by entering wavelength).
Chooses data-taking method 1, (2) or 4.
Sets next run number.
Enters object name (affects integration in progress).
Sets the interval between non-destructive reads (method 4), seconds.
Chooses processing method for mode 1 data
Takes data.
Gets out of repeat mode.
Sets slit wheel.
Ends observation when next cycle finishes.
Repeats multiple integrations.
Sets exposure time, seconds.
Sets upper filter wheel (OPEN, or narrow-band).
Puts specified run into XMEM dark slot.
Selects window IRIS128 (= ROCKWELL_128), IRIS31, IRIS25.

Notes on the commands

LENS. It is not always enough to drive the lens wheel into place. At present the upper grism must also be at the right setting. As of May 1993 you need to drive this on the separate terminal by typing HAND followed by the position. There is no echo when you type, but you will get a handshake when it gets there, and an error message if you type it incorrectly. We will install higher-level software to control this soon.

For direct imaging you need:

f/15 wide field      Lens: MIRROR   Hand: 1  (remember FLATT IN too)
f/15 intermediate          INTERM         1
f/15 polarimetry           INTERM         7
f/36 wide field              WIDE         6
f/36 intermediate          INTERM         2

The spectroscopic configurations are listed in section 7.1.

WINDOWS. Window options here refer to the chip, not the XMEM. Small windows are generally useful for standard stars. This is because the chip can be read out faster, avoiding saturation. The IRIS25 window spans the chip centre, (53-77, 53-77) but includes one bad pixel. IRIS31 is a clean portion of the chip about half way to the top right corner on the XMEM display (x: 69-99, y: 15-45). This is a good one to focus in if you want comparably small images all the way from the centre to the corners.

DARKS. There are degrees of darkness. Requesting DARK places the Flattener in the blank position, just above the detector, and takes an exposure. For darker conditions set SLIT BLANK too. The object name will be set to DARK for the run and then reset to what it was before. But the flattener is left at BLANK (except in ICL offset-runs, described below).

TIME. There are several times in the system. You'll find `exposure time' on the screen, and this is the one you set with the TIME command. Usually it will be dictated by the need to avoid saturation of the detector. The total time you spend integrating on a single field may be made up of many exposures, the number being determined by CYCLES. In this document exposure refers, not unreasonably, to the exposure time, and integration to the product of exposure and cycles.

Note that to avoid transients the first frame of mode 1 data is not used. For this reason don't use very long exposure times. If the signal is low, prefer method 4.

CYCLES. Cycles and repeat mode can be confusing. A single run may comprise many short exposures, usually to prevent saturation of the detector. These are stored successively in an XMEM buffer known as the 3-D stack. In method 1 one place will be used up for every cycle. At the end the results are averaged (with n-sigma clipping if you have selected it with the PROCESS command) and sent to the VAX. Thus a 100 second integration may be made up of 10 cycles of 10 seconds each, separated only by the readout time of the chip. It is important not to exceed the permitted size of the stack, which defaults to 256 cycles. Remember that in method 4 the stack comprises n+1 entries per cycle where n is tyhe number of periods in the integration time. Increase it with the command RESIZE n CYCLES.

Repeat mode, on the other hand, repeats the entire integration sequence, including writing it to the VAX, and bumps the run number appropriately. You can repeat integrations of many cycles.

Beware that cycles can be increased, but not decreased, once a run has started. To terminate early you must issue a STOP command.

PERIOD. When using method 4, your exposure will be made up of a number of periods. You should therefore select TIME to be a multiple of PERIOD. The hierarchy reaches its acme here --- the total time spent on an object will be (number of repeats)*(number of cycles)*(exposure time) and exposure time is itself broken up into a number of periods.

USE. You can call up any appropriate file you have generated, e.g. USE DISK$DATA:[IRIS]K_MEDIAN_SKY DARK However, if the name becomes too long, ADAM will not accept it. In this case, enter OBSERVER on another terminal and $ DEFINE/JOB Q DISK$DATA:[IRIS]

To IRIS `dark' is a general term for anything that is to be subtracted from the current exposure. Expect odd results if the dark was taken through a different window.

IMETH. Typically prefer IMETH 1 when you need to check for saturation visually or numerically, and IMETH 2 to see dark-subtracted data. If you want sky-subtracted data, take an image in METH 1, USE it as dark, and set IMETH 1.

DUMMY. Exposures are saved on DISK$DATA in directory [CCD_n.yymmdd] (n=1 or 2; yymmdd is the conventional date]. Files are named A, B, C ... in order of acquisition.

PROCESS. Determines whether sigma clipping is used by the Xmem when it compresses all cycles into a single image. Options are: PROCESS AVE_ER which averages all cycles and PROCESS SIGAVE_ER, the default, which applies 3-sigma clipping for every pixel, thereby rejecting cosmic rays, etc. Note that this takes longer. Method 4 data should be processed using FIT_NDR. The processing method is announced on the message section of the terminal at the end of each exposure. PROCESS NONE dumps the whole stack.

3.3 Setup requirements

There are relatively few requirements at the start of the night. Your night assistant will perform some snafus on the TV direct, and when light comes through to IRIS you should find the star near the centre of the image. Use the IRIS25 window to check, and if you want to tweak it ask the night assistant to set up aperture A under APOFF, driving the star to the chip centre.

If you are imaging it is useful also to APOFF aperture B to the middle of the IRIS31 window. Position the star about half way from centre to top right of the Xmem display on the IRIS128 window, then bring in the IRIS31 window and centre the star in it. Actually, you'll find it helpful to put the star about 3/4 of the way across the window (i.e. north of centre), for reasons that will become apparent when you read section 5.2.

Approximate offsets from chip centre (IRIS25) to IRIS31 window are given below. Add these to the APOFF values for axis A and use as a first guess at axis B.

                          x         y
  f/15 wide            +11.5     -18.1
  f/15 intermediate     +3.6      -6.0
  f/36 wide            +14.3     -19.2
  f/36 intermediate     +4.0      -6.1

In this IRIS31 window you should focus the telescope. Each filter has an appreciably different focus at f/15 wide, but they become sensibly the same as the pixel size decreases. Your night assistant will enter them into FOCOFF for you. If you are short of time it will probably be OK to use focus on one filter (Kn or K' are best) and apply previous offsets between filters, as given in section 10.1. Hints on focusing are given there too. Note also that at f/15 wide you will get slightly different image scales for different filters, as noted in section 10.1.

If you seek accurate photomety be sure to read section 10.2.

If you are planning spectroscopy, you will want to start with a standard star. Put in the appropriate slit, and the star and slit will be visible on the TV. Ask your night assistant to move the star to define two positions along the slit so that the star comes in about 1/4 of the slit length in from each end. This can be easily assessed on the Xmem display.


4.1 The default display

The display on the XMEM is perhaps excessively comprehensive. The default layout uses 11 windows (not to be confused with windows on the detector) arranged as below. Do not be alarmed by broad vertical bands on the small displays: these are caused by beating between the IRIS pixels and XMEM display pixels.

The upper-case names in each of these display boxes are used and recognised by the software. The processed image is not a recognised address, being the final version sent from the XMEM to the VAX for disk storage. Those commencing D_ are dark-subtracted, and appear only after a suitable dark frame has been loaded into the XMEM.

The top left display is active when the instrument is idling between integrations. This is where you see the star field appear when the telescope is set. It is overwritten during a method 1 or 2 run by the current exposure. When a run comprises many cycles you will see the latest component exposure appear here, and the average of the whole set appear to its right, in AVE_ER.

The three NDR positions are used only in method 4. The middle one, DELTA_NDR, is the difference between the most recent read and its predecessor. LAST_NDR is cumulative, showing the total integration, and AVE_NDR is the same frame divided by the number of reads. Again, the D_ versions are active if a dark exposure exists in memory.

All images have the same orientation, with pixel (1,1) at UPPER LEFT. This is a vertical inversion of the standard Figaro display package. When IRIS is in its normal orientation on the telescope, the orientation on the screen is:

W to top, N
to right

Thus the sky appears mirrored up-down on the XMEM display, but the field is correctly displayed by Figaro. In reduction, the Figaro command ROTATE will orient an IRIS frame conventionally, with north at the top, west to the right.

Except in method 4 you will tend to be disappointed by what you see in the raw data. Subtraction of at least a bias frame is important. The best results are usually had in imaging if a sky frame is taken first (or, if dithering around a small object, the first of its dither positions) and introduced to the XMEM as a dark exposure, with the USE command. This must be done after it has finished recording the relevant run to the VAX.

In method 4, sky subtraction is more tricky. Use a first observation as the dark, e.g. in spectroscopy a spectrum with the object in a different position on the slit. This subtraction will only be accurate at the end of a cycle, and you can watch the subtraction of the sky improve read by read.

4.2 Changing the display

The default display in each is autoscaled and compromise histogram optimised, similar to the options in the Figaro command IMAGE. This default can be modified by DISP(LAY) commands. The format of all DISP commands requires commas between all entries on a line except after DISP itself. For example:

replaces AVE_ER with the specified black and white levels.
produces a numerical display about the chip centre filling the D_LAST_NDR window as fully as possible.

The DISP(LAY) options are:

Revert to autoscaling
Kills a particular display
Display numerical values
Revert to image
Position image at chosen spot on XMEM screen (available size is 1024x768)
Set black/white range
Enlarge a chosen portion

SCALE (set to a suitable maximum) or NUM, are often used to estimate suitable exposure times. SCALE should be used whenever you feel a need for the true count level, and when autoscale is losing information in the bright parts of an image, as when viewing through the slit. SUBSET and SCALE is a useful combination when focusing.

Always give odd values to xstart and ystart when using POS.

You can clear the screen at any time with the engineering command: OBEYW XMEM MICRO
value> "clear 0" (`clear' must be lower case as this is unix)

There are various colour look-up tables to experiment with. Try GREY, GRAY (saturated pixels turn red), IRAS, GRJT, FALSE, JOS, PANDORA, STANDARD or CONTOUR. Be warned that if you are displaying numbers some of these tables make the writing vanish.

This suite of DISPLAY commands allows you to redefine the screen exactly as you wish. To overcome the tedium of typing all these commands you can set up a command procedure to do it, provided you are running under ICL (see section 6). The command is LOAD file where file has extension .ICL. The best way to set up a file is to borrow from DISK$USER:[DAA.IRIS]DISP1.ICL. You'll see from this that you need to know which CCD controller you are on; DISP2.ICL is set up for CCD_2.

These files give you a much simplified screen layout in which all frames are the full 256x256 pixels to avoid beating between display pixels and the odd-even columns. The layout is:


5.1 Darks, flats, biases etc.

Getting data with IRIS is easy. Getting good data is hard. The hybrid nature of the chip ensures that the idiosyncracies of each part are compounded. Thus you will find in your data:

The procedures for removing all these effects are now standard, though there still seem to be occasions when not everything is cleaned out adequately. In an ideal world you would dark and/or bias subtract to remove (i) and (ii), after which (viii) would be removed by a simple linearity correction as used in optical CCDs, while (iii), (v) and (vi) would flat field out by a wavelength-matched observation of a continuum source. The bad pixels [(iv)] one merely has to live with, and in fact by the standards of most IR arrays there aren't many of them: about 70. However, the real number depends on wavelength and background flux, and seems also to beincreasing with time.

In practice the last four effects intercombine to create a different nonlinearity at every pixel. This is compounded by effects that you probably won't notice in your data, namely that the Sr that precedes every method 2 and method 4 readout introduces several additional components to the bias that fade away at different rates. The impact of these components depends on the frequency of readouts in both methods.

So how to proceed? Here is the best recipe.

  1. Make sure that you have a dark frame taken with the same observing setup as the data frame, preferably on the same night. It might be as well to ask someone knowledgeable whether there is a recent SWIRL frame available; if not, see below.
  2. Take flats for every wavelength and image scale you use, and for the corresponding readout method. The counts in a flat should not exceed 60000 in method 1 data (or idle). To verify that they don't, set idle to IMETH 1 and examine the numbers. Remember that the NUM display defaults to the centre of the chip, which has about half the peak value at short wavelengths. Remember also when assessing this that the numbers you see refer to the IDLE TIME whereas the data you take are based on the EXPOSURE TIME. Dome flats are the most reliable, but these cannot be taken with broad-band 2-micron filters in summer due to saturation, so use the narrow-band 2.12 (for K'), 2.16 (for Kn) or 2.21 (for K) filters. J flats are also tricky because of the rapid change with wavelength. It is not practicable to match the flat perfectly to the spectral content of astronomical data, and we know no way round this. The floodlight is too bright for broad-band filters. Use the desk lamp at prime focus access to illuminate the white patch above the telescope (windscreen 22 degrees, dome 0 degrees). Control the lamp intensity with the control behind the IPCS desk (you must ensure the desk lamp is plugged in to the floodlight's power supply). Take all flats with and without the lamp on, and difference the result.
  3. Whenever possible, observe objects several times in different positions on the chip.
  4. Make sure you take enough sky observations to create good median filtered skies. If you are observing many fields throughout the night, then there is probably no need to take separate skies. If you are observing one compact object only, observe it in at least five positions. For extended objects, take at least five sky integrations in different places.
  5. Aim for about 10000 ADU above bias when possible. If you are displaying numbers on the central spot, remember that this is a below average region in sensitivity. Use IMETH 2 and a shortish ITIME, then make a rough estimate of the exposure time needed to attain around 8000 counts. You will still get pretty good data if you don't go this far, however.
  6. Do all these things for each window you use.

When you linearise your data, you will be asked to provide a SWIRL frame. One of these should reside in the IRIS default directory. However, the SWIRL does change a little from run to run. The change is small enough that it can probably be ignored, but if you do want to generate your own version, here are the steps to take.

5.2 Photometric standard stars

If you expect to get photometry from IRIS frames you will need to observe the standard stars quite frequently. The list of stars is given in Appendix 1. Most are so bright that they should be observed with a short integration time, 0.5 sec or less. Thus you must use the IRIS31 window. Hopefully you will have set up one of the telescope apertures on this window at the start of observing.

For each filter make two observations totalling 10 or 20 seconds each, with the star lying about 1/4 to 1/3rd of the way in from opposite edges of the window. Motion N-S is simplest as you can use a standard offset between the two positions. However, a new intermittent bad pixel at the left edge of makes it preferable to use diagonally opposite corners. Some observers like also to move the star out of the beam and take a third, sky observation. Appendix 2 includes a command procedure that will yield an approximate systemic magnitude for the star, enabling you to monitor changes in the zero point.

The systemic magnitude is defined as -2.5 log (ADU/sec). On this system approximate zenith zero points in the IRIS31 window are:

J H K,Kn,K'
~21.6 ~21.9 ~21.8

5.3 Calibrations for spectroscopy

Section 7.2 describes these.


In many observing modes you will feel the need to move the telescope frequently. This may be to take a number of sky frames, to build up a mosaic, or to grow a spectral map. For these activities an alternative way to log into IRIS has been developed. Follow the startup instructions in section 3, but instead of the command CCD1 (or CCD2) type

$ RVICL CCD1 (or CCD2)

Before you do so, ensure that the terminal connected to CONSOLE_1 is free. This is currently port 6 on the server. It must even be logged out of the terminal server. It is on this terminal that the usual IRIS screen will appear, and once it has done so you can drive the instrument from that keyboard. On the terminal you first logged into you will now see the prompt


This terminal is available to call up a file that issues commands to both the IRIS control task and the telescope. The file allows control of the following:

Dark exposure or run
Object name
Telescope offsets (RA, dec in arcsec from start position) Exposure time
Number of cycles
Period (in mode 4)
Number of repeats

The file is read by the command

ICL> OFFSET_RUN filename

The default directory is DISK$RAW:[OBSERVER], and .DAT is the default file extension. DISK$RAW is specific to the VAX 4000.

This command initiates a sequence of runs. At each run the next line of the file is acted upon. This may involve offsetting the telescope, taking a dark exposure, or changing one of the other parameters. Runs continue until the end of the file. Note that you cannot alter the optical configuration or mode of taking data: these must be set before you begin.

A sample file follows. Spaces are ignored, but DON'T USE TABS. Also, if you create it on some other directory ENSURE THAT IT HAS WORLD READ AND EXECUTE PROTECTION; failure to do so results in an error reporting failure to open file.

DARK _dark 0 0 TRUE 1.5 10 1 1
RUN north 0 100 TRUE 5 8 1 1
RUN west -100 0 TRUE 5 8 1 2

This file will take a 15-second dark run with exposure time 1.5 sec, then offset the telescope 100 arcsec north to take a 40 second exposure with 5-second exposures. Finally, two similar runs are taken 100 arcsec west of base.

Things to note here are:

This telescope control mode ensures much more reliable mosaicing and offsetting, and is easier on observers and night assistants, but there is an extra overhead between runs of about 10 seconds compared to conventional observing.

When running under ICL the messages announcing recording transfer of data to the VAX appear on the ICL terminal.

There are some tricks to watch in the event of a system crash when running under ICL. You can abort a telescope offset run by Ctrl C on the ICL terminal. But you should clean up by typing TIDY on the same terminal before you try to run again. It never hurts to type it twice. The telescope will be left where it last offset to, and you should resume the offsetting without driving back to the start position. But if you either log off from ICL (and on again) or slew the telescope, then the offset data sent to the telescope will be reset and you should then resume the next mosaic pattern with the telescope at the start position.

If you do have to restart the entire software, the procedure is:

  1. ICL> EXIT
    78a> LOGOUT
  2. Press the black cold start button on the xmem
  3. Ctrl C on the IRIS terminal. If nothing happens, wait a second or two and do so again. Eventually you will descend to the server:

    Enter username> type anything polite here

    Local> logout
    Local -020- Logged out port 6

    It is necessary to get completely out of the server before starting up OBSERVER under ICL
  4. Now startup as usual.


7.1 Configurations

There are several configurations for spectroscopy. In the present version some of these have to be selected in the cage, and it is necessary to be shown the ropes by a support astronomer or technical staff member.

At f/15 only grisms are available, covering the H or K windows, with spectral resolution 300, and pixels 0.6 arcsec along a slit that covers the full height of the chip (75 arcsec). The same grisms can be used at f/36 with the intermediate field, but a very narrow slit (0.4 arcsec), to give 0.27 arcsec spatial pixels, or with the wide field when they cover almost 1 micron of spectrum centred on the H or K window, and offer a resolution around 100 and 0.79 arcsec spatial pixels.

The configurations are (see section 3.2 re the Hand wheel):

Topend Field Disperser Resol'n Hand Slits

f/36 Wide IJ echelle 400 1 NARROW, LONG, WIDE f/36 Wide HK echelle 400 1 NARROW, LONG, WIDE f/36 Wide H grism 100 5 WIDE f/36 Wide K grism 100 4 WIDE f/15 or 36 Interm H grism 300 5 GRISMS, WIDE, NARROW f/15 or 36 Interm K grism 300 4 GRISMS, WIDE, NARROW

The slits have the following dimensions. Because their edges don't exactly land on pixel boundaries you will probably have to reject the extreme ends, so assume they are a little shorter than nominal. Figures in brackets are configurations you probably won't use. Standard configurations use slits that project to about 2 pixels; wider slits will degrade the spctral resolution appropriately.

SLIT Pixels | Arcsec INTERM WIDE | f/36 f/15 | NARROW 5.8 x 49 1.9 x 16 | 1.4 x 13 (3.5 x 30) LONG 6.2 85 2.0 28 | 1.6 22 3.8 52 WIDE 22.2 48 7.3 16 | 5.8 13 13.5 29 GRISMS 1.8 128 ( 50) | ( 40) 1.2 75* MILES (3.5 128) 1.2 128 | 0.9 100 (2.1 75*)
* Focus poor at ends; about 60 arcsec useful

Most of the slits are quite accurately concentric, but MILES id displaced about ten pixels from the others and therefore needs a separate flat field and wavelength calibration.

The wavelengths ranges are:

IJ echelle Order 10 0.860 - 0.953 (shorter than 0.86 filtered out)

9 0.905 - 1.049 8 1.008 - 1.168 7 1.138 - 1.318 6 1.310 - 1.515 HK echelle 8 1.439 - 1.704 7 1.622 - 1.913 6 1.855 - 2.184 5 2.170 - 2.537 H grism 1.4 - 1.8 (f/36 wide: 1.2 - 2.1) K grism 1.95 2.4 1.8 - 2.5

For all spectroscopy run with the UFILT and LFILT wheels open and FLATT out.

1000 seconds is a good exposure for a faint object with the IJ, though you can go longer. HK echelle data are limited by the thermal radiation at the long wavelength end, which depends on temperature. In winter you can integrate for 150 or even 200 seconds before that saturates, but in summer 100 may be too long. The H grism at f/15 handles exposures of several hundred seconds, depending on the airglow. The K grism saturates after a few tens of seconds.

Use method 4, so you can see what is coming in. Prefer a period of at least 5 seconds unless the exposure time is under 100. Remember that if some parts of the spectrum saturate the data are not lost in method 4, but you use only that part of the exposure up to saturation, so worsen the S/N in those regions.

During spectroscopy a visible object being observed can be seen reflected off the slit jaws with the TV on the MAIN FOCUS setting. The scale and field of view is:
f/36 0.95 arcsec/mm 18x25 arcsec
f/15 2.2 45x65
At high zenith angles consider using the routine LAMBDA (on the BANTAM terminal) to correct for atmospheric dispersion. Do so after optically aligning on the slit, and enter wavelength in Angstroms.

Make sure you avoid observing the sky without a filter of some kind, or you will have a remnant image of the slit on your first spectrum. This is particularly likely to occur as you reconfigure after centering in the slit. Insert the echelle BEFORE you open the lower filter wheel. Similarly, if changing between IJ and HK echelles, it is best to use FLATT BLANK during the move.

It is strongly recommended that compact objects be observed in two slit positions, one near each end. This helps to eliminate the effect of bad pixels but particularly to improve sky subtraction. For extended objects the echelles are not satisfactory.

7.2 Calibrations

Set up a flat field by taking a spectrum of the dome floodlight illuminating the white patch (the chimney lamp is probably OK too). Set the dome PA to 0 and the windscreen zenith distance to 22 degrees, and control the lamp from the panel behind the IPCS desk. As in photometry you will need a dark, which is better provided by an exposure with the lamp off. The desk lamp is a little faint, so plug the floodlight into the adjustable source.

You may want to calibrate the curvature of the echelle orders and the tilt of the slit, though these are sufficiently stable that you will probably be content to use standard values. The slit tilt is normally indicated by night sky lines (for faint objects) and in arc exposures. The curvature is best mapped by the flat field exposure.

For wavelength calibration you need to take several arcs. Night sky lines and astronomical emission lines provide partial calibration (for instance molecular hydrogen is good for the HK echelle order 5). In the afternoon you can use the chimney lamps and dome lights. Observe the white patch on the dome illuminated by the normal dome lights for mercury; put in the flap and both argon lamps for argon. To get xenon you use the flap but must also switch on the lamp at a small brown box mounted atop the electronics box on the side of the telescope. Make sure you switch it off afterwards. Access this from the west catwalk beside the Cass cage, at about shoulder height. Exposures times of order 1 minute in method 4 are best, with a few cycles. In the 2-micron region difference exposures with lamp on and off to improve contrast. The data should be flat fielded. The reduction guide IRIS.REDUCE describes how to handle mixed arc data.

Spectrophotometric standards are not well defined at present. Try using the time-honoured infrared standards listed on the CCS disk, but avoid the brightest few. Indeed, with the grisms only the very faintest avoid saturation. Be warned also that their fluxes are not presently known below 1 micron. They are listed in Appendix 1.

7.3 Source acquisition

Bright visible objects should be acquired and guided on the slit using the TV viewing system. However, at high zenith angles remember that the optical and infrared images will be displaced, so maximise the signal on the detector before deciding where to guide; and/or use LAMBDA.

Faint objects should be checked by direct viewing through the slit. This is a somewhat cumbersome operation, but quite effective, and it will be much improved when fiducial marks are available on the Xmem. The window IRIS25 is very helpful.

When you return the set up to spectroscopy, ALWAYS INSERT THE ECHELLE BEFORE YOU REMOVE THE BROAD-BAND FILTER. Failure to do so will saturate the detector with unfiltered radiation, causing remnance.


Although IRIS has gained one or two dimensions over IRPS and FIGS, this still is insufficient to satisfy some observers. Spectral maps, cubes comprising two spatial and one spectral dimension, are an attractive and powerful way to take data. At optical wavelengths Aspect has been developed to generate such data using the IPCS. For IRIS there is no established software to create such data cubes, but there are ways of gathering it. This section outlines how to do it.

The trick is to scan the telescope perpendicular to the slit direction, recording spectra at each of a number of positions. The number of frames taken is thus the number of positions required along the scan line, or a multiple thereof for multiple passes. The resulting data cube will have as the other spatial dimensions the 128 columns of the chip in the spectral direction (unless windowed smaller) and the 128 rows that include for each order spatial information along the length of slit used. In fact the latter will generally be longer than this for the echelle formats, either to allow for order curvature or to include more than one order.

There is no software to drive the chopping secondary, so the telescope must be moved perpendicular to the slit. The preferred procedure is to set up an offset_run file for the entire scan. It's cumbersome, but allows observation in method 4. You get one file for each position in the scan, and will want to combine them into cube using GROWXY in Figaro.

If your object is bright enough for method 1, you can also drive the telescope via its own controls. Differential tracking rates can be computed, in arcsec per hour (seconds of time/hour if you want to include an RA component in the scan). Or Trail can be used. Trail runs back and forth by a predetermined amount, and is useful if you want to make more than one pass across an image. It works in millimetres so you need to convert at 1.5 or 3.5 arcsec/mm for f/36 or f/15 respectively. Differential tracking just keeps going ad nauseam, and is best for faint objects where the integration time is long enough that you don't want to make more than one pass, at least without stopping to recentre the field.

To set up such a scan, follow these steps.

  1. Observe in method 1. Decide what integration time you want to use. This will be set by S/N considerations.
  2. Decide the steps between samples in the scan direction. A natural choice is the pixel size in the spatial direction so that there is no distortion of the image. This is also about half the slit width, so gives good sampling.
  3. From (1) and (2) deduce the scan rate desired.
  4. Decide how far you want to scan (and how many passes if using trail). This defines the total number of integrations to be taken. Add a few for safety and enter this as the number of cycles. The maximum is 500. Hence determine the total time. If using trail, set an appropriate period remembering that one period is a double pass, up and down, and discounting the few extra you added to the number of cycles.
  5. If using differential tracking, offset to one end of the scan. Start data taking, then start the differential tracking three or four seconds later. If using trail, start from the middle of the scan and begin both IRIS and trail at the same instant; both take about 7 sec to get underway.
  6. MOST IMPORTANT OF ALL. The image that is created at the end of the observation and written to the VAX is the sum of all the frames, and of minor interest. At the end of the integration the complete set of data resides in the Xmem. IT IS IMPERATIVE THAT YOU GET IT OUT BEFORE YOU START ANOTHER INTEGRATION. If you don't do this, you lose it. There are no second chances. Wait till you see the message that the resultant image has been written to VAX. Then type:
    Take care with the spelling. This will create a new run number. Wait for the stack to be written out before you start another observation. If you change the instrument setup during this readout you may find the new configuration in your FITS header. Note that 500 full frames is a large file, of order 50000 blocks, and takes about 5 minutes to write. If you plan a series of such stacks, an alternative is to issue before the runs the command PROCESS NONE. However, you would then need to reset the original process when resuming normal data taking (see section 3.2).

The result of dumping the 3-d stack is to create a cube in Figaro format with each plane being a separate frame, one for each position along the scan. The scan direction occupies the t-dimension of the final cube. This is identical to an offset_run scan built into a cube with GROWXY. XYPLANE will extract any one of the constituent frames. YTPLANE gives a map at a chosen x-position (= wavelength).

Your scan should always include regions of blank sky. Also make sure you take matching dark and flat exposures. Obviously you don't need to dump these as cubes. You can grow cubes out of images in Figaro (with GROWXY)


The polarimetric configurations within IRIS were developed and built by Prof J H (Jim) Hough of the University of Hertfordshire. He should be credited in any use of them.

9.1 Configuration

Imaging polarimetry with IRIS is available in the intermediate field configuration at either f/15 (0.6 arc sec pixels) or f/36 (0.25 arc sec pixels) using a MgF2 Wollaston prism polarizer which gives dual beam polarimetry of high accuracy. The polarizer is used in conjunction with rotating achromatic retarders covering the wavelength range 1 to 2.5 microns which are in a module external to the dewar, currently mounted on one of the guide probes in the A and G unit.

A modified configuration which will allow use of the grisms in conjunction with the polarizing prism should become available eventually.

9.2 Setup

Software - The polarimetry system uses a slight modification of the standard IRIS software which includes additional commands to control the polarimeter, and is otherwise identical to the standard system. Start up with the command $ SYSTEM IRISPOL instead of SYSTEM IRISCT and otherwise start the system as normal. Thus you will need to exit from the software and restart if you decide part way through a night that you want polarimmetry. However, if you startup IRISPOL you have access to all the usual imaging and spectroscopy.

Polarimeter Modules - There are two modules, one containing a half-wave plate for linear polarimetry, and the other a quarter-wave plate for circular polarimetry. The modules are not easily accessible once IRIS has been mounted on the telescope, so MAKE SURE THE RIGHT ONE IS INSTALLED FOR YOUR RUN. The module is driven into place from the telescope control computer using the GPROBE task. Select probe 1 using the button (marked CASS P1/CASS P2) and drive to the position X = 203.91, Y = 576.03 using the buttons on the console. Currently the X drive is disabled, but should be fixed at a position close to that required. If you revert to imaging or spectroscopy you can leave the waveplate in the beam, but will lose some light, so be sure you also leave it in place for standards and, possibly, for flat fields too. The focus does not change whenw the waveplate is inserted.

IRIS should be configured as follows:

The polarizing prism produces two images of any object corresponding to the ordinary and extraordinary polarization states. These are displaced by about a quarter of the chip width (N-S for the usual orientation). The precise spacing is slightly wavelength dependent as follows (J, 32.0 pixels; H, 31.6 pixels; Kn, 31.1 pixels). To prevent the two images overlapping, a mask is used which has two slots covering the full height of the chip and each about 32 pixels wide and spaced by 32 pixels. If you want an area larger than 128 by 32 pixels, the telescope must be moved to fill in the gaps between the slots.

9.3 Software

The polarimetry software contains the following commands in addition to the standard system.

- Drives the wave-plate to angle 'p' degrees.
CAL pos
- Drives the polarization calibrator to a specified position. The available positions are:
- Switches the system to polarimetry mode. In this mode a RUN command (or any equivalent command such as DUMMY or GLANCE) will take four observations at plate positions of 0, 45, 22.5 and 67.5 degrees. This set constitutes a complete linear polarization observation.
- Similar to POLAR but selects circular polarimetry observing mode. RUN commands take two observations at plate positions cz and cz+90 where the value of cz is set by the CZERO command (see below)
- Returns to normal (non-polarimetry) observing mode in which RUN commands take a single observation.
- Sets the starting plate position for circular polarimetry observations, in degrees

The polarimetry system can be checked out by allowing light to enter the system from the dome lights or the chimney quartz lamp (open the central dust cover and select Main Focus on the telescope console). On the XMEM display you should see four vertical bands across the chip, corresponding to the E and O images of the two mask slots. There may be slight gaps and overlaps, Type CAL WG to insert a calibration polarizer. Type POLAR to select polarimetry mode and take a DUMMY or GLANCE. On the XMEM you should see the E and O images modulate in an opposite sense as the half-wave plate goes through its four positions. To see this most clearly it is best to take a DARK first and look at the dark subtracted data.

9.4 Linear polarization observations

Linear polarization observations are taken using the half-wave plate module by typing POLAR to select polarimetry mode. Any RUN command will then take a sequence of four observations at the four plate positions which constitute a polarization observation.

The following calibrations should be obtained in addition to the usual dark/bias exposures:

Sky observations should be taken in polarimetry mode since the sky may be polarized (particularly at J if the moon is up). If the object is sufficiently compact it is best to switch it between two positions on the chip and subtract the two. For more extended objects it will be necessary to move off to a sky position. Frequent sky measurements are desirable to average out variations in sky brightness. If the object covers the whole of the chip, accurate sky subtraction can be difficult and can limit the accuracy of the polarimetry.

The window IRISPOL covers an area of 50 by 28 pixels across both E and O images near the top left corner of the chip. It is suitable for observations of standards. A slightly larger window IRISPOL36, which is 64 by 64 pixels, provides a little more room for use at the larger pixel scale of f/36.

9.5 Circular polarization observations

This mode has not been tested at present (this is probably no longer true - CGT 4/96!). It is used in conjunction with the quarter-wave plate module. To use this mode it is necessary to insert the circular calibrator (CAL IRCIRC) and then take a series of trial exposures to locate two plate positions 90 degrees apart which give the greatest modulation. Use the CZERO command to load the lower of these positions into the software. The CIRC command will then select a mode in which RUN commands will take observations at these two positions.

There are no useful standard stars for circular polarization. The only objects with high levels of circular polarization are the AM Her binaries and these are highly variable. However, an observation of one of these systems may be useful to check the sign of the circular polarization which is the only real uncertainty.


Here are listed a few useful snippets that don't conveniently fit elsewhere.

10.1 Focusing.

Although every attempt was made to keep everything parfocal, this was not possible. In particular, the different filters in a steeply converging beam cause focus differences. You and the night assistant must keep track of focus as you change filters.

Focusing is best done on a star that saturates slightly (you can adjust the exposure time to achieve this). Use method 1. Take a sky frame of the same exposure time off the star and use this as the dark (section 4). Centre the star as best you can in a single pixel on the dark-subtracted idle image. Then set DISPLAY SCALE,D_IDLE_ER,0,20000. Focus by minimising the brightness of the four pixels immediately adjacent to the saturated pixel. A numerical way is to type different offsets into FOCOFF.

When finished, remember to DISPLAY AUTO,D_IDLE_ER.

Focus offsets at f/15 wide field are as follows. All are offsets from the K filter, which focuses near 68:

1.08 microns +1.8 1.65 microns +2.8 J +1.6 2.12 microns +1.5 H +0.1 2.16 microns +1.5 K' +0.3 2.21 microns +1.5 Kn +1.5 2.34 microns +1.5
1.64 microns +1.5

These different focal positions introduce slight changes in image scale. For example, the scale is larger (stars further apart) at J than at K by about 0.8%, or 1 pixel across the array. If you want to combine data taken with different filters (to blink or make colour pictures) you will need to rescale using ISHIFT.

At f/36 wide the focus changes are small. Divide these offsets by 9. Unless the seeing is superb you can use a single average setting. The typical focus is about 51.5. At f/15 intermediate divide by about 12. At f/36 intermediate you will not see any focus change.

When using a spectroscopic slit, focus on the TV that views the reflection off the slit jaws. The focus is fairly close, but not identical, to the broad-band filters.

When using different filters, make sure you notify your night assistant when you change filters so that he will alter the telescope focus to suit.

10.2 Light loss between pixels.

Those seeking accurate photometry should be aware that light is lost between pixels. This is because some electrons and holes recombine on the longer route to the multiplexer from the corners than from the centre. Stars therefore appear fainter if their centroids lie near the corner of a pixel. The amount by which they are dimmed depends on the ratio of seeing to image scale. In good seeing, which gives the worst case, a rule of thumb is that the derived magnitude is lowered by 0.1*p/pixel, where p is the pixel size in arcsec. Thus in f/15 wide mode, where pixels are 0.19 arcsec across, a star landing exactly on the intersection of four corners (i.e. 0.71 pixels from the centre) will be dimmed by 0.71*0.19 = 0.14 mag. In practice stars very rarely land exactly at corners, so the worst case is usually 0.1 mag dimming, with an rms of about 0.03 mag from this cause. The situation is much less critical when smaller pixels are used.

If this causes you concern, then a test run is recommended. In DISK$USER:[DAA.IRIS] you will find files TEST06.DAT, TEST08.DAT and TEST19.DAT, for use with 0.6, 0.8 and 1.9 arcsec pixels respectively. These will take 16 paired standard observations (plus sky at start and end) with slightly different centres. You can then analyse the observations to determine the exact centroid and systemic magnitude for all 32 observations, and plot magnitude against distance of centroid from chip centre. This will give a measure of the magnitude of the effect, and hence a correction to apply to your data.

The TESTxx.DAT files assume you are using the IRIS31 window on a normal IRIS standard star that has first been centred in the window. Use method 1. The entire sequence takes under 10 minutes.

10.3 Saturation effects.

A very bright star that saturates the chip does not bleed its excess charge along rows or columns as in a CCD; the excess electrons recombine harmlessly. In method 4, however, saturated pixels lead to a faint brightening along the rows. You'll see this from hot pixels on a dark frame as well as from bright stars. frames. To reduce the glare around a bright star you might find the occulting bar useful. This is selected by SLIT OCCULT. As you will see on the xmem display, there are several square masks of different sizes behind which you can tuck a star. The central masks reflect the star back to the TV so that the night assistant can do his best to keep the star in place.

10.4 Remnance.

The remnance is acceptably good. Highly oversaturated images fade from significance in a minute or two. There is a long tail to the remnance, however, so that long dark exposures taken even half an hour later can show a weak afterimage.

10.5 Stray reflections.

Despite anti-reflection coating of the optics there are some odd reflections within the IRIS dewar. At present we can do no more than warn users to beware of unexpected effects. What look like reflections -- circular patterns that arise when differencing frames for sky subtraction -- are actually the out-of-focus images of specks on the filters. They are seen only if the filters move in their wheels, which is possible if the wheels themselves have been driven.

10.6 Dust on the Fabry lens.

A weakness of the design of IRIS is that the Fabry lens is not a long way out of focus, especially at f/36. Dirt and dust on the lens absorb at short wavelengths and radiate at long, and may additionally come and go. A flow of air across the lens usually keeps it clean, but if you find something new that you are suspicious of, draw it to someone's attention. It is possible (but not encouraged) to descend the chimney and clean the lens.

The characteristics of Fabry lens dirt are: at short wavelengths - absorption against flat field and sky; in K window - absorption against flat field lamp, invisible against dome thermal radiation, bright against sky.


Check that you have dark frames corresponding to all exposure times, methods and windows you used throughout the night, as well as a method 1 `bias' frame -- a dark of 1.5 sec exposure and at least 100 cycles. Also any flats that you want to use.

When you have all the calibrations you need, it is tidy to set FLATT BLANK and then EXIT. You must wait a minute or so for the account to perform its ablutions. Then log out.

There is no need to go near the instrument. It needs no attention.

Your data will have accumulated on DISK$INST in the CCD_1 (or CCD_2) account, subdirectory [.yymmdd] coded by the year, month and day. The data will generally be secure there until the end of your run, and will in any case be backed up to the archive. However, it's a good idea to back up each night's data as you get them.


Only approximate numbers are worth quoting since most depend on the seeing, telescope temperature, airglow intensity and readout scheme. But these should give a guide to the sort of sensitivity to expect.

Since many exposures are made up of numerous short ones, with telescope movements between, overheads can be costly. Allow 75% duty cycle for most observations. You will do better on long spectroscopic observations of faint objects, however, often exceeding 90% duty cycles.

As a working guide, at f/36 (wide field) or f/15 (intermediate field) if a star occupies 9 pixels you should get a 5-sigma detection in one minute of actual data taking (with no extra time for skies) as listed below. The table also shows the ADU/s expected from a zero magnitude star, and typical sky brightness figures.  The imaging sensitivity figures assume use of f/15 intermediate or f/36 wide modes.  f/36 intermediate may be more sensitive in good seeing for a point source (ie smaller measurement aperture).  f/15 wide is approximately 0.7 mags worse than these values.
Mode J H K K' Sensitivity
Imaging 18.5 18.3 16.8 17.4 5 sigma 60 sec, in a 2" x 2" aperture
Echelle spectroscopy 17 16 15 3 sigma 1000 sec, per pixel
Grism spectroscopy 16.5 15.5 3 sigma 1000 sec, per pixel

Band Zero magnitude star
Summer sky brightness
J 3.5e8 15.0
H 5.6e8 13.7
K 5.1e8 12.5
K' 4.7e8 13.7
Kn 5.2e8 13.5

There are 9.5 electrons per ADU.  Photon shot noise appears to integrate down for as long as 30 minutes total time, but beyond that systematic effects may take over.

The figures for spectroscopy are even more approximate since they are additionally highly wavelength dependent due to the structured and time-varying airglow spectrum and the steep thermal component in the K band. With the echelles, a working figure is 3 sigma per pixel after 1000 seconds at J=17, H=16, K=15. For the grisms the figures are about 0.5 magnitude fainter.  Relatively long exposure times are required at J with the IJ echelle to become background limited (it is not possible to be background limited below 1.2um because of the rapidly decreasing QE of the array).   It is not safe therefore to extrapolate the quoted J sensitivity to shorter wavelengths.  The current performance of the array in general shortwards of 1.5um is deteriorating because of an effective increase in the read noise.  We do not recommend IJ spectroscopy of faint objects.

For polarimetry, the required exposure times can be calculated assuming the need to achieve a certain signal to noise in each of the Q and U Stokes parameters (linear polarimetry), or in V (circular polarimetry).  Each Stokes parameter is calculated from two separate exposures with the waveplate shifted between them, so the error in Q, U or V is simply the error in a single exposure divided by 1.414.


A1.1 Photometric standards

Because IRIS saturates on bright sources, a new set of fainter photometric standards has been established. Virtually all the work has being undertaken by Brian Carter at SAAO; a few observations have been made on the ANU 2.3 metre and with the IRPS at AAO. The magnitudes are on the Carter system. Those quoted to 2dp are slightly less reliable. The stars can be called by name from the CCS disk.

Magnitudes and spectral type for all these stars can be found in Carter & Meadows, MNRAS, 276, 734.

Star J H K R.A. Dec HD590 9.195 8.969 8.947 00 08 -19 HD1274 8.781 8.412 8.352 00 14 -64 HD7644 9.487 9.239 9.199 01 13 -46 HD8864 8.631 8.510 8.474 01 25 +04 HD15189 8.460 8.159 8.126 02 24 -14 HD15274 8.490 8.276 8.250 02 25 +09 HD15911 9.452 9.462 9.472 02 30 -52 HD17040 9.401 9.141 9.107 02 41 -18 SA 94-251 9.131 8.426 8.305 02 55 +00 SA 94-702 9.246 8.434 8.289 02 55 +01 HD18847 8.992 8.686 8.653 02 59 -20 HD20223 8.824 8.673 8.663 03 12 -43 HD24849 8.802 8.530 8.489 03 54 -35 HD29250 9.465 9.361 9.354 04 33 -30 HD37567 8.997 8.990 8.993 05 37 -13 HD38150 8.210 7.913 7.880 05 39 -67 HD38872 8.136 7.985 7.960 05 44 -53 HD39944 8.465 8.144 8.104 05 52 -26 HD40348 8.926 8.895 8.892 05 56 -02 HD52467 8.637 8.659 8.699 06 58 -42 HD56189 9.141 8.910 8.873 07 13 -38 HD62388 8.727 8.691 8.670 07 41 -12 HD62998 8.593 8.323 8.294 07 44 -16 HD71264 8.612 8.565 8.538 08 24 -05 HD84090 8.568 8.506 8.493 09 39 -42 HD84503 8.849 8.638 8.579 09 43 -26 HD88449 8.522 8.257 8.223 10 09 -15 HD94949 8.267 7.968 7.938 10 55 +09 HD100501 9.305 9.119 9.087 11 31 -29 HD105116 8.153 8.050 8.032 12 03 -46 HD106807 8.707 8.645 8.637 12 14 -49 HD106973 9.148 8.875 8.836 12 16 -01 HD114895 8.252 8.135 8.101 13 11 -35 HD122414 8.536 8.281 8.224 14 00 -49 HD129349 9.077 8.861 8.802 14 40 -38 HD129540 8.466 8.283 8.247 14 41 -03 HD130035 8.648 8.461 8.427 14 44 -44 HD136879 8.659 8.560 8.528 15 22 -66 HD147778 8.453 8.151 8.088 16 22 -17 HD148332 8.480 8.218 8.174 16 25 -01 SA 108-475 16 34 -00 HD154066 7.962 7.845 7.812 17 01 -13 HD159402 8.102 8.113 8.140 17 33 -45 HD166733 8.495 8.190 8.130 18 10 -02 DM-59.7287 8.872 8.594 8.548 18 28 -60 HD177619 8.584 8.312 8.278 19 04 -56 HD193727 8.977 8.826 8.783 20 20 -16 HD194107 8.772 8.395 8.352 20 22 -46 HD202964 8.624 8.286 8.243 21 17 -39 HD207288 8.825 8.369 8.315 21 46 -18 HD 210427 9.245 8.895 8.852 22 08 -27 HD210863 8.590 8.272 8.234 22 11 -08 HD212874 9.077 8.921 8.890 22 25 +04 HD214497 9.237 8.951 8.906 22 36 -15 HD216009 7.988 7.949 7.947 22 47 -45 HD218814 9.334 9.242 9.227 23 09 -57 HD221462 8.630 8.282 8.219 23 29 -25

A1.2 Spectroscopic standards

The old IRPS/FIGS G dwarf spectroscopic standards are really too bright for IRIS, but are included here for completenes. But they are spectrally pretty featureless. Be warned that the calibration of these spectra shortward of about 1.1 microns as used in Figaro is extremely doubtful.

For most purposes it is probably better to use as well one of the photometric standards, and to interpolate across the Brackett absorption lines.

Star          R.A.   Dec      K
BS 88        00 20   -12     4.88
BS 448       01 31   -07     4.28
BS 772       02 35   -35     4.26
BS 996       03 17   +03     3.27
BS 1006      03 17   -63     4.00
BS 1294      04 07   -64     4.84
BS 2007      05 46   -04     4.45
BS 2290      06 19   -49     5.15
BS 2882      07 29   -37     5.18
BS 3421      08 36   -40     5.16
BS 4013      10 11   -33     5.01
BS 4523      11 44   -40     3.32
BS 4903      12 52   -44     4.51
BS 5384      14 21   +01     4.71
BS 5699      15 18   -48     4.10
BS 5868      15 44   +08     3.00
BS 5996      16 04   -14     4.87
BS 6094      16 21   -39     3.93
BS 6441      17 18   -19     5.18
BS 6748      18 03   -36     4.57
BS 7330      19 18   -35     4.97
BS 7644      20 01   -67     4.57
BS 8477      22 12   -42     4.70
BS 8658      22 43   -49     5.11


UNIX Procedures

Procedures to do quick reduction of IRIS data on the Sun/Sparc platforms are being developed and will be included here soon.

VAX Procedures.

Listed here are some DCL command procedures using Figaro which can be edited out and used to process data. In each case you will need to make changes before they can be used. This is because the file names are designed to be entered in shorthand format. A full file name would take the form:


but would be entered into the command procedure simply as 37.

The instructions are in the header of each file.

A2.1 Procedure to derive the systemic magnitude of a star.

This works entirely on raw data, so lacks the frills of linearization and flat-fielding. The full treatment typically changes the relative values by at most 2-3%.

$ ! This routine will return the raw counts in a star. $ ! It requires two observations in the IRIS31 window, sufficiently $ ! separated to allow 9x9 boxes around each position without $ ! overlap. $ ! If working at f/15 wide it is often preferable to take 7x7 boxes; $ ! to do this find the eight statements that end in +4 or -4 and $ ! replace each 4 by 3 $ ! $ ! The procedure should first be edited to replace all instances of $ ! the string `filename' by the appropriate directory and date $ ! e.g. `DISK$INST:[CCD_2.991225]25DEC' $ ! $ ! The input command is $ @STANDARD n1 n2 [time] $ ! where n1 and n2 are the run numbers of the two images $ ! and time is the exposure time of the observation $ ! The output is counts in the two images and their sum. $ ! $ ! If the optional exposure time is given, a systemic magnitude $ ! will be output. $ ! $ x:= 'p1' $ y:= 'p2' $ if x .gt. 9 then goto x10 $ if y .gt. 9 then goto y10 $ isub filename000'x' filename000'y' q $ return: $ ! define/user sys$output nl: $ istat q 1 31 1 31 med $ icsub q stat_median q $ let &xp=stat_xmax $ let &xm=stat_xmin $ let &yp=stat_ymax $ let &ym=stat_ymin $ x1=xp-4 $ x2=xp+4 $ y1=yp-4 $ y2=yp+4 $ ! define/user sys$output nl: $ istat q 'y1' 'y2' 'x1' 'x2' $ let &c1=stat_total $ x1=xm-4 $ x2=xm+4 $ y1=ym-4 $ y2=ym+4 $ ! define/user sys$output nl: $ istat q 'y1' 'y2' 'x1' 'x2' $ let &c2=stat_total $ slet c=c1-c2 $ p=f$locate(".",c1) $ c1=f$extract(0,p,c1) $ p=f$locate(".",c2) $ c2=f$extract(0,p,c2) $ p=f$locate(".",c) $ c=f$extract(0,p,c) $ write sys$output "Positive counts = ''c1' Negative counts = ''c2'" $ write sys$output " Total counts = ''c'" $ if (p3 .eqs. "") then goto quit $ t := 'p3' $ slet m=c/2.0/t $ let q.data_array[1,1] = 'm' $ ! define/user sys$output nl: $ ilog q q $ icmult q -2.5 q $ ! define/user sys$output nl: $ istat q 1 1 1 1 $ let &m=stat_mean $ slet m=m-0.005 $ p=f$locate(".",m)+3 $ m=f$extract(0,p,m) $ write sys$output "Normalised magnitude = ''m'" $ quit: $ purge q.* $ exit $ x10: $ if x .gt. 99 then goto x100 $ if y .gt. 99 then goto y100 $ isub filename00'x' filename00'y' q $ goto return $ x100: $ if x .gt. 999 then goto x1000 $ if y .gt. 999 then goto y1000 $ isub filename0'x' filename0'y' q $ goto return $ x1000: $ isub filename'x' filename'y' q $ goto return $ y10: $ isub filename000'x' filename00'y' q $ goto return $ y100: $ isub filename00'x' filename0'y' q $ goto return $ y1000: $ isub filename0'x' filename'y' q $ goto return

A2.2 Procedure to process a 3x3 mosaic array

including linearisation, flat fielding, and n-sigma-clipped sky subtraction, and to display the results with north at the top. You must have assigned an image display device first. The files are ready for subsequent combination by IRISMOS.

In order to reduce unneccessary scrolling of the screen, most output from Figaro is suppressed with the command $ ! define/user sys$output nl: However, in the event of an error the routine will hang awaiting input. The most likely cause of this is the failure to provide the sky9.dat file. If it seems to have hung, test it with a few Cntrl-Ts.

$ ! This command procedure reduces a 9-position mosaic of images taken in a $ ! clockwise spiral: centre, E, NE, N, NW, W, SW, S, SE. $ ! $ ! The string `filename' must be replaced by the address and $ ! date of your data, e.g. by `[CCD_2.991225]25dec' $ ! $ ! It is essential to have a file sky9.dat containing the entries $ ! z1 to z9 (one per line) in any order. $ ! $ ! The data are linearised, so a matching dark run must be specified $ ! They are also flat fielded before having the median sky subtracted. $ ! The display uses the range -100 to 1000. The data have quality arrays $ ! appended and are rotated to have north at top, as required by IRISMOS. $ ! $ ! The command line is: $ ! $ @PROCESS9 fieldname startrun darkrun flat $ ! and it assumed that the data occupy nine consecutive runs from the $ ! startrun number. The linearised files can be used subsequently. $ ! Their names have the form r***l, where *** is the run number. They $ ! are flat fielded to produce files r***f. The final frames, after sky $ ! subtraction, are retained with the fieldname (first input parameter) $ ! followed by an underscore and the position in the grid. $ ! $ n:='p2' $ m:='p3' $ if m .gt 999 then goto m1000 $ if m .gt. 99 then goto m100 $ if m .gt. 9 then goto m10 $ copy disk$inst:filename000'm'.sdf dk.sdf $ mreturn: $ q=1 $ write sys$output " Starting to linearise and flat-field files" $ loop: $ if n .gt. 9 then goto n10 $ ! define/user sys$output nl: $ irislin disk$inst:filename000'n' dk sw=swirl5 out=r'n'l $ nreturn: $ ! define/user sys$output nl: $ idiv r'n'l 'p4' r'n'f $ copy r'n'f.sdf z'q'.sdf $ n=n+1 $ q=q+1 $ if q .lt. 10 then goto loop $ write sys$output " Starting to form mean sky" $ ! define/user sys$output nl: $ medsky files=sky9 scale=y add=no nsigma=yes sigma=1.5 max=2 minmax=no out=sky9 $ n:='p2' $ write sys$output " Starting to sky subtract and display" $ ! define/user sys$output nl: $ idiv z1 sky9 q1 $ ! define/user sys$output nl: $ istat q1 1 128 1 128 median $ icmult sky9 stat_median q2 $ isub z1 q2 'p1'_c $ ! define/user sys$output nl: $ irisq in='p1'_c out='p1'_c nozero bad=iris_dir:iris.bad $ rotate 'p1'_c 'p1'_c $ image 'p1'_c xp=4 yp=3 atx=2 aty=2 lo=-100 hi=1000 er opt=0 \ $ ! define/user sys$output nl: $ idiv z2 sky9 q1 $ ! define/user sys$output nl: $ istat q1 1 128 1 128 median $ icmult sky9 stat_median q2 $ isub z2 q2 'p1'_e $ ! define/user sys$output nl: $ irisq in='p1'_e out='p1'_e nozero bad=iris_dir:iris.bad $ rotate 'p1'_e 'p1'_e $ image 'p1'_e atx=1 aty=2 lo=-100 hi=1000 \ $ ! define/user sys$output nl: $ idiv z3 sky9 q1 $ ! define/user sys$output nl: $ istat q1 1 128 1 128 median $ icmult sky9 stat_median q2 $ isub z3 q2 'p1'_ne $ ! define/user sys$output nl: $ irisq in='p1'_ne out='p1'_ne nozero bad=iris_dir:iris.bad $ rotate 'p1'_ne 'p1'_ne $ image 'p1'_ne atx=1 aty=3 lo=-100 hi=1000 \ $ ! define/user sys$output nl: $ idiv z4 sky9 q1 $ ! define/user sys$output nl: $ istat q1 1 128 1 128 median $ icmult sky9 stat_median q2 $ isub z4 q2 'p1'_n $ ! define/user sys$output nl: $ irisq in='p1'_n out='p1'_n nozero bad=iris_dir:iris.bad $ rotate 'p1'_n 'p1'_n $ image 'p1'_n atx=2 aty=3 lo=-100 hi=1000 \ $ ! define/user sys$output nl: $ idiv z5 sky9 q1 $ ! define/user sys$output nl: $ istat q1 1 128 1 128 median $ icmult sky9 stat_median q2 $ isub z5 q2 'p1'_nw $ ! define/user sys$output nl: $ irisq in='p1'_nw out='p1'_nw nozero bad=iris_dir:iris.bad $ rotate 'p1'_nw 'p1'_nw $ image 'p1'_nw atx=3 aty=3 lo=-100 hi=1000 \ $ ! define/user sys$output nl: $ idiv z6 sky9 q1 $ ! define/user sys$output nl: $ istat q1 1 128 1 128 median $ icmult sky9 stat_median q2 $ isub z6 q2 'p1'_w $ ! define/user sys$output nl: $ irisq in='p1'_w out='p1'_w nozero bad=iris_dir:iris.bad $ rotate 'p1'_w 'p1'_w $ image 'p1'_w atx=3 aty=2 lo=-100 hi=1000 \ $ ! define/user sys$output nl: $ idiv z7 sky9 q1 $ ! define/user sys$output nl: $ istat q1 1 128 1 128 median $ icmult sky9 stat_median q2 $ isub z7 q2 'p1'_sw $ ! define/user sys$output nl: $ irisq in='p1'_sw out='p1'_sw nozero bad=iris_dir:iris.bad $ rotate 'p1'_sw 'p1'_sw $ image 'p1'_sw atx=3 aty=1 lo=-100 hi=1000 \ $ ! define/user sys$output nl: $ idiv z8 sky9 q1 $ ! define/user sys$output nl: $ istat q1 1 128 1 128 median $ icmult sky9 stat_median q2 $ isub z8 q2 'p1'_s $ ! define/user sys$output nl: $ irisq in='p1'_s out='p1'_s nozero bad=iris_dir:iris.bad $ rotate 'p1'_s 'p1'_s $ image 'p1'_s atx=2 aty=1 lo=-100 hi=1000 \ $ ! define/user sys$output nl: $ idiv z9 sky9 q1 $ ! define/user sys$output nl: $ istat q1 1 128 1 128 median $ icmult sky9 stat_median q2 $ isub z9 q2 'p1'_se $ ! define/user sys$output nl: $ irisq in='p1'_se out='p1'_se nozero bad=iris_dir:iris.bad $ rotate 'p1'_se 'p1'_se $ image 'p1'_se atx=1 aty=1 lo=-100 hi=1000 \ $ write sys$output " Tidying up" $ purge dk.* $ purge q*.* $ purge z*.* $ purge sky*.* $ purge 'p1'_*.sdf $ exit $ n10: $ if n .gt. 99 then goto n100 $ ! define/user sys$output nl: $ irislin disk$inst:filename00'n' dk sw=swirl5 out=r'n'l $ goto nreturn $ n100: $ if n .gt. 999 then goto n1000 $ ! define/user sys$output nl: $ irislin disk$inst:filename0'n' dk sw=swirl5 out=r'n'l $ goto nreturn $ n1000: $ ! define/user sys$output nl: $ irislin disk$inst:filename'n' dk sw=swirl5 out=r'n'l $ goto nreturn $ m10: $ copy disk$inst:filename00'm'.sdf dk.sdf $ goto mreturn $ m100: $ copy disk$inst:filename0'm'.sdf dk.sdf $ goto mreturn $ m1000: $ copy disk$inst:filename'm'.sdf dk.sdf $ goto mreturn

APPENDIX 3 - Filter Profiles

Broad Band Filters

The figure below shows the J (blue), H(cyan), K(green),  K'(red) and Kn(yellow) filters.  A Postscript version of this figure can be found here.

Narrow Band Filters in H Window

The figure below shows the FeII 1.64µm(red) and FeII redshifted 1.65µm (green) filters.  A Postscript version of this figure can be found here.

Narrow Band Filters in K Window

The figure below shows the H2 1-0 S(1) 2.12µm (red), Br gamma 2.16µm(green) , 2.21µm continuum (blue), H2 2-1 S(1) 2.25µm (yellow) and continuum 2.34µm (cyan) filters.  A Postscript version of this figure can be found here.

Figaro and Fits format copies of the profiles are available.
IRIS Filter Profiles On-line
Filter Wave
Bandpass FITS Figaro


























FeII (z=0)


1%, 1.64µm



FeII at f/15 wide


1%, 1.65µm



H2 1-0 S(1)


1%, 2.12µm



Br gamma


1%, 2.16µm



H2 2-1 S(1)


1%, 2.25µm





4%, 2.21µm









IRIS is a complex instrument and it is easy to put the wrong element into the wrong beam. The instrument configurations outlined here are the most commonly used ones. Consult your support astronomer if you want to do something more unusual.

Imaging Configurations
f/15 Imaging
1.94" pixels SLIT OPEN
UFIL ????
LFIL ????
Filters Available
Filter IRIS
Bandpass Wheel
J J Standard LFIL
H H Standard LFIL
K K Standard LFIL
K' KP 1.9-2.3µm LFIL
Kn KN 2.0-2.3µm LFIL
FeII (z=0) 1.64 1%, 1.64µm UFIL
FeII at f/15 wide
1.65 1%, 1.65µm UFIL
H2 1-0 S(1) 2.12 1%, 2.12µm LFIL
Br gamma 2.16 1%, 2.16µm LFIL
H2 2-1 S(1) 2.25 1%, 2.25µm UFIL
Continuum 2.21 4%, 2.21µm LFIL
CO/Continuum 2.34 4%,2.34µm LFIL
NB: always check the actual filter name and
location at the telescope. Filters may be
moved between the two wheels.

NB: To leave a filter blank use the name

f/15 Imaging
0.61" pixels SLIT SQUARE
UFIL ????
LFIL ????
f/36 Imaging
0.79" pixels SLIT OPEN
UFIL ????
LFIL ????
f/36 Imaging
0.27" pixels SLIT SQUARE
UFIL ????
LFIL ????

Spectroscopic Configurations
f/15 Grisms (INTERM)
R ~ 300
H or K windows
0.61" pixels
75" slit
SLIT GRISMS, WIDE or NARROW.(GRISMS is the default, 2-pix slit)
f/36 Grisms (INTERM)
R ~ 300
H or K windows
0.27" pixels
34" slit
SLIT GRISMS, WIDE or NARROW. (GRISMS is the default, 2-pix slit)
f/36 Grisms
R ~ 100
1µm centered on
H or K windows
0.79" pixels
22" slit
SLIT LONG, MILES or WIDE .(LONG is the default, 2-pix slit)
f/36 Echelles (WIDE)
R ~ 400
IJ or HK passbands
0.79" pixels
13" slit

Spectroscopic Wavelength Ranges (µm)
IJ Echelle 0.86-0.953 Order 10 (Wavelengths below 0.86µm) filtered out.
0.905-1.049 Order 9
1.008-1.168 Order 8
1.138-1.318 Order 7
1.310-1.515 Order 6
HK Echelle 1.439-1.704 Order 8
1.622-1.913 Order 7
1.855-2.184 Order 6
2.170-2.537 Order 5
H Grism (INTERM) 1.4-1.8
K Grism (INTERM) 1.95-2.4
H Grism(f/36 WIDE) 1.2-2.1
H Grism(f/36 WIDE) 1.8-2.5
Spectroscopic Slit Sizes
Pixels on Chip Arcseconds on Sky
Slit INTERM WIDE f/36 f/15
NARROW 5.8 x 49 1.9 x 16 1.4" x 13" 3.5" x 30"
LONG 6.2 x 85 2.0 x 28 1.5" x 22" 3.8" x 52"
WIDE 22.2 x 48 7.3 x 16 5.3" x 13" 13.5" x 29"
GRISMS 1.8 x 128 DONT USE!! 0.43" x 40" 2.1" x 75"*
MILES 3.5 x 128 1.2 x 128 0.84" x 100" 2.1" x 75"*
* - Focus poor at ends. Only 60" useful.

Return to top of this page. |  Return to Documentation Home Page |  Return to AAO Home Page

This page last updated : 23 April 1996 by Chris Tinney