IRAC ISDS Usage

IRAC Science Data Simulator (ISDS)

Overview

Installation

Use

First, you're going to need a "truth image". This truth image is a representation of the real flux on the sky, in units of Jy/square arcsecond. How you make this is up to you. It is assumed that most users have at least some idea of what their targets will look like. If you're not completely modeling them from scratch, a good place to start for IRAC is the 2MASS Image Server at IPAC, since it's K-band observations are probably not so different from those expected from IRAC at 3.6 microns. Another possibility is the Digitized Sky Survey available at STScI.
Some basic guidelines when making truth images:
- It is a good idea to make the truth images with a finer pixel scale than IRAC, so as to more accurately simulate the IRAC dither patterns. A suggested value is 0.3", or 1/4 of an IRAC pixel. Note that if you oversample them too much, they will become very large.
- The image should be as close to noiseless as possible. Any noise in the truth image will be treated as "real" by the ISDS, and will contribute to the noise in the final output.
- The images should be so large that no part of any array will ever be "off" the truth image when dithering and mapping (remember that the two IRAC FOVs are separate).
- The images should be convolved with the IRAC PSF. However, this is most appropriate when the original (unconvolved) truth image has an effective beam size that is very significantly smaller than the IRAC beam. This will not be the case when starting with ground-based images (like 2MASS or the DSS) which have beam sizes comparable to that of IRAC. In this latter case, you might want to try convolving with different beams until you get something that resembles the IRAC PSF.
- In addition to scaling your images to get the correct target fluxes, you should be sure to get the background correct as well, since there is a significant infrared background at all wavelengths. As a first approximation, just set the truth image background equal to zero, and allow the ISDS to use it's internal model values. These are not based on your location on the sky (as they will be in real life), but rather are scaled according to the background values given in the Spitzer observer's manual.
- IRAC will detect many background stars and galaxies. For the stars, an approximation that we have been using is to access the 2MASS point source catalog for a given field to get a list of all the 2MASS stars. We then extend the stellar luminosity function down several more magnitudes, since 2MASS isn't very deep, and generate randomly distributed stars. We then insert the stars into our truth images with the artdata package in IRAF. In future releases we may include sample images of starfields, zodiacal structure, and background galaxies which users can tile and add to their truth images. Donations welcome!
- Your truth images will need the standard FITS coordinate system keywords in them, so that the ISDS knows where to point. These are CRVAL1/2, CRPIX1/2, CDELT1/2, and CROTA.
Get a Spot .aor file. Spot is the tool that Spitzer observers use to prepare observations. It is assumed that users are familiar with Spot, since all Spitzer proposals must use Spot for AOR submission. The ISDS will only read the first AOR from the .aor file, so users can either write the AORs out individually to separate .aor files, or can cut the pertinent sections from their .aor file into extra files for the ISDS to read. The former is preferable, because the ISDS relies on the formatting being fairly exact for the .aor file, and users cutting and pasting these files may introduce formatting problems.

We assume some familiarity with the IRAF package, and the use of IRAF commands (if not, see the tutorials provided by NOAO). At this point, log into IRAF:

usagi% cl

And enter the ISDS package. You should see something like:

cl> irac

       +------------------------------------------------------------+
       |               IRAC DATA SIMULATION PACKAGE                 |
       |                   Matthew L. N. Ashby                      |
       |                     2002 December 16                       |
       |                                                            |
       |    ***Development version 2.3, for IRAF V2.11p3***         |
       |                Shared-risk Beta Release                    |
       |    ***This package is being modified and will change***    |
       +------------------------------------------------------------+

                        IRAC Data Simulation Tasks

        AOREXPAND .... expand IRAC AORs into IRAC C&T commands
        IRACSIM ...... create simulated IRAC images, version 2.21Beta

      aorexpand   iracsim
ir>

The first thing you will need to do is run AOREXPAND. This will read your Spot .aor file and expand it into an IRAC command sequence. Your Spot file should have looked something like this. It contained a variety of switches controlling the AOR. After expansion you will get a .isds file like this (from an earlier version of AOREXPAND), which contains the individual commands and their arguments needed to carry out the requested sequence. AOREXPAND should work with output from Version 6.2 (and later) of Spot.
AOREXPAND takes very few parameters. You give it the name of the input .aor file, and the root name for the output .isds file. You also supply the position angle of the AOR (in degrees E of N). You can determine this in Spot by clicking on the "visibility window" button. Note that the current version of the ISDS has not had this feature fully debugged. In particular, the ISDS definition of PA differs from that of Spot by 90 degrees. Here you can input the actual date that your AOR will occur, and it will return the PA for that target on that date. This is an important point. The roll angle of the Spitzer spacecraft is dictated by many pointing constraints, most of which are sun-related. This means that for any given target the actual position angle of the detectors will be dictated by the date on which the target is observed, and this is determined by the schedulers at the Spitzer Science Center. Since in general users will not know a priori the date of observations, they are advised to either simulate a range of roll angles by performing multiple simulations, or to choose a "typical" roll angle.
```
PACKAGE = irac
   TASK = aorexpand

input   =      swire_small.aor  Input AOR to expand
output  =          swire_small  Output file name
rollangl=                 337.  Roll angle (deg E of N)
(mode   =                   ql)
```

Once you have created an expansion file, you perform the actual simulation by running IRACSIM. IRACSIM has many options, and users are advised to read Matt's document which has a detailed description of all the options.


Raw Ch. 1 image of NGC 2623 (left), and a BCD image (right). The raw image still has all the detector artifacts such as the flat-field and dark current signatures in it (it is also inverted due to the way raw data is packaged). The BCD image has had all of these effects turned off, and reflects only the expected noise and pointing properties of IRAC.

In general, you will probably be most interested in either generating raw data, which is what comes out of the camera, or BCD data, which is what comes out of the basic SSC pipeline. The former is done by turning on all the instrumental effects (here is an example parameter set). For the BCD data, one turns off most of the instrumental effects, but leaves on the poisson noise, shot noise, background, etc (another example parameter set). The BCD data naturally is idealized, since in reality the SSC pipeline will not perfectly remove all instrumental signatures.

Because the ISDS fully simulates the entire data-taking process, including the individual fowler reads, it is relatively slow. On a 400MHz Ultra 60 it takes about 1 minute to simulate a single IRAC frame. The usual tips apply. Run the ISDS on the fastest available computer. Set the "input" scratch area to be a local disk, which will speed up file access.