Matthew Ashby maintains a web page that details the version history of the ISDS. This page makes for good reading in understanding some of the issues with the simulation.

1. AOREXPAND does not include support for modes added since the time of launch, specifically stellar mode.

2. AOREXPAND may not work with .spot files from the current generation of SPOT.

3. While the ISDS allows inclusion of rotation via AOREXPAND, it has not yet been confirmed that the resulting fields have the expected orientation. The visualization tool in Spot should always be considered the "final word" in regards to where the arrays will actually be pointed on the sky.

4. Subarray mode is not implemented.

5. Instrument calibration reflects ground and not in-flight performance. This includes:
   • Flat-fields
   • Darks
   • Photometric Calibration (throughtput)
   • Read noise

6. Stray light is not modeled. Stray light effects may be an issue when bright stars glint off structures either in the Spitzer focal plane or near the detectors and their mounts.

7. The muxbleed application may be too idealized. In particular, it is algorithmically added to the data in exactly the same way that the pipeline currently removes it. This reflects the current limit of our understanding of this effect. In reality muxbleed is very complicated and difficult to remove well.

8. Column pull-down is not modeled.

9. Neither banding nor any other effects due to internal scattering in the Si:As arrays are modeled.

10. Muxbleed in the Si:As arrays is not modeled.

11. The pixel-phase effect in the InSb arrays is not modeled.

12. Because the PSF is not modeled and left to the user, the non-isoplanicity of IRAC will probably be ignored by most users. This is probably ok, since the level of non-isoplanicity in IRAC is very small.

13. There is no modeling of the variation in effective filter wavelength as a function of position, or of any other color effects. The input truth images for each filter are treated as monochromatic. This is significant for IRAC at the 10% level.

14. IRAC is very sensitive to external galaxies, and every IRAC observation (particularly in the very low noise short wavelength channels) will have a considerable level of contamination from these objects. The ISDS does not include these, and users must put them directly into their truth images.

15. Similarly, the ISDS contains no models of contaminating moving objects like asteroids.

16. The ISDS infrared background model is extremely simplistic (a constant level added to the entire image, which may be scaled by the user). It was included because it contributes to the noise and to the consumption of detector well-depth (particularly at 8 microns). If users desire something more realistic, they can model it themselves in their truth images and set background = no.

17. Stochastic pointing errors and jitter are simulated, but not pointing drift.

18. Cosmic rays are simulated by adding library frames containing cosmic rays randomly encountered during ground testing. While similar in appearance to what we expect in flight, these do not neccessarily have the same power spectrum as expected in flight. The cosmic ray incidence expected in flight is roughly one proton per second.

19. Fluctuations in the IRAC calibration are generally not modeled, although in practive they in fact do not vary. Stochastic fluctuations in the bias level are included, but do not follow the actual behaviour of the first-frame effect.

20. While the parts of the header which are derived from the instrument telemetry (voltages, times, etc.) are simulated, the header components created by the SSC are only in a rudimentary form. In particular, the pointing and distortion related keywords are not included, mostly because they have not yet been defined.

21. Image latency is modeled, but the model is not very sophisticated. In particular, long-term latents known to exist in channels 1 and 4 are not modeled.

22. Reprojection is not handled rigorously in the simulator. Whenever IRAC (or any imaging array) takes data on the sky, each individual image is a tangent projection of the celestial sphere onto the linear array, with a tangent point fixed in array coordinates (for on-axis optical systems this is near the center of the array). When a large number of these images are mosaiced, they must be reprojected onto a common projection. The truth image the simulator works with is a flat single projection of the curved celestial sphere, and the simulator does not reproject to get the proper tangent projection. This will probably be an issue only for very large survey areas.

23. The ISDS is not blindingly fast, as it is actually quite time consuming to simulate the fowler sampling. Typical speeds on an Ultra 10 are 1 simulated DCE per minute. However, even the largest of AORs can be simulated overnight.