Input file format

[bjricketts]

Currently, the input of the simulator is fits file based and uses the SIXTE conventions. Long-term we'd like to design our own data-cubes which are compatible with SIXTE so we can compare.

We should at some point discuss this further.

[bjricketts]

This is a bit more important now that we're building the Model GUI. Our previous implementation of this was to generate large arrays of an "image" (usually in Rg for the model in question), dependent on energy. This was pretty inefficient as they were very large, and also tended to be pretty inefficient for plotting (i.e. slow).

We can at least make the "images" smaller via binaries (numpy binary files are very efficient compressors for instance) instead of the text files we've been using. Building a similar GUI solution for the instrument simulator might be the best solution for the plotting problem, but I think this is low priority. Ideally, we should be able to use a lot of the same code from the model GUI as the plotting of the instrument simulator result is a lot "simpler" as it's simply an image.

I think the best way to start here is to simply checkout implementations of image FITS files in optical and infrared for instance. One issue we might have is that the expected resolution of the XRI is actually so high that we need something more similar to that of VLBI implementations (which are notoriously computationally hungry). We might just be in uncharted territory to a degree.

[bjricketts]

Relevant links and documentation

CFITSIO user reference guide
FITS defintion
WCSLIB
WCSTools
AST library
SIMPUT

Thoughts

We generally want to try and align with SIMPUT for eventual compatability with SIXTE. We can use the SIXTE FITS file format as a base:

• SRC_ID (required)
• SRC_NAME (optional)
• RA (required)
• DEC (required)
• IMGROTA (optional)
• IMGSCAL (optional)
• E_MIN (required)
• E_MAX (required)
• FLUX (required)
• SPECTRUM (required)
• IMAGE (optional)
• TIMING (optional)

Space co-ordinates

Looking into CFITSIO, there isn't any explicit requirements from NASA nowadays. For co-ordinates, the common practice is to use the World Co-ordinate System (WCS). The exact details now have been left up to external libraries linked above. The WCS is as follows:

For our purposes, this means transitioning from generated image pixels, to physical units, and then "project onto the sky" of the instrument simulator. In discussion, this likely means that we want to aim for intermediate world co-ordinates from the model simulator, which can generally be thought of as individual pixels having a dimensionless offset from some reference point on the sky that can be specified by the instrument simulator. We can then simply record the normalisation of the dimensionless offset in the FITS header. This means that we should be able to render arc-seconds to micro-arc-seconds with no issues other than a single normalising value.

This somewhat affects rendering - it must be in pixel form, but the size of the pixels should somewhat correlate to a reasonable size on the sky. This should probably be a setting the user can specify (very high resolution large FOV files can be saved at the user's own peril).

Time co-ordinates

We are currently considering time co-ordinates to be low priority. Currently, it seems most wise to simply render integrated time bins separately and produce distinct files. We can then perform the basic science showcases in post within the instrument simulator. Another possibility is to change the FLUX field to instead be a lightcurve that modulates the stored spectrum.

Spectra

Following in SIMPUT's footsteps would generally require us to have a single spectra. For extended models such as a rotating accretion disk, this seems somewhat insufficient but first implementation should align with this for now. A possible solution includes having multiple images that are split up by shared spectra. This requires some clever work with the rendering, but doesn't seem impossible. Outside of very high resolution images where every pixel has a different spectrum, this should be sufficient.

[bjricketts]

Another thing to investigate is how FITS files are composed for Integral Field Units (IFU)