search

mdschneider / JIF

Joint Image Framework (JIF) for modeling astronomical images of stars and galaxies as seen by multiple telescopes.
Other
6 stars 1 forks source link

Add filter throughput curves to hdf5 file #14

Closed wadawson closed 9 years ago

wadawson commented 9 years ago

Currently there are attributes that specify filter properties (e.g. central, FWHM; see c86ece1) but we should really just have another subgroup that specifics filter throughput curves and just have an observation specific attribute referencing a specific filter name.

jmeyers314 commented 9 years ago

Hey @wadawson, is the plan to put the throughput data directly into each image hdf5 file? Or to just reference a filename that holds the throughput?

Either way, when reading filter data, I'd suggest placing it into some kind of prebuilt-predebugged (or at least mostly debugged) container, like a galsim.Bandpass for instance, or possibly something from the LSST DMStack.

mdschneider commented 9 years ago

Good suggestion @jmeyers314. I see the available options for a galsim.bandpass instance are:

The input parameter, throughput, may be one of several possible forms:
1. a regular python function (or an object that acts like a function)
2. a galsim.LookupTable
3. a file from which a LookupTable can be read in
4. a string which can be evaluated into a function of `wave`
   via `eval('lambda wave : '+throughput)`
   e.g. throughput = '0.8 + 0.2 * (wave-800)'

So, perhaps we can store the lookup table information in the HDF5 file and then use this to instantiate a galsim.bandpass instance on load into Roaster (or similar classes)?

jmeyers314 commented 9 years ago

Yep, if we go with galsim.Bandpass, which I think makes sense since GalSim is already a dependency, the relevant code would look something like:

waves, throughputs = function_that_returns_waves_and_throughputs()
bandpass = galsim.Bandpass(galsim.LookupTable(waves, throughputs, interpolant='linear'))

Only tricky part is to get the units of waves right, which are nm by default, but can be interpreted as Angstroms if you set the wave_type='ang' keyword.

wadawson commented 9 years ago

I think galsim.Bandpass is great for what we are doing with JIF, thanks for the pointer. The two-column look-up table is our best option and closest to what is currently provided by most facilities and the format that is likely to be provided in the future. It is also trivial to store in the HDF5 file. I imagine doing so under a higher level group (e.g. ground/filters/i).

Once we get to the point of needing this it will be trivial to implement.