(Up to Development Notes

2007.01.21 EAM : notes

I've made lots of progress on psastro in both chip and mosaic mode. It basically works now; at this point, the remaining effort involves ensuring all I/O options are working and that we have enough information from the process written as either header data or graphics to understand the results. Some outstanding issues below:

• the input file format is not determined from the file; the user must set it in the camera config.

2006.10.19 EAM : notes

I've run a series of images from ngc7000 and triffid nebula:

• in the triffid nebula, the residuals are sometimes rather large. I suspect this is from the density of sources, esp in the nebula itself. too many mis-matches prevents the solution from converging very well. One possibility might be to use only the nearest sources to the reference source for the fit, or at least for the residuals. In these crowded regions, a single source will often match to many more than a single source.

• there were two segfaults which I have not investigated yet.

I have fixed up some of the issues below:

• getstar now takes a magnitude limit
• to enable this, relphot was updated to calculate the average magnitudes based on REF photcodes
• I added a significance test to the grid match to decide if the resulting match is probably a good match or not.
• log and error messages are better, but could still be improved.

2006.10.17 EAM : some notes on the current state of psastro

• runs successfully on images from ISP:
  • sep21_ngc7000_001_01.fits
  • sep21_trif_001_02.fits

• successfully uses stars from the DVO synthetic grizy bright star database
  • is somewhat slow to load the stars
  • getstar takes about 2 seconds for the field of ngc7000 and 5 seconds for triffid
  • the psastro read of the FITS table is very slow: 8 seconds for ngc7000 and > 100 seconds for triffid
  • we can probably put in magnitude limits to speed things up for now. the FITS->Metadata conversion will have to be better optimized.

• converges on a solution well from the initial ISP guess.
  • the fitted functional form is limited to linear at the moment. For the ISP, this leaves residuals of out 0.5 pixels. this can be improved with higher-order fitting (but I need to define the orders better, probably as I do for the pmPSF in psphot).

• a variety of other cleanups are needed
  • better log messages along the way
  • calculate and report residuals in arcseconds (currently in focal-plane units, probably pixels).

• the 'mosastro' version has not been tested.
• there were recent fixes to the polynomial / WCS conversions that i discovered for gastro2. I need to double-check that these are implemented for the psastro routines correctly as well. these were relevant only for very substantial distortion (eg, the ASCA images).