Date: 23 Sep 1997 23:35:52 -0400 From: David Huenemoerder To: melvis@head-cfa.harvard.edu, adam@head-cfa.harvard.edu, davis, dsd, wise, houck, joan@head-cfa.harvard.edu, jhk Subject: exposure map meeting notes cc: fap@sxg.physics.orst.edu, nadams@head-cfa.harvard.edu Below are my notes from the exposure/background map meeting. Most is from memory - so if you have different memories or wrote some additional items down, please feel free to amend or augment. I've sent this to the attendees, plus Nancy Adams who is working on the aspect histogram. Please forward to others who may be interested. -- Dave David Huenemoerder (617-253-4283; fax: -0861) Center for Space Research/AXAF Science Center MIT NE80-6023, Cambridge, MA 02139 http://space.mit.edu/~dph %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% Exposure Map Review/Planning Meeting Notes Friday, 19 September 1997 Attendees (Roles): M. Elvis (SDS Group Leader) F. Primini (ROSAT/Einstein veteran) D. Huenemoerder (exposure map scientist) J. Davis (exposure map programmer, science analyst) A. Dobrzycki (background map scientist, HRC science input) J. Flanagan (background map programmer) M. Wise (response/simulator scientist) J. Houck (exposure map science input) D. Davis (exposure map PSF science input) J. Kastner (ACIS science input) J.Davis presented overview of the mathematical formalism of the AXAF response, and definition of the exposure map. The formalism includes the spatial, spectral, and temporal response as well as the spatial (mirror PSF or grating diffraction) and spectral redistributions. We did not examine the diffraction, but concentrated on the spatial domain in order to review the basic principles. The resultant map has been defined as a spatially dependent response matrix (including mirror efficiency), determined by averaging the reponse over the time history of the observation (and assuming a point source and delta-function PSF). Two approaches for computing the map were discussed: the convolution of responses by the aspect histogram, or the integration over the time history. Each has advantages and disadvantages: Aspect histogram: advantages: does time integral once disadvantages: need to be careful with image-processing affects, such as resampling, aliasing. Harder to handle time-dependent affects --- require multiple aspect histograms for light-curve analysis, or due to transient bad pixels (as for ACIS bias corruption). Time integration: advantages: fewer image resampling issues light-curves (or time filters) are straightforward. disadvantages: integrate over time multiple times (for every sky bin) requires formal ability to invert a rotation matrix - realistic aspect solution is not yet available in order to assess this. [- D.Huenemoerder -] to talk to T.Aldcroft. Relative efficiency of the two methods requires more study, given realistic aspect history and calibration data. Trades are in sparseness of aspect histogram and number of operations in 2D spatial integration (convolution), vs. number of operations in temporal integral. [- D.Huenemoerder -] to define and prototype input products. [- D.Huenemoerder -] to specify "compression" of aspect history into stable aspect intervals. Both methods will need to handle multiple spatial resolutions. The detector QE uniformities are known on coarse grids (about 30x30 bins per detector element). Bad pixels and detector boundaries will have to be calculated at high resolution. This is of major concern as a user: whether to have a single map appropriate to any location in the field, or whether to only compute a high-resolution map for a specified region. (The low resolution map will ALWAYS be computed, and the high resolution pieces will ALWAYS be computed, but when and how to merge is TBD.) Some concerns which were brought up: - Need to handle secular drift, if any. Drift could make the aspect histogram too big, and for the convolution approach, it would have to be broken into segments (almost a slew). [- D.Huenemoerder -] to talk to T.Aldcroft. [- N.Adams -] to pre-compute region from aspect history. - Need to incorporate "fuzziness" of aspect solution. The solution specifies variances on each parameter (right ascension, declination, roll). These should be included in the aspect histogram (rotation matrix), or handled in the inverse rotation matrix. [- J.Davis -] to specify. - M.Wise is concerned with uncertainties in calibration products, wants an "exposure map variance map". This needs further discussion. [- D.Huenemoerder - M.Wise - J.Davis -] to discuss. We then had discussions about ROSAT and Einstein approaches and differences and similarities to AXAF. Differences of AXAF from previous missions: - Small PSF, relative to aspect motion. - Programmed dither pattern of aspect motion. - Hybrid detector arrays, present very different responses. - Spectroscopic (grating) modes. Similarities of AXAF: - Exposure map had similar definition, though usually integrated over spectrum. Differences of previous missions: - map dimensions not well determined a priori (hence oct-tree compression method in aspect histogram) - did not need multi-resolution maps. This left a little time to discuss BACKGROUND MAPS. Background maps are very similar to exposure maps, since we want to describe the response to a flat source on the sky, along with any "flat" contribution from detector-local signal. The essential difference between the two is that the sky source is vignetted by the mirror. One advantage of background maps is that they are only low spatial resolution. Otherwise, they can use the same aspect histogram convolution or time-history itegration as the exposure map. F.Primini pointed us to the FITS RDF format for existing background map file format and tools (by Cochran/GSFC). Implementation issues: Need a design and schedule for exposure maps. Develop as interface library, for use in other areas, like source detection, models, ASCfit. [- D.Huenemoerder - J.Davis -] to work on a plan. Other: J.Davis has a draft document on the response formalism. [- J.Davis -] make document available, update as required.