Polar CLEAN Strategies
- HSI_POLAR_CLEAN does not know about subcollimators or absolute spatial scales.
It only knows :
- the dirty map
- the Point Spread Function
- the visibility used to produce the dirty map (TBD)
- a function called to determine the stopping criterion (TBD)
- Advantages:
- the details of computing the PSF, which depends on the selection
of subcollimators and harmonics, etc, are computed externally.
- variations in the statistical-stopping criterion are done externally
- The same technique can be applied to MEM or Lucy-Richardson, etc.
with expected comparable speeds (Less than a minute per map).
- The PSF consists of a 3-D array of 2-D HESSI Point-Spread maps, where
- First dimension = solar azimuth
- 2nd dimension = solar radius
- 3rd dimension = the radius of the peak of the PSF
- (No dimension is required for the azimuth of the PSF peak, since
shifts of the 1st dimension suffice.)
- After the map center is selected, the appropriate PSF array is
read from disk (possibly after decompression).
The array, which
is of size 64 MB for maps of size 1024x128, is kept in
memory during a CLEAN process.
Advantages:
- SPEED -- (100 iterations of CLEAN take only 10-20 sec)
- Once the PSF is loaded, speed is independent of the number of
subcollimators used.
- One may use the same PSF in memory for all 4-sec time intervals of a
flare,
- and a wide range of energy bands),
as long as one stays with the same subcollimators, mapping scales, and harmonics.
Disadvantages:
- The restoration of a psf from disk takes on the order of 30 s, slow
compared to the entire cleaning time.
- A fairly large
library
of PSF arrays (~20) will be
required for total solar coverage. (CDROM storage??)
- Examples:
- One may envision a PSF disk library may contain a wide variety of
PSF arrays for all possible
- flare locations (radial distance from sun center),
- subcollimator combinations,
- and the most popular map sizes and resolutions.
- Recomputation for unusual mapping parameters will also be implemented.
- Disadvantages:
- All the dirty laundry is hidden in the conversion from counts to
visibility:
- Spatial variation of the subcollimator amplitudes.
- deadtime,
- transparency as a function of energy
- Handling the statistics is somewhat more complicated, requiring
large, but very sparse weight matrices.
(vis=weight # count)
- It probably can't work when the deadtime is 100% or for far off-axis sources like the Crab.
- PSFs using the two finest subcollimators may require more than 1024x128x128 (64-MB) arrays,
putting them out of reach of small-memory computers.
Although...
- One can gain a factor of 2 using PSF symmetry, and
- Various inline compression schemes may be concocted
Ed Schmahl
Last modified: Thu Jan 13 18:05:39 EST 2000