HI Prep Notes

Shutterless Correction

Since the camera does not have a shutter images are smeared across the camera during the process of clearing and reading out of the CCD. This operation is constrained by three numbers, the time the image is static on the CCD, the time to clear a row and the time to readout a row. The measured pixel value obs[i,j]=sum(actual[i,0:j-1]*lineclear) + actual[i,j]*exptime + sum(actual[i:j:n]*linereadout). Over the entire column this can be expressed as a matrix where all the cells on the main diagonal are the exposure time, all those above that are the readout time and all those above are the cleartime. Taking the inverse of the matrix and matrix multiplying by the observation removes the desmear and also coverts from DN to DN/s. Since the exposure time varies from row to row by roughly 5 seconds [due to the difference in the clear/readout times] it does not seem appropriate to covert this number into some equivalent DN.

Cosmic Rays

The full resolution images [_nxxxx.fts typically 9Mb in size] are not cleaned of cosmic ray effects. Cosmic rays will then appear as bright objects but do not smear, applying simple desmear to these will result in negative trails of the cosmic ray in the output data; more sophisticated image processing would be necessary to remove these effects. The summed [_s4xxx.fts typically 4Mb ] use a cosmic ray scrubbing process and should not be affected in this way. Usually the last pixels in the image contain a counter for each image in the sum of the number of pixels scrubbed, before the image can be corrected these a replaced by duplicating pixels from the row below, this will occasionally cause de-smearing artefacts in the same way as for cosmic rays.

Bleeding

When a pixel overflows the charge is constrained by the design of the CCD to bleed vertically up and down the columns. If this occurs to a small level contained within the CCD then the effect on shutterless correction is small, although the corrected image will still show the effects of this bleeding. Detecting saturated pixels is best done in the pre-processed data since the saturation level will be convolved with the desmear matrix in the prepped result. On rare occasions where there is strong bleeding into the readout register charge can leak sideways in the readout register and back up adjacent columns, see 20070114_000100_s4h1A where this happens with comet McNaught. The desmear process produces a false reduction in the output single in the few columns so affected. Even less often significant bleeding can occur as the image is being readout, in this instance you can see excess counts in the first row of an image, see e.g. 20070220_010000_n4h1B.

Alternative to desmear

Using desmear_off on hi_prep will compensate for the non uniform exposure across the ccd without doing the desmear, the results should be of order comparable to the desmear option but without some of the issues noted above. It is also interesting to look at the difference between the desmeared and the data returned with this option.