ROSETTA ROLIS DATA CALIBRATION DESCRIPTION

This data set contains the ROLIS descent images calibrated to the CODMAC3 
level. ROLIS CODMAC3 level data are in arbitrary units that have  been 
corrected so that values are proportional to radiance. No resampling is 
performed, so that edited data can be reconstructed.
Three steps of processing are used to produce the ROLIS CODMAC3 products:

1) Bias and dark current subtraction
2) readout smearing removal (de-smearing)
3) flatfield division

The ROLIS CCD and related readout electronics are extremely linear
within the whole 14bit dynamic range. For this reason, no linearity
correction is applied.


1)Bias and dark current subtraction
The digital numbers (DN) stored in the image file are a linear function of 
the incoming intensity I(i,j):

DN(i,j)= B(i,j) + exposure_time * (D(i,J) +  K(i,j) * I(i,j))

where B(i,j) is the image bias, D(i,J) is the dark current generation rate
and K(i,j) is the flat field image.
In the case of the ROLIS the descent phase, due to the very low temperature 
and short exposure times, the D(i,J) term is virtually zero. Furthermore, the
bias image is constant for all pixel. Therefore, for the ROLIS descent images,
bias and dark subtraction is achieved by subtracting the constant value 
of 211 DN.


2) Desmearing. 
The ROLIS CCD is used in frame-transfer mode to produce an image. The 
CCD architecture is characterized by an imaging region and by a
storage region. The image is exposed in the imaging region. At the end
of an exposure, the content of the imaging section is swiftly transfered
into the storage section. In the storage section, which is masked by
an opaque layer, the content of the image is slowly read out, without 
being affected by the scene illumination. This scheme implements 
the functionality of an electronic shutter, without the need of any
mechanical parts. However, during the fast shift phase, light continues
to be integrated by the detector, producing an effect called readout
smearing. As the smearing effect is proportional to the ratio of
image transfer time to exposure time, the effect is stronger for
shorter exposure times. 
On most frame-transfer CCDs, smearing occurs twice for an exposure: 
during the initial image clearing and during the image-to-storage 
region transfer at the end of an exposure. On the ROLIS CCD,
however, only the latter needs to be corrected, as the image clearing
is implemented by using the clearing capability of the anti-blooming gate, 
instead of the frame-tranfer feature. As a consequence, in the ROLIS images
smearing appears as a ramp of increasing brightness, increasing from line 0 
(where smearing is zero) to line 1023 (where smearing is maximum).
Smearing is corrected for ROLIS images with the following algorithm:

1) compute smearing factor as 
smear_fac=shift_time/(n_rows_total * exposure_time)
where shift_time=0.0032s and n_rows_total=1024.
2) compute and accumulate the amount of smearing for each line as
smear[j] = smear_fac * clean_image[i,j]
3) subtract smear for line j from all subsequent image lines
4) iterate for all lines.


3) Flatfield division

Flatfielding is used to correct for pixel-to-pixel inhomogeneities
in pixel sensitivity, fall-off of optics transmission at the edge
of the field, and particle contamination on the lens.
Flatfielding for the ROLIS descent images is achieved by dividing
the bias-subtracted, desmeared images by the calibration flatfield
ROLIS_IFL_FLAT_CLEAR.FITS and multiplying by the normalization factor 
c=11112.3. The calculations are performed with double float precision and 
the results are rounded to short int for storage in the data file.