CALSTIS Version 2.20 was released on 2006-Mar-27 in the new
STSDAS release, Version 3.5, which can be found on the
STSDAS Download Page.
Update: This version of CALSTIS was installed in the OTFR pipeline on
2006-Apr-10.
The following OPRs are implemented in the release of CALSTIS 2.20:
| OPR 55249: | CALSTIS6 correction for CTE loss for G750L and G750M |
| OPR 55248: | CALSTIS interpolation of error estimates |
OPR 55249: CALSTIS6 correction for CTE loss for G750L and G750M
The correction for charge transfer inefficiency (CTI, or CTE loss)
was improved for OPT_ELEM G750L and G750M. With these gratings,
the "red halo" effectively contributes to the background, and the
correction will be more accurate if the background term includes
the halo.
The new formula for the CTI (per pixel) is now:
CTI = 0.0562 * G**-0.820 * (0.218 * (t-51765.) + 1) *
exp (-3.0 * ((B' + 1.3 * halopar) / G)**0.18)
where:
halopar = max (0.0, (frac_halo - 0.060)) * NET
G = Gross counts in single spectrum exposure, as before;
B'= Total background level in spectrum (dark + sky + spurious charge),
as before;
t = observing epoch in Modified Julian Day Number, as before;
frac_halo = fraction of the PSF above the default extraction box
(of 7 pixels);
frac_halo = (h200 - 1) / 2, where h200 is the value from the PCTAB
for extraction height of 200;
NET = net counts in single spectrum
= NET column in _x1d spectra * TEXPTIME / NCOMBINE
OPR 55248: CALSTIS interpolation of error estimates
In calstis7 the interpolation of the error estimates (the ERR extension)
was modified. This applies to both spectral and imaging data.
With the previous algorithm, when
the interpolated value was an average over several pixels, the
result would have a lower error than the input value. The new
algorithm is based on the principle that when a user sums over an
area in an interpolated image and adds the errors in quadrature,
the resulting error estimate for the summed flux should be
essentially the same as the result that would be obtained for
an equivalent area in the original uninterpolated image. A new
parameter, ERR_ALG, was added to the x2d task to give the option
of using the previous algorithm; the default is to use the new
algorithm.
The original interpolation of the error array in calstis7 was:
err = sqrt ( sum {w[i] * (err[i])2})
the current version is:
err = sqrt ( sum {(w[i] * err[i])2} )
where i refers to one of the four pixels involved in the bilinear
interpolation, err[i] is the value in the error array at pixel number i,
w[i] is the weight for that pixel, and the weights have been normalized
so their sum is 1. The comment to PR 38751 mentions this change. The
distinction between these two can be described as weighting the variances
vs. weighting the errors.