Ensemble analysis: "new variables"

[Andrew Bennett <andrew.bennett@ns.sympatico.ca>]

Yes - I am still grinding away at Data Set 20.

The ensemble photometry has now improved to the
point where the noise floor for I magnitudes is
actually better than that without the ensemble
calibration. Perhaps 0.004 mags. The V magnitudes
still get worse with the ensemble calibration!

A major problem is the small number of stars in
accepted ensembles: I am using a very simple algorithm -
accept a star as a member of the calibration
ensemble if it is one of the 60 (say) brightest
stars within half a degree (say) and has a "good"
(no detected problems) V-I pair of measurements
matching every "good" V-I pair of the source to be
calibrated. I expected most of the bright sources
to get through this sieve but they do not: under
20% typically and a lot of those are fainter so the
scatter of the ensemble calibration is much worse
than expected. I haven't yet tracked down why the
acceptance rate is so low but bad pixels are a part
of it: I am currently rejecting a source for having
even one bad pixel on the edge of the fitting area.
There are 845 labelled bad pixels in V. So far.

In chasing down some of the problems, I came across
a couple of apparently "variable" stars.

#1 at 69.9225 6.9974 showed a remarkably smooth
dip in a series of 8 consecutive images (2199902 -
2199915) in MV but not in MI. This turned out to be 
not an eclipse but a transit across a CCD defect! 
This problem, affecting X1 = 1948,9,50 for X2 >= 456
(all counting from zero) got past the rather 
rudimentary checks in my Dark/Flat processing.
So there's another few thousand bad pixels to label ...

Does anybody else's pipeline do any better?

#2 at 69.2319 3.5466 shows a brightening by 0.8 mags
on one V image (2199939) with no corresponding
change in MI. Very close examination shows a small
shift in apparent position (2 arcsec) and a somewhat
large residual in the PSF fitting process - both well
within my acceptance criteria. I presume this must
be a cosmic ray hit but there is nothing suspicious
about the actual image. Nothing in nature is blue
enough for the observed brightening to be real, right?

Tom: processing all that data of yours is going to
be one hell of a problem!

Andrew Bennett, Avondale Vineyard