Re: long term variation and instrumental drift

[Stupendous Man <richmond@a188-l009.rit.edu>]

  Arne supplies some constructive criticism, and the only real
answer is "yes, of course, I'll do that."

  However, it may be informative to make a few small answers here :-)

>   First, any such transformation should be done properly
>   somewhere in the future.  You need to be sure that you can
>   back out the current transformation and insert a better one.

  If one saves all the output of the "photom" (final step) of
the pipeline, one has all the coefficients used in the reductions,
so one can back them out and start over.

>   Second, setting zeropoints is fine with Tycho stars; there
>   are enough of them to average out Bt/Vt->V,Ic transformation errors.
>   However, they are nowhere near good enough as "standards" and
>   should not be used for generating transformation coefficients.

  Agreed.  I will have to go back and find real standard stars
(a la Landolt) scanned in the course of Tom's observations, then use
those to find good color terms for his camera.  Of course, given the
Declination he's scanning (Dec = +7), there aren't going to be many
in the current dataset.  Which means that I'll have to go back to
older data taken with Tom's camera, when he was looking at the equator.
In an ideal world, Tom would go back to a few Landolt fields on the
equator every month or so.

>   Third, I assume you are setting a single zeropoint for each
>   image.  

  Correct.

>   Depending on the flatfielding, you can get quite radical
>   errors across a frame ...
>   The only
>   solutions here are to break the frame into subframes and set
>   subframe zeropoints; fit a low-order surface to the deviations;
>   or avoid nonphotometric nights.

  I have considered putting solution 1 into the pipeline, but it 
is tough, given the small number of even Tycho standards in one
tenth of the frame.  Solution 3 is easy after the fact, of course.

>   Fourth, setting transformation coefficients per frame rather
>   than keeping them fixed for long periods of time is very risky.

  The transformation coefficients are solved once per NIGHT, not
once per run.  I allow the zero-points to vary, but keep the 
coefficients fixed over the entire night.  Yes, garbage in will
still yield garbage out, of course.  

  One might run several (good) nights of data through the final photom
step together, I suppose, to improve the color terms.  I suppose
I ought to add the option to use fixed values in the reductions.

>   I can go on and on, sorry.  You have to be careful in both
>   astrometry and photometry once you want to place things on a
>   standard system

  Oh, yes.  Tom has said recently that we should all remember that
he considers the Mark IV to be still in the "commissioning" or
"engineering" phase.  The same is true of the pipeline.  I suspect
that the "final" reductions will have to wait until I (or someone)
can put something like DS23's entire data into a database,
look through it to find and discard the bad nights, search for
real standard stars, back out the initial transformations,
perform better color transformations, _and_ find and account for
systematic errors in astrometry.  I suspect this will take 
at least six months.

  On the one hand, had the pipeline shifted the magnitudes
for a zeropoint only, it would be easier to fix them for
color terms later.  On the other hand, not including any color
terms _now_ would lead to considerable systematic errors
in anything published for the next six months.  A tough call.
It's certainly dangerous to do a poor job of calibration, I admit.
Is the current one really that bad?  :-(  As mentioned earlier,
I hope to answer this question by the end of June in a TN...

                                       Michael R