Re: TN-107

[Tass Mailing List <tass@mail.alembic.net>]

Date: Thu, 25 Oct 2007 17:50:48 -0500
From: Thomas F. Droege <droege@fastmail.fm>
To: tass@tass-survey.org
Cc: arne@aavso.org
Subject: Re: TN-107

Hi Arne and all,

I am using the tass reduced data.  Reading the TN, you will see that the
fields range from 70 to 700 stars per sq degree.  The "cleaning process"
removes about half the stars in the most crowded fields and 10%  or so
in the least crowded.  I am removing stars that fall within an 16 pixel
box centered on the test star.  When another star is found in the box,
both are removed from the data set.

I don't understand why you propose that referencing a high resolution
catalog will make an improvement.  The tass pixels are 7.5 arc seconds
and we generally get 1 to 2 arc second positions when compared to high
resolution positions.  The PSF is typically 2-3 pixels.  We do use a
selected set of the Tycho2 catalog in the pipeline that produces the
positions.  About 150 Tycho2 stars are used for each 16 square degree
image for comparison.  These are selected to be in the tass magnitude
range.

When you start running ARNE, I will look forward to seeing what better
technique you find to process the data.  I have previously tried
selecting the good nights with not nearly as much improvement.  One has
to be careful or one can select down to nothing.  I find that what is
left after this "cleaning" has lots of stars that look variable.  I am
slowly getting everything set up and code snippets written so it is easy
to search the data.  I hope to be able make this available so that
others can have fun finding new variables from my data set.

As I previously said, to much criticism, it is like "shooting fish in a
barrel".  Is the AAVSO interested in lists of new variable stars?  Out
of the process there will be a lot of measurements of existing variable
stars.  These will be filtered values and will have error bars.  While
not as good as professional measurements, the measurements should be a
significant improvement over visual data.

Tom Droege


On Thu, 25 Oct 2007 14:36:40 -0700 (PDT), "Tass Mailing List"
<tass@mail.alembic.net> said:
>
> Date: Thu, 25 Oct 2007 17:20:40 -0400
> From: arne <arne@aavso.org>
> To: Tass Mailing List <tass@mail.alembic.net>
> Subject: Re: TN-107
>
> So it looks to me like your partial solution to the poor photometry
> problem is to eliminate the crowded fields.  How are you deciding
> if a neighbor is present?  You say all stars within 8 pixels of
> another star, but are those stars identified from the TASS images,
> or coming from another catalog like UCAC?  The better method is to
> use a high-resolution catalog like UCAC or Tycho to pick unblended
> objects.  You will of course eliminate most of the stars in the
> database, as the majority will be (a) faint, where crowding is
> more common and (b) near the plane of the Milky Way.  This is ok,
> if it gets you better results.
>
> What most of the exoplanet survey groups use is image subtraction
> techniques to eliminate the blending problem.  There, you just
> look at what varies and throw the rest of the stars away.  It is
> a complex operation, and works best on systems that stay stable
> (not trailed one night or out of focus the next), and you always
> want to work with clear nights so that you don't have differential
> extinction across a frame from clouds.  My guess is that some
> fraction of the noise floor is due to transparency variations, and
> if you could further select to remove nonphotometric nights, you
> might get better results.  In other words, you have lots of nights
> and plenty of data.  Pick out the 10percent that is really good
> and work with that.
> Arne
>
-- 
  Thomas F. Droege
  droege@fastmail.fm