Re: H*lp in quantifying an error

[droege@snapmail.us]

Rob,

One thing I do in the run procedure is to look at images.  If they look
OK, then they probably are OK.  The analyze feature of DS9 allows taking a
horizontal/vertical cut of the image.  I find this to be very useful to
detect shutter problems.  I have a problem from time to time where a
shutter fails to close completely.  This causes bright stars to leave
streaks.  I can rarely see any effect from this on the data since it is a
small change compared to other things going on.

Having said this, the pipeline does a pretty good job of removing such
things.  The problems in the data are elsewhere as we shall soon see.

Tom Droege

>
> I have a problem.  I never realized that when I received an error while
> operating the shutter, that the shutter was opening part way.  It just
> never
> occurred to me.  So, I have a lot of data in my database where some
> portion of
> the image was exposed 3 seconds or more than the rest of the image on the
> I
> camera.
>
> I believe I have enough information it fits headers, and thus the
> database, to
> find the errant images from the exposure times.  I'm thinking the best
> course is
> to just remove the potentially errant data from the database, and chalk
> this one
> up to stupidity...
>
> An example of the data I'm looking at (exposure time info in seconds):
>
>  night_id |       avg        |        stddev        |   min   |   max
> ----------+------------------+----------------------+---------+---------
>       126 | 89.7321237879475 |     2.41216849797424 | 86.5803 | 96.0406
>       125 | 90.1557695865631 |     1.79063103747542 | 86.7998 | 92.9699
>       124 | 89.5746297654073 |     2.29459277760115 | 83.7204 | 95.9905
>       123 | 89.8512607957951 |     2.50215442579959 | 86.7296 | 93.0198
>       121 | 89.3826369194135 |     1.55663320726759 | 86.5192 | 92.9899
>       120 | 90.5278317563169 |     1.33909685651333 | 86.8098 | 93.0599
>       119 |  89.872473349938 |   0.0343538758711632 | 89.8306 | 90.1105
>       118 | 89.8687515258789 |  0.00641751428538242 | 89.8603 | 89.9106
>       117 | 89.8676383416384 |   0.0263100235050021 | 89.6002 | 89.9608
>
> I believe nights 120 to 126 are suspect, while 117 to 119 are not.
> Overall,
> there are 31 nights out of 87 nights where the difference (max - min) >= 3
> seconds, which indicates potential problems...
>
> Thoughts?
>
> Sigh,
> Rob
>
> --
>  21:46:03 up 18 days,  4:32,  2 users,  load average: 2.08, 2.18, 2.16
> Linux 2.6.5-02 #8 SMP Mon Jul 12 21:34:44 MDT 2004
>