Re: Andrew's analysis of data set T

[Tom Droege <tdroege@veriomail.com>]

I think that Andrew is working on the roughly 65 disks of data sent out as 
the DRG set and was taken with TOM with fixed Declination.  This is a 
different set from that sent out to Arne which was taken with MICHAEL while 
moving the Declination axis.

The first thing to read is the .txt file on each disk which tells things 
like viewing conditions and the format for taking data.  I just looked at 
one.  Taken on 000203 the file is 1203.txt.

I thought all of this set was taken with roughly 100 second exposures.  A 
dark folowed by 4 sky frames, then a rewind to the starting position.  Then 
repeat through the night.  While I no longer have copies of the 65 disks, I 
do have a few later ones taken with the same procedure.   I looked at the 
log files and it looks like a typical exposure was 120 seconds.  I do not 
think the whole set varies by more than 20 seconds.

It is possible that the early data disks in the set were taken with a 
different program.  It should all be described in the .txt files.

OK, I just looked up the distribution list and looked at the log book.  The 
runs before 001002, JD1820, were taken under various conditions.  I can 
look up any run in my log book if you want, but the information is in the 
.txt file, I hope.  Starting with JD1820, all the runs were taken as 
above.  On 000928 (JD 1816) I adjusted the clock levels.  This could really 
affect the data.  Also note that the best data in the set was taken on 
001002 and after.

Remember the scale is something like 2.5 e-/ADU.  The manufacturer only 
claims a linear well to 80,000 (possibly 100,000 the old data book says 
100,000 but later spec sheets list 80,000) electrons in MPP mode.  Using 
2.5 gives a full scale count of 32,000 ADU above the zero position.  Zero 
is around -25,000.  So any reading above +7000 counts is suspect.  One can 
get much larger counts, but they may be saturated or may lose charge during 
read out or...

Seems to me an experiment is in order.  Throw out everything above some 
limit.  See how the paired data plots vary with the limit.  I would start 
with the +7000 count value as being pretty safe.  Then increase it and see 
what the plot shape does.

I would also work first with data from 001002 (JD1820) and later.  This was 
all taken with a nominal 100 second exposure and in the format described 
above.

Note that in non MPP mode, the full well is 400,000 e- .  This (I think) 
means that you can hold 400,000 e- for a short time.  Those electrons don't 
all recombine, so funny things can happen around something that is 
saturated.  It is also clear that large saturated values will appear in the 
first lines read out.  Just remember all the discussion about this.  Again, 
the clock levels were adjusted on 001002.

Tom Droege

At 12:05 AM 3/4/01 +0000, you wrote:
>On Sat, 3 Mar 2001 10:37:41 -0700, Arne
><aah@nofs.navy.mil> wrote:
>
> >Could Andrew or someone remind me of what the exposure
> >times were for the analyzed frames?  I don't see it in
> >his memo.
>Right. The times are not there; I remarked
>somewhere that I had not got round to parsing
>Tom's log files ... hoping somebody else would
>do it for me! I have used all the data: this includes
>runs with short and long exposures and the long
>exposures varied from time to time. A range of nearly
>2 magnitudes. Most were around 100 seconds, I think.
>
> > The I-band photometric error plot looks just
> >like the bright end contains saturated stars.  Andrew
> >says they are not, but does not give an explanation of
> >how he decides saturataion level (it isn't always
> >65535 ADU, for example).
>
>Again you are right- I need to go back and look at this.
>
>I looked at some earlier images and picked a value that
>was close to saturation in the worst case. In those days,
>saturation varied a lot across the images. Since then, Tom
>has improved things greatly and the saturation value varies
>much less than it did before. I have used the same constant
>value throughout. This should be very safe for all the recent
>images but may be marginal in one corner of the older images.
>As you say, it certainly looks like saturation.
>
> >  Another possibility is the transformation for the
> >Tycho stars used for the I-band calibration, since this
> >transformation is nonlinear and highly luminosity class
> >dependent.
>
>The standard deviation I measured is the internal value
>for the TASS data. It is quite independent of Tycho.
>
>Andrew Bennett, Avondale Vineyard