Hi Steve,

You refer to the SX285 linearity graphs and say you believe that the
Artemis linearity would look very much the same. Compare these two
links:

SX285: http://www.starlight-xpress.co.uk/Photometry.htm
Deviation from linearity: +/- 0.5 %.

Art285: http://www.eronn.net/astro/art285/lintest1e.png
Deviation from linearity: +/- 18 % (without corrections)

When the Art285 data are corrected with the 2nd order model the
deviation is reduced to be mostly within +/- 1.5 %. I don't know if
my source linearity was any more accurate.

I have assumed that the SX graph was calculated as the ratio between
the pixel output values and a linear fit to the data plotted versus
exposure intensity or exposure time. The same method was applied to
my Art285 data.


For reference, the original measured data can be seen here:
http://www.eronn.net/astro/art285/lintest1c.png
The green line is the linear fit used as reference above. The dashed
red line is the second order model that was used to correct the
data.

The linearity plot can be studied versus a logarithmic horizontal
axis here:
http://www.eronn.net/astro/art285/lintest1c.png


You have a point Steve when you ask about the motivation for this
work. I started out with noise measurements and drifted into
studying linearity, which is probably important only for photometry.
My motivation has been to learn about this fantastic camera and its
potentials, and I think I am learning a lot :-).

I am very impressed by your sky-survey site and the idea of using
Earths rotation for scanning the sky. However, I am not planning to
do a lot of photometry myself near future. If several other group
members express their interrest I may contribute to searching for
the source of the nonlinearity and/or to developing a robust method
for correcting for nonlinearity. Independent on other peoples
interrest I will try to finnish off what I have allready done by
doing a few more measurements and sumarize what I have found on my
web page www.eronn.net.

Erlend


--- In ArtemisCCD@yahoogroups.com, "Steve Chambers" <Steve@P...>
wrote:
> Hi,
>
> I think we are coming at this from completely different
directions.
> I hate the situation where a model is built then we hunt around
for
> a use for it.
>
> On the SX 285 linearity graph you link to TG (http://www.starlight-
> xpress.co.uk/Photometry.htm ). Its interesting that its not a
> straight line but looks to be +/-0.5% over the unsaturated scale.
I
> don't know the method SX used to get the data but I suspect that
the
> Artemis linearity would look very similar indeed if plotted on the
> same graph.
>
> Its interesting that SX do not then suggest fitting a higher order
> equation with no theoretical background to the data. Anyway
where
> would you stop. 3rd order would better than second, and 4,5,6
give
> even better r values.
>
> Please rest assured that in the same way Terry assures you that
the
> SXV is fine for photometery I can assure you that the Artemis 285
is
> likewise. There is no benefit to scaling the data in the fits
file
> but its one of a large number of factors to be considered when you
> start doing photometery.
>
> I think any model building must be objective driven to be
credible.
>
> If the objective is to do photometery on stars then the whole
system
> should be considered (camera, lens, filters, atmosphere, light
> pollution). I do have some previous form here. Four years ago I
> set up an automated sky survey using a NABG SBIG ST-7. Details at
> http://www.sky-survey.com
>
> Its still collecting data but the website has a few broken links
> after I needed to more servers. Anyway while setting up the
survey
> I tried many different methods to relate flux to magnitudes. The
> methods I have settled on is described at the end of this
message.
> You will see 2nd order equations featured as means to an end. The
> survey itself is able to catalogue the magnitudes of around
100,000
> stars on a clear night and monitors several hundred variables.
>
> I would suggest refocusing at this stage and decide what you are
> most interested in. A couple of suggestions that come to mind are,
>
> Investigating the source of differences in linearity between say
the
> ICX285 and ICX429. Probably looking at the effect of changing
> voltages etc to the CCD. Looking at charge transfer efficiency
for
> different charge levels. And CCD re combination losses etc.
>
> Or
>
> Developing methods to determine stellar magnitudes. Really
should
> include the analysis of some real sky images.
>
> Steve
>
>
> Message to sky survey users group
>
> Hello,
>
> Now that I have caught my breath from the survey update I will try
to
> explain what I have done and why!
>
> If it gets complicated jump to the end!!!!
>
> One of the problems I have is calibrating brightness (flux) to
> magnitude.
> The survey camera is not well calibrated and doesn't use standard
> filters so
> its not as straight forward as picking a couple of standard stars
to
> fix the
> mag flux relationship.
>
> The solution I have adopted is as follows. The stars on each plate
> that are
> identified as being singular (no other GSC stars within 45 arc
sec)
> are
> selected. Their log flux is plotted against GSC Mag. As there is a
> fair
> bit of uncertainty in the GSC mags anyway and as the GSC filter is
> not
> exactly matched by the PMDO camera there is a fair bit of scatter
in
> this
> graph. A 2nd order polynomial fitted to the graph which both
> averages out
> the variations and provides a smooth flux to mag relationship
which
> is much
> less dependant on limiting mag than the previous straight line
> approach I
> was using as well as providing a closer match with GSC mags
through
> out the
> range. This polynomial derived mag I am calling PMDO mag.
>
> Next a program was run to compare PMDO mags of the same star on
> different
> days. For this only the good quality plates were used. Stars that
> had been
> found on at least 3 occasions (usually many more) had there
average
> PMDO
> mag determined as well as the standard deviation.
>
> The next steps are used in the 'interesting star' tables on
> http://www.sky-survey.com/inter.htm
>
> Outliers (PMDO mags that are more than 2 SD's from mean) are
removed
> from
> the analysis (the number of outliers that have a high PMDO mag
than
> average
> are recorded for the stars that dip table). Then the mean PMDO
mags
> and SD
> are determined again.
>
> Next the mag to sd relationship is modeled with the polynomial (is
a
> kind of
> u shape with a minima around mag 10-11) so the average sd at any
> given mag
> can be determined. Then the ratio of a stars mag sd / average sd
at
> that
> magnitude is determined. The stars that have a high value for this
> ratio
> are reported in the 'interesting' tables.
>
> Next the average PMDO mag for the 'standard' stars is saved.
>
> Finally the individual plate stars are reloaded and the PMDO mag of
> standards present compared to the saved average PMDO mag. Now a
> relationship between RMS PMDO mag on plate-Pmdo standard mag vs
PMDO
> mag is
> fitted by a (you guessed it) polynomial. This line gives the
average
> error
> at any magnitude on that plate and is used to compute a SD to each
> star PMDO
> mag record.
>
> THE END
>
> Right got that!!
>
> On the page http://www.sky-survey.com/inter.htm you can find files
> containing stars with problems! The problem might be the plate
> solution,
> passing airplane cosmic rays etc. Or the star may genuinely of
> varied its
> magnitude.
>
> It would be good to understand what has happened. If you would
like
> to
> take a look could ask the survey for all the data on the given
star
> see
> http://www.sky-survey.com/eforms.htm and
> http://www.sky-survey.com/starhelp.htm
>
> Try plotting PMDO mag vs MJD with sd as error bars. (remember only
> quality
> 3 plates were used to assess variability)
>
> If we can get a better handle on what's happening to these stars
it
> will be
> easier to identify just the real variables. If any of the
> variability's
> look to be genuine we can arrange some more accurate measurements
> (about 25
> known variables have been pick up and are identified in the tables)
>
> Steve