UPDATE ABOUT DANISH TELESCOPE
JP Beaulieu & J.C. Donatowicz, May 12, 2006

We arrived on April 30 on the mountain.

Status on April 30 :
The CCD was cooled down. Telescope control system (TCS) was on. Filter
wheels were not properly initialized and filters not properly identified.
Dome was not rotating, nor the windshields were moving.

ESO is supposed to fill up the dewar with liquid nitrogen on a
daily basis, but it is worth checking.
There has been a need for a bit of carpentry work in the control
room to repair the table using tools provided by the swiss telescope crew.

Computers (TCS), PC to control CCD camera (BIAS PC), HP to view
images were on. Old PC (500 Mhz, called w1d5drl) for image processing was ok.
Fast PC for data processing was physically here (w1d5pl), but not installed.
Screen for Bias PC was dead, we found a spare one.
Data transfer between BIAS, w1d5drl, w1d5pl was not working (but was
finally fixed with the help of John Pritchard).
We installed both old and new pipelines, including various software such
as supermongo, gaia, cdsclient, ds9, wcstools.

On the night of May 1, everything was fixed. Both Dophot and ChC-pySIS
(called WISIS) pipelines are installed and automatically fed by the acquisition PC (from
May 2).

After that, we had a period of testing the various equipements and software :

- The telescope zero point is off by about -20 arcmin in RA. These RA DEC
positions are fed within the fits header and does not take into account
the correction that we have to do at the begning of the night.
It is of little importance since if one wants to do the
astrometry of the frame using the new tool available at the telescope
(ccsh_wcs) is ignoring the RA DEC provided in the header.

- The telescope is usually pointing with an accuracy of 5-10 arcsec, which is
good. We had only few occasions where the shift was larger than that.
There are some tracking problems on ONE target, likely linked
to a mechanical hard point (target OB-178). This might happen later in
the season for other targets. We point the target, wait for 2 min,
then start the exposure, this way, most of the time images are not trailed.
It is during the first min that the trailing is taking place, after it is
stable tracking.

- For 2 minutes exposure, in good conditions, we can expose without tracking,
and the resulting images are of good quality.

How the data processing is going on in Chile :

Images are transfered to a $RAW_IMAGES directory. Then, they are
flatfielded and nammed, one copy is written in the $IN_IMAGE directory
where it could be processed with DoPhot. One copy is written in the
$PI_IMAGE directory waiting for WISIS processing.
Dophot processing is done the standard way.

We have been running both pipelines in parallel at the begining, but over the
last days we are more focussed on WISIS. When we saw the deviation
of 207 on WISIS processing, we immediately checked the DoPhot reduction that
was also showing the same deviation.

For WISIS processing, we have a loop checking the $PI_IMAGE directory.
If the image corresponds to a new field, they are put on "wait" status,
waiting for the tree structure where the processing is done to be built.
Then, we run "process" to start the processing of this event.
If the image is from a field already being analysed, we have an "update"
that will process the image, and write the .pysis files containing the
photometry.

If one image is poor, then one just has to set the image to ".bad" and
run "update" again, the measurement will be removed.

WISIS keeps on refining the position of the target using the subtracted
frames. It means that once a bunch of poor frames is taken, you will
derive with LESS accuracy the centroid position, and you will then
screw up your photometry. If you have 10 good measurements, add 10 poor
measurements then the photometry of the 20 measurements might be screwed
up because the centroid is now bad. When you had only 10, it was good.
There is currently interaction in the "Nerd group" to solve this
problem.

We are having lots of interaction between Chile and Tasmania, about
the scripts we are running (they are the same ones). Michael, Andrew,
Pascal, John M, ChC and I are having nerdy discussions about various
aspects of the processing and procedures. Note that Andrew is running
the vanilla pyISIS, whereas the one we have in Chile and Canopus (the ChC
environement and modifications is WISIS).

Remarks about the data quality :

- WISIS is giving better scatter than the DoPhot processing.

- If we have bad images as templates, it is an absolute mess. Once good
images are acquired (and poor ones dumped) things are ok and straightforward.

- If you add more data (poor points), you might decrease the quality of
the full light curve you are producing (see above).

- It is important that people understand that with WISIS reduction, we will
have strongly deviant points : each time something will be screwed up somewhere
it will lead to one/few strongly deviant points. This will be much worse than
in the past. The scatter of good points might be smaller, but we will have
lots of deviant points. It is important that we all learn that, and do not
see planets with the same frenetic eye as in the 90s when we were learning
DoPhot and seeing planets everywhere dued to seeing correlations.

- We can upload data every two minutes or so, but given the amount of emails
triggered by one bad data point, we switched off the autoloader. We will
upload on an irregular basis during the night once deviant points are removed.

- At the end of the night we load ALL the images of the night to Paris.
We think that it would be good if other telescopes could do the same.
Therefore, if there is a deviant point... anyone of the collab can download
the bunch of images and reprocess them.

Some comments about the observing mode we adopted :

- Special attention to the focus, making always sure that images are
well focussed.

- Even for 60 sec exposure, we use the tracking with the guide star.

- If we have clouds and we are loosing the guiding, what we do is
we integrate without guiding. Therefore we are not looking
for a lost star in the guiding camera, which would result in trailed images.

- It is important for WISIS reductions to have high S/N images.
Therefore, we use usually about 50 flats and 50 biases. We integrate
rather long exposure to make sure to have enough photons in the targets.
Our exposure times are ranging from 60 sec (OB94) to 400 sec (OB207).
We try to put an emphasis on image quality.

- when an image is bad, we just discard it.

I think that there is only one thing to keep in mind

IMAGE QUALITY : WELL FOCUSSED, WELL TRACKED, GOOD S/N, GOOD FLATFIELDING.
It is more critical than ever with pySIS processing.