
X-Original-To: oxborrow@dsri.dk
Delivered-To: oxborrow@dsri.dk
From: Niels Lund <nl@dsri.dk>
Subject: SDAST input
To: oxborrow@dsri.dk (Carol Anne Oxborrow)
Date: Thu, 14 Aug 2003 12:20:56 +0200 (MEST)
Cc: nl@dsri.dk (Niels Lund)
MIME-Version: 1.0
Content-Transfer-Encoding: 7bit


SDAST meeting, Helsinki, 11-12 August 2003
Niels Lund

Presentations:

*********

*** INTEGRAL instrument status:

*** Spacecraft: 

The overall telemetry rate increased by XX%.
The operation was generally successfull allthough the MOC data 
handling system capacity was marginal for the new rate and this 
has caused some delays in the data deliveries to ISDC.
The data rate increase was extremely important for the two large
instruments on INTEGRAL, since with the increased background,
particularly for SPI, there simply was not telemetry available for 
SPI to download all its events, and IBIS was allways dangerously
close to filling its allocation completely, in which case the data
stream would begin to show "holes", which would make timing analyses
much more difficult. With the new telemetry allocations both IBIS
and SPI are providing good and continous data.

The spacecraft is otherwise performing very well, and the reserves
of gas for the attitude control system are adequate for more than
15 years of operation.

An extension of the lifetime of the mission from 2 to 5 years will 
be discussed at the SPC meeting in November.


*** SPI: 

The instrument is performing well, a second annealing of the
germanium detectors has just been completed.

An update to the on-board software to impprove the quality of the
data has been prepared and is planned to be implemented soon.


*** IBIS:

ISGRI is performing very well, it is the results from ISGRI which
so far has been the most visible from INTEGRAL.

PICsIT have serious background problems. I am not aware that a 
solution of these problems have been found.

*** OMC:

Is performing essentially without problems. Nice light curves for
pulsating and oscillating stars have been produced.


*** JEM-X:

We still only operate with JEM-X2. Re-activation of JEM-X1 will be
discussed in the beginning of September. The anode loss issue seems
to have been settled (see Jeromes presentation), but we have been
asked to investigate another peculiarity, namely a steady increase
in the gas gain which we have observing since the beginning of the 
mission. The rate is surprisingly high, about 1% per 4 days. 
Operationally we counteract the gain increase by reducing the high
voltage, since the last Crab calibration we have reduced the voltage
3 times, corresponding to a total gain change of about a factor 1.5.

A few weeks ago we had a brief opportunity to check the gas gain in 
JEM-X1 during a short diagnostic exposure to Sco X-1, the gas gain
in this unit had essentially not changed since it was last used 
during the Crab calibration in February. The gain increase is 
therefore connected to the active use of the detectors.

We have carried out a laboratory test with the JEM-X spare model and
a strong radioactive source (Sr-90, a beta emitter with a maximum
energy of 2 MeV), and these tests confirm that the gas gain does 
increase at about the rate observed in space when the detector is
under irradiation. We have not yet found an explanation for this
behaviour.

We are working on a patch for the on-board software which will
reduce the background at high energies while maintaining the low
energy efficiency at the present level. This patch will possibly
only be implemented in the spring of 2004 - we want to have it
calibrated on the Crab after uploading, and we are probably now
too late for the current Crab calibrations.

***************

*** Source modelling, finding and extraction

*** Source modelling

I derive my source models for a 1x1 mm grid in the shadowgram. 
For a given source position the source position are specified in 
detector coordinates as continous variables (not on a fixed grid).
I first identify all those detector pixels which can be illuminated
from the source. In this step the collimator shadowing is taken into
account - this becomes very important for off-axis sources. Also the 
background region relevant for the source is determined, here only 
the edges of the detector, of the collimator and of the mask are
considered. The collimator lamellae and the mask code pattern is 
disregarded.

The next step is to calculate the distribution of detection points
corresponding to a photon arriving at a given pixel at the detector 
window. Here I consider the distribution of the absorbtion depths
for the photon and the limted position resolution of the detector.
Both of these quantities depend on the photon energy. I normally
divide the JEM-X energy range into four energy bands.

I end up with a map (a shadowgram) giving for each pixel on the
detector the probability for observing a photon from the given
source. The model is normalized to correspond to one arriving photon.
A map showing the region of the detector relevant for the background 
for this particular source is also prepared.


***  Source finding

The source finding is done with the "midisky" program. (Currently
midisky8). 

Midisky takes an observed set of four shadowgrams (corresponding to
four energy bands) as input.

It uses several auxiliary input files of which the most important 
are:
1) A set of four shadowgrams (four energy bands) from an "empty field" 
   observation. I use these shadowgrams also to indicate regions 
   of the detector I dont want to use (dead or noisy anodes, high
   background levels near the edge or near calibration sources).
   "Dont use" pixels are indicated by a negative value of the
   background pixel count.
2) A large file providing source models calculated for a 255x255 grid
   covering the full JEM-X field of view (13.2 degree diameter).

Midisky uses the "iterative removal of sources"  philosophy. In the 
first step it only fits the background - using just the empty field
shadowgrams plus a costant term. After this step, and for all the
following iterations the residual shadowgram contains "zero photons" 
i.e. the residual shadowgram contains both positive and negative 
pixel values (float values).

After fitting and subtracting the background from the shadowgram
through the first fit (I shall discuss later the problems encountered
in this step) the program proceeds to the backprojection of the 
residual shadowgram. The fact that the residual shadowgram has a 
mean of zero counts means that in the backprojection stage of the 
iteration loop I dont need to worry about "balancing the 
reconstruction", the mean value of the reconstructed skymap is also
close to zero.

The program then searches for the highest peak in the skymap. It uses
the pixels values around the highest peak to derive a weighted peak 
position which does not necessarily fall on one of the grid points in
255x255 search grid. (The position of the source candidates are 
floating point numbers).

A model for a source at the derived position is then evaluated and
this source candidate is included in the fit for the next iteration. 
The source modelling is not perfect, so it frequently happens that 
the iteration loop find the same source again - only with a small 
shift in the position. In such cases the position of the source 
candidate in the direction indicated by the second or third 
occurrence of the source. For really bright sources even this is not 
enough to totally get rid of the source signal. So after the fifth 
repetition of a source a  blind spot is placed on the skymap. 
A 5x5 pixel region around the source position is blanked out for
all following iterations.

The iterative loop can terminate for one of several reasons:
- too little significance of the source candidates found
- too many sources (usually I take 10 sources as a limit)
- too many iteration loops (usually I accept at most 20 iterations)

When the loop is terminated the source candidates found are saved 
in a list and the program proceeds to the next energy interval.

When all four energy intervals have been processed the four lists
with source candidates are compared and only those candidates which 
have been seen in more than one energy range are accepted. (The
candidates which have only been seen in one energy band are saved for 
possible verification by compparison with similar lists from other 
science windows). 

The final step in the Midisky program is to do a fit in all four
energy bands using only those sources which have been detected more
than once. This fit provides the final source strengths for the
confirmed sources.

The source models used in the final fit are output, they are needed
as input for the programs extracting light curves or spectra for
individual sources.


*** Source extraction

Based on the source models generated by Midisky for a specific science 
window the source extraction program will identify an optimal set of
pixels from which the source (+background) can be obtained and another
set of pixels from which the background (with a minimal source 
contamination) can be derived. The optimization process tries to
maximize the signal to noise ratio for the extracted source signal.
This is done by first sorting the pixels in the source model in order
of falling probability for detecting the source. Then the integrated
probability for detecting the source photon is evaluated for a number 
of threshold levels of the probability per pixel, starting from those
pixels providing the highest probability. The signal obtained from
such a set of pixels necessarily includes some background. In order
to subtract this background we want to define a second set of 
shadowgram pixels containing as little signal as possible (but still
containing many pixels so the background count rate can be reliably
estimated. An algorithm has been written for determining the best pair 
of such probability thresholds - a high one for the signal sample and 
a low one for the background sample.

