<>Checking WFS Images
TopGui allows the user to grab WFS images and save them as fits files.
This has to be done frame at a time so is unsuitable for anything but
the most rudimentary measurements. Images can be grabbed by selecting
the WFS Display Page on TopGui. A button at lower right labelled
'Save as Fits' will save a single WFS frame. This takes a few seconds
to respond. The image is given a filename that contains the date and
time. The save directory on sextans/lpss94 is :
/opt/Electra-save-dir/Grabs/WFS
The images in this directory will then need to be ftp'ed to another
machine for display and analysis with IRAF and ds9. When using ds9 it
is important to know that the image orientation is different from that
of the TopGui. The diagram below shows this:
<>
<>
<>
By default, images are saved with the bias already subtracted. This is
also the way they are displayed on TopGui. Since the images are
actually composed of 4 seperate quadrants read out through seperate
video channels they will inevitably have different bias levels and this
can produce a messy looking image when displayed and can make faint
detail hard to see. When analysing engineering images however it is
more useful not to have the bias subtraction applied. There are two
commands for turning on and off this bias subtraction feature:
WFS SetCCDtestMode 2 : no bias subtraction
WFS SetCCDtestMode 1 : with bias subtraction.
Measuring Noise
the fits files can be measured using IRAF using imexam. The cursor is
clicked over each quadrant and the 'm' key pressed each time. This will
display the noise in ADU. If the images were taken in the dark, the
noise value should be around 8 ADU. The gain of the system is about
0.55e per ADU.
Taking Flat Fields
A flat field can be a useful diagnostic of camera problems. It has been
found that by placing a white paper in front of the WFS entrance
aperture such that it reflects diffuse room light into the WFS and
dimming one set of Grace lights down to minimum (the other set being
off) that a decent flat field can be taken with a 5ms exposure.
Troubleshooting
On two occaisions there have been problems with the WFS SDSU controller
that required changing a video card (the card with a 25 way D connector
and 2 micro coaxes along its edge). These appeared as large sensitivity
drops on two of the quadrants of the chip. The clearest way to diagnose
this kind of problem is to first switch off the bias subtraction
feature and grab a fits frame in the dark. Next the room lights are
switched on and a flat field image grabbed. We then subtract the bias
frame from the flat field and display on ds9. By doing horizontal and
vertical cuts through the resultant image any gross changes in
sensitivity from one quadrant to another should be visible. If
everything is working well then only a small step in response should be
seen at the quadrant boundaries. The quadrants are processed in pairs.
Quadrants A and B are handled by the SDSU video board that is mounted
closest to the middle of the controller. Quadrants C and D are handled
by the video board close tot he right hand size of the controller. A
problem was once seen that affected quadrants A and B equally ( they
appeared to have very low sensitivity), it was then clear that the
problem must have been their common video board which was subsequently
replaced.
Measuring Gain
Measuring the system gain of the WFS is a rather complex procedure. The
nominal value is 0.55e/adu for all quadrants. If one quadrant
experiences a problem it would be useful to remeasure its gain value ,
however, the flat field method is probably the best way to spot any
relative changes between quadrant sensitivities. If an absolute value
for the gain is required then the following procedure must be followed:
1) Turn off the automatic bias subtraction feature
2) take two identical flat field exposures which give about 10,000 ADU
of signal
3) take a bias frame in total darkness.
4) measure the ADU noise in the bias in each of the four quadrants :
NoiseA,NoiseB,NoiseC,NoiseD
5) subtract the bias from the first flat field and then measure the
mean signal in each quadrant : MeanA,MeanB,MeanC,MeanD
6) subtract one of the original flat fields from the other and measure
the standard deviation of the signal in a small box positioned within
each quadrant: STDDEVA,STDDEVB,STDDEVC,STDDEVD
7) The gain can now be calculated for each quadrant:
gain in
electrons per ADU = (2 *Mean)/(STDDEV^2-2*(Noise^2))
This is quite difficult to get right. One must avoid measurement boxes
too close to the quadrant boundaries. One must also avoid using
too-deep expsoures since the pixel data wraps round to zero above 14
bits and plays havoc with the statistics.
