Among the pending items to be delivered by the VLT Astronomical Site
Monitor (ASM) is the ASM All Sky Imager (ASI), with the following
Objectives: -To detect and locate all types of clouds, including thin fragmented
cirruses (note that 50% of the non photometric nights at Paranal are of
-To establish sky brightness maps during moonlight nights
-To determine the extinction during photometric nights
The field covered is the operation field of the VLT (+/-60 degree around
The sky map (2x2 degree pixels) should be updated once a minute.
The system should -whenever affordable- follow the VLT standards for
An informal study group (detector: C.Cavadore,
image processing: A. Wicenec, astronomy: H. Pedersen) investigated
the field (
Study Report -ps file- ) and came with the following proposal:
The solution proposed is a photometric analysis of 1x1 arcmin resolution
untracked sky images, compared to a standard photometric catalogue. The limiting
magnitude (and thus the exposure time) is shuch that the density of
reference stars is aproximately uniform over the sky.
For this task, 8kx8k pixel input map is necessary. The single detector
(monolithic or butted) solution leads to unreasonably expensive optics
(input P. Dierickx, ESO). The optimal configuration using commercial optics
is a 2kx2k detector with 14mu pixels coupled to a 50mm apochromatic
objective giving an unvignetted 30 degree FOV. An assembly of 4 such
detectors is attached to a 4 positions scanning device to cover the useful field.
sketch ). A complete cycle including 3 colours (12 images per detector) is
accomplished in about 3mn with a CCD readout time equal to 5s and a
maximum exposure time of 10s.
Advantages of the segmented
-The segmented field approach has an enormous advantage in moonlight
nights: only one image (ie 1/16th of the field) is saturated.
-The 2kx2k detector is available on the market. Sample images can be
obtained on short notice to start testing the software.
-Limited specific hardware development is required (a scanning mount with
speed such that the next position is reached in a time smaller or equal to the
CCD readout time)
-The motion of the assembly from one sub-field to the next is done during the
chip readout thus without additional overhead.
-The time available for processing one 2kx2k image in the optimal configuration
is realistic: it is set by the sum of the exposure time and of the readout time,
ie. about 15s.
Sep-Dec 1999: finalize design, development of prototype software on test
Jan-Apr 2000: procurement and assembly of one subunit
Jun 2000: commissioning
July-October 2000: procurement and assembly of one complete unit (4 subunits)
Dec 2000: commissioning
Aug.99: Test images acquired by C. Cavadore using his home made controller to readout the
2048x2048 Thomson THX7899 chip: one
10s and one
7s exposures taken under a photometric sky and under a slightly cloudy sky (guess which
is which!). The images were taken with a 50mm objective at f/3.5
Sep.99: Test Bench Processing with SExtractor (A. Wicenec):
3.8s on linux with a 550MHz PentiumIII
Oct.99: Inclusion of calibration frames (flat, dead pixel map) for testing astrometric
accuracy: 185 stars of the Guide Star Catalogue (0 <= mag <= 5.5) were identified
with an rms error smaller than 3.8 arcsec.