Commissioning of the NOTCam science array
The NOTCam science array, which has been stored in a bank box
in Santa Cruz for 4 years, was installed in NOTCam on October
the 20th. The installation was made with the assistance from
Preben Nørregaard from CUO, and the procedure is now
for future reference.
The first warm tests on the 20/10 showed that we had contact
with all four quadrants of the array, and that we could read
the detector. Much more could not be tested warm. NOTCam was
closed and set to pumping, moved to the dome, more pumping,
and finally cooled down according to normal procedures. NB!
The array controller was powered off during cooling down!
The first cold tests were made as soon as NOTCam was cold on
the 25th. The new array has much better cosmetics than the
engineering grade array. Some time was spent adjusting the
bias and reset voltages for each quadrant to make sure we
have the reset level within an optimum range. If it is too
low, we get more zero level pixels, and if it is too high we
get a small dynamic range before entering non-linear regime.
We selected to use a level of ~ 5000 ADUs.
NOTCam was mounted on the telescope on the 26th, but the weather
did not allow opening the dome. We got the data stream for the
quality control observations of darks and domeflats. Because of
the bad weather and the presence of our detector experts we took
the opportunity to study the strange darks more in depth.
We refer to the
NOTCam Science Array page for pictures and further information.
Here follows a short summary of the commissioning results.
Read noise and gain
The read noise was found to be 14-15 electrons, a bit higher
than the earlier 10-12 electrons. The gain is 2.2e/adu compared
to earlier 2.8e/adu.
The new array starts to saturate around 54000 adu (before: ~
40000 adu). The 1% deviation from linearity starts at ~32000
adu (before at 16000 adu). Taking the different gain into
account this gives a larger linear regime range than for
the old array.
The science array looks much better cosmetically than the
engineering grade array.
The number of cold and zero pixels is about 1% alltogether.
The number of hot pixels ~1.4% for long dark exposures, but
< 0.2% for normal exposures. The bad pixels are practically
all located along the edges, but there is one region inside
that has an elevated number of individual bad pixels.
The behaviour of the dark level with time is similar
as for the engineering grade array, but still very poorly
understood. Firstly it is too high, and secondly it is highly
non-linear and variable. For the engineering grade array it was:
~ 7-9 e/s/pix for short darks, but much smaller (< 1 e/s/pix) for
longer darks taken with the reset-read-read mode. For the science
array it seems to vary from 2-9 e/s/pix for the ramp-sampling mode,
while it is 2-3 e/s/pix for short exposures in reset-read-read
mode and < 1 e/s/pix for longer exposures in this mode. The offset
level is variable, perhaps depending on the time since last reset.
The problem is briefly described under the
engineering grader array and the recommendation given in the
NOTCam User's Guide was not to do separate dark subtraction.
Earlier a dependency of 400 adu per degree change of the detector
temperature was found, but the detector temperature is more stable
since Apr-03 when we started using the PTR also when mounted on the
telescope - although the temperature we actually monitor is at the
location of the array only, so we do not monitor the internal array
temperature as it might vary depending on its active or dormant
state. We suspect the dark level behaviour might be related to the
way the on-chip FET has been disabled, and investigations are on-going.
The detector flat field is flatter than before. For Ks filter
dome flats (lamp ON-OFF) the stddev in 20x20 pixel boxes is
~2% and the deviation across the whole array +-5% for both the
HR and WF camera. For the engineering grade array the Ks sky
flats had ~3-4% stddev in 20x20 pixel boxes and a deviation
across the whole array of +20% (in the bright band) and about
-15% in the darkest corner.
The memory (or charge persistency) effect is 1% or less, and goes
away after one dummy read (e.g. dark 0) for non-saturation. For
levels above saturation, however, it is more persistent and seems
to produce transient changes of the gain (~ 50% lower) for the
affected pixels. The gain recovers after one dummy read, but the
memory persists as a positive signal after 6 dummy reads, although
at a low level of 0.03%.
The sensitivity of the new array has been tested on calibration
lamps only so far. If we can rely on the stability of these, it
seems the array is at least 30% more sensitive than the previous
one. The old array has a QE of about 55%, so 30% more is > 70%.
This will be checked on standard stars as soon as we get clear
sky on a NOTCam night.
The alignment of the array looks OK. Internal focus
has changed by only 50 units, and the array is 1.6 degrees
rotated with respect to the old one (unavoidable, as it goes
like this into its socket). Judging from the calibration arc
line widths there is no serious tilt of the array, the fwhm of
the lines vary by 0.7 pix over the array in the y-direction and
by 0.2 pix in the x-direction (i.e. along slit). These are the
same values as for the old array.
Still to be done:
- The quality control pipeline s/w will be adjusted for the
new array to have proper boundaries and limits etc.
- Sensitivity must be found from standard stars, and the SIGNAL
code (exp time calculator) must be updated.
- A time-efficient clean-array command should be implemented.
- Understand the origin of the strange dark level behaviour and
Comments to Anlaug Amanda Djupvik