<pre>
From tmca@ctiosz.ctio.noao.edu Tue May 28 16:59:41 2002
Date: Fri, 17 May 2002 23:16:37 -0400 (CLT)
From: Tim Abbott <tmca@ctiosz.ctio.noao.edu>
Reply-To: Tim Abbott <tabbott@ctio.noao.edu>
To: scistaff@ctiosz.ctio.noao.edu
Cc: Ramon Galvez <ramon@ctio.noao.edu>,
Ricardo Schmidt <rschmidt@ctio.noao.edu>, Oscar Saa <osaa@ctio.noao.edu>,
dmaturana@ctio.noao.edu
Subject: Mosaic linearity test
In an effort to get at least something useful out of this miserable
week, I experimented with using the flat field lamps & white spot for
linearity testing of the mosaic. Nonlinearities in the CCD response
and changes in illumination level can never be completely separated,
but the drift in illumination can be estimated by taking a series of
exposures of ascending and descending interleaved exposure times.
This week's data suggest that the combined variation in illumination
level and detector nonlinearity is <~1% for all output amplifiers and
CCD nonlinearity is probably considerably less than this. Basically,
the flat field lamps appear quite stable enough for general linearity
testing.
Readout noise and gain match reasonably well with the numbers given on
the Mosaic web pages, although there are some differences. The
individual results are tabulated below:
Amp Gain RON RON Nonlin Bias
e/ADU ADU e % Slope (ADU)
1 2.75 2.17 5.97 ~<1%
2 2.86 3.3 9.4 ~<1% ~10
3 2.84 2.47 7.0 ~<1% ~5
4 2.30 2.2 5.06 ~<1% ~8
5 2.45 2.9 7.1 ~<1% ~10
6 2.37 4.2 10.0 ~<1% ~12
7 2.53 3.3 8.3 ~<0.8% ~15
8 2.37 2.6 6.2 ~<1% ~8
9 2.03 3.5 7.1 ~<1% ~-5
10 1.95 3.5 6.8 ~<1% ~-5
11 2.24 6.5 14.6 ~<1% ~-5
12 2.13 3.8 8.1 ~<1% ~-5
13 2.63 2.3 6.05 ~<1% ~-2
14 2.79 2.9 8.1 ~<1% ~-2
15 3.35 2.6 8.7 ~<1% ~-5
16 3.49 2.5 8.7 ~<1% ~-5
All bias frames show some degree of slope in the vertical direction
away from the serial register. Comparing with the overscan (flat) and
some dark images shows that this is not dark current. The first
explanation that leapt to mind was that it is spurious charge --
charge introduced in the CCD by impact ionisation if the vertical
clock transitions are too fast, too sharp, or too large. Chris
confirms this and that it is apparently incurable.
The bit biases in most ADCs are larger than I am used to, and as much
as 20% or more in some cases. This is an intrinsic problem in many
ADCs and its effects are impossible to predict. Generally, however,
only the lower few bits are affected and any consequences will be
small.
A 4 sec exposure shows an anomolously low count rate, hinting at a
possible problem with the shutter at short exposure times, but a
subsequent sequence of exposures from 1 to 10 sec shows the shutter to
behave well in this regime with a maximum shutter error of 0.01 sec
(inclusive of possible illumination drift).
I attach a <a href="2002-05-17-plots.tar.gz">tar file</a> within which you will find the linearity, transfer
and bit-bias curves for each amplifier.
Cheers,
Tim
--
Tim Abbott, tabbott@noao.edu, www.ctio.noao.edu/~tmca
CTIO, Casilla 603, La Serena, Chile, +56 51 205200
or NOAO/CTIO, P.O. Box 26732, Tucson, AZ 85726-6732, USA, +1 520 318 8259
[ Part 2, "" Application/X-GZIP 266KB. ]
[ Unable to print this part. ]