UNIVERSITY OF CALIFORNIA OBSERVATORIES / LICK OBSERVATORY
Introduction and Explanation of Detector Data
Gain |
Amplifiers |
MPP |
Traps |
CTE |
Linearity |
Readnoise
Baseline (bias) |
Offset |
Return to Detector Table of Contents
Lick observatory detectors are identified by the number or name of the dewar in which they are installed (e.g. Dewar #6 or Kastred). Dewars are occasionally upgraded with new detectors--in fact, most Lick dewars have contained more than one chip during their lifetimes. When a detector is replaced, the dewar designation adheres to the new device. Only the data for the current occupant of a dewar are given in these pages (a dewar's history may be learned by contacting the Lick CCD Lab).
Most of the measurements given in these pages have been made at the lab in Santa Cruz, and while they represent the best available data on how a dewar will perform at the telescope, some differences are inevitable, due to different hardware and operating conditions on the mountain.
N.B. On-chip binning--an operating option with most of our
detectors--affects the gain when the device is being run with a
Lick controller, the gain being reduced by four when the chip
is binned by 2 in rows and columns. However, CCD's operated with Leach
controllers maintain the same gain irrespective of binning.
In MPP mode all of the clocks are held
at the negative level during an exposure. A special negative implant is
used under one of the parallel clocks to define a charge barrier.
Because of the details of how dark current is generated, there is much
less when all of the clocks are held at a sufficiently negative level.
This is the only reason for MPP, but the barrier
provided by the implant isn't as strong as the barrier provided in
non-MPP mode and the maximum charge that can be held in a pixel is
often smaller.
Some types of CCD's aren't fabricated with the MPP implant, so they can't be
operated in MPP mode. All of our Loral and Orbit 2Kx2K CCD's have the
implant, but some of them aren't run in MPP mode because of the reduced full
well.
For linearity at high count rates, a safe rule of thumb is to keep counts
10-15% below the chip's full well. (If the ADC reaches its maximum before
the chip's full well is reached, the question of high-level linrearity is
moot in any case.)
The standard Lick data-taking programs allow the readnoise to be
checked in real time (note that this is not the case for MOS which uses
the "Expose" data-taking system). To check readnoise, select Z-4 from
the data-taker main menu. This invokes a "test" submenu. Selection "B"
in this submenu sets the readout to "fast" or "slow" (an option with
Lick controllers which, when "fast" is selected, halves the readout time
at the cost of some additional readnoise--most calibration and science
frames are made with "slow" readout).
Selection "D" displays the horizontal clock noise in either fast or slow
modes, depending on which has been chosen with selection "B." The mean of
the horizontal clock noise, displayed at the bottom of the screen, is the
readnoise in DN. Use the gain for the CCD being tested to convert to
electrons. The readnoise may vary somewhat from the published figure.
If it seems unduly high, contact a technician.
By default, most Lick data-taking systems perform an automatic baseline
subtraction, so that the baseline is not retained in the raw data pixels.
The overscan region is not saved, but the baseline level is stored in the
the highest number column of every image. The overall baseline level is set
in the controller, and should be in the range from 300 DN to about 2000 DN.
Baseline subtraction may be disabled in Lick data-taking systems under the
Z-3 option in the main menu. However, if this is done, the CCD window must
be made larger to include the overscan region in the raw frame. To do so,
set the number of columns in the CCD window to a number greater than the
actual number of columns on the CCD, using selection "D" in the data-taker.
It is also necessary to create the file /u/ccd/engineering.
Note that MOS dewars do not allow automatic baseline subtraction, and that
the default for the LIRC-2 NICMOS device is "baseline subtraction disabled."
Gain
Gain is expressed as the number of electrons per Digital Number (DN).
Gain is usually meausured with our Fe55 Xray source. Each absorbed
xray releases 1620 electons. This gives us a known amount of charge
which can then be compared to the DN recorded. Their ratio is the gain.
Amplifiers
For the 2Kx2K CCD's there are two serial registers on opposite sides of
the chip. One side is called the "A" side and one is called the "B" side.
Each serial register has two amplifiers, one is the left or "L" amplifier
and one is the right or "R" amplifier. So amplifier AR, for instance,
is the right amplifier on the A side.
Multi-Phase Pinned (MPP) and non-MPP Modes
In non-MPP mode the three parallel clocks are used to define the pixels
along each column. During an exposure one of the clocks is held at the
positive voltage, and charge collects under that phase. The other two
parallel clocks are held at the more negative voltage and act as charge
barriers to define the pixels.
Traps
Traps--or bad columns--are identified by taking a very low level
exposure and then shifting the charge
back and forth on the CCD before it is read out. Charge trapping is
dependent on tiny imperfections and is different depending on which
way the charge gets shifted. The number of traps identified reading
from the A side and the B side are often different, therefore traps
are identified in the dewar data as belonging to one side or the other.
Charge Transfer Efficiency (CTE)
CTE is the measure of the fraction of charge moved from pixel to pixel
per each clock phase during readout. The vertical CTE measures this
quantity along columns, the serial CTE along rows.
Linearity
A series of exposures with increasing exposure time are done to see if
the measured signal increases in proportion to the exposure time.
We usually test only at low count levels since that is where problems
usually appear if there are any.
Readnoise
Readnoise is expressed as e-/pixel. Readnoise is independent of exposure
time. It is measured from the detector's overscan region.
Baseline (bias)
A baseline (or bias voltage) is added to each pixel in an image, on a
row-by-row basis. The value for each row's baseline is determined from the
overscan region.