Astronomy Principles and Practice Fourth Edition.pdf

302 Detectors for optical telescopes
indicates typical voltages that might be applied to the components of each pixel during the exposure
phase. During this time, the generated photoelectrons accumulate under the right-hand well of each
pixel according to the local illumination. The operation of the detector is controlled by computer. This
includes the clocking of the voltage levels, the interrogation of the pixels and the organization of files
to record the mapped illumination levels.
At the end of the exposure, the accumulated charges are successively transferred along the
columns of the detector matrix by cyclically switching the voltages as indicated in figures 18.12(a)–
(d). The technique is referred to as the ‘bucket brigade’. The contents of each pixel arrive in turn at the
extreme column, the charge flow is amplified and re-transformed to a digital value by an analogue-todigital
(A-to-D) converter and temporarily stored in the computer’s memory, according to the location
of the originating pixel on the chip.
Inevitably, during the transfer process, charge is left behind from one stage of the register to
another. In the operation of the detector, it is very important that the efficiency of the transfer process
is extremely high. Suppose that
N 0 = number of electrons under a gate
and
N t = number of electrons in the adjacent gate after transfer.
The charge transfer efficiency (CTE) can be defined as
( )
N0 − N t
CTE = 1 −
.
N 0
To see just how important it is to have extreme high values of CTE, consider what happens to a starting
charge value of 1000e − under a gate following 100 transfers through the detector columns with a CTE
of 99%. After the transfers, the readout of the charge would be
1000e − × (0.99) 100 −→ 370e − .
Thus, any image would be severely degraded, with the electron accumulations at one side of the
chip being very much weakened with respect to the pixel columns which are immediately closer to the
readout column. In practice, CTEs are ∼0 · 999 990 but this depends on temperature and on the speed
or clocking at which the transfers are made. Rather than quoting the CTE, a figure of merit which is
important in calculating the noise associated with high precision photometry is the pixel readout noise.
Typical values for this are about ±10e − per pixel. As CCD technology has continued to improve, this
figure has been progressively reduced and is now generally lower than the photon shot noise associated
with the number of photons collected in each pixel.
A regular required calibration procedure arises because the individual pixels have differing
responses to unit intensity or, in other words, the quantum efficiency is not uniform over the pixel
array. This is addressed by subjecting the detector to a source of even illumination and, according to
the responses, all other records are adjusted by a direct ratio, pixel by pixel. This process is referred
to as flat-fielding and it is common practice to point the telescope with a CCD camera attached to
the twilight sky at the beginning and end of the nightly observations. In order for this process not to
introduce its own noise, it is important that the flat-field frames are repeated a large number of times to
obtain accurate values for the relative sensitivities of each pixel.
CCD detectors suffer also from background signals which accumulate during the exposure. For
short to medium exposures, these can be reduced effectively to zero by housing the chip within a
sealed unit and cooling it to liquid nitrogen temperatures. It is standard practice to record frames with
a shutter in the optical path so that any effects of the thermal background can be subtracted from any