Residual Bulk Image
RBI effect is inherent to the CCD detector itself and it has
nothing to do with any particular camera model. It appears on
Front-Illuminated Full-Frame detectors (do not confuse it with
full-frame 24 × 36 mm detector format used in classical
photography, Full-Frame denotes architecture of the CCD regardless
of its physical dimensions). So-called Back-Illuminated (or
thinned) CCD chips do not suffer from this effect and RBI is
suppressed during CCD clear operation in the case of
Interline-Transfer CCDs. Unfortunately, Back-Illuminated CCD
detectors are prohibitively expensive compared to
Front-Illuminated variants (often by several orders of magnitude),
so they are out of reach for many applications. Interline-Transfer
detectors on the other hand do not offer high quantum efficiency
and suffer from higher dark current, while their advantages
(electronic shuttering, which allows very short exposures) bring
no use in astronomical applications, where exposure times are
quite long. So Front-Illuminated detectors are widely spread and
its RBI effect must be taken into account.
The basic principles of how CCD detectors work are generally
known—photons incoming to silicon chip
release electrons, which are accumulated in the potential wells
(areas of the chip, surrounded by negative walls, which do not
allow negative electrons to escape) during exposure. Such
potential wells are also called pixels. Individual packets of
electrons are then moved around the CCD (shifted to neighboring
potential wells) during the detector readout until they are
finally moved to the output node, which converts charge to
voltage. Output voltage is then digitized by camera electronics
and a number corresponding to each pixels charge is sent to the
But situation is always more complicated in reality compared to
idealized model. Not all photons are captured within pixels, some
of them (especially the ones with longer wavelengths) penetrate
deep into the chip substrate, where they also release electrons.
Majority of these electrons are drained by the substrate, but some
of them are captured in charge traps, which appeared in the CCD
substrate during the CCD manufacturing. The charge accumulated in
these traps slowly disappears, unfortunately some of the electrons
move back to the image pixels. So if the CCD is read (even without
the access of light), a signal is observable on the areas, where
electrons remained in the charge traps.
Let us take for example RBI effect of the KAF-16803 CCD
detector. The following image shows a portion of 180 s long dark frame of this CCD, taken at
Portion of correct KAF-16803 CCD dark frame
This dark frame was acquired before the observing session,
while waiting for complete darkness. The camera was kept in the
dark (shutter closed) for approx. an hour prior to dark frame
exposures. Dark frame exposures were started so they finished in
time for actual observation begins.
The following dark frame was acquired on the end of observing
session, after many hours of imaging of the same field of view.
Although it was taken at the same temperature and with the same
exposure time, it shows higher level of dark current (frame mean
value is higher) and also contains clear traces of stars. The
corresponding portion of the star field is shown prior to the dark
frame for comparison.
A portion of the star field (left) and a dark
frame acquired immediately after light frame series
RBI effect is clearly visible on the dark frame. Slightly
higher background level corresponds to weak RBI effect caused by
sky background. But the number of electrons trapped in substrate
is much higher in the areas where the light from bright stars hit
the CCD, which results to star ghost images.
How prominent is RBI effect? Even if the image is stretched so
the RBI is well visible, especially in our case of KAF-16803 the
mean value in the brightest ghost star images is only 14
ADU above the background level on 180 s long dark frame. This is approx.
1.5 times the detector read noise RMS. We can say
without hesitation that the RBI is more an aesthetic problem than
a real issue for photometry or astrometry in this case.
Even more prominent is the RBI after the flat fields were
acquired. The whole CCD is often saturated many times when
acquiring flat fields (especially when we wait for proper sky
brightness at twilight—number of images
are saturated before the flat field brightness is around 1/2 of
the CCD dynamic range). An example of such flat field and
subsequently acquired dark frame are on the following images:
Typical flat field image (left) and immediately
acquired dark frame (right)
The dark frame taken immediately after the flat frames were
exposed shows RBI effect uniformly all over the detector area.
Clearly visible arc structures demonstrate intensity of charge
traps within the CCD substrate.
Let us summarize the RBI effect features:
RBI is inherent to Front-Illuminated Full-Frame CCD chips
and cannot be affected by camera construction. Of course camera
design may be so bad, that image noise hides the RBI effect. But
this can hardly be a desired goal and cannot be considered a
solution of this problem.
RBI effect intensity depends on the particular CCD
detector. Generally CCDs with bigger pixels suffer from more
intensive RBI effect.
The charge captured in traps slowly disappears and the
shadow image fades. The speed of fading again depends on the
particular CCD type as well as on the detector
Unfortunately with the lower temperature the
speed of RBI fading also slows.
Because the charge is trapped under the pixel structure,
the speed of RBI fading is not affected by CCD read/clear
operations. However, the CCD must be of course kept in dark if
the RBI has to disappear.
How much the RBI effect affects observations? This depends on
specific observation program and on the possibilities of its
adaptation. It also depends on particular camera model (the speed
of RBI fading) and on the time available for waiting for RBI to
The experience shows that when a single field of view is imaged
and the observing program is adopted (particularly dark frames are
captured after some time during which the camera is kept in dark),
RBI does not affect observations at all. The precision of
photometric measurements is not compromised if the image does not
move around the CCD detector too fast (exposures are guided). As
already stated, intensity of RBI on many minutes long exposures
are only a few electrons (or tens of electrons) per pixel, which
is typically well below the influence of other disturbing effects
(e.g. high clouds) and measurement uncertainties.
Similar situation is in the case of astrophotography—RBI does not affect observations if the single
object is imaged and exposures are properly guided. Even if the
filters are changed, traces of RBI from previous filter are not
detectable. When we take into account that very long exposures are
the trend in astrophotography (everybody wants the best images
with very high S/N ratio), moving from one object to another
during single night is not so common.
In situations when RBI is really unwanted and disturbing (e.g.
objects like M31 galaxy and almost empty variable star
fields are on the observing list), it is necessary to deal with
the RBI caused shadow images.
The first possibility is naturally t wait for RBI to fade. When
a camera with small pixels (~7 μm) is used, it is enough
to wait for 10 to 20 minutes and RBI disappears.
Because the speed of RBI fading increases with
temperature, it is possible to raise the detector temperature to
e.g. 0°C during wait and then to cool it to working temperature
again. Because the dependency is exponential, RBI fades much
faster at higher temperatures.
Another alternative to deal with RBI is so-called NIR
preflash. In fact this method does not remove RBI, but the
opposite is true—it creates the RBI
uniformly all over detector before each exposure (something like
if you cannot defeat the enemy, make it ally). Uniformly
does not mean that the RBI intensity is the same over the entire
detector area. Different parts of the chip substrate are capable
to accommodate different amount of charge, which makes the RBI
intensity to vary. Arc-like structures are typical, they are
related to production process of the silicon monocrystal, from
which the CCD is made. These structures are well illustrated on
the dark frame captured after the flat field (see image above),
which means after the complete detector is saturated.
The whole CCD is flooded with light during Preflash, so its
substrate is completely saturated. It is better to use near
infra-red light (this is why NIR Preflash abbreviation is
used), which penetrates to the substrate more easily. The detector
is cleared after flooding with light to remove electrons from
active detector areas (image pixels, horizontal register, output
node, ...) so the regular exposure can be started. Because all
portions of the CCD were completely saturated, CCD clear operation
must be repeated several times, one clear is not enough.
If the above described procedure (saturation + repeated clear)
is performed before every single exposure (dark, light, flat),
initial conditions are always the same and calibration (dark frame
subtraction) eliminates RBI. Ghost images are not visible, any
residual charge left after previous exposure is overrun during
light flooding to maximal value and removed during subsequent dark
The fact the NIR Preflash electronics is present in the camera
does not mean it must be used. This function is programmatically
controlled and it is enough to define 0 s Preflash time and camera hardware will
not perform light flooding (and subsequent clear operations).
NIR Preflash control is available in the Exposure
tab of the SIPS CCD Camera tool window.
Preflash control in CCD Camera tool of the
SIPS software package
Optimal values of both parameters depend on specific model, CCD
temperature etc. NIR LED within the Gx cameras used for light
flooding are powerful enough to saturate the entire detector after
a fraction of second. So Preflash time 2 to 3 s is enough with wide margin. Number of
clears should be at last 2×, but
specific number is again necessary to adjust, 3 to 4 clears still
changes the intensity of dark frames acquired after Preflash.
It is necessary to take into account time delays when NIR
Preflash is used. Especially in the case of small cameras
(G2-0402, G2-1600, ...) with digitization time from a fraction of
second to several seconds, the delay of several seconds for
Preflash and subsequent clear operations significantly increases
interval between exposures.
NIR Preflash is also supported in Gx CCD camera drivers for
other software packages.
NIR Preflash support in configuration dialog boxes of
general ASCOM driver (left) and MaxIm DL specific driver
Using of NIR Preflash brings also other disadvantages in
addition to increased delays among exposures. The charge
accumulated in the chip substrate slowly dissipates and partially
moves into image pixels as added noise. This added signal behaves
very similarly to dark current, so we can say that the dark
current increases after IR Preflash, when the substrate is
saturated with electrons, compared to dark current of the detector
with empty substrate. Even after the dark frame is
subtracted, higher dark current always means higher background
noise. CCD signal RMS corresponds to square root of the signal
level, so higher background signal also means higher deviations,
which are not removed by calibration. The solution is the same
like in the case of natural dark current. If NIR Preflash is used,
CCD must be cooled to limit dark current as much as possible. This
is why not using of NIR Preflash when it is possible and e.g. to
adopt the observing program is a better option.
NIR Preflash control is implemented in SIPS v2.1 and newer
versions. Older versions do not support NIR Preflash.
NIR Preflash support in drivers for other software packages is
stated in the particular driver documentation in the History of
changes section. It is recommended to download the latest version of
the used driver from this www site.
RBI overview is well described in Richard Crisp article
Residual Bulk Image: Cause and Cure, published in Sky and
Telescope, May 2011, pages 72-75.
We thank to Richard Crisp for advices and consultations regarding