Then writes...
The error seems to accumulate,
since the lines gain strength from left to right.
Then writes...
The lines are only a
problem in long >30min dark slide scans with very long exposure.
Sorry for hacking your quotes about Markus, but when these three
statements are presented together the problem you are reporting comes
into perspective.
I do not believe that you have a faulty Minolta scanner, so I do not
believe that replacing the scanner will eliminate this problem. In
fact, I would hazard the view that every user of the Minolta (and most,
but perhaps not all, other desktop film scanners) will experience
similar problems under these conditions.
As Fernando explained, each cell in the CCD has a unique dark current,
which is the current that flows into the storage well even when no light
is present. When the CCD is illuminated by an image, the photocurrent
generated by each cell flows into the storage well together with this
dark current. The resultant total charge packet accumulated in each
storage well is then transported to the device output, where it is
dumped onto a small capacitor to produce an output voltage sequence
proportional to the total packet of charge accumulated at each cell
during the exposure.
This raw voltage sequence, once digitized, would produce a very poor
image indeed, since the lowest levels would be determined by the dark
currents in the cells, whilst the contrast would be determined by the
photoresponse, which is also unique to each cell. Consequently the
scanner goes through a calibration step, as Fernando described, to
determine how the CCD responds to no light (measuring the dark current
in each cell) and full light (the difference between this and the no
light condition being a measure of the photoresponse). This data is
then used to compute a clean image from the raw data.
However, there are a couple of problems that arise.
1. The "calibration" is only accurate for the particular exposure used.
2. The dark current is a strong function of the CCD temperature.
There are other problems, such as the stability of the dark current,
even under a controlled, fixed temperature - so called 1/f noise, but we
can ignore these for the moment, since you are reporting something that
appears to be driven by these first two problems.
So far as problem 1 is concerned, the question is what exposure is used
in the calibration stage, and is it the same as the long exposure used
with the very dark slides you are experiencing the problem with? I
suspect it isn't - and consequently under these dark slide conditions,
the dark calibration is probably in error.
I don't have a Minolta scanner so I can't try this, but in another
thread Bart mentioned that you can force a calibration by using
Ctrl-Shift-I, to re-initialise the scanner. It may be possible, you
will have to check, to lock the exposure to that required for your dark
scans by auto-exposing on them first and then switching off the
autoexposure function. Re-initialising the scanner *may* then perform a
calibration with the same exposure as you are using for the scan and
thus create a more accurate dark current calibration reference.
I emphasise that I don't have a Minolta to try this with and frankly,
while worth trying, it really is grasping at straws! This is because
the calibration sequence also includes a response measurement, which
requires an exposure with the scanner illumination source on. With a
long exposure, this is likely to saturate the CCD. The Minolta, like
many other scanners, uses a cold cathode light source and this cannot
switch on an off fast enough (ie. in less time than the actual exposure
itself) to prevent saturation with long exposures.
It is one of the advantages of the Nikon series (and the newly announced
Mk-II Minolta) that they use LED light sources, which can be switched
very fast indeed, so such calibration under representative exposure
conditions is possible. I wouldn't like to suggest that this is why
Minolta have changed to an LED source on the new scanner, but it may be
a contributory factor.
So the best solution may actually be to lock the exposure with a normal,
*light* slide in place and then use that exposure to scan the dark slide
- correcting the density afterwards in the digital image, with all that
this entails in terms of increased noise etc.
So far as the second problem is concerned, I suspect that this is why
the intensity of your lines change across the image - 30 minutes is a
very long scan time to keep the CCD temperature stable! In a
professional imaging system, the CCD would be mounted on a
thermo-electric cooler, typically a Peltier device, not only reducing
the dark current dramatically but, using a very precise control system
to maintain the device temperature over long exposures, stabilising the
dark current. None of the desktop film scanners on the market have such
devices though, and rely on the device temperature remaining stable.
Roughly speaking, the dark current in a CCD cell increases by about a
factor of 2 every 7 or 8 degrees Celsius, so it doesn't take much
temperature change on the CCD to cause a significant change in dark
current. Fortunately, dark current is generally quite low in normal
usage and its change with temperature isn't too much of a problem during
the scan time. However, in dark scans with long exposure, it
contributes a much greater proportion to the total charge packet than it
would in normal use, so the change in dark current as a result of device
temperature change are consequently more significant also.
Since we are talking about device temperature here, rather than package
or case temperature, the main cause of temperature change during the
scan time is actually the power dissipation of the CCD itself. Usually,
the change in temperature is limited by operating the CCD for a couple
of seconds prior to beginning the actual scan, however this may not be
adequate for long exposure runs. So one suggestion to minimise the
problem, although you are unlikely to eliminate it entirely, is to start
a scan and let it run for, say, 5 minutes or so, then cancel and
immediately restart the scan. This may help to get stabilise the CCD
temperature better than the few seconds that the scanner uses for normal
exposures.
Summing up then, there are a couple of things you can try to alleviate
the problem, but very little that will eliminate it. I suspect you are
pushing the device quite a lot beyond its original design envelope. Not
much help in resolving your problem other than perhaps understanding the
issues better, which may lead you to avoid the situation in future.
.
