--- Pat Perez <patdperez@yahoo.com> wrote:
> This is a wild-ass guess, but maybe memory at the
> byte level isn't being
> accessed or allocated or released properly, and what
> appears as a band is
> the result of regular 'overflows'.
I don't think that is the problem. If there would be
overflow you would see completely wrong values as the
MSB will be cut off. If the error accumulates over
multiple pixels until it overflows the pixel values
would gradually increase and then fall a lot. If
memory is not properly accessed you probably would get
an assertion or at least similar errors in all pixels
(assuming you write only for example 8-bits in a
12-bit word with the 4 LSBs not initialized), etc.
Unfortunately, I do not have the email with the scan
anymore but it seemed to me that the banding happens
at constant pixel spacing. Therefore, I do not believe
that it is a problem with the CCD itself because it's
quite unlikely that the sensors are bad in a equal
spacing. One thing I could imagine is the amplifier.
In order to reduce noise due to fast read-out times
and to allow somewhat faster scanning there might be
more then 1 amplifier per CCD line. Assuming they use
32 amplifiers, i.e. pixel x goes to amplifier 'x MOD
32' and assuming that the gain for one of these
amplifiers is off then you would see such banding.
That's a pretty wild guess, though.
The original poster said that he saw the banding only
when adjustment were done. Have all other parameters
been the same? For example I have heard some issues
with multi-scanning on Polaroid scanners which could
lead to soft images. If I remember correctly Nikon
scanners have some HW support for multi-scanning. So
instead of soft images an artifact could be banding.
Hint: Wild guess!!!
One thing you could check is if the banding always
happens at the same place. For example do a scan of a
picture that has some clear sharp lines. Scan it and
record where the banding happens relative to this
line. Repeat it to check for consistency. Then move
the picture to be scanned a little bit within the
holder. If the banding does change relative to the
line it is quite likely a HW issue. Otherwise it could
be, but doesn't have to be, a SW issue.
Robert
__________________________________________________
Do You Yahoo!?
Get personalized email addresses from Yahoo! Mail
