Apache-Talk @lexa.ru 

Inet-Admins @info.east.ru 

Filmscanners @halftone.co.uk 

Security-alerts @yandex-team.ru 

nginx-ru @sysoev.ru 




      :: Filmscanners
Filmscanners mailing list archive (filmscanners@halftone.co.uk)

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

[filmscanners] Re: FINALLY, why Nikon LS-8000 bands

About a month ago, I had a eureka moment as a result of another
discussion which seemed to have little interchange at the time.  It was
in regard to something refereed to as "channel misalignment", which
manifested as a shift between the three color channel separations by as
much as a pixel or more.  The discussion revolved around high end medium
format scanners, but of course was a possible problem with 35mm scanners
as well.  The reason for it to be more likely in medium format was due
to the size of the film, and the fact that in most scanners the image
information at any one line across a frame is recorded at a different
time and film position for each channel.  (see diagram "A" below).

================= >>>film moving >>>>>>>>
   ^     ^     ^    <Stationary tri-line sensor (cross section)
RED  GREEN  BLUE  < filter color on each sensor line

As a result, should the film frame distort in any manner during
movement, (or even should the stepper motor or film carriage slip or
provide uneven stepping) by the time the "blue" information is captured,
for instance, these factors could cause the information to be misaligned
relative to one or both of the other channels, causing this "channel
misalignment".  Further, it could be variable, meaning there would be no
way to correct it via a channel shift in something like Photoshop.

At the time, someone mentioned that they wanted to test the LS-8000 with
superfine scan mode to see if this problem existed with it.

I started to answer that posting that you could not test the LS-8000
for channel shift with the superfine mode, because that mode used only
one line of the sensor, so no channel shifting could occur, when it
suddenly (like lighting) struck me that you couldn't check the LS-8000
for channel shifting under any circumstances because it doesn't work
like the other scanners.

Unlike all the non-Nikon scanners, the LS-8000, even when using all
three lines of the tri-line sensors, doesn't compose the image in the
same manner.  It uses it's sequencing LED light source to make the three
color separations, during which time the three lines capture the info in
three monochrome separations, and the film does not move until the next
capture or all three separations.  Therefore, the information for all
three color channels is exactly superimposed upon itself by the same
exact sensor line in the same exact position.  In other words, the R, G,
and B are all captured by the same CCD sensor line at the same time for
the same image data.  Since the three sensor lines are spaced by a
certain number of "pixels" these captures are then interleaved together
to make the full image. (see diagram "B" below)

============== >> film movement >>>>>>>>>>>>>>>
    ^   ^   ^    triline sensor lines (cross section)
    R   R   R   < capture color due to LED color in use
    G   G   G   < capture color due to LED color in use
    B   B   B   < capture color due to LED color in use

This, in theory, is a great design, as it eliminates the channel shift
problem, as it is recording all the information from one "line" of image
data at the same time by the same CCD sensor line at the same film
position without any movement of the stepper motor or the film.  IT IS

Other scanners have three sensors each capturing only one color channel,
at different locations in the film movement (see diagram "A" again)
This means that slight amounts of shifting of stepper motor, film
expansion or distortion, carrier slippage, etc, get hidden as the three
channels are superimposed, since even if a slight gap or overlapping
were to occur it would be hidden by the information from the other two
color channels.  Same for slight variations in calibration errors
of individual sensors or even if one full tri-line showed a slight
variation of sensitivity or response.

Now look at diagram "B" again.  In the case of the Nikon LS-8000
ANY variation in those factors (stepper motor, film dimensional change
or distortion, carriage slippage, or either individual sensor
calibration variation, and certainly if one sensor line was calibrated
or responded differently than another) would be AMPLIFIED by three.  Any
gap or overlap would be identical on all three color channels. ANY
variation in a sensor or sensor line would by visible on ALL THREE
channels in the exact same location on the image.

In other words, if it is true that the Nikon LS-8000 only calibrates one
of the triline sensor, unlike if this were the case it the other designs
used by all other CCD sensor companies, which would show up as one full
color channel being lighter or darker or variation between sensors on
each line, on the Nikon LS-8000, every scan line captured with that
sensor line would be lighter or darker for ALL THREE color channels in
the exact same place.  Further, since Nikon's LS-8000 uses an interlace
or interleaving capture method, when all three sensor lines are used in
the mix for all three color channels, this banding would be very regular
in spacing.

Finally, I realize why there is NO FIX, for this problem.  The ONLY
answer is to use one sensor line, because then the sensor line
calibration differences are eliminated, and any variation in problems
with mechanical spacing is likely to be lessened because the stepper
motor is slowed down to steps one third the distance, reducing
sloppiness in the mechanics (it probably helps to eliminate kickback and
other factors as well by slowing down the movement).

So, in conclusion, I don't think necessarily the other scanners don't
have some of the same mechanical problems the Nikon manifests, it is
just that the design of the other scanners tends to HIDE these problems
by the way they capture and reconstruct the image, while, in fact, the
Nikon LS-8000 would drastically EMPHASIZE these defects by its design
and manner of reconstruction the image.

In a perfect world, where the film didn't distort, or change
dimensionally with temperature, where no vibration existed in mechanical
devices, where stepper motors could move objects rapidly and perfectly
precisely, and where no slippage between surfaces occurred when desired,
and where all sensors and sensor lines responded identically, the Nikon
design would be far superior to the others for capturing images.

Unfortunately, so far, at any cost approaching reasonable, these factors
apparently cannot be designed out of scanners yet, and Nikon, it would
appear, did not factor these physical limitations into the LS-8000.  In
fact, this banding problem showed up long before the LS-8000, in some of
the LS-2000 models.  I suspect it was further aggravated by the size of
the structures needed in the LS-8000 to handle larger film sizes.  I
also suspect this banding problem will become more noticeable with aging
of the device, as tolerances get sloppier still.

And, now, I open this for discussion and/or debate on what I have
brought up here.  Remember, as some others are so apt to point out, I am
NOT an engineer (thank goodness) and I do not own a LS-8000 (thank

I am UNWILLING to respond to:

My motives for posting this
My love or hatred for Nikon as a company
My intelligence or lack thereof
If the word "prat" should be banned from this list
Why my ASCII diagrams aren't animated.
If I knew that my last name translates from higher German to "Eureka!"
or "Enlightenment".



Unsubscribe by mail to listserver@halftone.co.uk, with 'unsubscribe 
or 'unsubscribe filmscanners_digest' (as appropriate) in the message title or 


Copyright © Lexa Software, 1996-2009.