Re: Film scanner experiment & questions

["Phil.Lippincott" <phil.lippincott@xxxxxxxxxxxxxxxx>]

Gary,
   Your testing is very reasonable to experiences I've had testing other
scanners. Your procedure and findings are totally reasonable.   I have used the
Kodak ST-34 target with ISO Status A density values printed on the original for
density testing.   My Nikon 8000ED  testing showed (in fine mode) it starting to
create a lot of noise at about Status A density of 2.6 and go up to 3.0 with
nothing but noise.  My POlaroid SprintScan 45 testing indicated it lost density
at 2.3 with noise to 2.6.   As my Nikon testing has not included the SC4000 nor
Polaroid the 4000;  I would like to consider being able to reference your
results.  Could you please more concisely and definitively restate your findings
and confirm (the detail source) of your stated test film density values.    Also
do you have a means to provide a comparable optical resolution finding like a
scan of the 1951 U.S. Air Force target showing visable line pairs per millimeter
and therefore optical resolvable resolution unlike PPI or dpi resolution.

  On a separate note per your request. The SpecsManShip of scanners and scanner
software runs deeper than most people suspect.  For example just because a
manufacture claims and / or there maybe a 12 bit Analog to Digital converter
doesn't mean that the software takes advantage of it even if you can get 16 bit
tiff out of the scanner.  The output data stream can be padded from 8 to 16 bit
or 12 to 16 bits with Zeros.  The real data is based upon both the particulars of
the look up table data formatting and the raw scanner output data formatting both
under the control of the scanner software not the hardware.  So scanner software
(No Particular Names Mentioned) which commonly supports lots of scanners can
easily do so by capturing the data either in the lowest common denominator 24 bit
mode for all scanners; or even raw data only and do ICC color bending from a
single input scanner specific profile to various output common correction per
film type.  Both of these approaches are throwing valuable and irrecoverable
color fidelity away and saves a lot of time dealing with the specific issues of a
particular scanner and how to control it's strengths and weaknesses.  One very
interesting example of the above is a famous very high end drum scanner where the
Dmax is reduced to only 3.0 because all scans are converted via ICC profiles on
capture to CIE Lab for subsequent output to either CMYK or back to RGB. This
profile to profile conversion itself limits the results because it filter all
scans be no better than the best media specific profiles.  The is a simple
example where the scanner is better than available interface.  The reverse can of
course also be possable.
  So I strongly suspect that your Polaroid versus Nikon comparison are both
software control method comparisons and hardware.  Not only 8, 12 or 16 A to D
converter comparisons.  I hope this helps.  I look forward to hearing from you
more specifics abourt your testing.

Best Regards,
Phil Lippincott
Aztek

"Gary L. Hunt" wrote:

> Finally got around to trying to compare my Polaroid Sprintscan
> 4000 with my new Nikon SC4000.  I don't claim these results
> mean a lot--if anything, I have more questions now than
> before.  But I scanned a 21-step step wedge target (0.15 density
> steps) negative in RGB mode on both scanners.  I varied the
> exposure setting on the Polaroid and the analog gain on the
> Nikon for 8-bit scans [to be sure I wasn't clipping the "highlights"],
> and also scanned in hi-bit mode for both scanners.  I had the
> vague idea this would at least let me use Photoshop to compare
> the usable density range of the two scanners, and I guess it does,
> but not so easily as one might think.
>
> The target has a total density range of 3.0 (i.e. 0.05 to 3.05),
> which is within the manufacturer's claims for "dynamic range"
> for both scanners.  (Not that that means anything, as we have
> been brought to know.)  And in fact neither scanner approaches
> being able to distinguish the full range of the target even in high
> bit mode.  The Nikon comes closer--by expanding the dense
> regions with PS levels, I can separate the steps to about level
> 18 (19 in the green channel), although only by sampling them--the
> histograms don't show distinct peaks beyond level 14 due to noise.
> The 8-bit steps are detectable to maybe level 16 (17 in the green
> channel), consistent with 8-bit resolution.  But even the 14-bit scan
> only covers a density range of about 2.55-2.7.
>
> The Polaroid numbers don't actually come out much different at
> level 16/17 for 8-bit scans, but the 12-bit scan (where no scanning
> adjustments are possible with Polacolor Insight) is extremely
> compressed; nothing beyond level 14 or so (density 2.0) is detectable.
> The change in step "density" is very pronounced in this mode, which
> probably explains why my high-bit scans always looked so overly
> dark.  In addition, the denser steps are very noisy--they look really
> grainy.  (Actually they look this way for all the steps in this mode.)
> Maybe my particular sample isn't working right, but it sure looks to
> me like the 12-bit mode has less information than the 8-bit mode.
>
> The Nikon scans are all much "smoother" (less grainy or noisy) looking.
> (I used NikonScan 3.1 with all the processing options turned off.)  I also
> made a scan with VueScan, which appears visually to have a very slightly
> greater range in 8-bit mode.  This scan is smoother-looking yet, although
> that could be due to some default setting in VueScan that I didn't notice.
>
> Now for the questions:
> (1) can anyone tell whether the density ranges I am getting with these 2
>      scanners are in the ranges one might expect if everything was set right?
>      (Never mind the specs--I knew they were a pipe dream.)
> (2) if not, is there something obvious I am missing in doing the scans?
> (3) can someone suggest a reason why my 12-bit Polaroid scans are
>      so compressed?
> (4) any idea why the steps on the Polaroid scans look so much more grainy
>      than the Nikon ones, even for relatively low density steps?
>
> Gary Hunt <glh@datawav.net>
>
> -
> Turn off HTML mail features. Keep quoted material short. Use accurate
> subject lines. http://www.leben.com/lists for list instructions.

-
Turn off HTML mail features. Keep quoted material short. Use accurate
subject lines. http://www.leben.com/lists for list instructions.