Re: new Nikon scanner
Re: new Nikon scanner
- Subject: Re: new Nikon scanner
- From: Roberto Michelena <email@hidden>
- Date: Fri, 27 Apr 2001 13:21:23 -0500
You all have to take into account that transparency scanning is what
everything is geared to, not scanning of artificial samples. In that
context, many technologies besides actual sensor improvement can
influence the result. Most of them are relatively new.
Directly influencing capture, we have multi-sample averaging, CCD
cooling, sup-pixel movement, variable light output, etc.
Then in software there's grain reduction, sharpening algorithms, etc.;
they can influence in captured density detail too.
Totally agree that inflating specs is a sadly common practice; I wonder
if manufacturers even test their equipment to get a spec, or just figure
out a number based on what they suppose the new
parts/additions/technologies will amount to. I believe the latter.
Most CCD high-end scanners used to quote 3.7-3.9 DMax and 3.7 range, with
0-0.1 Dmin; among these all the Scitex (pre-Supreme) and such.
When they came out with better A/D convertors they tended to extend the
range or the Dmax (why!) or both but staying under 3.9-4.0.
Some went to say 4.0 Dmax by taking the glass and mirrors from the
equation (direct light-emulsion-lens-sensor paths); for example Purup's
(now ScanView) F series and Imacon's Flextight. Purup also cools the CCD
(with a peltier) I believe, and does multisampling.
Scitex was the first to claim 4.2 ; for that, they added cooling (that
they didn't have) and multisampling. But they didn't take out the glass.
Nikon claims 4.2 on more or less the same grounds. Cooling,
multi-sampling, and I believe variable light output too. In fact, VLO
combined with multisampling does have indeed the potential to improve
density (both min and max) and range a lot.
Imacon tried to do so in the FlexTight. They implemented and advertised
VLO in it, but kind of let the claims die because they were not able to
vary it more than a small amount. That rendered it useless, and combined
with the potential color temperature change (given the light source they
were using), better not to use it.
I did some extensive tests comparing a Scitex Eversmart Supreme with a
Purup-Eskofot ScanView F10 and with a Purup-Eskofot ScanMate 11000 (drum).
The Supreme had the most accepted color rendition in saturated and mid
(flesh) tones.
Both ScanMates had better neutrals (probably due to their IT8
calibration) but somewhat dull saturated colors. Nothing you can't
improve with better output ICC profiles, I believe.
In shadow detail, the drum did better. Specially the grain structure, it
was less "colorful" and thus less noticeable, and grain had softer edges.
Both the Supreme and the F10 had the best shadow detail I had seen in
CCDs; the Supreme had, again, better built-in automatic gamma which
showed more detail, but a quick manual gamma fix on the F10 brought it to
the same.
On both CCD scanners, if you didn't use "multisampling" then you got
christmas lights on the shadows. On both of them enabling multisampling
alleviated the problem almost completely, leaving only some noise, more
than in the drum but much less than in any other CCD I had seen.
I wonder how the Nikon will compare... I believe it'll be at least at the
level of previous (3-4 years ago) crop of high-end scanners, those that
claimed 3.7-3.9; and probably with less shadow noise. But color rendition
depends on software and separation tables mainly, and that seems still
unmastered by all the manufacturers that are not traditionally high-end.
best regards,
-- Roberto Michelena
EOS S.A.
Lima, Peru