Howdy, folks,
Well, I finally got around to getting a film scanner. My plan is to
get my (primarily color negative) film developed only, then scan it and
do any printing I want on the inkjet printer. This will pay for the
scanner in two or three racing events, best case, or four or five,
worst case.
THE SCANNER
-----------
The scanner is a Canon CanoScan FS4000US. It does both USB and SCSI,
and provides either 8-bit or 14-bit color at 4,000 dpi. It handles
35mm slides and negatives, and comes with an APS handler in the package
(rather than a cost-added option). It also comes with a USB cable and
a warning not to plug it in through a USB hub.
The 35mm negative carrier holds strips up to 6 frames long; the 35mm
slide carrier holds four slides; the APS holder, of course, takes a
developed APS roll. The FilmGetFS and VueScan software can both batch
process all of the 35mm frames and up to 40 frames on a roll of APS
film.
This scanner includes IR dust removal, which Canon calls FARE. It
works very well, but I haven't compared it to the Nikon version (ICE?).
It has both USB and SCSI interfaces. I'm using USB, at least for the
moment (more below).
Scanning at full 4000 dpi resolution takes a _long_ time and results in
120-135 MB of image data. Turn on FARE and it takes even longer. On
my system, it takes about ten minutes to scan a frame at full
resolution, 42-bit color, with FARE enabled. Some of this is inherent
to the high resolution of the output, but see the notes below about my
system and throughput.
THE COMPUTER
------------
My Windows system is a 733 MHz Pentium III with a 133 MHz FSB, 256 MB
of RAM, and ATA-66 drives. I've already ordered extra memory to max my
motherboard out at 512 MB, and that may still not be optimal. 1 GB or
more would make life easier. Depending on PhotoShop's appetite, I may
end up upgrading my motherboard or computer so 1GB or more is possible.
I'll probably switch to using the SCSI interface before it's all over
with to get the image transfer speed up coming from the scanner to the
computer. I may also end up switching to SCSI hard drives to increase
throughput between the computer and drives. I need to research
high-speed A/V drives, which are optimized for long sequential
read/write operations (like video), which may be able to further
enhance throughput for the image data files.
It will be interesting to see where the scanning speed ends up. It's
obvious on my current system that the swapper is thrashing it's brains
out when I've got more than one of these images in memory. It's also
obvious that these drives take a while to write that much data. I've
still got to defrag the drives, though, to give them a fair appraisal.
THE SOFTWARE
------------
I'm mostly using the FilmGetFS interface that came with the scanner,
because I've had problems with VueScan. Basically, the images all come
out with washed out colors and an overall light sepia tone. I'll be
contacting Hamrick, and I'll expand on this later, when there's
something useful to pass on.
NB. I've figured out that my problems with VueScan came
from improper calibration. The first time I tried
VueScan, I had already tried FilmGetFS, so the
negative carrier had a strip of negatives in it
already. When I hit the "preview" button on
VueScan, it said it couldn't find the calibration
file and asked if I'd like to calibrate the scanner.
I said "Sure". So, it proceeded to calibrate itself
on the negative in the first frame position in the
negative carrier. It didn't warn me, and I didn't
consider, that it had to be calibrated on an empty
negative carrier. :-|
So, now I'm getting much more reasonable scans from
VueScan. It still doesn't seem to do as good a job
as FilmGetFS (colors seem washed out), but I haven't
done in depth tests yet. When I get around to it,
I'll report back, since VueScan seems like it has a
better work flow for my tastes.
I capture the images into PhotoShop using the "Import" menu item on the
"File" menu, and selecting the Canon scanner. This invokes the
FilmGetFS "program" provided with the scanner. As far as I can tell,
there's no way to use FilmGetFS without some sort of graphics program
driving it (presumably via TWAIN). If it could be run standalone and
scan to a disk file, the pain of memory use during batch scanning could
be greatly reduced.
VueScan can do batch processing directly to files on the hard drive.
This, by itself, should greatly reduce the memory needs while batch
scanning proceeds. It doesn't do much for PhotoShop's needs while
editing the images, though. If it does, I'll do detailed comparison of
scan quality between VueScan and FilmGetFS, and maybe switch.
THE EXPERIENCE
--------------
I'm finding that my negatives contain a lot of data that never made it
onto the print. Some was lost in the paper's dynamic range. Some was
lost in the extra layer of optics and focusing inaccuracy in the
minilab. Some was lost in the minilab operator's choices of
filtration. Some was probably lost other places. But it was gone in
the scans of 4" x 6" prints.
Plus, there's the additional resolution, color and spatial. The
flatbed scanner only does 24-bit color, while the film scanner does
42-bit color (as 48-bit image data).
Furthermore, even scanning a 4" x 6" print at 600 dpi results in an
image with about as many pixels as a 2,400 dpi scan of the negative, so
the 4,000 dpi of the film scanner is a substantial increase, spatially.
And that 600 dpi scan is trying to capture four to nine times the
amount of information the paper actually holds (200 to 300 dpi is the
conventional wisdom).
Using the "auto levels and colors" capability of FilmGetFS results in
"Velvia-ish" oversaturated images. Leaving it set to "flat" color
response results in scans that (usually) closely match my memory of the
event and the ("NNNNN") prints from the minilab.
The ones that don't match always turn out to be ones where there had to
be a good bit of filtration in the minilab to get the print they sold
me. IOW, they're on the money, but I'm going to have to do the work in
PhotoShop that the minilab or its operator did when creating the print
from the negative.
Oddly enough, the FilmGetFS previews don't seem as sharp as the images
TWAIN-ed into PhotoShop. I guess there's something "wrong" with the
viewing code in FilmGetFS. The bottom line, though, is that you can't
see how sharp the image is until you get the full scan over into
PhotoShop.
I'm seeing some signs of what I believe to be grain aliasing. So far,
I'm only noticing it in large areas of light gray. Large areas of
other colors seem OK. And it seems to be mostly on certain types of
film, in particular the Kodak GOLD series. I don't see as much of it
in scans from Royal Gold negatives. I've got some Fuji Superia and NHG
400, and Kodak Portra 400 NC, negs coming up in the next few days of
scanning and we'll see how they do.
OPEN QUESTIONS
--------------
1) Does anyone have suggestions for "PhotoShopping" away the artifacts
of "grain aliasing"? In my experiments, it looks like a Median Blur
with a radius of about four pixels does a decent job of smoothing
things out, but doesn't get rid of the "speckles". It does, though,
make the image easier for USM to handle. It helps reduce emphasizing
the speckles, and it seems to give USM data it "likes better" for the
actual edges in the image. I'm really wondering if there's a way to
clone the area into a new layer, then combine the layer with the
original layer to get rid of them or knock down their intensity or
something.
2) Does anyone know of any resources on scanning and printing accuracy
in relation to the solidity of the mechanical support of the scanner or
printer. IOW, we all know about blurry pictures and wobbly tripods ...
but how do wobbly tables play into the accuracy we're getting from our
scanners and printers?
I once interviewed with Intergraph, who would have put me on a mapping
project. The maps were to be scanned on a 72" bed scanner at something
like 1,000 dpi. The scanner itself weighed something like a ton,
largely ballast and two huge tables. In addition, the spec called for
it to be mounted to a free floated block of concrete weighing over 20
tons. They said it was to maintain the spatial calibration of the
scans regardless of the vibration of the ground, etc.
TTYL, DougF KG4LMZ
-
This message is from the Pentax-Discuss Mail List. To unsubscribe,
go to http://www.pdml.net and follow the directions. Don't forget to
visit the Pentax Users' Gallery at http://pug.komkon.org .
