Flatbeds max out at 2000dpi for negs...??

brian wrote:

"Ivan" wrote in message ...
Did I read it here recently that regardless of the claimed max resolution of
high end flatbed scanners ie. Epson 4870, that when scanning negs the max
resolution is 2000dpi because of some issue with the glass?? Is this true?
If so then it stands to reason that MF is the logical choice of format
because it's the largest format that can be scanned on a dedicated film
scanner. A bigger format than MF will have to be scanned on a flatbed
anyways right? So unless you're shooting in a very large format (6x9
inches) you won't gain anything by scanning on a flatbed. Am I right? If
so then it's one more reason for me to invest in a few choice lenses for my
rb67.
Ivan

The reason many flatbeds come up short of their claimed resolution is
that the scanning lens isn't good enough. Manufacturers will often
install a higher density CCD without changing the lens.

Brian
www.caldwellphotographic.com

Doesn't focusing distance count? Flatbeds are supposed to scan both from
the surface of the glass (opaque targets) and a little elevated
(transparencies). In high resolution scans they should have issues with
depth of focus.

-- Lassi

"brian" wrote in message
m...

The reason many flatbeds come up short of their claimed resolution is
that the scanning lens isn't good enough. Manufacturers will often
install a higher density CCD without changing the lens.

So all we get is more noise, correct?

jjs wrote:

So all we get is more noise, correct?

No, since the electronic noise is not limited by the optical path you
get to see it at a much higher resolution. Now, isn't this progress?

Ralf

--
Ralf R. Radermacher - DL9KCG - Köln/Cologne, Germany
private homepage: http://www.fotoralf.de
manual cameras and photo galleries - updated April 29, 2004
Contarex - Kiev 60 - Horizon 202 - P6 mount lenses

jjs wrote:

So all we get is more noise, correct?

No, since the electronic noise is not limited by the optical path you
get to see it at a much higher resolution. Now, isn't this progress?

Ralf

--
Ralf R. Radermacher - DL9KCG - Köln/Cologne, Germany
private homepage: http://www.fotoralf.de
manual cameras and photo galleries - updated April 29, 2004
Contarex - Kiev 60 - Horizon 202 - P6 mount lenses

"Ralf R. Radermacher" wrote in message
...
jjs wrote:

So all we get is more noise, correct?

No, since the electronic noise is not limited by the optical path you
get to see it at a much higher resolution. Now, isn't this progress?

Great!

On Sun, 8 Aug 2004 23:14:41 -0600, "Ivan"
wrote:

Did I read it here recently that regardless of the claimed max resolution of
high end flatbed scanners ie. Epson 4870, that when scanning negs the max
resolution is 2000dpi because of some issue with the glass?? Is this true?
If so then it stands to reason that MF is the logical choice of format
because it's the largest format that can be scanned on a dedicated film
scanner. A bigger format than MF will have to be scanned on a flatbed
anyways right? So unless you're shooting in a very large format (6x9
inches) you won't gain anything by scanning on a flatbed. Am I right? If
so then it's one more reason for me to invest in a few choice lenses for my
rb67.
Ivan



There are a few high-end flatbed scanners that double
as film scanners. What sets these apart is that when
they scan film they do so through a separate optical
path than what's used for transmissive targets.

A few models to consider: Agfa/Microtek 2500,
Umax Powerlook 3000, Creo/Scitex Eversmart or
IQSmart. These can generally be found on eBay
or at dealers of used/demo prepress equipment.

For that matter, check eBay for used drum scanners.
Generally these go for a lot less $ than I spent on
my LS-8000, 3+ years ago.

IMO, it never pays to scrimp on a film scanner.
Film scanning is time consuming. In terms of
overall image quality, the scanner is no less
important than the camera and lens that were
used to capture the image in the first place.


rafe b.
http://www.terrapinphoto.com

"jjs" wrote in message
...
"David J. Littleboy" wrote in message
...

My guess is that 4x5 scanned on a 4870 would look better than 6x7
scanned
on
a Nikon 8000, but there may be problems scanning 4x5 on a 4870 in that
(4800
x 4 x 4800 x 5) x 3 = 1382 MB, and it's hard to configure a PC to handle
that. A Mac with 4GB might work.

A small technical aside: WindoZe XP can handle the same max-size
application
of 1.7GB RAM as the Macintosh can. You can install 4gb but the application
cannot use more than 2 (irl 1.8) on either platform. Not yet, anyway.

There's a version of XP that can use 3GB, but it may be expensive since it's
a "server" version. I was under the impression that the OS-X didn't have
that limitation, since I seem to recall hearing a Machead claim that. Oops.

David J. Littleboy
Tokyo, Japan

"jjs" wrote in message
...
"David J. Littleboy" wrote in message
...

My guess is that 4x5 scanned on a 4870 would look better than 6x7
scanned
on
a Nikon 8000, but there may be problems scanning 4x5 on a 4870 in that
(4800
x 4 x 4800 x 5) x 3 = 1382 MB, and it's hard to configure a PC to handle
that. A Mac with 4GB might work.

A small technical aside: WindoZe XP can handle the same max-size
application
of 1.7GB RAM as the Macintosh can. You can install 4gb but the application
cannot use more than 2 (irl 1.8) on either platform. Not yet, anyway.

There's a version of XP that can use 3GB, but it may be expensive since it's
a "server" version. I was under the impression that the OS-X didn't have
that limitation, since I seem to recall hearing a Machead claim that. Oops.

David J. Littleboy
Tokyo, Japan

"Neil Gould" :

Recently, David J. Littleboy posted:

(The reason for bringing up dSLRs, is that I want a common standard
for comparison.)

(cough...) why not use drum scanners as a common standard for comparing
results from other kinds of scanners? At least the objectives would be
similar.

Because I find images straight from scanners to be soft, even drum scanners.
I don't recall seeing even a single scan on Rafe's site that would make an
acceptable print at 300 dpi. So I like to "normalize" to an image that looks
decent on the screen and makes a nice 300 dpi print.

My guess is that 4x5 scanned on a 4870 would look better than 6x7
scanned on a Nikon 8000, but there may be problems scanning 4x5 on a
4870 in that (4800 x 4 x 4800 x 5) x 3 = 1382 MB, and it's hard to
configure a PC to handle that. A Mac with 4GB might work.

Oh? I haven't had any difficulty doing so for well over a decade.

Have you actually worked with 1.4 GB files?

And, until OS-X, virtual memory has always worked better on a PC.

Understatement of the last two decades.

Still, Windows NT is (essentailly) limited to 2GB for virtual address space
per process. I forget the details, but there are versions of the OS that
allow 3GB (or 4GB???) processes. But they're not the standard versions.

Such problems
with "large" scans were typically caused by the way that some scanner
drivers were written, especially so on consumer-level units. The better
scanners and software were (and still are) only limited by maximum OS file
sizes, not RAM.

Being able to handle things, and being able to handle them in a way that
makes one willing to actually use the system, are very different.

FWIW, both the image editors I have at hand (Photoshop and Picture Window)
need enough virtual memory to hold two images (basically a work image and a
result image). So that makes doing anything with over 1GB files seriously
problematic. Photoshop might do it (since it does it's own memory
management), I suppose.

Anyway, from the experience here with 645 and 6x9 scans at 4000 dpi, I
really wouldn't want to fool around with 4x5 scans at 4800 dpi without a
machine much larger and faster than this 2GB and 3GHz (XP) machine.

My assumption here is that scanning at 4800 dpi and downsampling to
2000 dpi would look better than scanning at 2400 dpi and downsampling
to 2000 dpi.

Well, that flies in the face of accepted resampling practices and typical
algorithms (e.g. bicubic, etc.). The larger the step, the worse the
results.

I didn't specify how you did the resampling: I'm concerned with whether
scanning at 4800 dpi actually scrapes more info off the slide than scanning
at 2400 dpi does, and would make a better looking 2000 dpi file.

David J. Littleboy
Tokyo, Japan

Recently, David J. Littleboy posted:

"Neil Gould" :
Recently, David J. Littleboy posted:

(The reason for bringing up dSLRs, is that I want a common standard
for comparison.)

(cough...) why not use drum scanners as a common standard for
comparing results from other kinds of scanners? At least the
objectives would be similar.

Because I find images straight from scanners to be soft, even drum
scanners. I don't recall seeing even a single scan on Rafe's site
that would make an acceptable print at 300 dpi. So I like to
"normalize" to an image that looks decent on the screen and makes a
nice 300 dpi print.

Fascinating. I guess drum scans are quite different over here, as I would
hardly call those that I've gotten "soft". ;-)

As for "on screen" and "300 dpi prints", both of those are quite a bit
lower resolution requirements than most any drum scan delivers, so I'm a
bit confused as to how your drum scans wound up being soft.

Have you actually worked with 1.4 GB files?

Actually, yes, but why do you ask?

Still, Windows NT is (essentailly) limited to 2GB for virtual address
space per process. I forget the details, but there are versions of
the OS that allow 3GB (or 4GB???) processes. But they're not the
standard versions.

Processes and file sizes are not the same thing. One doesn't need a 2 GB
process to work with a 2 GB file. In fact, I'm unaware of an application
that has a 2 GB process, though I'm sure somebody somewhere has one. On a
PC, maximum file sizes are determined by a couple of factors; sector size
and the number of sectors the OS can handle. You don't have to buy
anything special to be able to handle 4 GB files, but you do have to use
the right file system and format your drives to accommodate them.

Such problems
with "large" scans were typically caused by the way that some scanner
drivers were written, especially so on consumer-level units. The
better scanners and software were (and still are) only limited by
maximum OS file sizes, not RAM.

Being able to handle things, and being able to handle them in a way
that makes one willing to actually use the system, are very different.

Well, on this we can completely agree! ;-)

FWIW, both the image editors I have at hand (Photoshop and Picture
Window) need enough virtual memory to hold two images (basically a
work image and a result image). So that makes doing anything with
over 1GB files seriously problematic. Photoshop might do it (since it
does it's own memory management), I suppose.

IIRC, Photoshop needs enough virtual memory for _3x_ the image size. This
is not a problem on a PC, even with "1.4 GB" image files. It will be slow
with Photoshop, but, those are quirks of that particular application, not
a general requirement of image editing. Other applications used a more
efficient approach, for example using proxy editing of screen-resolution
portions of the image and writing the changes to an edit decision list
batch file that processed the actual image on "saves". Those applications
were fast and consumed few system resources while retaining the ability to
work on images as large as the OS' file size allow.

I for one, was quite happy when Adobe stopped foisting Photoshop's
internal memory management on us PC-based users. It was one of the more
fragile parts of the application, and completely unnecessary. Fortunately,
we've had other image editing applications that were as good or better.
However, now that Photoshop is king-of-the-hill, one just needs to ante-up
w/r/t system resources. ;-)

My assumption here is that scanning at 4800 dpi and downsampling to
2000 dpi would look better than scanning at 2400 dpi and
downsampling to 2000 dpi.

Well, that flies in the face of accepted resampling practices and
typical algorithms (e.g. bicubic, etc.). The larger the step, the
worse the results.

I didn't specify how you did the resampling: I'm concerned with
whether scanning at 4800 dpi actually scrapes more info off the slide
than scanning at 2400 dpi does, and would make a better looking 2000
dpi file.

I see. The higher resolution scan should give one more to work with.
Retaining those details in a 50% reduction is the challenge.

Regards,

Neil