New Epson flatbed scanner...

[Colin D]

http://www.macworld.com/news/2006/02/22/perfection/index.php

Sounds yummy...

Comments?

I recently bought a Canon 9950F scanner, similar specs but 9600 x 4800
optical res, with dual lamps, one either side of the ccd sensor to
eliminate shadows from textured copy. Also film, 30 frames of 35mm, 8
frames of 6x4.5, or 6 frames of 6x6 or 4 of 6x9, and 1 frame of 5x4

Scan speed 11.1 msec per line. here is a review:

http://www.photo-i.co.uk/Reviews/interactive/Scanners/Canon_9950F/page_1.htm

It appears, by this reviewer's testing, the Canon is a whisker ahead of
the Epson. Me though, I'm rapt.

Colin D.

 

[Colin D]

I recently bought a Canon 9950F scanner, similar specs but 9600 x 4800
optical res, with dual lamps, one either side of the ccd sensor to
eliminate shadows from textured copy. Also film, 30 frames of 35mm, 8
frames of 6x4.5, or 6 frames of 6x6 or 4 of 6x9, and 1 frame of 5x4

Scan speed 11.1 msec per line. here is a review:

http://www.photo-i.co.uk/Reviews/interactive/Scanners/Canon_9950F/page_1.htm

It appears, by this reviewer's testing, the Canon is a whisker ahead of
the Epson. Me though, I'm rapt.

Colin D.

Addendum: The 120 film holder will also hold a pano frame of 6 x 17 mm
in one shot.

Colin D.

 

[Rob]

Addendum: The 120 film holder will also hold a pano frame of 6 x 17 mm
in one shot.

Colin D.

Do you think that its a 4800dpi optical and the 9500dpi is the tricked
up bit.(double stepped)

 

[Colin D]

Do you think that its a 4800dpi optical and the 9500dpi is the tricked
up bit.(double stepped)

The specs say 4800 x 9600 optical, so unless they're stretching the
truth, it's not double-stepped or interpolated. I tend to believe them,
though. I can't see Canon laying themselves open to charges of false
representation.

Colin D.

 

[rafe b]

The specs say 4800 x 9600 optical, so unless they're stretching the
truth, it's not double-stepped or interpolated. I tend to believe them,
though. I can't see Canon laying themselves open to charges of false
representation.

Lots of ways to stretch the truth without
outright lying.

Let's put it this way. If the Nikon film
scanners are 4000 dpi, these film/flatbeds
aren't anywhere close to that in terms of
sharpness.

IOW... if you expect a correlation between
scanner dpi and sharpness, the film/
flatbed scanners almost always come up
short, compared to dedicated film scanners.



rafe b
www.terrapinphoto.com

 

[Don]

Some people in this NG have mentioned that some defocussing might be
beneficial to reduce grain aliasing because it reduces the spatial frequency
content of the image to better match the sampling rate of the scanner.

Exactly! But by reducing the spatial frequency image data is thrown
away which was my point.

Djon's basic premise was that the amount of grain in a scan is a
function of "skill" which is wrong. Grain is a function (an intrinsic
part) of film image and reducing grain means reducing image data.

Don.

 

[Djon]

"Djon's basic premise was that the amount of grain in a scan is a
function of "skill" which is wrong. Grain is a function (an intrinsic
part) of film image and reducing grain means reducing image data."

No, that intentionally twisted what I said. Film grain is itself a
function substantially of skill, with or without a scan.

Scan results are irrelevant without considering reproduced images,
whether litho, print, or monitor display. And the qualities of all
those reproductions depend on skill as much as upon technology.

Photographers make images, not "image data."

Grain was traditionally addressed in wet darkrooms by grain dissolving
developers such as D23, and by diffusion enlargers. Honorable,
beautiful methodology for people who liked grainless images.

I preferred point light source, or at least optically coated Durst
condenser enlargers with glass carriers in the wet darkroom. That's
simply another approach, entailing different/equal skills. My work
showed sharp grain because I wanted to show it, was skilled enough to
show it sharply. Ansel's students didn't want to show it, were taught
how to avoid it: Skills that are just as applicable today, with
scanners, as they were in Ansel's day.

Human skills are crucial to photography. Digital technology is not
crucial to photography. Grain is controllable by skill.

 

[false_dmitrii]

Sounds too good to be true: very high DPI (oversampled CCD? -- good to
suppress grain aliasing), new optics, large batches for 35mm, high
reliability. As an owner of a just broken KM 5400-II I can value the latter.

From past reviews and personal experience with the 4870, I have the

impression that while Epson's marketing claims substantially inflate
expectations, the scanners themselves are still excellent performers
and a great value. So if you expect it to outdo your (working) 5400,
you'll probably be disappointed, but if you expect the best consumer
flatbed you've seen, there's a good chance you'll be satisfied.

false_dmitrii

 

[Chris]

The specs say 4800 x 9600 optical, so unless they're stretching the
truth, it's not double-stepped or interpolated. I tend to believe them,
though. I can't see Canon laying themselves open to charges of false
representation.

Colin D.

When the Epson ???? was released they quoted something like 800x1600
resolution the resolution in which the 1600 was double stepped to appear
like a 1600 scanner. But still the true resolution was still 800 dpi.
(There must be an explanation somewhere on the web how they come to
these figures.)

I always only accept the smaller number as being the true optical
resolution. Still 4800dpi is more than enough resolution for 35mm film.

I very rarely use 4800 (normally 2400/3200) for medium format film
anyway as I can't see any difference in the final output.

 

[David J. Littleboy]

When the Epson ???? was released they quoted something like 800x1600
resolution the resolution in which the 1600 was double stepped to appear
like a 1600 scanner. But still the true resolution was still 800 dpi.
(There must be an explanation somewhere on the web how they come to these
figures.)

I always only accept the smaller number as being the true optical
resolution. Still 4800dpi is more than enough resolution for 35mm film.

The smaller number is still a funny number. Epson uses an "offset" CCD that
consists of two half-resolution CCDs offset by 1/2 the pixel pitch. The
result is that instead of the theoretically optimal point sample, it samples
each pixel with an area much larger than the pixel size for the nominal
resolution. The good news is that this should reduce grain aliasing, the bad
news is that the resolution MTF is abysmally low above 1/2 the nominal
resolution, so the appearance is a lot more like what you'd expect from a
point sampling scanner with 1/2 (or less) the resolution.

The new Epson scanner appears to be playing the same game.

http://photo-i.co.uk/News/Feb06/Epson_V700_scanner.htm

(I almost always find that those guys get much better results from Epson
scanners than anyone else does. Still, it looks pretty good.)

Also, the 6400 dpi is good through 4x5, so maybe 4x5 users with 64-bit PCs,
tens of GB of DRAM, and a lot of time will be amuzed...

David J. Littleboy
Tokyo, Japan

 

[rafe b]

The new Epson scanner appears to be playing the same game.

http://photo-i.co.uk/News/Feb06/Epson_V700_scanner.htm

(I almost always find that those guys get much better results from Epson
scanners than anyone else does. Still, it looks pretty good.)

Also, the 6400 dpi is good through 4x5, so maybe 4x5 users with 64-bit PCs,
tens of GB of DRAM, and a lot of time will be amuzed...

I read at least one post on photo.net saying that the
V700 and V750 use the same CCD chip as the 4990.

So anyway -- from the dual lens trick, we'd still get
a 33% increase in linear resolution, if not MTF.

First person to get their hands on one of these: I
want a sample for the "scan snippets" site.



rafe b
www.terrapinphoto.com

 

[Rob]

The smaller number is still a funny number. Epson uses an "offset" CCD that
consists of two half-resolution CCDs offset by 1/2 the pixel pitch. The
result is that instead of the theoretically optimal point sample, it samples
each pixel with an area much larger than the pixel size for the nominal
resolution. The good news is that this should reduce grain aliasing, the bad
news is that the resolution MTF is abysmally low above 1/2 the nominal
resolution, so the appearance is a lot more like what you'd expect from a
point sampling scanner with 1/2 (or less) the resolution.

The new Epson scanner appears to be playing the same game.

http://photo-i.co.uk/News/Feb06/Epson_V700_scanner.htm

(I almost always find that those guys get much better results from Epson
scanners than anyone else does. Still, it looks pretty good.)

Also, the 6400 dpi is good through 4x5, so maybe 4x5 users with 64-bit PCs,
tens of GB of DRAM, and a lot of time will be amuzed...

David J. Littleboy
Tokyo, Japan

Yep that could be correct with the quoted resolution. Just knew they
played with the figures to make them look better than the actual
straight optical resolution. (Not taking away quality but its the
consumer deception of how the figures are calculated)

 

[Noons]

resolution. Still 4800dpi is more than enough resolution for 35mm film.

Perhaps if it is real 4800. Tricked up is not the same.

I very rarely use 4800 (normally 2400/3200) for medium format film
anyway as I can't see any difference in the final output.

Oh, I can! At least with the 4990. When it's working: been in
warranty repairs at Epson for over a month now, they can't find
what's wrong with it.

<venting>
Came back today, still not working, going back tomorrow: what is
it with service departments that do NOT test a repair to make sure
the darn thing is working? One would think that turning the
blessed thing ON and TRYING a scan would be the very
least one could test after a repair?
</venting>

 

[Ernst Dinkla]

I read at least one post on photo.net saying that the
V700 and V750 use the same CCD chip as the 4990.

So anyway -- from the dual lens trick, we'd still get
a 33% increase in linear resolution, if not MTF.

Rafe,

It is a 33% on both axis. The 9600 stepping rate over the
length is in both SPI resolutions a heavy oversampling
considering the well size. But the longer lens makes the
scanned area (at one sample) covered by one well 33% smaller
in both directions so the oversampling with the longer lens is
reduced compared to the wider lens.

There's another concept change I think. Unlike the 2450 up to
4990 models the lid isn't meant to be taken off for reflective
scans of originals larger than A4. The DOF of the wider lens
that covers the A4 scan bed no longer needs to have best focus
at 1 mm above the glass for the film holders and still
acceptable focus at the scan bed and the 2,5-3 mm higher rim
around the scan bed. The longer lens takes care of the film
holders and will have less DOF. The wider lens will focus on
the scan bed and will have less DOF too. The scanning of
originals larger than A4 is sacrificed for that with less
sharpness.

The matrix CCD and the over-sampling still requires sharpening
but less than with the lens system used on the older models.

I'm not surprised that photo-i gets so excited and I expect
more than the 33% quality gain.

Ernst
--
Ernst Dinkla


www.pigment-print.com
( unvollendet )

 

[rafe b]

It is a 33% on both axis. The 9600 stepping rate over the
length is in both SPI resolutions a heavy oversampling
considering the well size. But the longer lens makes the
scanned area (at one sample) covered by one well 33% smaller
in both directions so the oversampling with the longer lens is
reduced compared to the wider lens.

There's another concept change I think. Unlike the 2450 up to
4990 models the lid isn't meant to be taken off for reflective
scans of originals larger than A4. The DOF of the wider lens
that covers the A4 scan bed no longer needs to have best focus
at 1 mm above the glass for the film holders and still
acceptable focus at the scan bed and the 2,5-3 mm higher rim
around the scan bed. The longer lens takes care of the film
holders and will have less DOF. The wider lens will focus on
the scan bed and will have less DOF too. The scanning of
originals larger than A4 is sacrificed for that with less
sharpness.

The matrix CCD and the over-sampling still requires sharpening
but less than with the lens system used on the older models.

I'm not surprised that photo-i gets so excited and I expect
more than the 33% quality gain.

The snippets now showing at photo-i don't tell us very
much, since we don't know the actual film areas involved
in the hi-res samples.

Have you ever seen the lens in one of these film/
flatbed scanners? Have you ever looked at the lens
in an LS-8000/9000?

I really know very little about optics, and am
seriously puzzled at the complexity of the Coolscan
lens. According to Nikon, it has 14 elements in
six groups, and six of those 14 elements use ED glass.

Why? I mean, this is probably more sophisticated
than any *camera* lens I own.

In any case I can assure you that the lenses in the
Epson 700/750 are nothing like this. Also, in the
Epson, there are probably several mirrors in the
path between the film and the imaging lens. Those
extra reflections can't help the image quality.


rafe b
www.terrapinphoto.com

 

[Ernst Dinkla]

The snippets now showing at photo-i don't tell us very
much, since we don't know the actual film areas involved
in the hi-res samples.

Have you ever seen the lens in one of these film/
flatbed scanners? Have you ever looked at the lens
in an LS-8000/9000?

I really know very little about optics, and am
seriously puzzled at the complexity of the Coolscan
lens. According to Nikon, it has 14 elements in
six groups, and six of those 14 elements use ED glass.

Why? I mean, this is probably more sophisticated
than any *camera* lens I own.

In any case I can assure you that the lenses in the
Epson 700/750 are nothing like this. Also, in the
Epson, there are probably several mirrors in the
path between the film and the imaging lens. Those
extra reflections can't help the image quality.


rafe b
www.terrapinphoto.com

Rafe:

Yes, I have seen the lens assembly in my Nikon 8000.
Impressive lens. One mirror.

The Epsons have a different concept and the lenses are not in
competition with the Nikon lens. The use of the Matrix CCD +
over-sampling + a lot of light, is another route. It would
work with a pinhole lens if needed. I have wondered if they
use the lens at diffraction to get the DOF needed, the wide
angle covered, the light fall off compensated and keeping good
geometry. You need sharpening at the end with all the
ingredients of that concept. Now they emphasise in the V
models the higher speed of the lens so I think they went for
more sharpness and less DOF. At least for the longer lens.

The Nikon uses it rows of CCDs for speed, the light is reduced
if compared to the Epson and the sharpness of the scan is
right there after the scan. There's no quality loss if only
one row of the linear CCD is used. With the 8000 there's even
less quality with all the CCD rows in action. Multi-sampling
not counted.

There is a picture of the Epson lenses at the Japanese pages:
http://www.i-love-epson.co.jp/products/colorio/scanner/gtx900/technology.htm

I expect that the Nikon 8000/9000 still beats the Epson V750-M
Pro in MF and 35 mm quality but the Nikon doesn't scan 4x5 and
panorama sizes (and possibly 5x7's) at 3000 PPI, doesn't do
reflective A4 and 8x10 film at 2000 PPI, doesn't come with wet
mount carriers from the factory.

The first German magazine tests of this scanner will tell
more. Photo-i is always fast but not very consistent in its
testing and it is also very manufacturer friendly. ISO
standards tell more.

On the test crops:
Of the crane there is a full size pic in the 4990 review. It
is a pity that he didn't use the same crop.


Ernst

--
Ernst Dinkla


www.pigment-print.com
( unvollendet )

 

[rafe b]

There is a picture of the Epson lenses at the Japanese pages:
http://www.i-love-epson.co.jp/products/colorio/scanner/gtx900/technology.htm

I expect that the Nikon 8000/9000 still beats the Epson V750-M
Pro in MF and 35 mm quality but the Nikon doesn't scan 4x5 and
panorama sizes (and possibly 5x7's) at 3000 PPI, doesn't do
reflective A4 and 8x10 film at 2000 PPI, doesn't come with wet
mount carriers from the factory.

The first German magazine tests of this scanner will tell
more. Photo-i is always fast but not very consistent in its
testing and it is also very manufacturer friendly. ISO
standards tell more.

I was actually impressed at the degree of attention
and technical detail given by Epson on the Epson-
Japan website. Eg., showing the lens arrangement,
down to elements and groups. Maybe that's a good omen.

Another good omen is their specific references to
wet-mounting. Unless Epson is just being devious
and cynical.

I can tell you that a "typical" flatbed scanner
has a lens on the order or 20 mm f/8 or so, and
that even the 4990 is probably not much better
than this. By comparison: the MF Coolscans use
a lens that's around 75mm f/4.

So, we'll just have to wait and see. If nothing
else, it's comforting that Epson still has not
given up film, and (most surprisingly) on medium-
and large-format film.

Apparently they've noticed a "hole" in the market
and decided to address it. I only wish Nikon would
try to compete in that same niche.

If it turns out that the new Epsons really offer
a significant upgrade to the 4990, I will probably
buy one. Even at $500 or $750, these are still
in what I'd call the "dirt cheap" category, for
a MF or LF film scanner.


rafe b
www.terrapinphoto.com

 

[Ernst Dinkla]

I was actually impressed at the degree of attention
and technical detail given by Epson on the Epson-
Japan website. Eg., showing the lens arrangement,
down to elements and groups. Maybe that's a good omen.

Another good omen is their specific references to
wet-mounting. Unless Epson is just being devious
and cynical.

I can tell you that a "typical" flatbed scanner
has a lens on the order or 20 mm f/8 or so, and
that even the 4990 is probably not much better
than this. By comparison: the MF Coolscans use
a lens that's around 75mm f/4.

There have been lens drawings on Epson's site for the older
models. There's some difference visible between the lenses on
the Japanese V700 site and Norman Koren's 2450 and up pages.

Ernst


--

--
Ernst Dinkla


www.pigment-print.com
( unvollendet )

 

[Bart van der Wolf]

I
want a sample for the "scan snippets" site.

Would be interesting.

It would also help if that person could scan a Slanted Edge test
object, such as a razor blade mounted in a 35mm slidemount at a slant
of approx. 5-6 degrees (90 degree rotation for horizontal and vertical
scan resolution). I'm offering to quantify the effective limiting
resolution as can be deducted from an MTF curve that's calculated from
a crop of the slanted edge.

That would take a lot of speculation out of the discussions...

If that person happens to be located in the (central portion of)
Netherlands, we could use my Stouffer T4110 step wedge to test the
claimed 4.0 dynamic range, and improve the MTF result's accuracy.

Bart

 

[Bart van der Wolf]

"rafe b" <rafebATspeakeasy.net> wrote in message
SNIP

I really know very little about optics, and am
seriously puzzled at the complexity of the Coolscan
lens. According to Nikon, it has 14 elements in
six groups, and six of those 14 elements use ED glass.

Why? I mean, this is probably more sophisticated
than any *camera* lens I own.

The main design problems are; Flat plane projection, reduction of
aberrations, and (presumably) large physical aperture to reduce
exposure time. The tolerances are small and the projection result is
'magnified' by the fine sampling by the sensor, so there are many lens
elements needed to reduce small amounts of residual aberrations from
preceding lens elements.

Bart