Scanner bake-off: initial wave of results compiled and published...

[Jim Hutchison]

....at http://www.jamesphotography.ca.


pls read my disclaimers before ripping my head off.....


jim h


~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

http://www.jamesphotography.ca

-free downloads
-prize for every 1,000th visitor
-scanning service

Even a bit of humour...

 

[Wilfred]

...at http://www.jamesphotography.ca.


pls read my disclaimers before ripping my head off.....

Thanks for the great work. It seems I will have to start using the
manual-focus function of my DSE 5400 when I compare Bart's results to mine.
BTW (1) I used AF but I did perform some manual action by specifying a
focusing point in the central area. I guessed this would be slightly
better than 'auto-autofocus'.
BTW (2) You mentioned that, for now, you ignored scans made with Apple
Macintosh computers. I forgot to tell you that my scans were produced on
a Mac. But do you really think it makes a difference?

 

[Simon Waldman]

Thanks for the great work. It seems I will have to start using the
manual-focus function of my DSE 5400 when I compare Bart's results to mine.
BTW (1) I used AF but I did perform some manual action by specifying a
focusing point in the central area. I guessed this would be slightly
better than 'auto-autofocus'.

I've always used autofocus on mine... I figure that once I'm resolving
grain there isn't really any point in worrying about making it sharper.

 

[Bart van der Wolf]

Thanks for the great work. It seems I will have to start using the
manual-focus function of my DSE 5400 when I compare Bart's results to mine.
BTW (1) I used AF but I did perform some manual action by specifying a
focusing point in the central area. I guessed this would be slightly
better than 'auto-autofocus'.

My manual focussing effort was also based on the center of the target, but
the MTF measurement was done more towards the edge of the target. That means
that film flatness also plays a role.

When not in a hurry, I usually allow the film to acclimatize in position for
scanning for a while, occasionally letting VueScan focus (Menu bar
Scanner|Focus) until the position only changes very little. That will also
reduce the chance of the film moving during the scan. For this mounted slide
it worked out fine, it even resolved a 1 pixel (1/5400 inch, 4.7 micron)
wide scratch closer to the corner
<http://www.xs4all.nl/~bvdwolf/main/downloads/Minolta_DSE5400_5400_scratch.j
pg>.

BTW (2) You mentioned that, for now, you ignored scans made with Apple
Macintosh computers. I forgot to tell you that my scans were produced on
a Mac. But do you really think it makes a difference?

I think that was to reduce the effect of different gamma. You apparently
produced a JPEG file with something close to gamma adjustment of 2.0 to 2.2,
so that would be in line with the intentions. A more rigorous test would
require further gamma testing and exposure- and Black/White point
synchronization. The complexity would probably have scared a few people off.

Anyway, the results are not absolute, but indicative, and more independent
test results would increase statistical significance, but Jim already put in
a lot of his time to get to this result, which is much appreciated.

Bart

 

[Wilfred]

Bart van der Wolf wrote:

My manual focussing effort was also based on the center of the target, but
the MTF measurement was done more towards the edge of the target. That means
that film flatness also plays a role.

When not in a hurry, I usually allow the film to acclimatize in position for
scanning for a while, occasionally letting VueScan focus (Menu bar
Scanner|Focus) until the position only changes very little. That will also
reduce the chance of the film moving during the scan.

Thanks for the tip. I will try it.

I think that was to reduce the effect of different gamma. You apparently
produced a JPEG file with something close to gamma adjustment of 2.0 to 2.2,
so that would be in line with the intentions.

VueScan's default color space for the Mac is, for some reason, Apple
RGB, but I scanned to sRGB and removed the embedded color space from the
final file.

 

[Bart van der Wolf]

SNIP

VueScan's default color space for the Mac is, for some reason, Apple
RGB, but I scanned to sRGB and removed the embedded color space from the
final file.

Yes, that puts it in sRGB slope-adjusted gamma 2.2 space.

Bart

 

[Roger Halstead]

I've always used autofocus on mine... I figure that once I'm resolving
grain there isn't really any point in worrying about making it sharper.

I keep hearing about resolving grain, but I've never seen my LS5000 ED
resolve grain even at 4000 dpi. I'll not argue there's grain in the
images, but the only grain I find comes from the scanner. It's even
and smoothly distributed. ASA 400 film should show grain clumps,
strings, and all kinds of crappy looking stuff when blown way up. I
just don't see that.

Roger Halstead (K8RI & ARRL life member)
(N833R, S# CD-2 Worlds oldest Debonair)
www.rogerhalstead.com

 

[David J. Littleboy]

...at http://www.jamesphotography.ca.

Looks good. Two comments. (Oops. Make that 3.)

1. How about a 100x or so photomicrograph of the slide (and its MTF 50
value) to see what we're all missing.

2. As a Nikon 8000 owner<g>, Bart's 5400 scan looks sharper. Why is the
background in the 5400 scan so dark? Presumably, the darkness of the
background is what makes the MTF50 value so low.

3. Hmm. Under 24 lp/mm at MTF50. Depressing. An MTF10 value would be
interesting as well. Even if only for a couple of scanners.

David J. Littleboy
Tokyo, Japan

 

[Philip Homburg]

2. As a Nikon 8000 owner<g>, Bart's 5400 scan looks sharper. Why is the
background in the 5400 scan so dark? Presumably, the darkness of the
background is what makes the MTF50 value so low.

I wonder what the effect is of gamma. Are the pictures (and the data used
for the computations) linear, or gamma corrected.

On the other hand, the LS50 from Mike Engles also scores higher than Bart's
5400, and that picture is really dark.

3. Hmm. Under 24 lp/mm at MTF50. Depressing. An MTF10 value would be
interesting as well. Even if only for a couple of scanners.

To what extent are we measuring the film or the scanner? In
<http://groups.google.nl/[email protected]&output=gplain>
Bart posted 76.1 lp/mm for 'a medium high contrast target'. That
translates to 76.1*2 / (5400 / 25.4) = 0.72 pixel. The graph in
<http://www.jamesphotography.ca/bakeoff2004/minolta_DSE_5400_Bart_van_der_wolf_5400_MTF.jpg>
is a the bottom way before that.

 

[Jim Hutchison]

All good discussion.

1) I'd LOVE someone to step up and offer to do a photomicrograph of
the original target. When I loupe the slide, it's obvious even the
best scanner doesn't come close to resolving the knife-edge sharpness
of the image.

2) The reason the 5400 appears sharper is because of the DPI. But
the edge profile isn't as sharp. Look at each of the edge profile
graphs, and it's obvious the Nikon's is sharper. The 5400 "looks"
sharper, but that's because of the extra 1400 DPI. Dark or light
background doesn't matter, as exemplified by Mike Engle's LS-50.


The images in the right-hand column can be misleading for these
reasons, so please consider the DPI as well when judging the MTF
numbers to be off target.





jim h


~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

http://www.jamesphotography.ca

-free downloads
-scanning service

Even a bit of humour...

 

[Mike Engles]

2. As a Nikon 8000 owner<g>, Bart's 5400 scan looks sharper. Why is the
background in the 5400 scan so dark? Presumably, the darkness of the
background is what makes the MTF50 value so low.

I wonder what the effect is of gamma. Are the pictures (and the data used
for the computations) linear, or gamma corrected.

On the other hand, the LS50 from Mike Engles also scores higher than Bart's
5400, and that picture is really dark.

3. Hmm. Under 24 lp/mm at MTF50. Depressing. An MTF10 value would be
interesting as well. Even if only for a couple of scanners.

To what extent are we measuring the film or the scanner? In
<http://groups.google.nl/[email protected]&output=gplain>
Bart posted 76.1 lp/mm for 'a medium high contrast target'. That
translates to 76.1*2 / (5400 / 25.4) = 0.72 pixel. The graph in
<http://www.jamesphotography.ca/bakeoff2004/minolta_DSE_5400_Bart_van_der_wolf_5400_MTF.jpg>
is a the bottom way before that.

--
The Electronic Monk was a labor-saving device, like a dishwasher or a video
recorder. [...] Video recorders watched tedious television for you, thus saving
you the bother of looking at it yourself; Electronic Monks believed things for
you, [...] -- Douglas Adams in Dirk Gently's Holistic Detective Agency

Hello

Now I have to confess that I forgot to convert my scan fron Apple
Compensated to sRGB. I use Windows, but a Apple colourpace and gamma.
The difference is gamma 1.22

I can send a corrected one and we can see if there is a difference.

Mike Engles

 

[MW]

Bravo, Jim and participants! You should submit this to Consumer
Reports. It's amazing what these scanner makers get away with in
marketing hype.

Matt

 

[Philip Homburg]

All good discussion.

1) I'd LOVE someone to step up and offer to do a photomicrograph of
the original target. When I loupe the slide, it's obvious even the
best scanner doesn't come close to resolving the knife-edge sharpness
of the image.

The Provia 100F datasheet lists 140 lines/mm for 1000:1 contrast, which
translates to 3360 lines/ph. From the same datasheet, MTF50 is about
40 line pairs/mm or 1920 lines/ph.

Maybe I am reading the wrong datasheet (or drawing the wrong conclusions),
but it is not supposed to be knife-edge sharp (maybe Fuji is sandbagging).

 

[Bart van der Wolf]

SNIP

Looks good. Two comments. (Oops. Make that 3.)

1. How about a 100x or so photomicrograph of the slide (and its MTF 50
value) to see what we're all missing.

That might be interesting. Although the target slide is not as sharp as it
could be, all scanners were given the same target, let's say a real life
target, with subject and lens and film limitations.

2. As a Nikon 8000 owner<g>, Bart's 5400 scan looks sharper. Why is the
background in the 5400 scan so dark? Presumably, the darkness of the
background is what makes the MTF50 value so low.

My 5400ppi scanner produces more pixels so, at same size output, it is
sharper than a 4000ppi scanner. I avoided clipping in the (paper) white. The
Nikon 8000 image had some clipping which makes the measurement less
accurate. The algorithm used is quite capable of extracting MTF of lower
contrast edges, in fact the ISO suggests a much lower target contast (to
avoid clipping).

There is a pitfall when judging per-pixel quality, especially where the 5400
ppi scan has more pixels and requires less magnification for same size
output. To illustrate:
Sample , MTF20 , 10-90% rise
===== , ===== , =========
Marty's LS8000 , 0.2762 cy/px, 2.84 px
Mike's LS-50 , 0.2490 cy/px, 2.88 px
Bart's DSE-5400 , 0.1963 cy/px, 4.04 px

If you translate that to on-film-lp/mm, you get:
Sample , MTF20 , 10-90% rise
===== , ===== , =========
Marty's LS8000 , 43.5 lp/mm, 18.0 micron
Mike's LS-50 , 39.2 lp/mm, 18.3 micron
Bart's DSE-5400 , 41.7 lp/mm, 19.0 micron (but requires 35% less
magnification, leading to 29% higher resolution at MTF20, or a 22% narrower
10-90% edge rise, than Marty's LS-8000).

As always, the devil is in the detail.

3. Hmm. Under 24 lp/mm at MTF50. Depressing. An MTF10 value would be
interesting as well. Even if only for a couple of scanners.

A few remarks.
1. The target (which as I mentioned was not as sharp as could be) tests the
combination of lens (+focus+camera shake) and film, not scanner alone. For
a relative comparison it is adequate, but it doesn't show the limit of what
these scanners can do.
2. The scans were not sharpened, which affects MTF50 a lot with a larger
radius USM.
3. An MTF20 is given in the graphs, which is much closer to the limiting
resolution. An MTF10 is sometimes difficult to evaluate accurately,
especially when only a small part of the film was measured (the readout
becomes more statistically accurate with a longer edge). An MTF10 would
correspond more to the limiting human visual resolution.

Bart

 

[Bart van der Wolf]

SNIP

I wonder what the effect is of gamma. Are the pictures (and the data used
for the computations) linear, or gamma corrected.

Yes there is some influence, and the software used can be adjusted for image
ganna. That gamma would have to be calculated separately, but the target had
no robust provision for that.

On the other hand, the LS50 from Mike Engles also scores higher than Bart's
5400, and that picture is really dark.

As mentioned in response to David, don't forget the difference in PPI. The
resolution on sensor may be close, but the DSE produces 35% more pixels and
needs less magnification for equal size output. That 35% magnification will
cost output resolution for the 4000 ppi scanners in comparison with the 5400
ppi scanners.

To what extent are we measuring the film or the scanner?

Jim's film target was not as sharp as could be, but all scanners had to scan
'the same' target, so as a relative test it suffices.

In <http://groups.google.nl/groups?selm=3f2a86d3$0$49106$e4fe514c@news.
xs4all.nl&output=gplain>
Bart posted 76.1 lp/mm for 'a medium high contrast target'. That
translates to 76.1*2 / (5400 / 25.4) = 0.72 pixel. The graph in
<http://www.jamesphotography.ca/bakeoff2004/minolta_DSE_5400_Bart_van_der_wo
lf_5400_MTF.jpg>
is a the bottom way before that.

The number I mentioned was:
1. Of a higher resolution target on film (I shot my similar contrast target
with a *very* good, even at approx. 3 meters distance, macro lens),
2. I measured the absolute limiting resolution.
3. I didn't focus my scanner on Jim's target at the point that was measured,
although film flatness was good enough to be close. I did focus the scan of
my film on the target, so it was optimal.

That demonstrates that you can't resolve with the scanner wat is not in the
film. It is a small consolation for me, for the lugging around my tripod,
and only shooting with fixed focal length lenses.

Bart

 

[Bart van der Wolf]

SNIP

The Provia 100F datasheet lists 140 lines/mm for 1000:1 contrast, which
translates to 3360 lines/ph. From the same datasheet, MTF50 is about
40 line pairs/mm or 1920 lines/ph.

Maybe I am reading the wrong datasheet (or drawing the wrong conclusions),
but it is not supposed to be knife-edge sharp (maybe Fuji is sandbagging).

The Fuji data sheet ONLY measures film MTF. It's the combination with a
lens' MTF that sets the challenge for the scanner (with it's own MTF). After
scanning, you look at the combination of the entire imaging chain's MTF,
before output that is.

Bart

 

[Philip Homburg]

If you translate that to on-film-lp/mm, you get:
Sample , MTF20 , 10-90% rise
===== , ===== , =========
Marty's LS8000 , 43.5 lp/mm, 18.0 micron
Mike's LS-50 , 39.2 lp/mm, 18.3 micron
Bart's DSE-5400 , 41.7 lp/mm, 19.0 micron (but requires 35% less
magnification, leading to 29% higher resolution at MTF20, or a 22% narrower
10-90% edge rise, than Marty's LS-8000).

Somehow I have the feeling that you count de difference in resolution twice.
Your DSE-5400 gets an MTF20 value of 0.1963. This results in
0.1963 * 2 * 4844 = 1902 line/ph.
The LS-8000 gets 0.2762 * 2 * 3616 = 1997.

No matter what you do, you won't get more that 1902 lines (at MTF20) whereas
the LS-8000 gets 1997 lines. Scaling doesn't introduce new lines, and
should not drop lines. Unless, you are saying that higher scaling factors
would somehow result in a serious loss of sharpness.

 

[Philip Homburg]

The Fuji data sheet ONLY measures film MTF. It's the combination with a
lens' MTF that sets the challenge for the scanner (with it's own MTF). After
scanning, you look at the combination of the entire imaging chain's MTF,
before output that is.

Hopefully the lens on the camera doesn't have much effect. However, the
main thing is that this test doesn't say much about the differences between
high-end scanners.

The thing that surprises me is the big difference in scans on the same
model.

 

[Philip Homburg]

The number I mentioned was:
1. Of a higher resolution target on film (I shot my similar contrast target
with a *very* good, even at approx. 3 meters distance, macro lens),
2. I measured the absolute limiting resolution.

What is the difference between the calculation you used and the calculation
in this test? I assume you followed the ISO standard. But I can't find
any on-line resources that describe that basic idea behind the ISO standard.

In your test, your scanner scores quite a bit higher than the LS-4000.
Did you somehow take the extra resolution into account?

 

[Ralf C. Kohlrausch]

All good discussion.

1) I'd LOVE someone to step up and offer to do a photomicrograph of
the original target. When I loupe the slide, it's obvious even the
best scanner doesn't come close to resolving the knife-edge sharpness
of the image.

Hi,

I can do that and will send it in with my results.

Just looked at the slide through my micrscope:

1/4 is still separated quite clearly, but much less so than 1/3. 1/5 is
very murky, extremely low contrast. 1/6 is still softly seperated in the
vertical orientation and there is an slight indication of an idea of
seperation in the horizontals, but I'ld rather not be put under oath for
this one. 2/1 is not seperated any more.

I refocused to compensate for field curvature and I checked at 50, 120
and 500 mags. I will photograph with a LH 10/30 lens and a 20mm plössl
astronomical eyepiece (better flatness than the eyepiece that came with
the microscope) using eyepiece projection with my Micronikkor 2,8/60
well stopped down.

Greets
Ralf C.