Epson are announcing the F-3200 for the Photokina

[Bart van der Wolf]

SNIP

Interesting... I keep getting different figures for fast scan
and slow scan directions.

Small differences in scan directions like the ones below are not very
meaningful. A slightly different crop from the same scan will also
produce small differences.

I have built a better target now.
My new measurements:

Minolta DSE-5400
(5400dpi; Nyquist frequency = 106.3 cy/mm)
==========
Fast scan test:
10-90% edge rise = 1.83 pixel
MTF50 = 63 cy/mm
MTF at Nyquist = 0.24

Slow scan test:
10-90% edge rise = 1.79 pixel
MTF50 = 64.2 cy/mm
MTF at Nyquist = 0.244

Still/again pretty amazing performance for the DSE-5400.

SS120 (with misaligned CCD, unfortunately)
(4000dpi; Nyquist frequency = 78.7 cy/mm)
==============
Fast scan test:
10-90% edge rise = 2.27 pixel
MTF50 = 39.8 cy/mm
MTF at Nyquist = 0.125

Slow scan test:
10-90% edge rise = 1.83 pixel
MTF50 = 63 cy/mm
MTF at Nyquist = 0.254

The SS120 slow scan direction has overall unexpected high values
compared to the fast scan.

Bart

 

[Fernando]

Small differences in scan directions like the ones below are not very
meaningful. A slightly different crop from the same scan will also
produce small differences.

Yes... I found that the difference between dark and light areas counts
a lot, too: much more than I was expecting. So I think I'll re-run the
tests under more controlled conditions. I have registered ImaTest, so
now I have to justify the expense! ;-)
Given the problems with Vuescan handling the D.R. of the 5400, I end
up with a contrast of about 3:1 instead of the 6.5:1 I have on the
SS120 (with Vuescan, too). Of course with the same target slide, and
comparable exposure (same "white" level, but much higher "black" level
on the 5400).

The SS120 slow scan direction has overall unexpected high values
compared to the fast scan.

Yes... with the misaligned CCD, my SS120 lost quite a lot of its
amazing ability of capturing low-contrast details without exagerating
grain.
It used to have a really high resolution along both scan directions,
before this issue (didn't try the S.E.T., but I did many tests with
simpler slide targets).
I don't know if I'll send it to the repair facility once again (it
keeps misaligning every 6 months or so, and it's a very costly
repair), but I'm tempted. It's a very, very capable machine.

I'll try the Epson 2450 in a couple days; will try both Vuescan and
Epson Scan, both at 2400x2400 and (Epson Scan only) at 4800x4800 to
see if the slow-scan resolution gains something (I doubt it, with that
lousy lens; but I'll try).
Bye!

Fernando

 

[Bart van der Wolf]

"Bart van der Wolf" <[email protected]> wrote in message

Yes... I found that the difference between dark and light areas counts
a lot, too: much more than I was expecting. So I think I'll re-run the
tests under more controlled conditions. I have registered ImaTest, so
now I have to justify the expense! ;-)

The contrast between the lightest and darkest area (within limits)
shouldn't have much influence. The SFR/MTF quantifies the contrast
*loss*. What does have an influence is (clipping and) the shape of the
edge profile. The shape is undoubtedly influenced by VueScan's D-max
problems with this scanner model. So indirectly it will influence the
MTF less if Dmax contribution is reduced. The Minolta Scan utility
doesn't have that problem, so the MTF is not affected by contrast.

You can try that, now you've got a licenced version of Imatest, on an
artificial slanted edge that you can make in your photoeditor. High or
low contrast gives similar results. I even tried a 127/128 edge
contrast. Of course, after ignoring the Imatest warning, the result
was different due to loss of accuracy, but as long as there is at
least 20% contrast in the edge target all should be fine.

Given the problems with Vuescan handling the D.R. of the 5400, I end
up with a contrast of about 3:1 instead of the 6.5:1 I have on the
SS120 (with Vuescan, too). Of course with the same target slide, and
comparable exposure (same "white" level, but much higher "black" level
on the 5400).

Yes that'll be one of the effects the VueScan D-max bug will cause.

SNIP

I don't know if I'll send it to the repair facility once again (it
keeps misaligning every 6 months or so, and it's a very costly
repair), but I'm tempted. It's a very, very capable machine.

There should be a reasonable time warranty on the repair. They must
'permanently' fix it so you'll be able to use the product as intended.
Depending on local jurisdiction, you may want to stress that such
recurring need to fix it constitutes a problem caused by inadequate
repair, and should be fixed for free (excl. parts and shipping).

Bart