Scanner Resolution test

[Andre]

I'm seeing several post about scanner resolution. Some of them have already
been done by german photo magazines Color Foto and C't.

They were tested to ISO 16067 standards.

Resolution slow scan-fast scan:
Epson 3200: 1200 dpi
Epson 4870: 1600-1700dpi
Nikon Coolscan LS50 (V): 3600-3600
Nikon Coolscan LS5000: 3600-3600
Minolta 5400: 5000-4000 dpi

 

[Wilfred van der Vegte]

I'm seeing several post about scanner resolution. Some of them have already
been done by german photo magazines Color Foto and C't.

They were tested to ISO 16067 standards.

Resolution slow scan-fast scan:
Epson 3200: 1200 dpi
Epson 4870: 1600-1700dpi
Nikon Coolscan LS50 (V): 3600-3600
Nikon Coolscan LS5000: 3600-3600
Minolta 5400: 5000-4000 dpi

I've read similar results about the Epsons in MacUP (also a German
magazine). But these results you present here make me wonder: In the
first place, the Epson 4870's results are better when scanning faster.
That's remarkable.
In the second place, the Minolta 5400 is tested at two speeds. I have a
5400, but it has no setting for speed. Is the slow scan achieved with
the grain dissolver, or would slowing down the scanning process by using
a slower computer improve the resolution?

 

[Wayne Fulton]

I've read similar results about the Epsons in MacUP (also a German
magazine). But these results you present here make me wonder: In the
first place, the Epson 4870's results are better when scanning faster.
That's remarkable.
In the second place, the Minolta 5400 is tested at two speeds. I have a
5400, but it has no setting for speed. Is the slow scan achieved with
the grain dissolver, or would slowing down the scanning process by using
a slower computer improve the resolution?

In this case, fast-scan means in the direction of the CCD sensor array,
and slow-scan means in the direction of the carriage motor motion.
I.E., horizontally or vertically, as appropriate. Not a speed difference.

I'm wondering what the results "dpi" means in this case?
A wild guess is lpi x 2 ?

Otherwise some cases seem extremly high, according to Nyquist. I suppose a
ranking can use arbitrary units however.

 

[Andre]

I rechecked the figures and the numbers for the Epson 4870 are inverted.

For the Minolta 5400, I can only see that a normal (1x) scan was compared to
a multi-sampling (16x) scan.

 

[Douglas MacDonald]

Andre...
These tests are rubbish. I have the results to prove that. If the 4870 was
such a low grade scanner it would not be capable of higher quality work than
my coolscan was or my Canon film scanner is.
http://www.technoaussie.com/judgment.htm
This is the smallest of several prints I made - some 2400 mm long - from 6cm
x 9 cm negatives. You simply cannot do this with a 1600 dpi scanner.

All the technobabble you lot are going on about must surely have had some
foundation but I tell you now... As a user of many different scanners -
scanning film and photos for a living, the test results published by that
magazine are absolute rubbish.

If they were correct, I simply could not do what I am doing and my results
would be so poor, my customers would go elsewhere... I now have repeat
business for work from this scanner. Personally, I think there might be 1
possibly 2 of you augmenters who actually have a grasp on reality,. The rest
are cut and pasters.

Douglas

 

[Andre]

Andre...
These tests are rubbish. I have the results to prove that. If the 4870 was
such a low grade scanner it would not be capable of higher quality work than
my coolscan was or my Canon film scanner is.
http://www.technoaussie.com/judgment.htm
This is the smallest of several prints I made - some 2400 mm long - from 6cm
x 9 cm negatives. You simply cannot do this with a 1600 dpi scanner.

I'm not there to judge the quality of your print, so I cant comment on it.
Some people cannot understand why one would make prints larger than 16x20
from a 35mm negative while others go as high as 50"x75". This is all
subjective.

I had an Epson 3200 scanner and scanned everything from 35mm to 6x8.2cm. The
scanner test results reflect what I experienced with the Epson 3200.

 

[Bart van der Wolf]

Andre...
These tests are rubbish. I have the results to prove that. If the 4870 was
such a low grade scanner it would not be capable of higher quality work
than my coolscan was or my Canon film scanner is.
http://www.technoaussie.com/judgment.htm
This is the smallest of several prints I made - some 2400 mm long - from
6cm x 9 cm negatives. You simply cannot do this with a 1600 dpi
scanner.

Well, let's see. Assuming, for the sake of argument, the 1600 ppi is correct
and given that 6 x 9 cm is approx. 2.36 x 3.54 inch including some border,
that gives 3780 x 5669 resolved pixels and then some... For 2400 mm long
(94.5 inch) output that equates to 60 ppi, which seems hardly adequate for
close viewing. However, is it???

I think the difference is in the RIP used for printing.

I have a mounted inkjet print measuring 900 x 1340 mm (35.4 x 52.8 inch)
which I printed from a 1949 x 2901 pixel scan from 35mm film I did almost 10
years ago. I scanned with a Kodak RFS 3570 scanner, did perspective
distortion, cropped, and printed with PosterJet as RIP on an HP wide-format
inkjet printer (300 or 600 ppi setting, I don't recall).
This equals 55 ppi, not too different from Douglas' resolved results, still
assuming 1600 ppi is correct for the 4870 scanner.

Although I prefer printing at 600/720 ppi, because that result can hardly
get any better, I can tell you that the enlargement looks, not razor sharp
but, stunning! Really, I had an unsuspecting professional photographer
assuming it originated from 4x5inch film. What one could call a jaw dropping
experience for him...
I just did a small crop reprint with Qimage as RIP, and it looks even
better!

So I think Douglas may be right, and the limiting resolution test as well.
Just make sure you use a very good RIP for printing.

Bart

 

[Robert Feinman]

Just to add to the noise level in this resolution discussion...

It is important to apply unsharp masking to scans made with
models like the Epson 4870. This is not "cheating".
I have a discussion of this on my web site.
Since the spread function of the scanner is fixed it is possible
to partially compensate for it. In the digital domain the software
can apply "negative" density to one edge to help reconstruct the
original image. This can't really be done in the analog domain.
It's the same basic technique used in improving satellite and
astronomical images and the mathematics are well known.
As can be seen from the examples I posted with the Epson 4870
the overall results will look "better".