Achievable resolution with Epson V700/V750/V800/V850

[Ted Baker]

For what it is worth...

More than nothing, because you make a excellent point. Another way to say what have said is that oversampling can produce a better image, with less artefacts. Which goes back to OP original point, how do you get the best out of what you have, in a practical way.

 

[philosli]

Very interesting work.

How do you decide where to put those red arrows?

The V7XX scan seemed to me that the scan didn't have proper focus. Even the numbers are blurred. Unless the scanner is really crappy.

My experience is different then yours and I consider myself pretty competent with post tools.
Here are scans from an Epson V500, V7XX and Coolscan of the same frame of 35mm film (Fuji RVP - ISO50) shot using optimal setup with a 4 X 4 arrangement of test charts. The Epson were shot at various resolution settings and I also scanned with and without ICE.

Coolscan full res -> http://www.fototime.com/02BB797801DCA89/orig.jpg

Epson V500 full res -> http://www.fototime.com/33269E445D10043/orig.jpg

Epson V7XX full res -> http://www.fototime.com/11F59FA46FF9497/orig.jpg

When using test targets and shooting under optimal conditions the differences in achievable detail - provided the detail has been captured on the frame of film, is very obvious and cannot be manufactured in post. If you don't use good film, don't take care in taking the shot, don't use good equipment or the target itself doesn't have the detail then the difference may not be so noticeable. Incidentally, no post sharpening were applied to any of these.

My use of these Epsons is not optimized as intended by the OP so I don't know if these could have been improved to equal that of the Coolscan.

I was also curious to compare just how much detail can be resolved by the Coolscan relative to DSLR scanning so I used a Pentax K20D+Autobellows (14.6MP) and Nikon D800+Autobellows (36.3MP) and Coolscan on the same frame of Kodak Techpan shot at ISO25 and processed in Kodak Technidol and got the following results shown below.

• Bottom left shows test target arrangement. Center area bounded in red are shown as 100% crops.
• Above it is 100% crop from Pentax K20D.
• Above that is 100% crop from Coolscan.
• Above that is 100% crop from Nikon D800.
• To the right is an optical magnification (4.5X) that clearly shows real detail not resolved by these methods.

Full res version -> http://www.fototime.com/8372250EA44CB06/orig.jpg

BTW, you'll notice that even though the D800 has more pixels then a 4000dpi Coolscan file, there is very little difference between the two. Of course there is no dust and scratch removal using DSLRs and the Coolscan+Nikonscan ICE is the most effective tool available.

 

[Les Sarile]

Very interesting work.

How do you decide where to put those red arrows?

The V7XX scan seemed to me that the scan didn't have proper focus. Even the numbers are blurred. Unless the scanner is really crappy.

The red arrows are positioned to the point where I can't see between the lines.
The V7XX does look out of focus by comparison of course.

 

[alanrockwood]

The red arrows are positioned to the point where I can't see between the lines.
The V7XX does look out of focus by comparison of course.

Have you experimented to find the best film holder height?

 

[Les Sarile]

Have you experimented to find the best film holder height?

I have not.

 

[alanrockwood]

I have not.

It would be worth experimenting with the film holder height. Each individual specimen of the Epson scanners is a little different, and it is not unusual for there to be a 1mm or 2mm difference in optimal film height from one scanner to another. It often makes a noticeable difference to tune it up for the optimal film height.

 

[alanrockwood]

I found an article that addresses the question in my original post:

Dead Link Removed

In that person's hands an Epson V/700 achieved a resolution of about 2800 line pairs per inch. That was after finding the optimal film height, and apparently it was also after finding optimal sharpening settings. That person also found that a scan resolution setting of 3200 gave the best resolution. A higher setting produced lower resolution, and a lower setting also produced lower resolution.

Without sharpening the resolution was around 2400, which is about what other people usually report.

2800 isn't too bad, maybe slightly better than 70% of what a 4000 dpi scanner can do, but it isn't great either. However, do you remember when people used to say that 2700 is all you need? I do. The claim was that going higher would only result in resolving grain. Of course, that theory was wrong, as was demonstrated when good 4000 dpi scanners came out. Then people said there was no point in going higher because going higher would only result in resolving more grain. Then the higher resolution Minolta scanners came out and proved that theory wrong as well.

My somewhat wild guess (thought not completely wild) is that if one had the most carefully crafted 35mm image, photographed under the best conditions, using the optimal lenses and film choice, using the optimal aperture setting you could probably see an increase in quality (though a subtle increase) up to about 8000 dpi. That would correspond to a Nyquist limit of 167 line pairs per mm, which is probably a bit more resolution than the lens/film system could produce, but I suspect that the ability to to better resolve the grain under those conditions would probably give a slight increase in quality, perhaps through a lower level of grain aliasing.

 

[doctorpepe]

In looking at which sampling rate to use in scanning, doesn’t it depend on the resolution of your final product? If your printer prints at 600dpi, for example, and you are scanning an 8x10 neg to make an 8x10 print, then you should scan at 600dpi.

Likewise, if you are scanning 120 film to print a 10 x 10 print on the same printer, then the 120 film would need to be scanned at 2” x 5= 10 or 600 x5 =3,000 dpi.

The print should show no scanning artifacts as long as the scanner scans (without interpolation) at the required dpi.

And so on. If your printer is better than 600dpi or less, you would adjust the scan dpi according to the size of your final product.

 

[dokko]

In looking at which sampling rate to use in scanning, doesn’t it depend on the resolution of your final product?

yes, the final product of course changes scanning requirements. a 4x6" print needs much less (5 times less) resolution than a 20x30" print of the same negative.

If your printer prints at 600dpi, for example, and you are scanning an 8x10 neg to make an 8x10 print, then you should scan at 600dpi.

here is where things get fuzzy and it's useful to differentiate between ppi and dpi.
in the digital file, the image is described in ppi, where each pixel can have several million (or trillion) different colors.
in the printer, the dots are described as dpi, where each dot comes from one ink, so usually 4 to 12 colors.
that means to mix one pixel in the image, you usually want to use several dots in the printer.

in the file, you usually want 200-400ppi for a high quality print, with 300ppi often quoted as the standard.
in the printer, most people aim for more than 600dpi for a high quality print, usually around 1200-2500dpi.

Likewise, if you are scanning 120 film to print a 10 x 10 print on the same printer, then the 120 film would need to be scanned at 2” x 5= 10 or 600 x5 =3,000 dpi.

ok, let's use this as an example:
If we aim for a a 10x10" print of highest quality, let's say for a epson inkjet printer, I'd say we should have a file with about 360ppi (the printer will print at 1440 or 2880dpi).

so that means a file with 360 x 10 = 3600pixels width

a 6x6 neg is 2.2" inches wide, so we'd need to scan with 3600 / 2.2 = 1640ppi on the scanner.

in practice, no scanner will reach the full resolution at native scanning, so if I want absolutely best results, I would aim for about 20-30% oversampling, i.e. 2000-2400ppi scanning resolution.

In practice the term resolution is even more complex. for example 1800ppi on an Epson V850 will look noticeably softer than 1800ppi on a Nikon 9000 or an Imacon 848 or my dokko scanner.
there are also other factors like scanning noise, dynamic range, color etc, which are hard to put in simple numbers (or rather, every manufacturer uses the method that makes their product look better than it is).

but in practice, the standard version of aiming for around 300ppi on print size, and using a high quality scanner and printer with careful processing will give excellent results.

 

[loccdor]

2800 isn't too bad, maybe slightly better than 70% of what a 4000 dpi scanner can do, but it isn't great either.

Resolution as we experience it is better measured as an area (square) than as a line. So a 2400dpi to 4800dpi jump is really a 4x in total detail. A 2800dpi to 4000dpi jump is a 2x.

I found that 2400dpi for a 35mm frame was good for printing up to about 10 inches but fell apart at 15.