Epson V750 native resolution

[rdenney]

The issue is not the spacing of the sensels, but their size. The sensels on an Epson see a big fuzzy spot. The higher the scan resolution, the more those big fuzzy spots overlap. Each one might still be subtly different, but at some point the fuzzy spots overlap so much that they don't add anything useful by being more plentiful.

If you evaluate the resulting file in terms of line-pairs/mm or MTF using a proper test target, then the spatial frequency of the fuzzy spots won't matter, as long as there is enough of them to fill in the gaps. All that will matter is the spatial frequency of the detail.

Someone already did this (was it Ben Syverson?), and found that the Epson does not provide a good modulation transfer at spatial frequencies greater than about 1500-2000 effective pixels/inch. But I do seem to recall that one still needed to oversample the negative to attain that outcome. So, it's reasonable to think we might overlap those fuzzy spots closely together and then summarize them into crisper spots using downsampling, or oversample and then limit enlargement to make sure we don't go beyond the scanner's capabilities. But I would never think that the scan resolution should be the same as what we expect the spatial frequency at good MTF to be. I routinely sample at 3200. Do I get more than if I sample at 2400? I dunno. But I'm sure that I don't get less, and it provides some resolution headroom so that I can be sure that I'm getting everything the scanner can deliver.

I also don't enlarge from Epson scans more than 4X happily. Let's do the math. If we get 1500 pixels/inch at good MTF, that is 60 pixels/mm, or 30 line-pairs/mm. If we want 8 line-pairs/mm in the print to provide the most sharpness a viewer can see even on close inspection, then we can print at 30/8=4x. If we set that standard at 5 line-pairs/mm in the print (which is as good as I can see in good light even with the bottom lens in my trifocals), we can go up to 30/5=6x. Most here are pretty confident in their Epson scans (at least in terms of resolution) up to a 4x enlargement, so my math lines up with that experience exactly.

But we still may choose to scan at a higher sampling rate as a strategy to ensure that we get the highest spatial frequency at good MTF possible. I really don't have a problem sending a file to the printer with a pixel density of 1000 or 1200 pixels/inch. My hard disk is big enough.

Rick "who often scans at 2400 and gets good 4x enlargements" Denney

[architorture]

FWIW, just last week I did another quick run-through of the different scan DPI settings with a recent 4x5 negative. I scanned a small portion at 2400, 3200, 4800, and 6400, then downsampled in PS with bicubic sharper to 2400dpi.

The results unfortunately confirmed my previous observation: with my scanner, 4800spi gives the best results for detail, evaluated subjectively of course. I have no idea what actual dpi it is resolving (other than that I believe it looks better than the Imacon's 2000dpi scan).

If I were to rank each setting in order, from best to worst, it would be as follows: 4800 > 3200 > 6400/2400. Interestingly, the 6400spi setting looks worse.

I was really hoping that 3200 would be indistinguishable from 4800 so that the scans would run faster, but alas, it was clearly not up to snuff

Here's a thought: anybody want to do a "collaborative scanner resolution chart test" similar to the collaborative test that is on the site's main page? If someone has a resolution chart they are willing to lend, we could send it around between people and test many machines quantatively... then we could really be talking apples-to-apples!

[Brian C. Miller]

Here's a thought: anybody want to do a "collaborative scanner resolution chart test" similar to the collaborative test that is on the site's main page? If someone has a resolution chart they are willing to lend, we could send it around between people and test many machines quantatively... then we could really be talking apples-to-apples!

I like that idea. My "test target" is a sign on a crane, which is legible at the 6400spi setting. The other easily obtainable target is the USAF test target. There's also a number of other test target PDF files out there, but the problem is that they are subject to being printed on my printer. (An Epson 2200 can make a nice print, but the test targets don't look that good)

Anybody remember the Spirograph? (Apparently the current models are not up to the standard set by the original) Doodle a pattern with one of those and photograph it along with the USAF target.

[Leonard Evens]

You have to distinguiish between the sampling rate and the resolution. 6400 ppi means that the scanner collects 6400 samples per inch. That would correspond to (6400/2)25.4 ~ 126 line pairs per mm. But the optical system of the scanner can't actually resolve that well. As others have pointed out, it is more like 2200 ppi or (2200/2)/25.4 ~ 43.3 lp/mm. Since modern large format lenses and films can do a lot better than that, it follows that the scanner is the limit on the resolution you can expect.

Assuming that an average person can resolve 5 lp/mm at about 10 inches, that seems to say that if you viewed an 8 X enlargement of a 4 x 5 negative at 10 inches, the results would still look sharp. Of course some people can do considerably better than 5 lp/mm, and the calculations ignored the resolution limits of the printer. So taking those factors into consideration, it would seem that someone looking closely at a 16 x 20 print made from a scan might possibly find it sharp. Of course, if the viewer got back to a reasonable distance, a larger print would still look sharp.

[Kirk Gittings]

FWIW. I know this thread is about resolution, but res is only half the issue with these scanners. I own a 750 but virtually never use it for files for exhibition prints. Why? Because besides resolution limitations these scanners lack the ability IME to extract all the shadow/highlight info from a negative processed for normal silver printing which is my basic MO. You can exp/process a neg pretty flat and they will scan OK, but I want a neg that will print well on silver too. So since I have access to an IQSmart, I use that for files aimed at exhibition or for a rally primo scan I send it to Lenny.

But back to resolution limitations. I put a ton of work into a file and while I rarely print above 16x20 why would I want to limit my print size to that size given all the work involved (I actually am never happy with a 750 scan for prints above 11x14-they just look mushy to me unless you over-sharpen them and that adds another look I don't like? So I archived a worked up file at 30x40 from an IQSmart or Lenny to cover any eventualities.

[Ken Lee]

If 4x5 isn't good enough on an Epson, then try 5x7 or 8x10: there are plenty of pixels to choose from :-)

I like Tim Parkin's piece about the questionable utility of scanning LF at resolution greater than 2000 spi,

[Kirk Gittings]

If I understand him I don't really agree with him. Understand he is talking about scanning at 2000 vs. larger ON A DRUM SCANNER not on an Epson-two different propositions entirely and not relevant really to this thread as there is nothing you can do on an Epson to scann actual grain on a 4x5. IME with drum scans when you have actual grain clearly defined IMO you need less overall sharpening and end up with a more traditional look to the image that if you scan below the grain resolution and end up sharpening grain clumps in an to attempt to achieve the same sharp feel. One looks traditionally sharp and the other looks digitally sharpened.

[sanking]

Resolution is what it is. If real resolution of 2000 dpi is not high enough to see grain clumps with an Epson flatbed then it will not be enough with a drum scanner. With most slow and medium speed films you need a minimum of about 3000 dpi to see grain clumps. You might be able to see grain clumps at 2000 dpi if scanning a high speed film, or a film that has been over-exposed or over-developed.

Sandy

[Sylvester Graham]

Boy, am I glad I don't print over 8x10, mostly.