real resolution of flatbed scanners

[Eric Brody]

For the last 25 years I have worked happily in my 4x5 black and white darkroom. I make decent prints and enjoy the workflow but it is slow and going back to reprint an old negative is difficult, even with my notes.

I am now beginning to look at flatbed scanning of my LF and MF negatives. I have no illusions that this hybrid digital work will be lightning quick but at least, if and when I get something decent, it can be repeated with some ease, eg not having to set up all the darkroom chemistry.

I read with interest Ted Harris's comment on this forum that the main advantage of the new Epson 700 and 750 scanners over the previous generation 4990 is when they are used at full resolution, eg 6400. If my math is correct this will produce a staggeringly huge file, even in black and white.

My question, however, is ...if I were to scan at 1600 or 1800, will I really get the quality of an 800 or less dpi scan? Given the "generous" resolution assessment of flatbeds by their manufacturers as opposed to their performance in real tests, if one scans at a resolution which generates a reasonable file size, does the resolution suffer proportionally, eg is it reasonable to "down-res" these machines to get reasonable file sizes. Does one just have to suffer with a huge file to get decent quality even if most of it is wasted bandwidth?

I would appreciate the thoughts and comments of my digital expert colleagues.

Eric Brody

[paulr]

It's not exactly that your actual resolution is half the stated resolution. With a typical flatbed, you'll get the resolution you ask for up to 2000ppi or so. If you tweak it out, by finding the true plane of focus, wetmounting, etc., you might get as much as 2400ppi. But at nominal resolutions beyond that, you won't get any more actual resolution.

You might get some advantages, though. If you scan at double the resolution and then downsample, you can slightly reduce noise.

I don't think you have to suffer too much with huge file sizes unless you're printing huge color. 6400ppi seems entirely excessive for large format. You could probably downsample that file to an 8th the file size before you'd see a lot of printable resolution disappear.

Beware of the promise of the "lightning quick" workflow. If it's out there, I sure haven't fount it!

[Brian Ellis]

I don't think resolution is proportionately reduced. IOW, if the scanner specs say 4000 and the true optical resolution is say 2000 then you're just wasting file size by scanning at anything greater than 2000. But if you scan at say 1000 then you're getting 1000, it doesn't go down to 500 just because the actual spec is half the stated spec. At least that's my understanding, as always (especially when I venture into an area that smacks of high school science) I could be wrong.

[Eric Brody]

Thanks for the info.

What you say, Brian, makes sense empirically. I have an ancient (in digital time) Epson 2450. Recently, I looked at two scans I made from the same 6x6 color negative of Whitney Portal, at 1600 and 2400 respectively and I honestly cannot see any increased detail in the 2400 scan despite the fact that it is twice the file size.

Eric

[Ron Marshall]

I have an Epson 4990, and an Epson 2200 printer. I scan so that I will be able to print at 12.5 x 15.5 at 360 dpi, a 1200 ppi scan is sufficient.

The learning curve for Photoshop is fairly steep at first. Make sure you get a good book. I spent many painful nights before I found my current guide: Photoshop Studio Techniques, by Ben Willmore.

Now that I am proficient, even with a very difficult negative, it takes only a few minutes to scan, correct the scan, and print. But it takes a while to get to this point.

[Keith S. Walklet]

My tests with my 4870 seem to produce a different result than most people. I shared prints of the tests with folks at the recent View Camera Conference in hopes that I could give some credence to my claims. When I undertook the tests, I actually set out expectingn to confirm the opposite of what I found.

What my tests show is that there is a marked difference in qualilty with each increase in optical resolution of that scanner, up to 4800 spi, beyond which interpolation kicks in and the quality drops off dramatically.

With the 4870 scanner, the highest I can scan a 4x5 RBG file is 3200 spi (buffer limit), otherwise I would scan it at 4800 spi. This 4800 spi is not equivalent to 4800 with a drum scanner (different capture technology), but the quality did improve noticeably with each incremental boost of the spi until I got beyond the claimed optical limit of the scanner (4800).

The initial scan file size is a bit cumbersome even at 3200 spi (1 gig!), but in my workflow, I save the raw scan for later, downsize a working version of it to apply all of my corrections. I've covered this in some other threads on this subject (the Layer Transfer Technique) before.

But, one thing I did confirm in recent tests (using Leigh Perry's collaborative scan tranny) is that it IS better NOT to sharpen at the scan level, (which I had experimented with) because sharpening at the scan level introduced noise in the shadows.

So, I now scan at the maximum resolution the scanner will let me (3200 spi for 4x5 color) and don't sharpen, and I don't even use ICE. I just do a really good job of cleaning the film and scanner. I am still amazed at the results that are achievable.

[Doug Fisher]

There also seems to be quite a bit of confusion about the "twin lens" system out there. Many people seem to be implying that you must scan at 6400 for the lens to be scanning at the highest resolution. From our tests, this does not appear to be true. So far, it seems that the way the lens is switched is by the the choice you make in the software (film frame or film holder). Choose "film holder," and our results indicate the scanner will be scanning scanning with the 6400 "designated" lens even if you choose 4800 or 3200. When you compare the same film scanned at 3200 and 6400 and then view them at equivalent maginfication levels, I think you will be hard pressed to see any difference. If there is a difference, I would not be surprised if it was due to some sort of processing difference that occurs somewhere in chain of events between the collecting of data at the CCD level and the final output on your screen.

Doug Fisher

[Ted Harris]

Doug, according to my conversations with Epson the higher resolutiuon lens is switched into place by the software link when you choose the 6400 setting in the scanning software. Any otehr setting and it is not switched into place. Our tests with AIG's T20 Test Target seemto confirm this as the scans of the target done at the 6400 setting in either Silverfast SE or Epson Scan software are in the range of 2400 - 2800 and those done at the 3200 setting are in the range of 1800 - 2000. This is a large difference and we can only explain it in terms of the change in optics. Happy to hear any other possible explanations.

[Ciaran Brennan]

Ted, are the resolution values you measured MTF50 values or extinction resolution values? MTF50 of 2400 - 2800 would be quite good, about the same as a film scanner like the Nikon 9000. An extinction resolution of 2400 - 2800 would be less impressive.

[Doug Fisher]

Ted, sorry that I failed to reply to your previous post. Here are some thoughts on why I came up with my conclusion about the lenses besides just looking at the scan output. I don't claim to be an expert by any means, and this is just a friendly discussion, so no doubt maybe you can point out some faults in my reasoning/assumptions/methodology...

The information I have indicates there are two lenses, each is fixed focus. Therefore, their focus shouldn’t be able to be adjusted by any software setting. The way I understand it, you are either using one lens with its optimal *fixed* focus point or the other lens at its optimal *fixed* focus point (3.0 mm for one, right near the glass for the other). The 6400 ppi lens is supposed to focus at 3 mm and certainly seems to have less depth of field (for lack of a better term) than the other lens which focuses near the glass.

There are two settings in EpsonScan’s software parameters when scanning film – “Film (with Film Holder)” and “Film (with Film Area Guide)”. VueScan and Silverfast also display similar parameters when these scanners are being used. At this point, I am still thinking this is how the lens is switched/selected.

If you scan a test image (we are going to assume the same image for each test scan) at 3200 dpi in the film holder (so it is suspended at 3 mm) with the “Film (with Film Holder)” setting, is it definitely more sharp than the same image scanned (still in the holder at 3 mm) with “Film (with Film Area Guide).” Something definitely changes when you vary between these settings. The same thing is true if you scan the film at 6400 dpi and switch between the two film holder/film area guide settings (but still keep the film in the holder for both scans). This leads me to believe the changing of the film holder/film area guide settings is what determines which lens is used. One scan is sharper than the other.

Since the focus of the second lens for 6400 ppi cannot be varied AND our tests have shown that a scan at 3200 (or 4800) using “Film (with Film Holder)” is definitely sharper than a scan using “Film (with Film Area Guide)”, I came to the conclusion that 3200 ppi and 4800 ppi film scans (probably other resolutions, too) made “Film (with Film Holder)” are using the same lens at the same fixed focus as when using the 6400 ppi highest quality film scan set at “Film (with Film Holder)”. Since there are only two lenses, and both are fixed focus, that narrowed down the possibilities in my mind. As I said before, there may be slight differences in the final scans, but I am guessing the dissimilarities are due to any processing differences (throwing out pixels, downsampling algorithms or whatever).

Ok, let me know if I am off-track on this. Meanwhile, back to working on the new film holder for the V 7xx series...

Thanks,
Doug