Film's true DPI?

[Dante_Stella]

Ok - I just got finished scanning 5,300 black and white negatives on a Kodak F-235 plus, which effortlessly kicks out 3000x2000 scans at the rate of about 800 an hour (assuming, of course, you can get negative strips out of pages that fast). This is an amazing little machine that has a smaller footprint than a small flatbed. Too bad that most of the minilabs that used them knew nothing of the RAW mode.

On going through the results (directed toward my project of cataloguing negatives), I was not surprised to see that 400-speed black and white film shows grain at that resolution. The real surprise was that 100-speed film (TMX) did as well. I checked the grain on TMX, PX, TX, TMY, TMZ, XP1, XP2 Super, and every variety of Kodak chromogenic that exists (or ever existed). So it's not likely that it's grain aliasing.

Are we deluding ourselves that there is any more than 6Mp of image data in a typical 35mm frame?

My suspicion is yes; the conventional wisdom before scanners was that a typical 35mm frame was not much good above 8x10 (viewed close up). For reference, it takes about 240 dpi on the paper to make a final 8x10 dye sub or Frontier print comparable to one optically printed. This brings us back to a 3000x2000 scan. My suspicion is that lab scanners never got past this because their manufacturers (Kodak and Fuji) decided that there was nothing to be gained - at all - by resolving grains more precisely.

Next, given my experience in scanning b/w at 4000 dpi on my Sprintscan, I have noticed that grain aliasing is an issue that seems to get bigger with higher scan resolution and finer-grain film. And if there is only 2000 dpi of information in there, why are we quadrupling the required storage and processing power?

Thoughts?

 

[David Goldfarb]

While the question is valid, I don't think you can make that judgment without looking at drum scans made by a competent operator. Aside from the DPI of the sensor and Dmax of the system, there are issues like the quality of the optics of the scanner, the scanner's actual focusability, film flatness, and wet mounting that all contribute to getting the maximum resolution from film.

West Coast Imaging (http://www.westcoastimaging.com/) seems to think that the maximum useful limit of a drum scan is 5000 DPI, but some people using the latest Aztec drum scanners say they get better results at 8000 DPI (see http://www.eigerphoto.com/services_technology_ep.php/).

 

[retro]

Develop with Pyro and have infinite "DPI" 🙂

 

[amateriat]

Some time back, someone out here on the Internets did some fairly thorough testing along these lines, working his way from a garden-variety desktop film scanner to a drum scanner. His conclusion, which seems to jibe with my scanning experience, is this: assuming fine-grained 35mm film, exposed with excruciating care, with the camera atop a rock-solid tripod, with all the stars aligned, the absolute upper limit in terms of usable scanning resolution is about 6000dpi. (I wonder if Minolta took this into account when they created the 5400, which comes in about 600dpi short of that.)

Of course, most of our film isn't shot this way, under these conditions...


- Barrett

 

[mh2000]

first of all, if you accept that you are seeing grain aliasing, then you have to conclude that you probably have to at least double your scanning resolution... aliasing occurs simply because you are working beneath the Niquist frequency.

I did some simple tests a long time ago comparing negative scans verses high resolution scans of optical prints... making some assumptions of how many pixels it would take to represent the larger Tri-X grain clumps I think I came up with something like that 6000dpi value... but that was based on Tri-X in D76 and only a 3x3 array of pixels being required to capture a visible grain clump (from a 8x10" print... for larger prints it would require a higher scanning resolution).

Again, anytime you see grain aliasing at 100% you have not scanned at a high enough resolution to capture the film grain... grain has actual shapes... you should see them.

What does it all mean? ... nothing really... what really matters is that you are able to get prints that look good.

On prints, I can say that my Coolscan V scans give much better prints, even at 8x12" than a 6MP DSLR will give... in crude terms, based on output, I'd say 35mm scanned negs from my 4000dpi home scanner come close to Canon 5D DSLR results... no better... just different.

 

[Roger Hicks]

What does it all mean? ... nothing really... what really matters is that you are able to get prints that look good.

Quite. Any attempt at an absolute value for comparison is screwing the inscrutable.

In megapixel terms, Ilford's estimate is that a well-exposed, low-speed slide film, taken with a top-quality lens, with the camera on a tripod, equates to at least 18 megapixels.

An imaging software expert of my acquaintance further added that for some subjects -- as he picturesquely put it, 'a portrait of Art Garfunkel with every hair sharp and no jaggies' -- it might even top 30 megapixels.

But as others have pointed out, most pics aren't like this. My own experience is that 6 megapixels is the starting point for decent quality in digital, and about the minimum you'd expect from a neg that was half-competently exposed.

Cheers,

R.

 

[varjag]

A lot of voodoo, bad math and handwaving commonly found around such estimates. Overall I'd side with Roger on that.

Very few people pay attention that resolution should be measured as lp/mm with image through the lens for any meaningful comparison. This will account for Bayer interpolation as well. Taking the sensor resolution figure from datasheet is meaningless.

Below is an old TMax 400 in Rodinal (not the lowest grain combo), and a 200% blowup fragment. You can clearly see my 3600 DPI scan does not approach to resolve all the detail (and I confirm that after checking the neg with microscope).

 

[ruben]

I am of rather low understanding in the issues involved, but it seems to me that when we scan a film negative, we are pixelating it and that means degrading it real potential.

I would say that the best test will be to look side by side a small piece of a negative film converted into an analog print, versus the same small piece with a digital camera and converted into a print too. Print vs print of both small areas can be finally flat bed scaned and shown.

Cheers,
Ruben

 

[Ray Nalley]

All that microscope gazing, though, becomes meaningless when you then print your image on an inkjet printer. Even in analog printing days, transfering any ultimate resolution of the negative to paper was unlikely.

But none of that is important, really. The human eye has a limit on what it can resolve in the real world. The digital camera replaced medium format long before the 1DsMIII came along. All that math doesn't explain the real world results.

 

[David Goldfarb]

Small sensors still have a small sensor aesthetic--DOF characteristics, local contrast, lens availability, etc.--apart from issues of the resolution of the sensor, print resolution, or the resolution of the human eye, so, no, medium format hasn't been replaced by anything yet, except for the newest medium format digital backs, which have large sensors, and for static subjects, large format scanning backs, which have been around for a while now.

 

[varjag]

All that microscope gazing, though, becomes meaningless when you then print your image on an inkjet printer. Even in analog printing days, transfering any ultimate resolution of the negative to paper was unlikely.

It transfers just fine. This image is fine to print into 30x45cm sheet, bigger than that I just don't have suitable trays for chemistry.

 

[varjag]

I would say that the best test will be to look side by side a small piece of a negative film converted into an analog print, versus the same small piece with a digital camera and converted into a print too. Print vs print of both small areas can be finally flat bed scaned and shown.

Ruben, problem is that most flatbed scanners suck, and commodity inkjets just as well. Not to mention the processing required in such a chain.

 

[like2fiddle]

Ok then, so for the average joe who uses film and then scans it on an average consumer flatbed to be printed on an average consumer inkjet, what is the bottom line? What would be the recommended scan resolution to use in order to print an 8 x 10 that would appear, to the naked eye, to equal a wet print?

 

[literiter]

Right now it looks as if both systems were at a tie. One having marginal advantages over the other. However, just in case it hasn't happened yet, I'd say digital is about ready to wallop film in as many categories as one could suggest. DPI could just be one of them.

I love the argument but it almost seems futile. It is a rather like trying to prove the existence of the Ether or of a bearded Deity.

I prefer the aesthetic and permanence of film.

 

[varjag]

Ok then, so for the average joe who uses film and then scans it on an average consumer flatbed to be printed on an average consumer inkjet, what is the bottom line? What would be the recommended scan resolution to use in order to print an 8 x 10 that would appear, to the naked eye, to equal a wet print?

Honestly, no idea. But thing is, average joes have quit film for digital some time ago now. And if you have, for whatever reason, affinity towards film, there's not much reason to do sloppy job. IMHO.

But true, there is no lower bound for poor results 🙂 Just as earlier if one used shabby enlarger, poor film processing and inadequate camera technique, the results can be just terrible.

 

[varjag]

Right now it looks as if both systems were at a tie. One having marginal advantages over the other. However, just in case it hasn't happened yet, I'd say digital is about ready to wallop film in as many categories as one could suggest. DPI could just be one of them.

I agree this is most definitely true, and just as you are am into film not for resolution, sensitivity, latitude or whatever.

 

[Ray Nalley]

If you don't print, then all this DPI stuff is of no importance to you. 72dpi and 300,000 dpi all look the same on the web or on your monitor screen.

 

[varjag]

Same goes for digital imaging though. VGA cameras are enough for everyone 🙂

 

[nzeeman]

i dont think they are tie. they are totally different things. we cant compare those things.
for example some people like vw beetle , they dont care if there are faster , more comfortable car etc. i think we should stop thinking about resolution and we should use whatever we like.

 

[squirrel$$$bandit]

I've been scanning negs on my Epson V500, and 11x14 prints (from Mpix, admittedly; I don't have a printer) at 2400 are completely indistinguishable from ones scanned at 3600. Both look pretty good to me, by the way.