6x9 Film vs Digital

Here's 100% crop from the digital shot.

And the same crop from 6x9 film.

But if you search the scene very hard it's possible to find areas where 6x9 film can show its greater resolving power. Whether this actually matters is a different question, in an A3 inkjet print you just can't see this level of detail. Here's the digital shot rezzed up.

And here's the film version.

Digital also has other issues to contend with. The Canon 1Ds has moire very well controlled, and after taking thousands of shots I've hardly had any problems at all. But it's not been eliminated entirely. And on the this shot you can see a trace of moire in the roof tiles, in the final print it's hardly visible, but it is there.

Another digital issue is colour fringing out in the corners with wider lenses. There's some evidence of fringing here. A 24mm shift lens is a torture test in this respect, and to put it in perpective it's hardly intrusive in an A3 print. However, I can't pretend fringing doesn't exist.

However, colour fringing can also be found with film. Here's a crop from the Reala negative and there's some fringing here too. A lot less, but it's there.

Thanks for the comparison!

A few thoughts:

-I very much like that you compared them on paper and not only on screen. The fact that they is very little difference at 11x14 is a better data point than just about anything else that can be said.

-About the "ultimate" resolving power: without a doubt film wins. While the image may have been slightly oversharpened for screen use, it seems to me however that digital offers more contrast in medium-sized details. Worth mentioning that film loses when it comes to grain at large enlargements.

-Color fringing: I think it's time that everybody realize that there is some color fringing everywhere, but that until the advent of high-resolution scanners and digital cameras it had never really been visible. I see color fringing on my 10D with a 28/2.8, sometimes more than a pixel wide, i.e. about 10 microns, or about as much as on the 828. Worse, I see color fringing of at least 2 pixels on 1600dpi scans of large format slides (admittedly with a cheap lens), which translated to 30 microns (and that's when shooting at f/16).

Interesting, thanks. Looks like your conclusion is valid -- there's far more detail on the 6x9, but it's detail that's basically unnecessary at 11x14 or smaller.

Some people will surely ask, how did you scan the 6x9? One interesting test would be to make your best large darkroom print from the negative and your best inkjet print at the same size from the digital file, and then scan both on a flatbed.

Nice, objective comparison.

Superb job.

Thanks for your report Gary!

a few questions:

What speed is the Reala film?

How did you scan the film?

Did you sharpen the D1s image? It looks like it in the alarm insert (see the fringing)

Did you sharpen the film scan?

Interesting. I, too, have two questions:

- What is the difference in the ability to do movements (both tilt and shift) with the Schneider XL 58 mm and the Canon 24 mm T/S?

- Will you be using 6x9 after this? At 11x14 it seems to me that digital would be much more practical with similar quality, but at larger sizes film might have an advantage (at least in resolution, if the color rendition is good.)

Up to 11x14 I would tend to say that the difference is minimal, but its still there. It's for sizes larger than this that digital starts to fall apart when compared to mf film.

Digital's minimum acceptable print resolution is around 240dpi, while film is at 300dpi at the scanner's real output resolution, not the advertised resolution. For example the Nikon Coolscan 8000 real resolution is 2850dpi.

Dslr's have a long way to go before they can equal mf film in term of print size. For practical purposes, there are just about equal at sizes up to 11x14"

Let me quality this further: any resolution less than those that I've stated would be at less than optimum resolution.

Glad you found this useful, answers as follows,

-Reala is nominally rated at ISO 100 but I tend to rate it at ISO 64 unless there's the time to fully analyse the shot and make sure the shadows receive full exposure.

-Film was scanned on a Nikon 8000 with the glass plate negative carrier. I've been scanning for four or five years but I'm the first to admit I'm not the world's best, in fact the more I learn the more I realise there is to learn. Scanning's difficult!

-Some sharpening is done automatically and unavoidably by the Canon 1Ds, so I also did a little capture sharpening on film to try and bring it into line.

-From memory Canon's 24mm Tilt & Shift lens has an image circle of about 65mm, the Schneider's image circle is about 160mm.

-I'll be carrying on using both film and digital. I still can't get a glossy black and white print out of an inkjet that I'm satisfied with (matte B&W is a different story), so that's a good reason to stay with film and a darkroom. A photo library that's kind enough to carry some of my stuff prefers 6x6 to 6x9 transparencies, so that's what I give them. Often I use Reala for that extra couple of stops of latitude. And there's some shots that cry-out to be printed big with masses of minute detail, I'm doing a series at the moment of sweeping rooftop shots of European cities, and anything less than 20"x30" would just short change the subject.

Gary – this is a very nice report and very nice practical approach. i think for someone with an 11MP camera and a scanner for 120film this is very informative - and for the rest of us as well.

however, i think one should be very careful to interpret the results in the way that the test shows the capabilities of a 6x9 system.(someone might do that even though you were careful not to do so) i think if you print selected areas from the 120 film at higher magnification than shown from the scans and compare this to your images with the digital camera you will find that the 120 film has an clear advantage. i bet you a case of oranges :-) for example the prints from the 120 film will not show the noise from your scanner. also the 120 film can show more details than what you have shown.

what you show to us -and that is certainly worth our attention - is that a top DSLR camera can make 6x9 plus scanner obsolete under conditions that are favorable (dynamic range) for the digital chip recording. I would complete agree with that from my own experience (i also use 120 film and a LS8000 as well as DSLR imaging) except for very wide angle lenses where film based cameras still have a pretty major edge. (i will therefore not mention to compare say a Mamija7 equipped with one of the legendary wides for backlit landscapes – this would be unfair. )

could you agree to that?

cheers walter

http://www.nikonusa.com/fileuploads/pdfs/8000.pdf

Let's do the math one more time. 4000 dpi equals 2000 lp per inch. (one black and one white)

Divided by 25.4 gives you 78 lp/mm

The ccd manufacturers say and objective testing proves that .7 times that (the Nyquest frequency) is the best that they can do, so that is 55 lp/mm.

Objective testing of film shows 80 to 200 lp/mm.

If someone really wants to compair film to digital at least do what I did and get a 12000 dpi drum scan using a photo multiplyer tube.

This is just another very detailed compairison between digital direct capture and a scanner.

This is like trying to measure your own weight with a postage scale and concluding that you weigh 25#

Neal, if I understand you correctly you're saying that only a 12,000 dpi drum scan can do justice to film, basically because film's resolution can be up to 200 lp/mm.

I'm not an expert, but I understand that stratospheric film resolutions like these don't have much in common with real world photography. They're derived from contact exposing film against micro-engraved glass plates or using contrast ratios of 1:1000. Neither are techniques that have much relevance for me and my work.

The MTF's of high quality film posted on the manufacturer's web sites seem to show contrast transfers of about 30% at 60 lp/mm. That's a more practical benchmark. Most authorities describe 30% transfer as "poor" or "marginal", and say that 20% represents an effective floor for usefulness. Doesn't that suggest film's practical resolution advantage maybe isn't that great after all?

Over the years I've had plenty of drum scans made. And they usually deliver better results than I can achieve with a Nikon 8000. But that advantage is generally most evident in terms of colour, tonality and shadow detail. True there's sometimes a resolution benefit, but on the final print it's usually pretty marginal, and it's the rare image where that fractional gain is the difference between success and failure. Even Zeiss, who've forgotten more than you and I will ever know, concluded that above about 40 lp/mm there's little increase in perceived image quality.

I accept that 6x9 film delivers more detail than an 11MP sensor in larger prints, I think that my post makes that conclusion fairly clear. But the tone of your reply suggests that I was short-changing film in some way. I disagree. However, I've an open mind and if there's evidence that a 12,000 dpi scan can liberate up to 200 lp/mm of resolution from film, then I'd appreciate seeing it.

Detail is only useful if you need it. Most photographs don't need the difference between a high end digital camera and film.

However, if you want to compair detail then use methods that have been used for almost 100 years or come up with your own that address the goal of your test.

You can download test charts off the internet for free, or use a daily newspaper.

If you continue to back away from the target, at some point in time for the same field of view the newspaper will no longer be legable on the digital and will be legable long aftwards on film, even for 35 mm.

You can buy 100X microscopes at the pawn shop or the acution site for $100 there is no reason to muddy up your results with any type of scan.

I took a 4x5 slide and had it scanned at 12000 dpi however, and could not see anything under the microscope that I couldn't see in the scan.

However, I could read construction signs at a range of 1 mile where the type was less than 1/10 of a mm tall. If you assume that it takes at least 8 to 10 rows of pixels to reproduce type with jaggies than your results are off by a factor of at least 6 as my tests showed that 35 mm film was about 150% as good as 10 meg.

The FBI's tests showed 16meg for 35mm.

http://www.fbi.gov/hq/lab/fsc/backissu/april2002/swgitfield1.htm

This link talks about how many pixels it takes to archive printed material.

http://www.dlib.org/dlib/october96/cornell/10chapman.html

This link shows some of the tests that I have run.

http://truckgenerator.com/subdomain/sueandneal/digital__film.htm

However, the tripod was staked into the ground, I used a mid range f stop, lens shade and focused very carefully.

Beyond my tests, Photo Techniques magazine just did a Canon Digital to Canon film test that seems scientifically supportable. Doesn't it strike you as a bit odd that when people don't use scanners in there tests that they get such different results from yours?

I really don't understand why when people make these Digital vs film claims they don't use some type of target that can be read objectively and don't read the film optically.

Or maybe I do?

Nice report! Gary I would have expected the 6x9 to do much better, considering a Digital sensor does not respond to light from very oblique angles (i.e wide angle lens) well. (But that would happen towards the edges..). Would you kindly post some sample from the edges?

So what happens beyond A3 inkjets?

"You can buy 100X microscopes at the pawn shop or the acution site for $100 there is no reason to muddy up your results with any type of scan."

I think the using the scanner is a good method because if compares a method of getting to print. Sure, there may be lost detail in the film, but who displays a piece of negative film (or slide) in a frame on the wall? By necessity, film must be duplicated to make a print and information will be lost. From my experience, scanning on the nikon 8000 seems to make a more detailed print than an optical enlargement from the negative. If not, the results are certainly close.

Gary's comparison shows that shooting 6x9 film and scanning on a ccd scanner is a good alternative to shooting with an $8000 digital camera. Sure it takes more time, and cost more per frame by a long shot, but for myself, I'm not shooting in a professional capacity and going through a lot of film. I would be satsified with the Canon 1ds if the price were substantially lower as I'm sure it will be sooner than we expect.

-bruce

but I understand that stratospheric film resolutions like these don't have much in common with real world photography.

They don't, unless your profession is looking for terrorist bunkers from 60,000 feet. Otherwise, I dismiss anybody that bothers scanning film of any sort beyond 2000dpi and then bragging about how much wonderfull resolution it has. If you can justify a quality factor with film scanned beyond 2000dpi, you must not object to the image quality of small sensor digicams shot at ISO 800.

I took a 4x5 slide and had it scanned at 12000 dpi

Wow, that yields a file 48,000 x 60,000 pixels.

Doesn't matter what format film you shoot because the resolution is the same, and beyond 2000 dpi film start to break up rapidly regardless of what format of film you use. Under that it can compete with digital capture at 1:1 with no arguement.

Thank you for the answers. This is an interesting comparison, particularly since I tend to shoot more larger film sizes these days, using movements.

I think that maximum scannable resolution depends a bit on the film. For practical photographical purposes (ie. overall image is excellent), I feel that the sweet spot lies between 2000 and 3000 dpi, depending on film and lens. Scanning at higher resolutions can with some combinations be useful and then one can always downsample if there are aliasing artifacts present in scans at smaller resolutions.

Is color fringing a new term for Chromeric Aberration? which largely is a lens charactistic, and the way liaght reacts differently with different emulsion layers. And really has always exited but few understood it except that expensive lenses are sharper than cheap and this color blur at edges of things was mostly thought of as lens sharpness/quality more than something to do with the negative or digital file.

As Kermit The Frog said in The Muppet Movie, "I don't understand any of this."

For thoses interested read here http://www.luminous-landscape.com/ for articles on film vs digital. Michael Reichmann does extensive comparisons and practical tests on the subject. Larry