Film - The Color of Truth

[Daniel Browning]

By the way, I didn't give any resources above for the color accuracy stuff, so here's one:

"The Science of Digitizing Two-Dimensional Works of Art for Color-Accurate Image Archives - Concepts Through Practice", Berns, Roy, 2000, Munsell Color Science Lab Journal.

They tested 24 colors, the average delta E was 2.8 for digital but 4.9 for film. Colors that were difficult for film, like reds, had three times the delta E.

Are you saying film has better dynamic range, resolution, and contrast than digital?

Essentially, yes, I was saying that they are valid reasons for shooting film if it happens to be true in the given set of circumstances. I guess the underlying assumption is that it's *possible* for it to be true, and everyone agrees that it's possible under the right circumstances. But the truth is that I don't really have enough knowledge to make a judgement one way or the other for the "general" circumstance, so I probably should have refrained from stating such an underlying assumption.

I think I read somewhere that film has 9-10 stops of dr.

Good negative films have 14 stops of dynamic range. 16 (!) for carefully processed B&W. The 9-10 stops may have come from Roger Clark's film-vs-digital test, which was badly flawed in several ways. (The clarkvision site has a lot of reliable data and few flaws, but that is one of them.) I don't have any references on this, but I'm pretty confident about it, and I would like to think that GIYF could do it for me if pressed to action.

Doesn't that mean that digital has surpassed film here, too (5D II has 11 stops at iso 200)?

Film's dynamic range is sort of fixed, no matter the film size or print size, due to its nonlinear nature. However, on digital, one of the advantages of linear response is that print size can be traded for dynamic range. In other words, exposure can be reduced, then increased nonlinearly in post production (increasing noise power at the nyquist frequency), then the nyquist frequency is moved by using a smaller print size, resulting in the same noise as the original frequency.


Using this method I can easily get more than 14 stops with my 5D2 for wallet-sized prints. I think dynamic range in digital will *really* improve when someone finds a way to simulate larger full well capacities (e.g. read-reset seamlessly inside a single exposure). Read noise will continue to improve as well.

As for resolution, I have the figure "100 lines/mm" in my head for film... maybe that's an old number, or maybe it is wrong, or maybe better film does better. But at 100 lines/mm, 35mm film has about 9 megapixles (no bayer array, though, obviously...).

I don't really have anything interesting to contribute to a discussion about resolution because I haven't tested this myself, nor have I researched it thoroughly. The only thing I will say (and this is not an assertion) is that the various film tests I've read seem to place 35mm film close to what 17 MP Bayer would be with a good OLPF. I don't have any references handy, so I'm not going to stand behind that.

Whatever the actual resolution of film is, everyone would readily agree that it's multiplied by the format size; so that 645 has more than 35mm and 8x10 has more than 645. Of course, the standard assumption should be that we're talking about 35mm film.

What are limitations on contrast for digital?

Contrast is such a nebulous and all encompassing term that I can now make up any definition I want and apply it retroactively [6]. I was actually thinking of the tone curve, which is automatically built-in to film and has well known and documented characteristics. With film, you get a certain attractive tone curve, with a very good amount of highlight headroom and nice midtones/shadows.

Whereas on digital you can have any tone curve you want, but sometimes it's difficult to get a good one, especially if your idea of "good" is film-like. The reason it's hard is that all the JPEG engines and raw converters are tuned for just 2.5 - 3.5 stops of highlight headroom, which is not enough to get smooth film-like highlight rolloff. The reason they are this way is that they were designed when digital was more noisy, and have not been updated (in most cases) with improvements in cameras. So the difficulty is in getting more stops without losing contrast in the midtones, color casts in any zone, or too much noise in the shadows/midtones. Some software is better at this than others.

In short: film isn't dead yet, but the last will and testament has been signed.

[Jon Ruyle]

Thanks, Daniel, for another informative post.

However, on digital, one of the advantages of linear response is that print size can be traded for dynamic range.

Yeah. I got my dr data from dxomark.com. I'm not sure, but I think they mean "per pixel", ie, each pixel has 11 bits. But measuring per-pixel seems to favor cameras with low pixel densities, for the reason you just mentioned. Even so, the highest rated camera in dr is the d3x: even though it has 24 megapixles it beats out the d3.

I think dynamic range in digital will *really* improve when someone finds a way to simulate larger full well capacities (e.g. read-reset seamlessly inside a single exposure).

If you could do that fast enough, you could pretty much have all the dr you want, I suppose. I talked to some guy who said he worked on a special purpose ccd (or some sensor) in which something like that was done for longer exposures. Not sure I believe him, but such a feature sure would be cool for astro (where you really could use the dr).

In short: film isn't dead yet, but the last will and testament has been signed.

Sure. But hey, I know guys in good health and under 40 about whom you could say the same thing []