Hacker News new | past | comments | ask | show | jobs | submit login

He's talking about anti-aliasing filters (http://en.wikipedia.org/wiki/Anti-aliasing_filter), which essentially blur the image enough to mask moire (http://en.wikipedia.org/wiki/Moire).

Because this is the internet, I feel obliged to point out the inaccuracies of some of his other statements:

>>Film was and is to some degree still capable of more resolution, more dynamic range, better color fidelity and less prone to weird errors in the way that a Bayer sensor does<<

First of all, film and digital sensors of comparable sizes have nowhere near the same amount of resolution, dynamic range, or color fidelity. Digital sensors simply murder film in every category. (You can make the argument that B&W film still has some an edge in terms of dynamic range, but considering you're matching it against RGB sensors, I'm not sure that's fair.) Even 8x10 film is now out-resolved by medium format backs.

In terms of color fidelity, there is no film emulsion that will get you more accurate results than properly color-managed digital equipment. Which is exactly why museums and people who care about accurate color reproduction use digital capture.

Finally, the anti-aliasing filters aren't exactly a big deal. Yes, they degrade image quality. Yet, even with them on most consumer sensors, those sensors still massively out-resolve, and produce sharper images, than comparable film. Also note that most medium format backs don't have AA filters.

So, while somebody might like using film (or may need to in certain circumstances) the arguments presented aren't exactly the best justifications for doing so.




Just 'cuz I can't let this pass without comment, same as you can: let's take a nice color film and compare this. Say, Ektar 100; in the past I might have used Ektar 25 for this but 100 is a nice speed good for outdoors shooting or controlled lighting conditions. I'm using a Kodak film here both because we're talking about Kodak and because they offer superb information on their products. I'm using a color film because technical B&W films exist and are superb for their application but are not really suitable for general photography, plus are more directly comparable to digital.

Film is an analog medium, so we need some way of measuring resolution. The accepted way is line pairs per mm, or cycles per mm; a dark line next to a light line. A Bayer sensor because of the way the color filter works requires four linear pixels to resolve this same object.

Now Kodak's papers on the subject indicate that Ektar's color response desynchronizes around 20 lp/mm and reach 20% contrast at around 65 lp/mm for the red channel and 80+ for the blue and green channels. Since we're dealing with black and white lines here, we use the blue and green filters as they will still be visually distinguishable. We can argue whether further than 20% contrast is relevant but it seems like a decent enough stopping place. Fine. So the film can resolve 80 lp/mm.

Actual gate size varies a little but generally it is considered to be 36mm x 24mm. So that translates to 11520 x 7680 or an 84 megapixel camera, under perfect conditions.

In realistic use because of diffraction you are unlikely to get 80 lp/mm on the film, the film is not likely to be flat, you probably can't hold the camera perfectly rigidly, etc. It is generally accepted by archivists that 4000 DPI scanning is good enough for archival, which translates to about a 5700 x 3800 image or about 20 MP. In 8x10 that translates to about 1.2 gigapixels, although when you get to that point 4K DPI is probably overkill due to the diffraction limit reached at typical 8x10 working apetures.

As for dynamic range, Kodak doesn't show when Ektar 100 shoulders off but it has minimum 11 stops of dynamic range. I've used Ilford's XP2 when shooting a wedding before, and that has a bizarre shoulder with something like 13-15 stops of dynamic range. Digital cameras are typically more like 7 without using HDR, and HDR is not well suited to anything that isn't completely static.

Finally we get to color fidelity. Kodak worked very, very hard to be as precise as possible with their films and largely succeeded; you can find in the Ektar 100 tech document precise filter recommendations to cancel out the oddity of florescents if you know what sort of florescent you're working with. If you are fortunate enough to use flash, all flashes are designed to emit a precise blackbox radiative spectrum that corresponds to daylight, where the films are balanced. EPP, which is a long obsolete reversal film that I have worked with, was used a lot in catalog photography where it was vitally important that the color used in the catalog must match precisely the color that was photographed, and it was very good for that. Digitals are theoretically more flexible and thus have the potential for better color fidelity, but to do that you basically need to know precisely which lights you are using and at that point you're in the same boat as film users.

Now, is all this really compelling reasons to go to film? Not really. But I spend all day every day working in front of a computer and I'll be damned if I'm going to spend my hobby time calibrating my monitors and messing with scanners and printers (each in a dead heat for most miserable fucking peripherial ever), so I do traditional darkroom stuff, occasionally including some color work.


1. I didn't check your numbers, but I'll assume you're correct. However, in my experience, film does not even come close to approaching those theoretical lp/mm numbers. This test, done with Ektachrome instead of Ektar, reflects my own findings: http://www.luminous-landscape.com/reviews/cameras/iq180_vs_8...

2. Ok, you're right. Some films will certainly outperform some digital. But, the new MF backs will shoot 13-15 stops as well, so who wins? I'd argue that, on average, digital trumps film. I've shot mostly velvia (the most popular landscape/wildlife film for 30 years?) and provia, and those can't compete with even my 3 year old dslr. (see http://www.clarkvision.com/articles/dynamicrange2/ for tests done with an 8 year old dslr)

3. Get a color checker (http://xritephoto.com/ph_product_overview.aspx?ID=1192). No filters necessary, and precise color no matter what lighting conditions. And, you'd be able to eliminate one of those miserable fucking peripherals.

FWIW, going digital has tremendously simplified my workflow and allowed me more shooting time with less dicking around with equipment time, and much better quality prints (and I make really big prints). Cheers.


Oh, no question that you have a better MTF from an Alpa than you would from an 8x10 at f32 (a perfectly reasonable working aperture for 8x10). I think the 8x10 can still out resolve the Alpa but you're getting into that ugly 20% part of the MTF curves. The rule of thumb I've had for a while is that digital has fewer, better pixels; the digital back hits a wall beyond which you cannot proceed any further, but at and up until that wall the results are superb. I think the current sweet spot is speculated to be 6 cm x 9 cm cameras, mostly because of diffraction and film flatness issues. And reversal film has notoriously poor dynamic range, and mediocre color correction; that's one of the things that pushed me to doing all negatives in my photography.

A professional is probably only going to use film these days if they want a specific effect; say, infrared, or sometimes people fart around with Holgas and call it art. For an amateur, the $20,000 digital back that can compare with a 4x5 with some decent lenses costing a tenth that much is a much harder sell. And that's pretty much where I sit; I'm not going to spend $50,000 on a top of the line MF outfit. I might spend the $1.2K for an X-100, though, still mulling that over. A couple thousand buys a lot of film and developer.


regarding hobby stuff - have you looked into computational photography at all? coded apertures, plenoptic cameras, etc?




Registration is open for Startup School 2019. Classes start July 22nd.

Guidelines | FAQ | Support | API | Security | Lists | Bookmarklet | Legal | Apply to YC | Contact

Search: