Where cameras stand in Dynamic Range: Film vs Digital

Many of you might have seen The Great Camera Shootout 2011 from Zacuto. Episode one, The Tipping Point, dealt with Dynamic Range: Film vs Digital:

Exclusive Bonus: Download my free swipe file on how to shoot night scenes well (PDF file optimized for mobiles and tablets).

The results of the test surprised many, and were supposed to herald the beginning of the end for film. Two years later, Kodak is out of business, but is film dead? Take a look at this year’s ‘important’ Oscar nominees:

Camera Film Stock DI Master
1 Panavision Panaflex Millennium XL2 Lincoln Kodak Vision3 4K
2 Anna Karenina Kodak Vision3 2K
3 Django Unchained Kodak Vision3 4K
4 Arricam Les Misérables Kodak Vision3 2K
 5 Argo Kodak Vision3 4K
6 Silver Linings Playbook Kodak Vision3 2K
7 Arriflex 16mm Beasts of the Southern Wild Kodak Vision2/3 2K
8 Arri Alexa Amour Arriraw 2K
9 Skyfall Arriraw 4K
10 Life of Pi Arriraw 2K
11 Zero Dark Thirty Arriraw 2K
12 Red Epic The Hobbit: An Unexpected Journey Redcode 2K

This table might not be entirely accurate, and is based on imdb.com.

Documentaries and shorts have been using video cameras to tell their stories since they first came out, so I haven’t included them in this list. The purpose of the list is to only show that many filmmakers still choose to shoot on film for aesthetic reasons.

The DI Master column was added to show that both film and digital are being mastered at 4K when needed. Note: Skyfall was shot on 3K but mastered on 4K – a perfectly acceptable interpolation.

Kodak Vision3

Kodak Vision3 DR Chart

Vision3 is the latest (last?) stock from Kodak, and is available for 8mm, 16mm, 35mm and 65mm film. Kodak presents detailed data sheets which charts MTF, granularity, color temperature and sensitivity information, like this one for 5219.

Kodak states Vision3 has a maximum dynamic range of 13 stops. The Zacuto test pegged this figure at 14.5 stops. I’d rather believe Kodak, they’ve been doing this for a hundred years. However, that is not to say Zacuto was wrong. Why?

The point is, sensors are linear devices, while film is non-linear in its exposure response to light. As I’ve explained in Dynamic Range of the Human Eye, the human eye is strictly neither (or more than both) linear or non-linear – it has its own rules. Furthermore, the way in which film is developed and scanned has a huge role to play in its contrast response. In this respect it behaves closer to the eye, which too, can change its behavior to the intensity of light.

For this reason, when comparing film to digital tests, it would be wise to give film an extra stop of ‘latitude’, if you will. Therefore, I have equated Vision3 with a dynamic range of 14 stops.

Dynamic Range: Film vs Digital

Who’s the boss?

Arri Alexa vs Red Epic DR

Red carried out a test between the Epic and the Alexa back in 2010, when everyone (including me) was going gaga over the impending demise of film. In this test, the Red Epic checks in at an impressive 13.5 stops of dynamic range, while the Alexa does 14 stops (to my eyes at least).

Before we get to the list, let me get in a word about DSLRs. In my RAW vs video for DSLRs test, I noticed that RAW still images have a 3-4 stop advantage over the video mode. I have seen this time and again in practice, and the following table takes this into account.

Here is a list of cameras and their ‘claimed’ DR:

Camera Stops of DR
Red Dragon Sensor 18
Sony F55/F5 14
Arri Alexa 14
Kodak Vision3 14
Red Epic 13.5
Blackmagic Cinema Camera 13
Canon C300/C100 12
Canon C500 12
Sony F3 11.5
Nikon D800 11.4
Nikon D600 11.2
Sony FS100/FS700 11
Sony A99 11
Nikon D4 10.1
Panasonic AF100 10
Canon 6D 9.1
HDC1500 9
Canon 5D Mark II 8.9
Canon 1DX 8.8
Canon 1DC 8.8 11.5*
Canon 5D Mark III 8.7
Canon 7D 8.7
Panasonic GH2 8.3

*See comments below post

Without splitting hairs over half a stop or so, let’s try to analyze the results. Only the Alexa actually comes close to Vision3. Whether or not the Epic ‘beats’ film is debatable. The F5/F55 have just landed and have not yet been tested for their ‘claimed’ 14-stop latitude.

My personal opinion? After having watched five movies shot on the Epic (the last of which was the Hobbit), I can say that the Epic still doesn’t have the kind of response film does – at least to my eyes. The Alexa comes damn close. While watching Skyfall, only on a few occasions were I reminded that this movie was shot on digital. I can’t say the same about the Epic.

In any case, even if one were to equate all the ‘RAW’ cameras, mainly the Alexa, the Epic and the BMCC, to film, it’s still a draw.

So, in the battle for dynamic range, film holds its own. It also has two advantages over digital that many find hard to disregard:

  • The best skin tones
  • Beautiful roll-off to highlights

Digital on the other hand, wins with:

  • Resolution
  • Shadow detail
  • RAW images

What about the Red Dragon Upgrade?

Red Dragon DR Chart

If the Dragon upgrade proves as capable as advertised, it will truly be the end of film. Finally, we’ll have a camera that matches the ability of our eye. We’ll see soon, won’t we?

Exclusive Bonus: Download my free swipe file on how to shoot night scenes well (PDF file optimized for mobiles and tablets).

21 replies on “Where cameras stand in Dynamic Range: Film vs Digital”

  1. Very good article Sareesh.
    I like Arriraw (Uncompressed Raw 12 bit) footage very much.
    The CinemaDNG (Uncompressed Raw 12 bit as well) comes in second after the Alexa.
    Life of Pi is great, and with lots of DR and color fidelity approaching film.
    Cesar Rubio.

    1. CesarRubio I was actually disappointed by Life of Pi – not the imagery, the movie. Frankly, I found Skyfall a better test for the Alexa – or maybe I’m biased because my favorite cinematographer shot it!

  2. Curious: where did you get your info on the 1DC’s DR of 8.8 stops? For what it’s worth Shane Hurlbut pegs it at 12.5 stops. Thanks.

      1. Sareesh Sudhakaran davidjcain again, not saying you’re wrong but if you could actually point me in the direction of an alternate publication I’d appreciate it. The only thing I’ve found online is Hurlbut’s which clearly contradict’s what you’re saying. And I’m inclined to go with him – en lieu of my own testing – since he’s used the camera extensively.

        1. davidjcain DXOMark score for the 1DX: http://www.dxomark.com/index.php/Cameras/Camera-Sensor-Database/Canon/EOS-1Dx It tells you the dynamic range is 11.5 stops – for RAW stills. Here’s another independent test: http://legault.perso.sfr.fr/1DX.html

          In my own tests of Camera RAW vs H.264 video, https://wolfcrow.com/blog/dynamic-range-comparison-of-raw-vs-video-mode-on-the-canon-550d/, we see that the typical difference between the two is about 2-3 stops of DR. This is what I would expect for the 1DX.
          On the other hand, the 1DC records JPEGs in 8-bit, not RAW 14-bit, so it is reasonable to assume it might lose a stop in encoding, which would give it 10.5 stops max. Even at 11.5 stops, it is still in FS100/700/C300 territory.
          You tell me how the video can be 12.5 stops if the sensor in RAW mode is at best 11.5 stops? Also, let’s look at it another way. Vision3 is 13 stops, can anyone guarantee that at 12.5 stops they could shoot a movie with it and no one would know the difference? Film-based content transferred to Vimeo looks good, so why blame (not saying you did, but people do) encoding when video-based content is transferred to Vimeo?

        2. davidjcainCredible? I don’t think so, not when it comes from Canon. Manufacturers don’t provide measurement details and testing specifications, and are notorious for providing MTF charts based on who knows what. 
          Just so we’re clear, I love Shane’s work and his blog, and I believe him. If at all anyone is at fault, it is most likely to be me.

          In the real world, a camera always deals with flare, motion, temporal artifacting, encoding, and many other things. I know the C300 claims 12 stops, and I see it every day. The 1DC could be better, but it is not reflected in the footage – that’s all I’m saying. What’s the point in looking at the C300? The 1DX has the same sensor as the 1DC. Both of which have a RAW maximum of 11.5 stops.

          Here’s a test I published on Technicolor Cinestyle vs Neutral for DSLRs: https://wolfcrow.com/blog/technicolor-cinestyle-vs-neutral/ Maybe one of these days I’ll compare Canon Log with Rec. 709 – I don’t use log because I have found a negligible difference (maybe half a stop at best). 
          Log does not give you more dynamic range, just a ‘flatter curve’, if you will. The half-stop advantage is merely an ‘impression’. In real testing, if I meter a scene 12 stops apart on the C300 there’s no way this camera can handle it gracefully. Not complaining, but that’s the reality. This is what I shoot with, and I can live with that. It pays my bills.

          You are lucky to have the 1DC in hand. Do you see a 12.5 stop scene rendered like film? Like I said, I’m dying to know, so I can order mine!

        3. Sareesh Sudhakaran davidjcain
          Sareesh, 
          It’s difficult to discuss the issue if you disqualify Canon’s published material as a marketing strategy, but allow what looks to be a homemade site on astrophotography as definitive proof. Sorry, I’m inclined to go with a manufacturer who stands to lose way more in publishing “fudged” facts than a French shutterbug. But, again, I’m looking for proof not trying to win a debate. So if Canon is lying then shame on them and I was naive to trust them…
          Also, I am not clear what exactly you “believe” about Hurlbut. I’ve quoted him as saying the 1DC’s DR is 12.5 stops @ 400 ISO in Canon log gamma mode. That clearly isn’t what you believe  since you say the max DR for the 1DX is 11.5 stops (presumably once again basing everything on Thierry’s test). And since you further discount the Canon log gamma, I have no recourse but to simply say I think you’re wrong to do so. Isn’t the C-log gamma there specifically to improve DR? I ask it as a question to be polite, but, frankly, it’s its raison d’etre! And, yes, it flattens the curve… that’s its “characteristic”, meaning that it smooths out the highlights so they don’t clip, giving it a film-like characteristic, thereby extending the shoulder roll off. I know you know this, so apologies. But here’s another thought: whether that’s as you claim an “impression” or evidence of real DR… who cares? If everyone “sees” it that way than I’m happy to go with a camera that gives the impression.
          And speaking of impressions, yours of the 1DC price is like an Alexa. I’m lucky I guess, but it’s not as though your C300 was any less money. In fact it cost more, no? At B&H $16K vs $12K for the 1DC. Even with the extra CF cards (in my case 4 x 128G) you’re still spending less.

        4. davidjcain I admit, I could have been clearer in my responses, and less condescending. Sorry about that.
          Here’s how I approach the problem:
          DXOMark is a known testing site and company – the only one that has a rigorous testing methodology. Astro-photography is probably the most demanding of any sensor. A quarter of a stop decides whether they can catch that star or not. Don’t you find it odd that a ‘French shutterbug’ did a comprehensive
          test on a sensor and took the pains to publish varying ISO-DR tests,
          while many others don’t?
          In any case, I must mention that I’m (was) an engineer, so well tabulated data appeals to me.

          Shane says it’s 12.5 stops. I believe him, why shouldn’t I? I’m in awe of what he does and says. But I don’t make purchase decisions based on what others say, at least in this field. Ask my wife!
          So, I look deeper. DXOMark is sort of a standard, so one can’t discount it. A quick qoogle search shows every test for the 1DX. I collect all the data I can.
          So, whom do I believe, David? I tend to follow my own experiences. I
          shoot a lot with 5Ds and 7Ds, too, and I believe the 1DX is similar in
          DR. So I base my final decision on a host of factors. I don’t discount
          any data, including Canon’s un-scientific chart. And anyway, that chart
          is for the C300, not the 1DC. Here’s another interesting test: http://www.dvxuser.com/V6/showthread.php?301116-Canon-1DC-ISO-vs-Dynamic-range
          I’m sure you’ll agree that if I’m writing a blog my readers deserve to hear what I have to say, and that’s what I say. It would be too easy to pass along Shane’s or DXOMark’s or anybody else’s data rehashed. This discussion we’re having is very helpful, and I thank you for it.
          That’s that. Here are some thoughts about the rest of what we’re discussing:

          Don’t you find it odd that Canon’s chart has the same DR across the entire ISO stretch – except at one ISO setting 640? Look at tests for any other camera, including the 1DX – usually dynamic range is highest at the ‘native ISO’ and drops on either side. I’m sure you know this. Now, I know for a fact that the C300 does not hold DR across the entire range.

          Secondly, digital SLRs and cameras don’t have an 18% middle grey. Middle grey for these sensors fall elsewhere, usually at 10%. Here’s an article I wrote about middle grey: https://wolfcrow.com/blog/notes-by-dr-optoglass-color-bit-depth-middle-grey-and-white-balance-of-the-human-eye/ You tell me, why would I want to expose a scene at 18%, if my sensor middle is at 10% (or whatever it turns out to be)? To avoid such blunders, I use a calibrated Sekonic light meter.

          Log Gamma doesn’t improve DR, David. It’s just a gamma curve. The white and black points are fixed. It redistributes the tones to get a flatter look, so you can ‘pull’ whichever way you want in post production. But the end points stay fixed. Here’s a primer I wrote on that: https://wolfcrow.com/blog/what-is-display-gamma-and-gamma-correction/
          When I used the word ‘impression’ in quotes, I was actually being condescending. Guess it backfired!
          The C300 is still a workhorse, isn’t it? Let’s assume the 1DC is 12.5, and I am wrong (gladly). Half a stop isn’t anything really, is it? The C300 has many other tools that make it a video camera in full. I just need to switch it on and shoot. Can’t say the same for the 1DC. At the very least no XLR and no SDI makes life difficult, for me.

        5. Sareesh Sudhakaran davidjcain Sareesh… I won’t retread what’s the 1DC’s DR terrain. Suffice to say we both think it might be between 11-12.5 stops. I see however you still have it as 8.8 in your review. This is a blog, not a peer-reviewed site, so you can do what you wish. But –  having admitted you got it wrong initially – it would be nice to see you correct the mistake.
          By improve the DR I simply mean that the c-log is a better S-curve than the other 1DC picture profiles if your goal is to get the most latitude from the camera’s DR.  To help (hopefully) illustrate my point: Rec 709 is a six stop curve. Therefore if your camera has a DR of 12 stops a Rec 709 monitor will limit your contrast by half. That particular gamma is not giving full expression to the camera’s DR potential. Make sense? In any case I’m not sure we really disagree on this point, but here’s a clear-headed article on what I’m trying to say.

          And I agree the C300 is a far more production-friendly camera in a traditional sense. It has industry standard video/audio connectors, a flip-up monitor with proper metres and built-in NDs. But it’s bulkier, doesn’t shoot stills, doesn’t shoot 4K and is less discrete (just let me have this point). Moreover, I’ve grown accustom to using DSLRs and have a variety of work-arounds so I’m comfortable. I’ve switched to Redmere HDMI cables and have had zero issues with them crapping out. Really the only thing I’d like to have are built-in NDs. But I can’t get too ornery: other high-end cameras like the Alexa and Epic don’t either. And of course, no film cameras. My guess is they’re a broadcast camera invention. A good one for those of us who want/need something fast.

        6. davidjcain It’s being corrected. Thanks for pointing it out.

          Here’s a test of Canon C-log vs S-log: http://www.xdcam-user.com/2011/12/canon-c-log-on-the-c300-compared-to-s-log/ 
          Not a rosy picture, is it? It’s exactly what I’ve seen in practice. All said and done, luckily manufacturers lets us choose, so no one’s getting hurt.
          Rec. 709 has nothing to do with dynamic range. Just so you are aware, here’s the entire Rec. ITU-R BT.709-5 specification: !!PDF-E.pdf
          A little knowledge is a dangerous thing. But let’s save this for a boring rainy day. It’s not relevant to this discussion.

          Regarding the C300 being ‘bulkier’, you need to read https://wolfcrow.com/blog/canon-1dc-vs-sony-fs700-which-is-the-better-camera-for-4k-tv/
          The C300 weighs 1450 grams in the EF mount, while the 1DC weighs 1360 grams, without XLR or viewfinder. Funny you should mention ND filters. Add one, and you’re even. If you compare the actual dimensions of both cameras from the front, an interviewee would find the 1DC more imposing. 
          Nothing wrong with using DSLRs, but don’t be under the misconception that they are smaller or lighter. Maybe, if you shoot with a T4i or GH2, but not a 5DIII or 1DC. More on my thoughts about ergonomics here: https://wolfcrow.com/blog/comprehensive-guide-to-rigging-any-camera-2-ergonomics/

        7. Sareesh Sudhakaran davidjcain Thanks for Alistair’s article. Of course It’s no mystery the 1DC is an 8 bit camera. Didn’t know that was even relevant to our discussion. I’d love it to be more. But I’m fine with it. As I said from the start I love the Canon image. Not so much Sony’s. So all’s rosy here. 
          And I gather you didn’t read Art Adam’s piece in your rush to condescend. Too bad. It’s a good one. But I’ve zero interest in trading insults with you. I was only attempting to be clear on my point.
          Good luck. No hard feelings. We just disagree.

        8. davidjcain I was not being condescending, David. Not this time anyway. I wonder why you have that impression.
          The fact that a camera is 8-bit or 10-bit has nothing to do with either its dynamic range or its color space. Here’s a question: why do cheap DSLRs produce stunning 8-bit JPEGs in-camera? I actually love Canon’s decision to go 8-bit – it took guts.

          I can understand if you don’t want to continue this discussion, but I don’t want you to believe I was treating this conversation as anything other than a friendly chat.
          All the best with your new purchase! If you can make it work, teach me. I’ll be grateful.

Comments are closed.