C500: C500 II Streaking

[modernhuman]

Yes, oversampling is gonna get closer to nyquist. But in real world, especially considering motion blur, distance to screen and individual visual acuity, resolution beyond a certain point matters little, compared to how far we have to go with dynamic range. Show the average person a 4k shot with 10 stops dynamic range and a 2k with 14+, and I bet they'd prefer the latter. Maybe it's me and my blurry vision haha, but I love dynamic range and natural contrast. Resolution has reached a critical point, but why is every consumer camera still not at the level of an Alexa, going 10 years now? Surely, if any of these companies wanted to make a sub $20k even $10k 3.2k oversampled s35 or even a 4.5k ff oversampled camera with a true 14+ dynamic range they could. It's marketing and Alexa knows their market and even undersells the specs, because only professionals use it and know it's good.

 

[Guest]

Yes, oversampling is gonna get closer to nyquist. But in real world, especially considering motion blur, distance to screen and individual visual acuity, resolution beyond a certain point matters little, compared to how far we have to go with dynamic range. Show the average person a 4k shot with 10 stops dynamic range and a 2k with 14+, and I bet they'd prefer the latter. Maybe it's me and my blurry vision haha, but I love dynamic range and natural contrast. Resolution has reached a critical point, but why is every consumer camera still not at the level of an Alexa, going 10 years now? Surely, if any of these companies wanted to make a sub $20k even $10k 3.2k oversampled s35 or even a 4.5k ff oversampled camera with a true 14+ dynamic range they could. It's marketing and Alexa knows their market and even undersells the specs, because only professionals use it and know it's good.

Doesn’t Arri have the patent on the two stage overlay that allows for wider dynamic range?

 

[NorBro]

Sensor is 6k, with a 4k downsample.

For sure, just wanted to double-check based on the comment below.

I’ve not tested an FX9, but 6k bothers me a bit.

 

[Barry G]

Yes, oversampling is gonna get closer to nyquist. But in real world, especially considering motion blur, distance to screen and individual visual acuity, resolution beyond a certain point matters little, compared to how far we have to go with dynamic range. Show the average person a 4k shot with 10 stops dynamic range and a 2k with 14+, and I bet they'd prefer the latter. Maybe it's me and my blurry vision haha, but I love dynamic range and natural contrast. Resolution has reached a critical point, but why is every consumer camera still not at the level of an Alexa, going 10 years now? Surely, if any of these companies wanted to make a sub $20k even $10k 3.2k oversampled s35 or even a 4.5k ff oversampled camera with a true 14+ dynamic range they could. It's marketing and Alexa knows their market and even undersells the specs, because only professionals use it and know it's good.

Just to be disagreeable... .. show the average person a 4K shot on the average rec 709 tv they own, and they’ll tell you it looks flat. (Because the only way to show that many stops in a system that can only display 6-7 stops is to map it down to those 6-7 stops. ) Show them a 14+ stop image on that monitor, and they’ll tell you it looks really flat. The best broadcast hdr is maxing out at about 10 stops, and all the cameras we are discussing are capable of producing well beyond that —12.5 - 14stops of usable dr. Show somebody a 14+ stop image on a fictional monitor that can display it, and they are likely to say “that hurts my eyes”.

This extra DR is not about a 14 stop image looking better, but how much headroom or bottom end you have to work with in post. These cameras routinely test to be within about a stop of each other in latitude. If DR is the most important thing to you, there is one camera that routinely beats all the others. The others are all exceptionally close to each other.

 

[jcs]

John, I really don’t want to start this up again, so maybe you could help by quantifying how much “much better” is in terms of highlight detail. 1/2stop?.. 1stop?... more?less? For weeks you demanded that others produce tests (skin tone ramps) refuting your comments (instead of the more typical scientific approach of.. “here is what I think, here are the results of tests I’ve conducted that prove it”). Several weeks ago you got ahold of both cameras. Did you conduct that test? I’ve really hoped to see your results.

I’ll say that, from the Side by side tests I’ve seen, the fx9 has about 1/2 stop more latitude in overexposure, and at least as much less in under exposure. I’d say the fx9 is a little better in the highlights, and a little worse in the shadows. Does that seem right to you?

The DP's that set up the test wanted non-flat lighting, so both cameras started clipping around +2. I would have liked to replicate the REDUser test (flat lighting) but we ran out of time. When exposed correctly, the cameras perform very similarly, however in this case the FX9 appears to produce more accurate color (recall the color accuracy test I mentioned in the other thread), where the C500 II had a magenta bias in our test (HMI (hot side) + diffused LED (soft side) of subject's face). That said I've seen some slight green issues with the FX9. I don't like how Canon is cutting color and blowing to white early compared to the FX9 in areas that should be 'valid' skin tones. It's fine to pull saturation approaching clipping ('the film look'), however Canon is pulling saturation and going white early in skin tones, well before clipping, and is also pulling color in the shadows early (perhaps to deal with color noise). This shows a valiant attempt to hit the quoted 'paper' stops, however as we've seen from the ramps, those stops aren't usable in the real world. I'm less concerned with the shadows and shadow noise, more concerned with shadows with accurate color, and no e.g. green shifts (C500 II).

Regarding highlight behavior on the C500 II- I did some tests and found I could put some color back into the white regions on skintones in post, so perhaps a LUT can address and/or a sat-lum curve can fix (or a FW update). So in the end, both cameras can look very similar when exposed correctly and I think it's fair to start rating cameras on SOOC performance, as many of us are tired of fixing camera issues in post. That's another reason why ARRI is so popular on the high end: great, accurate color SOOC with legit, usable DR as provided by the manufacturer. Doesn't it bother you when manufacturers lie about specs?

On that note, I contacted (busted) ARRI about calling LF large format with language like "magical / more 3D etc.", when really it's just full frame (+ a tiny bit more, but not even close to medium format), and FF isn't magical or more 3D, just easier to get shallow DOF based on available lenses. The person I chatted with was pretty cool, and didn't have much to say other than that's the marketing folks. They did verify the 1.5x oversampling & Nyquist. Which is interesting as that means they are well aware the Alexa (m)LF isn't producing True (full Nyquist-limit) 4K (the Alex65 does). For an epic outdoor film I can see the Alexa 65 being used for wide landscape shots and the (m)LF being used for everything else. Red 8K footage can also be breathtaking for wide angle landscape shots (viewed in 4K). While I have 4K monitors on my desks, I'm still using 1080p panels for big screens (Sony & Samsung). Each time I check out the latest new big screens (top of the line), I leave the store unimpressed. Maybe when the quantum dot / mini LED tech matures more?

 

[Guest]

Just to be disagreeable... .. show the average person a 4K shot on the average rec 709 tv they own, and they’ll tell you it looks flat. (Because the only way to show that many stops in a system that can only display 6-7 stops is to map it down to those 6-7 stops. ) Show them a 14+ stop image on that monitor, and they’ll tell you it looks really flat. The best broadcast hdr is maxing out at about 10 stops, and all the cameras we are discussing are capable of producing well beyond that —12.5 - 14stops of usable dr. Show somebody a 14+ stop image on a fictional monitor that can display it, and they are likely to say “that hurts my eyes”.

This extra DR is not about a 14 stop image looking better, but how much headroom or bottom end you have to work with in post. These cameras routinely test to be within about a stop of each other in latitude. If DR is the most important thing to you, there is one camera that routinely beats all the others. The others are all exceptionally close to each other.

I thought you were against misinformation? This is just using information to be misinformed.

 

[jcs]

Yes, oversampling is gonna get closer to nyquist. But in real world, especially considering motion blur, distance to screen and individual visual acuity, resolution beyond a certain point matters little, compared to how far we have to go with dynamic range. Show the average person a 4k shot with 10 stops dynamic range and a 2k with 14+, and I bet they'd prefer the latter. Maybe it's me and my blurry vision haha, but I love dynamic range and natural contrast. Resolution has reached a critical point, but why is every consumer camera still not at the level of an Alexa, going 10 years now? Surely, if any of these companies wanted to make a sub $20k even $10k 3.2k oversampled s35 or even a 4.5k ff oversampled camera with a true 14+ dynamic range they could. It's marketing and Alexa knows their market and even undersells the specs, because only professionals use it and know it's good.

Sure, past a certain threshold, resolution doesn't matter. I'm currently shooting mostly on the EOS R, in pretty low quality 1080p (soft, aliased), however after upscaling and sharpening to 4K along with shallow DOF, it looks ok on YT (YT 4K look much better than 1080p, so best to upscale 1080p and sharpen to 4K for YT delivery). Also shoot in the studio on the 1DX II (4K), which looks better (~1.3 crop), and used to own a C300 II, however the EOS R is the right size+weight, has a flippy screen, and the D.IS works surprisingly well. Looking forward to the EOS R5 (aka the A7S III with a Canon badge ).

Usable DR is clearly more elusive and it's unclear why the CamCartel can't come anywhere near the ancient ARRI tech unless the only viable way known is patented (around 10 more years on the ARRI dual gain simultaneous read out patent (no other sensor is doing exactly this process: https://patentimages.storage.googleapis.com/98/69/eb/50ec67d99a56b5/US7952629B2.pdf )). It's not just the sensor- ARRI is producing superior color performance as well, especially exposure invariant color (see the exposure ramp tests- amazing compared to CamCartel cameras).

 

[cpreston]

I thought you were against misinformation? This is just using information to be misinformed.

I think the real problem with working with higher dynamic range cameras is that people are still trying to get low key, high contrast images without recognizing that they are fighting against the increased dynamic range of the camera. What used to take a 3 stop difference to look right on the monitor now takes a 6 stop difference due to viewing it in wide dynamic range profile. All of those extra stops can be nice if you are trying to shoot in natural lighting that is backlit, but people are making it much more difficult on themselves trying to light a 14 stop scene vs a 7 stop scene due to not understanding the LUT they might or might not be viewing.

 

[Guest]

I think the real problem with working with higher dynamic range cameras is that people are still trying to get low key, high contrast images without recognizing that they are fighting against the increased dynamic range of the camera. What used to take a 3 stop difference to look right on the monitor now takes a 6 stop difference due to viewing it in wide dynamic range profile. All of those extra stops can be nice if you are trying to shoot in natural lighting that is backlit, but people are making it much more difficult on themselves trying to light a 14 stop scene vs a 7 stop scene due to not understanding the LUT they might or might not be viewing.

This is spot on.

 

[Barry G]

The DP's that set up the test wanted non-flat lighting, so both cameras started clipping around +2. I would have liked to replicate the REDUser test (flat lighting) but we ran out of time. When exposed correctly, the cameras perform very similarly, however in this case the FX9 appears to produce more accurate color (recall the color accuracy test I mentioned in the other thread), where the C500 II had a magenta bias in our test (HMI (hot side) + diffused LED (soft side) of subject's face). That said I've seen some slight green issues with the FX9. I don't like how Canon is cutting color and blowing to white early compared to the FX9 in areas that should be 'valid' skin tones. It's fine to pull saturation approaching clipping ('the film look'), however Canon is pulling saturation and going white early in skin tones, well before clipping, and is also pulling color in the shadows early (perhaps to deal with color noise). This shows a valiant attempt to hit the quoted 'paper' stops, however as we've seen from the ramps, those stops aren't usable in the real world. I'm less concerned with the shadows and shadow noise, more concerned with shadows with accurate color, and no e.g. green shifts (C500 II).

Regarding highlight behavior on the C500 II- I did some tests and found I could put some color back into the white regions on skintones in post, so perhaps a LUT can address and/or a sat-lum curve can fix (or a FW update). So in the end, both cameras can look very similar when exposed correctly and I think it's fair to start rating cameras on SOOC performance, as many of us are tired of fixing camera issues in post. That's another reason why ARRI is so popular on the high end: great, accurate color SOOC with legit, usable DR as provided by the manufacturer. Doesn't it bother you when manufacturers lie about specs?

On that note, I contacted (busted) ARRI about calling LF large format with language like "magical / more 3D etc.", when really it's just full frame (+ a tiny bit more, but not even close to medium format), and FF isn't magical or more 3D, just easier to get shallow DOF based on available lenses. The person I chatted with was pretty cool, and didn't have much to say other than that's the marketing folks. They did verify the 1.5x oversampling & Nyquist. Which is interesting as that means they are well aware the Alexa (m)LF isn't producing True (full Nyquist-limit) 4K (the Alex65 does). For an epic outdoor film I can see the Alexa 65 being used for wide landscape shots and the (m)LF being used for everything else. Red 8K footage can also be breathtaking for wide angle landscape shots (viewed in 4K). While I have 4K monitors on my desks, I'm still using 1080p panels for big screens (Sony & Samsung). Each time I check out the latest new big screens (top of the line), I leave the store unimpressed. Maybe when the quantum dot / mini LED tech matures more?

Thanks for the explanation. A few things -- questions, really...sort of...

-- when you say SOOC, what does that mean for each camera...Log/LUT? Cinetone/WideDR? IDT/ACES?
I ask because typically, SOOC means a strait Rec709 LUT, which of course doesn't really make sense with a 13-14 stop camera.

-- Magenta bias in C500II -- We've found that the canon supplied LUT's produce a slightly red/magenta skin tone, but a simple contrast and saturation grade from the LOG or Raw footage doesn't, indicating its a characteristic of the LUT (which, btw, when displayed on a consumer OLED, looks more balanced than what we're seeing on our production monitor, indicating that canon is "playing the broadcast game" with its LUT. Anyway, what was your workflow in arriving at that conclusion..it's really just so we know what you're looking at.

--Re Canon skin tones going to white, early. First...What Color Matrix did you use for the test? Canon supplies 3 presets. One of them is more desaturated in the highlights than the other two. Second, again, the LUTs canon supplies -- do quite strongly desaturate the skin tones above a certain point, yet we don't see that at all when grading without a LUT. (and again, those desaturated highlights seem to look better once we get to a rec709 output source).

Other than that, it does seem interesting that the two cameras clipped at the same point in your test, I would have expected that the Canon would have clipped earlier....a "little" earlier, certainly not "much" earlier. I don't know if you shot RAW for the Canon as many people seem to think it's fairer to the Sony to not shoot the Canon at its best setting. If you did, it might be worth it to see if grading without the LUT changes your opinions about the highlights. (I'm seeing plenty of color in mine

 

[Barry G]

I thought you were against misinformation? This is just using information to be misinformed.

You'll have to explain that to me. What am I misinformed (misinforming?) about? Is there a 14+ stop monitor on the market that I missed (one that doesn't require a welding hood for viewing . Is there a way to display 14+ stops on a rec709 monitor without it looking flat, or do you have to compress all but the 6 stops at the center of the curve to get it to "look" right...and If so...when you do that, are you still looking at 14+ stops? and would it look significantly different (better) than a 10 stop image treated the same way? Modernhuman says it would, and I respectfully, and disagreeably...disagree. (I will, however take as much DR as you can give me, for overhead, but will continue to light and monitor in Rec.709 and P3.)

 

[jcs]

Thanks for the explanation. A few things -- questions, really...sort of...

-- when you say SOOC, what does that mean for each camera...Log/LUT? Cinetone/WideDR? IDT/ACES?
I ask because typically, SOOC means a strait Rec709 LUT, which of course doesn't really make sense with a 13-14 stop camera.

-- Magenta bias in C500II -- We've found that the canon supplied LUT's produce a slightly red/magenta skin tone, but a simple contrast and saturation grade from the LOG or Raw footage doesn't, indicating its a characteristic of the LUT (which, btw, when displayed on a consumer OLED, looks more balanced than what we're seeing on our production monitor, indicating that canon is "playing the broadcast game" with its LUT. Anyway, what was your workflow in arriving at that conclusion..it's really just so we know what you're looking at.

--Re Canon skin tones going to white, early. First...What Color Matrix did you use for the test? Canon supplies 3 presets. One of them is more desaturated in the highlights than the other two. Second, again, the LUTs canon supplies -- do quite strongly desaturate the skin tones above a certain point, yet we don't see that at all when grading without a LUT. (and again, those desaturated highlights seem to look better once we get to a rec709 output source).

Other than that, it does seem interesting that the two cameras clipped at the same point in your test, I would have expected that the Canon would have clipped earlier....a "little" earlier, certainly not "much" earlier. I don't know if you shot RAW for the Canon as many people seem to think it's fairer to the Sony to not shoot the Canon at its best setting. If you did, it might be worth it to see if grading without the LUT changes your opinions about the highlights. (I'm seeing plenty of color in mine

• SOOC meaning, other than any required LUT(s) (e.g. IDT + 709/2020 or just one LUT, e.g. Canon or Alexa709 (popular) or to Canon Wide DR (PP CC raw panel))
• Agree the C500 II color bias is minor, however the goal is accurate SOOC (using ARRI as the current metric as a goal to match/surpass)
• The owner of the C500 II chose to shoot raw, so I used PP CC's raw panel and selected Canon WideDR for output (I used to use Production Matrix on the C300 II to match ARRI for PP CC's Alexa LUTs (CLog2 closely matches LogC gamma). Note a matrix is a simple linear transform for color (vs. a 3D LUT which is nonlinear and can affect the gamma curve))
• I didn't check clipping closely- just noticed both clipping around +2 and didn't look further (note the FX9 was shooting 709: S-Cinetone vs. C500 II's raw + Wide DR). I did find the tests showing S-Cinetone doing so well at +6 to be hard to believe vs. cameras shooting log and raw. To really get the max useful DR for the FX9 (and best color, from what I've seen IMO) requires using SLog3. This is based on reviewing footage online- we didn't have time to also shoot SLog3 tests.

Again, the cameras are very close, with certain features being more appealing based on existing gear owned (e.g. lenses) and use cases. For a vs. comparison for IQ, Sony's FX9 gets the win this time. Last iteration the C300 II won vs. the FS7 (II caught up a bit in color). For someone already owning a C500 II and looking to add another view and/or take advantage of FX9's features (e.g. digital vari-nd) and/or save (at least) $5K, the FX9 would pair well.

Remember, I've been comparing these cameras out of curiosity, having owned pro/cine Sony and Canon cameras (now shooting exclusively on Canon: 1DX II and EOS R). For the camera test, my friends were curious to hear what I had to say for both cameras (given my background in camera tech, software (cameras/imaging/games/sims/VR/AR/AI), and filmmaking (I've never worked full time as a DP or cam op: most of what I've shot/produced is for internal projects)). Future cameras will be much smaller, more modular, and more flexible. I'm done buying any camera I can't shoot comfortably with one hand (and also get stable results). Thus the EOS R / A7S III / Komodo sized cameras and features are appealing moving forward.

 

[Guest]

You'll have to explain that to me. What am I misinformed (misinforming?) about? Is there a 14+ stop monitor on the market that I missed (one that doesn't require a welding hood for viewing . Is there a way to display 14+ stops on a rec709 monitor without it looking flat, or do you have to compress all but the 6 stops at the center of the curve to get it to "look" right...and If so...when you do that, are you still looking at 14+ stops? and would it look significantly different (better) than a 10 stop image treated the same way? Modernhuman says it would, and I respectfully, and disagreeably...disagree. (I will, however take as much DR as you can give me, for overhead, but will continue to light and monitor in Rec.709 and P3.)

You know what Modernhuman meant.

 

[modernhuman]

Wider dynamic range means cleaner images with more flexibility in post and the ability to achieve more natural roll off and contrast. Yes, everything is compressed to Rec709, but it would still have a more pleasing and natural result. In highly controlled lighting scenarios (like Deakins), camera dynamic range might not matter as much, but lighting has to be plentiful and super precise. I’m using Terrance Malick, natural light films, The Revenant. A lot of Lubezki’s work, would destroy a lot of these cameras, probably all. His curves are very wide and reveal a lot of information at the extremes. The shadows stay way cleaner on Alexa and the highlights, go further. Also, agree with jcs that Alexa color is most accurate and best. It’s that simple. Lubezki, was quoted to say cameras need even more dynamic range. That’s what he valued the most.

 

[NorBro]

Wider dynamic range means cleaner images with more flexibility in post and the ability to achieve more natural roll off and contrast. Yes, everything is compressed to Rec709, but it would still have a more pleasing and natural result. In highly controlled lighting scenarios (like Deakins), camera dynamic range might not matter as much, but lighting has to be plentiful and super precise. I’m using Terrance Malick, natural light films, The Revenant. A lot of Lubezki’s work, would destroy a lot of these cameras, probably all. His curves are very wide and reveal a lot of information at the extremes. The shadows stay way cleaner on Alexa and the highlights, go further. Also, agree with jcs that Alexa color is most accurate and best. It’s that simple. Lubezki, was quoted to say cameras need even more dynamic range. That’s what he valued the most.

The desire for more dynamic range is nothing new; we can find posts going back 8+ years in which yesteryear posters are asking the same questions. (Not just on this forum.)

We've seen an overall improvement over the years, but it feels like everyone's plateaued within the past 4-5 years.

I haven't followed every RED camera and not sure what they are claiming these days, but if most of these companies cannot get past that true 14-15+ stop range then it's either because there's a lack of knowledge to do so (you don't imagine that for these larger-than-life corporations but it does exist), or they are simply not allowed to by using existing protected information and technology (as mentioned previously).

 

[modernhuman]

Yeah, agreed that’s true. I’ve heard of the patent argument, too. It’s just sad that nothing can be done. So pumping more pixels into these cameras is the race again. Just like it’s been since late 90s. Film is still great looking and its resolution is sub par. No actor wants an 8k-16k close up and a lot of cinema lenses try to creatively lessen resolution to soften the digital cameras. Most still cameras still don’t even mention dynamic range. But we’re getting 50+mp sensors. It’s dumb I tell ya. The crop in post argument only matters for some things. But not the things I’m filming. Oh well, at least we have cameras that are close, even if they’re all a little over priced. Not meaning to rant so much. Thanks everyone for your passion and knowledge about cameras. It’s been fun to discuss.

Kind regards.

 

[Barry G]

Wider dynamic range means cleaner images with more flexibility in post and the ability to achieve more natural roll off and contrast. Yes, everything is compressed to Rec709, but it would still have a more pleasing and natural result. In highly controlled lighting scenarios (like Deakins), camera dynamic range might not matter as much, but lighting has to be plentiful and super precise. I’m using Terrance Malick, natural light films, The Revenant. A lot of Lubezki’s work, would destroy a lot of these cameras, probably all. His curves are very wide and reveal a lot of information at the extremes. The shadows stay way cleaner on Alexa and the highlights, go further. Also, agree with jcs that Alexa color is most accurate and best. It’s that simple. Lubezki, was quoted to say cameras need even more dynamic range. That’s what he valued the most.

I wish it were that simple. That interview Lubezki gave with a couple of young, DR obsessed writers from DPreview, came at a time when 14+ stops (at a single, low ISO) had appeared on the stills landscape in the form of Nikon's implementation of a Sony still sensor. If you dig deeply enough into what Lubezki is saying, you'll see that he's describing a need for more DR to allow him to shoot without lighting, but nowhere is he implying that he has a way of displaying that extended range. Those "wide curves" you're talking about are monitor LUTs...designed to assist in seeing an image that can't be seen well on a 7 stop, or even a 10 stop monitor. They are indeed "flat" profiles designed to sit somewhere between a log profile, and the 709 (pseudo 2020) monitors that existed at the time. The discussion they are having refers to the particular problem these DR obsessed writers had to grapple with, which was that exposing the "new" way, meant you couldn't see the image on the monitor(because they seemed only interested at the time in exposing for sunspots).

So yes, Lubezki indeed did trade lighting gear for hours of secondary color correction and contrast masking in the DI, but all that extra DR was headroom beyond what he could display on the best theatrical projection at the time (8 stops?) and for broadcast/blueray distribution (7 stops). I think we actually agree on this. Your introduction of Terrence Malicks work doesn't mention his typical production schedule that involved only shooting at certain times of the day, when the light fit into the range suitable for the camera, and the presentation mediums.

Interestingly, one of the points Lubezki is making re: the Revenant, is that he needed to make sure that there was enough shadow detail available to get a good hero shot of Mr. DiCaprio when the sky, or sun was in the shot. There are two points I'd like to conclude with around that. From my experience, all of these cameras are capable of doing this in most situations these days (4 years after that interview), but in fact Arri, arguably does it better than the rest. When you try to dig Leo's face out of the murky bottom, we are talking about a camera's shadow range. This is the range that's important to Lubezki in his work, as in this "no light" technique, he agrees with the DPreview writers, that one can underexpose these cameras to preserve a lot of highlight detail that will be rolled off extremely in post, (rather than the traditional approach of placing middle grey where it belongs) and still salvage very important detail (the movie star) that would normally be buried in the shadows.

A lot is made of Arri's highlight range, but I've often looked at how it renders shadows differently than almost every other camera on the market. It's noise in the deepest depths of the image is remarkably clean...remarkably uniform...remarkable --I'd say, in that it doesn't look like "real" sensor noise. You'll also notice that in those deeper stops of a xylaa chart, that while the noise seems consistent in stops 12-17 stops, the edges of the stops get blurrier and blurrier. This would indicate to me (and I have no intrinsic evidence to support this) that Arri is applying significant noise reduction beyond a certain point, and then masking that NR with a reasonable amount of applied noise in-camera. (canon did this with the original C300 to mask some of it's 8 bit issues). Great technique...but also not particularly novel...as Colorists have been doing this, in post, for years. (lets not forget where the Alexa became popular, in Single camera TV drama, at a time when post-production was generally held at a minimum, thus Arri had provided an important tool in the HD broadcast workflow that was often delivering in 720p and 1080i.

I went to a presentation of the C300II at it's launch, and the presenters were rather candid that Canon was preserving noise at the bottom of the DR, because that also preserved detail at the bottom end. It's a difference in approach (a more traditional approach, actually). One could easily argue that specific noise reduction and "re-noising" should be done in post, especially in a situation like Lubezki's as he's often placing the movie star, and the thing that's in focus, in the most precarious part of the image. I am certainly not arguing that the C300II would have been a better camera on the Revenant than the Alexa65. I'm just saying it's not as simple as 14+ stops looks better than 10 stops, or 12 stops, or 13.5 stops(where most of these cameras, except Arri, live.)

From my experience with the camera, the C500II is preserving remarkably more detail in the "Lubezki Range" than the C300II that preceded it, and that my experiments with denoising, and re-noising make those stops very usable. I've actually been a bit surprised how well the FX9's deepest shadows, which have way more problems, clean up these days, although there is some loss of detail compared to the C500II. As I don't have access to an Alexa, I can't test as to which camera preserves more shadow detail, but I have my suspicions.

The reality is that the new cameras from Sony and Canon don't play in the same sandbox as the one we all want to compare them too. Different toys for different children. I'm as big of a fan of shooting with natural light (in practice) as anyone, but we should look at Lubezki's approach as a relatively rarified one, compared to the situations most owner operators using these lesser cameras typically find themselves in. It requires buy in from the director, producer et al, and per Lubezki, months of testing, followed by a much more tedious post-production regimen than is typical on small productions.

Great discussion, though, ModernHuman. It's the kind of thing we should be talking about.

 

[Barry G]

I haven't followed every RED camera and not sure what they are claiming these days.

Red claims 16.5 stops on Gemini, and Helium, and 17+ stops on Monstro. It's funny that Canon and Sony get so much crap for only over-stating by 1-1.5 stops.

 

[NorBro]

That's maybe with HDRx if that still exists, IDK.

 

[Guest]

Red gets crap for it. They are often credited for starting the more exaggerated specs.

Red is also credited for being the company that started releasing cameras in a not yet ready state, with a lot of bugs and issues.

DR is one spec that is difficult to rate to within 1.5 stops depending on some user’s tolerance of noise or NR. And in the case of Geoff Boyle, sometimes people rate camera DR based on a feeling. It is just weird seeing Canon getting into Red territory with marketing BS.

Red Gemini and Monstro are no slouches. So, while they are overrated on a specs sheet, they are often underrated by professionals. Fanboiz still drink the cool aide, but i see people scoff at Reds and the last time they used one was the Epic MX. Which is a deserved reputation from Red. They deserve all industry hate. But their cameras are pretty pretty good.

Under 15stops DR, EVERY stop or half stop counts. The midtones have been fine for a long time. Human vision hasn’t changed that much in the last 100years. Cameras can handle the midtones, as that is a small range. And film got away with less because the medium itself has so much analogue personality, that when things go bad, there is still something. Digital just goes empty. Pure white. So having the roll off and DR is more important for digital than film. That is why digital can’t just meet film specs, but has to surpass film by a good margin.