EPIC MX Highlight Rolloff thread....

[David Battistella]

Here is the first (of I am sure many way more excellent) examples of nice smooth EPIC rolloff.

I will share my exact setup, a lighting diagram and even an R3D in the spirit of trying to showoff how great the MX sensor already is.

David



Here is a simple little frame showing how nice EPIC can handle things, when you want wide latitude and smooth highlights.

 

[Martin Stevens]

This looks wonderful David.

 

[António Fagundes]

Here is the first (of I am sure many way more excellent) examples of nice smooth EPIC rolloff.

I will share my exact setup, a lighting diagram and even an R3D in the spirit of trying to showoff how great the MX sensor already is.

David



Here is a simple little frame showing how nice EPIC can handle things, when you want wide latitude and smooth highlights.

Although this is a very beautiful frame I fail to see where the rolloff is, as the small highlight areas are very tight and controlled. I guess, what I'm trying to see as a successful example of good rolloff is if that window in the top right corner would be a much larger part of the frame and see how the rolloff would behave if that highlight would spill completely into the room, with objects or a face right into the middle of that spill.

Thanks for posting this David.

 

[David Battistella]

I will look for other examples, but maybe some other people will chime in to this thread as well.

David

Although this is a very beautiful frame I fail to see where the rolloff is, as the small highlight areas are very tight and controlled. I guess, what I'm trying to see as a successful example of good rolloff is if that window in the top right corner would be a much larger part of the frame and see how the rolloff would behave if that highlight would spill completely into the room, with objects or a face right into the middle of that spill.

Thanks for posting this David.

 

[David Battistella]

OK,

Here is a nice frame from spiderman...which I think was underrated for cinematography.

 

[Blair S. Paulsen]

Roll off? Or halation?

Roll off? Or halation?

I'm getting the impression Antonio is looking for halation, a bit of veiling and perhaps even some diffuse flare to suit his definition of "roll off". As David Mullen noted on the other thread, the physical properties of film itself are the genesis of a significant portion of this "effect" - so much so, I posit, that the anti-halation properties of modern film stocks to improve contrast is one of the "tells" when comparing films shot in the 70's vs those shot post 2000.

Not to be overly pedantic but perhaps "glow" or "halation" is a better descriptor than roll off. David B's image above fits my definition of "roll off" (or highlight compression if you prefer) similar to the more "log-ish" native response properties of silver halide in film emulsion. The shots David M posted in the other thread from Bob Richardson's JFK OTOH, would fit my definition of using the interaction of extreme overexposure with a particularly non-linear portion of the response curve of the film stock he was using to purposefully to create a stylized look.

If the core point of contention is whether every organic nuance of film response characteristics is duplicated by Epic MX, or any other digital camera, I agree with David M. - no. If the issue is whether one can finesse the shoulder of the RAW development curve to very closely mimic the roll off characteristic of film, then David B's image would indicate you can.

Cheers - #19

 

[Blair S. Paulsen]

Just saw the Spiderman shot, I'd guess there's a pretty fair amount of haze/fog in the air to spread the photons, but the results are more typical of what I would call halation. If David Mullen rejoins this thread at some point, he can likely delineate the proper terminology per "look" with greater accuracy than I.

Cheers - #19

 

[Luis Otero]

David,

Excellent example. Very appropriate screenshot for this discussion.

 

[Brian Boyer]

After having read this thread and the thread where this convo started, I'm going to try to reframe what I think Antonio is saying because I believe his point was lost on a difference in interpretation of the terminology. Hell, I might not be using the terms the way he was, but I think I get his meaning.

I'm not taking sides one way or another, I'm just trying to clarify for the sake of my understanding and others'.

Antonio, as best as I can tell, there are two separate issues you are referring to that people are combining into one - the first being the clipping point of the sensor (where the image breaks and artifacts start to appear) and the second being the rolloff, or shoulder.

You'll surely correct me of I'm wrong, but I think your issue with RED sensors is that they always show you exactly where they clip and that clipping produces unpleasant looking (to some) artifacts. You see the results of the break.

Another point I think you're trying to make is, if all you have to do is fix this in the grading suite, why don't more movies shot on RED have large white areas that are artifact free with smooth rolloff if it's so easy to obtain in post? Your conclusion is perhaps people avoid them because they're harder to deal with, because of the sensor, than people want to admit.

I believe what you meant when you said the Alexa and F65 don't clip is that we rarely, if ever, SEE them clip. They roll off to a white that (presumably) stops just short of clipping the sensor.

That's what you meant by saying the white in the Oblivion shots aren't from the sensor clipping, they're just white, where some people were suggesting white necessarily indicates clipping.

Maybe the sensor is clipping but those cameras just doesn't show you anything past a certain white value. There's no retrievable picture info there but you don't have to deal with the artifacts clipping causes.

(ASIDE: Maybe this isn't correct at all and it just seems like it's how it works simply because of the smooth rolloff and how they go about "painting" or otherwise compensating for the artifacts in the clipped whites.)

For example (I'm merely speculating here, and this could be totally wrong), it seems you could conceivably have a silky smooth rolloff right up until the clip but still have artifacts at the clipping point, if you actually show the clip. There could still be funkiness in the white area. I know the Alexa had an issue with purple dots in the hotspot and the BMCC had the black dot issue.

What you're proposing sounds a lot like what a limiter does in the audio world. It keeps you from ever going above a certain level even though the input signal is well above the clipping point, but still preserves dynamic range when the signal is below that point. Yes, it lops off info the higher the signal gets, but it does it in a pleasing manner that you don't notice when done expertly.

When done with cameras, you always get well-behaved whites with no artifacts, even if the whites clip hard. Then, if RED decided to have a complimentary and smoother rolloff response in camera with Dragon, this might go a long was towards being able to pull of those shots that MX seems to struggle with, if I'm understanding you correctly.

When you said DPs don't want to have to worry about highlights and what they're doing (paraphrasing), you meant if you know the worst you're going to get from the large, bright area is just a smooth transition to white without artifacts, every time, you can carry on normally when that's the desired look.

On the other hand, if you're putzing around trying to avoid or treat that large, bright area because you're afraid of what it might look like when the artifacts come out to play, it just unnecessarily adds to all of the other concerns that crop up on set.

As people mentioned, you could take steps to avoid clipping but what you're is calling into question is if MX is capable of producing that Oblivion cockpit shot with the same aplomb as the F65, even after taking those steps. And, how much better will Dragon be at it?

Maybe I've just made a mess of your argument.

 

[Rob Castiglione]

Nice images David.

I am struggling to understand what Antonio is talking about. I am constantly impressed with how the Epic X handles properly exposed highlights. They are just beautiful to my eyes - soft and creamy. It is somewhat frustrating listening to someone talking about a non existent problem. He needs to post specific examples. Unfortunately I am away so can't post some images at the moment but if this thread is still going when I get back home I will try to give some examples.

 

[Björn Benckert]

As people mentioned, you could take steps to avoid clipping but what you're is calling into question is if MX is capable of producing that Oblivion cockpit shot with the same aplomb as the F65, even after taking those steps. And, how much better will Dragon be at it?

Maybe I've just made a mess of your argument.

My guess, those white parts in the cockpit where not filled photon buckets on the sensor, not filled to the rim but to bright gray... And then in post they where brought towards the white point. In other words I do not think they shot such shot with any of the channel over exposure lamps lit or the right goal post even slightly showing. But I could be wrong.

 

[Mark Wilkinson]

I think Bjorn's got it right.

Which would mean that the image was properly exposed to protect the highlights and then graded to appear white to the eye - often around 230 8-bit or 90 IRE in broadcast.

Rather ask Alex Carr what the histogram looked like as the ftg came to him on the set of Oblivion. I'd be curious as hell how the daylight exposure went down on the sensor.

 

[Blair S. Paulsen]

I think an audio limiter circuit is a fair comparison, creating the sound amplitude equivalent of highlight compression. I can even see the argument that building in a subtle shoulder between say 97% of full well and full well (digital equivalent of D-max) into the file would be both popular and advantageous for the many gigs where there is no time or budget for the RAW development step. That said, if a trip through the virtual darkroom is not on the agenda then RAW capture might not be ideal to begin with.

I think its worth pointing out that a 16 bit linear image file is inherently different than a 10 bit log file. Their respective response to gamma adjustment for starters. There is also the question of whether one uses a video style lift-gamma-gain pipeline with the original linear file or ventures into a log space, like RLF, then grades the image using log pivots in a more film scan optimized DI/color/finishing tool.

FWIW, with proper preparation, on most shows, the DP and DIT could manage the "development" of the RAW file into an RGB format to get the intended look without significant impact on production schedules or budgets.

At the end of the day, choosing between Epic or Alexa should be a horses for courses decision based on the actual differences between them. The frustration for a lot of RedUsers comes when the choice is skewed by erroneous conclusions about system capabilities based on results achieved in particular instances with a myriad of other variables.

Cheers - #19

 

[David Battistella]

If David Mullen rejoins this thread at some point, he can likely delineate the proper terminology per "look" with greater accuracy than I.

Cheers - #19

I hope David can bring his wealth of expertise into this thread and perhaps guide us in terms of terminology so that we can begin the discussion on terms of highlight rolloff or halation.

I really started this thread because that other thread was being derailed (although the discussion is interesting) and has now been closed. Hopefully we can continue it here.

Thanks for posting into this thread.

I'll keep looking for examples.

FWIW, My theory is that I think the whole skintone highlight talk are cleverly selected. Why pick on these two things? 1. Highlight control has always been a thing with digital cameras 2.Producers pay big money for name actors and a nice way to discredit a camera is to say that the money will be thrown away because a camera can not render skin (we all know it is bullshit that EPIC can't "do" skin tones)

 

[David Mullen ASC]

I dropped out of the other thread for obvious reasons, I didn't want to be drawn into an argument with someone who has formed a strong opinion and then dares everyone to prove him wrong. You can't win those kind of arguments.

Also, I'm not an engineer so don't expect me to be the final word on how terms are defined. Some of you are more technically knowledgeable than I am.

But first of all, I think the halation issue is a bit misleading, it's a sideshow -- halation is just the glow around a point of overexposure, usually created by filtration though some older optics can create it. Halation is easy to get, just use some sort of mist filter and have a point of overexposure in the frame.

The main discussion and controversy is over the clipping point and the gamma curve of the highlights leading up to the clipping point, the knee and how compressed it is. Film negative has a wide DR -- maybe fifteen stops or so -- and it rolls off into undetailed white (D-max) gradually because at some point, you don't get as much silver formed compared to the amount of exposure the film gets so the curve gets flattened. To replicate that in digital, you need a camera with a wide DR and you need enough stops of highlight detail that you can apply a gamma curve to compress the knee. In theory, in order to create that flattening curve, you almost need one more stop of highlight detail than film has so that you have flexibility to bend that curve. Of course, you can get that extra highlight information by underexposing the image a bit more, at the risk of more noise.

I don't see why anyone is making such a big deal over the fact that the Alexa and F65 have about a stop more dynamic range, maybe 1.5-stops, than the Epic with the MX sensor -- that's been well-documented. It's a difference that can be moderated through lighting and exposure, and if you really want more overexposure latitude with the MX sensor, you just need to rate it faster. I can't see why a stop is enough for someone to say one camera's images are acceptable and the other camera's images are unacceptable, that seems oddly nitpicky. Either way, the Dragon sensor will bring the Epic to par and beyond the competition, so if someone doesn't like the way that the Dragon sensor handles exposure, then I guess they can just stop shooting because there isn't anything else on the market that can do better.

Sure, more DR is better, I also think the Alexa and F65 handle overexposure a bit more gracefully than the MX sensor does, but when I say "a bit" I mean "a bit", and it seems that the Dragon sensor will match or surpass the competition in DR so this whole discussion will soon be moot once Dragon is out there for everyone to use.

As for skin tones, that's more of an issue of color science and color-correction -- as others have noted, it can be easier to nail skin tones right out of the gate with the Alexa and F65, but you can match the Epic to either. Just the other day, I saw a trailer for a movie where I thought it had "Red camera skin tones" only to find out it had been shot on the Alexa! It was just a trailer so I don't want to assume that the final movie will have that look, but it shows how color-correction can screw skin tones up into something unnatural. And I've seen plenty of Epic-shot movies lately to have good skin tones, like "The Great Gatsby" and "The Amazing Spider-Man" did.

 

[David Battistella]

David,

Thank you for posting into this thread.

I agree that the discussion was getting more than playful in the other thread.

The other thing we are dealing with here is the subtle difference between an analog and digital process. Anyone into hi-fi audio has heard a million times the argument between "warm sounding tubes" compared to "transistors". Almost the same applies here. Light hitting silver and that reaction is never going to be the same as light hitting a photosite on a sensor so it becomes a discussion about semantics.

The DR/skintone discussion gets boring fast as we all know that EPIC is more than capable in each department.

I am more interested, and hope, this thread can be a bit more about how people have attained "looks" and best practices.

OF course it will all change with DRAGON, but that is the fun part, seeing and learning how that sensor "sees and responds" to light/lens/filter combinations.

David

 

[António Fagundes]

After having read this thread and the thread where this convo started, I'm going to try to reframe what I think Antonio is saying because I believe his point was lost on a difference in interpretation of the terminology. Hell, I might not be using the terms the way he was, but I think I get his meaning.

I'm not taking sides one way or another, I'm just trying to clarify for the sake of my understanding and others'.

Antonio, as best as I can tell, there are two separate issues you are referring to that people are combining into one - the first being the clipping point of the sensor (where the image breaks and artifacts start to appear) and the second being the rolloff, or shoulder.

You'll surely correct me of I'm wrong, but I think your issue with RED sensors is that they always show you exactly where they clip and that clipping produces unpleasant looking (to some) artifacts. You see the results of the break.

Another point I think you're trying to make is, if all you have to do is fix this in the grading suite, why don't more movies shot on RED have large white areas that are artifact free with smooth rolloff if it's so easy to obtain in post? Your conclusion is perhaps people avoid them because they're harder to deal with, because of the sensor, than people want to admit.

I believe what you meant when you said the Alexa and F65 don't clip is that we rarely, if ever, SEE them clip. They roll off to a white that (presumably) stops just short of clipping the sensor.

That's what you meant by saying the white in the Oblivion shots aren't from the sensor clipping, they're just white, where some people were suggesting white necessarily indicates clipping.

Maybe the sensor is clipping but those cameras just doesn't show you anything past a certain white value. There's no retrievable picture info there but you don't have to deal with the artifacts clipping causes.

(ASIDE: Maybe this isn't correct at all and it just seems like it's how it works simply because of the smooth rolloff and how they go about "painting" or otherwise compensating for the artifacts in the clipped whites.)

For example (I'm merely speculating here, and this could be totally wrong), it seems you could conceivably have a silky smooth rolloff right up until the clip but still have artifacts at the clipping point, if you actually show the clip. There could still be funkiness in the white area. I know the Alexa had an issue with purple dots in the hotspot and the BMCC had the black dot issue.

What you're proposing sounds a lot like what a limiter does in the audio world. It keeps you from ever going above a certain level even though the input signal is well above the clipping point, but still preserves dynamic range when the signal is below that point. Yes, it lops off info the higher the signal gets, but it does it in a pleasing manner that you don't notice when done expertly.

When done with cameras, you always get well-behaved whites with no artifacts, even if the whites clip hard. Then, if RED decided to have a complimentary and smoother rolloff response in camera with Dragon, this might go a long was towards being able to pull of those shots that MX seems to struggle with, if I'm understanding you correctly.

When you said DPs don't want to have to worry about highlights and what they're doing (paraphrasing), you meant if you know the worst you're going to get from the large, bright area is just a smooth transition to white without artifacts, every time, you can carry on normally when that's the desired look.

On the other hand, if you're putzing around trying to avoid or treat that large, bright area because you're afraid of what it might look like when the artifacts come out to play, it just unnecessarily adds to all of the other concerns that crop up on set.

As people mentioned, you could take steps to avoid clipping but what you're is calling into question is if MX is capable of producing that Oblivion cockpit shot with the same aplomb as the F65, even after taking those steps. And, how much better will Dragon be at it?

Maybe I've just made a mess of your argument.

Brian, this is an amazing post. I was getting so exhausted of 'fighting' my way through an argument that couldn't be understood as a legitimate concern of mine. It's extremely difficult to get answers in here without getting punched a few times. But your post was able to say it much better than I could ever articulate!


Thank you.

I dropped out of the other thread for obvious reasons, I didn't want to be drawn into an argument with someone who has formed a strong opinion and then dares everyone to prove him wrong.

David, I didn't start any of these treads or discussions, someone else did, and I just joined in because I share the same doubts. I thought that was the point...

But it's funny that you see me as brick wall and then stop talking to me, because I don't share your opinion until I'm proven wrong. Since I haven't been proven wrong yet, it seems to me that it is you that has a strong opinion and dismisses anyone who dares to question that opinion...

Just sayin

 

[Rob Castiglione]

Antonio. I think that you posted a still from one of Mark's Toia's torture tests which showed a young fellow at sunset still reasonably exposed and the the sun behind him blown out. You drew a red circle around the blown sun as if this was a problem. Forgive me if this was not in fact you - it was one of the more comedic moments in the numerous responses that followed so perhaps it was not you. If it was you, do you think that this is a "legitimate" concern?

 

[David Battistella]

I think we have to look at music videos for and example like this where you have strong, blown out backlight with a forground subject.

Antonio, repost some of your examples of what you are talking about from the other thread.

I know you have asked for examples, but I have not seen any examples as posted by you, perhaps shot on film or whatever, that to you seem correct, or that EPIC should be able to emulate.

Share the images which you think RED can not accomplish.

David

 

[António Fagundes]

I think we have to look at music videos for and example like this where you have strong, blown out backlight with a forground subject.

Antonio, repost some of your examples of what you are talking about from the other thread.

I know you have asked for examples, but I have not seen any examples as posted by you, perhaps shot on film or whatever, that to you seem correct, or that EPIC should be able to emulate.

Share the images which you think RED can not accomplish.

David

Ok David, here's a shot we did on Red One MX. This is a very difficult shot for any camera. See the harsh rolloff as the sky fades into blue from the sun. All I want to know is how can I do this kind of shot without the clipping and the artifacts. I understand that I can use HDRx on the Epic. But are there any other options?

And please, if you don't see anything wrong with this frame grab, don't confront me by telling me I can't expose or grade or I'm not competent enough to shoot on Red whatever else you think of. I'm really trying my best here, if I start to get shit from people I will just leave the thread.