EPIC MX Highlight Rolloff thread....

[Gunleik Groven]

thx david!

nikons can be good

btw: my definition of "crap picture" is a picture that will not function in the intended context....

 

[Antonio Forjaz]

 

[António Fagundes]

Found some pretty cool examples on the web of how the Epic looks like in difficult lighting situations and if exposed accordingly will look stunning.

By the way David, I only posted about the Zone System because that's the standard way photographic light levels are read. Even Red's false color system is a version of the Zone System, but instead of numbers it gives you different colors to indicate the different luminance values of the scene. You can only read a light meter if you understand the one system and so forth...

Saying that, I think the best DPs are those that light by eye and not by numbers... BUT the more informed your eyes are, the better your judgment and work will be...

Thanks Phil Holland for your post. Those shots look damn impressive!

I'm feeling much better now regarding my concerns from MX's highlight control. From this thread people were kind enough to put up with my questions and doubts without ever having to bring up the usual behavior of trashing. I'm sure this thread will be very informative for a lot of people.

And thank you David Battistella for starting this thread.

Still... I wish I didn't have to worry about clipping...

Ah ah :cornut:

 

[David Battistella]

Very nice indeed

Thanks Mark!

Feel free to post any Dragon frames into this thread. Cheers mate!

David

 

[David Battistella]

Thanks Phil Holland for your post. Those shots look damn impressive!

..snip


And thank you David Battistella for starting this thread.

Still... I wish I didn't have to worry about clipping...

Ah ah :cornut:

Antonio. Thanks for posting these other great examples. I agree with you about the ZONE, it's just good knowledge to have.


David

 

[David Battistella]

Mr. Forjaz,

Yes!

Nice one.

David

 

[David Battistella]

Nice examples everybody. Fun thread to go through.

Hey Phil,

Thanks for posting in here. Really nice having your examples in this thread!

David

 

[Brice Ansel]

Some nice pics there.
It is turning into a sun world tour.

 

[António Fagundes]

Ok, just to further enhance this discussion, I'm posting two photos shot on a Canon 5D, these are photos, not video frame grabs.

They are both very similar in light, composition and difficulty for the sensor. In my opinion, one photo managed to capture the specular highlights from the sun in a 'perfect' way, whilst the other didn't. It has artifacts, no rolloff and the sun is bleeding (literally)...

Both photos are from the same photographer, the amazing Henry Liu, http://media-cache-ak0.pinimg.com/originals/32/28/ca/3228ca5d83b34f10c85e17ac8d13297e.jpg

I'm posting these because it's not about Red only, I think this is a problem with digital. I say problem because that's how I see it. I understand that others don't see this as a problem. This photos were shot on RAW and using grad filters. Both shot at f22, ISO 100.

 

[Scott Crawley]

These are both apparently so heavily manipulated I don't know how to draw many useful conclusions about what the original exposure or the sensor did not capture. We can talk all day about what remains, but any information that is lacking could have been discarded in post processing.

That is true of any image that has been processed after the sensor without regard for whether it was done inside or outside of the camera, not just these examples. In other words, any image that is not uncompressed RAW is less than a complete representation of what the sensor captured. This is why the photographer and his art have a larger bearing on the image than the tool in his hands. It all begins with his exposure, which the artist hopefully sets in full knowledge of how his tool will respond. Just as in Toia's recently posted test shots, the images only offer clues to what has actually taken place in the camera. They do not tell the sensor's whole story.

Ok, just to further enhance this discussion, I'm posting two photos shot on a Canon 5D, these are photos, not video frame grabs.

They are both very similar in light, composition and difficulty for the sensor. In my opinion, one photo managed to capture the specular highlights from the sun in a 'perfect' way, whilst the other didn't. It has artifacts, no rolloff and the sun is bleeding (literally)...

Both photos are from the same photographer, the amazing Henry Liu, http://media-cache-ak0.pinimg.com/originals/32/28/ca/3228ca5d83b34f10c85e17ac8d13297e.jpg

I'm posting these because it's not about Red only, I think this is a problem with digital. I say problem because that's how I see it. I understand that others don't see this as a problem. This photos were shot on RAW and using grad filters. Both shot at f22, ISO 100.

 

[David Battistella]

Scott,

It's true, those are pretty well finsihed graded and most likely photoshopped images, so tough to judge on that one.

I'm going to share this closer shot we did on the shoot for Brice's film this past week. Thanks Brice.

THis was on my Leica 50mm Sumilux

David

 

[Brian Boyer]

I wish I didn't have to worry about clipping...

Antonio, I still think people aren't quite getting where you're coming from so I'm going to take another stab at it by posing some questions.

Firstly, I think the term "clipping" has been so ingrained in us to mean a certain thing that, for the sake of this discussion, it's confusing the issue. The MX sensor is a piece of electronic equipment, so I'm going to use the term "overload" instead.

Several people have shown that smooth rolloff is not only achievable with MX, it can be a thing of beauty (my observation has been the MX in Epic/Scarlet is better behaved at rolloff than MX in the RED ONE is).

Antonio's other concern is overloading a sensor's capacity to capture information.

In any given digital sensor, it has a finite, numerical dynamic range and what you might refer to as a usable dynamic range, which takes into account things like noise floor and what personal tastes deem acceptable.

Also, for any given sensor, the upper limit of that range has a value, x, at which the sensor overloads in each of the R, G and B channels, whether simultaneously or individually.

There seems like there's a value, x-1, in whatever and however many numerical increments are available that make up the total dynamic range, that would represent the point just before the sensor channels overload.

This value, x-1, is a white pixel the sensor actually captured as being white and is part of what we would consider recoverable detail.

The value, x, is a white pixel the sensor could not actually capture as color and is a result of what happens when the sensor exceeds its capacity, or overloads.

So, the questions are these:

If the pixel you get just before you overload the sensor is captured white and the pixel you get when the sensor overloads is white (for a different reason) but runs the risk of producing artifacts, why would you ever go past the white that doesn't produce artifacts (x-1)...ever?

What if every value of white in the image that is overloading the sensor (x) gets filled in with the white just before the sensor overloads (x-1)? As far as the image is concerned, it still looks like blown-out highlights, right?

If you end up wanting to reduce the ISO but still want the blown highlights you would simply adjust the curve in post, which is just as easy to do in order to "fix" an artifact producing overload of the sensor.

Lowering ISO dramatically wouldn't yield any more detail in the areas filled in by the x-1 value but those areas would dim proportionately to the rest of the picture (like non-overloaded pixels do now) and wouldn't leave you with a grayish blob.

In the case of RED cameras, the traffic lights would still light up as normal and you'd still have an accurate representation of when and where you're overloading the sensor, you just wouldn't be capturing the overloaded values.

What's the benefit to the user of allowing the sensor to overload in a manner you're going to want to correct eventually anyway (unless you're Steven Soderbergh)?

From a technical, logistical or design standpoint, is it easier for the sensor maker to let the sensor overload and then pretty it up versus limit the signal on the front end?

Is what I'm saying pie-in-the-sky nonsense?

 

[António Fagundes]

Antonio, I still think people aren't quite getting where you're coming from so I'm going to take another stab at it by posing some questions.

Firstly, I think the term "clipping" has been so ingrained in us to mean a certain thing that, for the sake of this discussion, it's confusing the issue. The MX sensor is a piece of electronic equipment, so I'm going to use the term "overload" instead.

Several people have shown that smooth rolloff is not only achievable with MX, it can be a thing of beauty (my observation has been the MX in Epic/Scarlet is better behaved at rolloff than MX in the RED ONE is).

Antonio's other concern is overloading a sensor's capacity to capture information.

In any given digital sensor, it has a finite, numerical dynamic range and what you might refer to as a usable dynamic range, which takes into account things like noise floor and what personal tastes deem acceptable.

Also, for any given sensor, the upper limit of that range has a value, x, at which the sensor overloads in each of the R, G and B channels, whether simultaneously or individually.

There seems like there's a value, x-1, in whatever and however many numerical increments are available that make up the total dynamic range, that would represent the point just before the sensor channels overload.

This value, x-1, is a white pixel the sensor actually captured as being white and is part of what we would consider recoverable detail.

The value, x, is a white pixel the sensor could not actually capture as color and is a result of what happens when the sensor exceeds its capacity, or overloads.

So, the questions are these:

If the pixel you get just before you overload the sensor is captured white and the pixel you get when the sensor overloads is white (for a different reason) but runs the risk of producing artifacts, why would you ever go past the white that doesn't produce artifacts (x-1)...ever?

What if every value of white in the image that is overloading the sensor (x) gets filled in with the white just before the sensor overloads (x-1)? As far as the image is concerned, it still looks like blown-out highlights, right?

If you end up wanting to reduce the ISO but still want the blown highlights you would just adjust the curve in post, which is just as easy to do in order to "fix" an artifact producing overload of the sensor.

Lowering ISO dramatically wouldn't yield any more detail in the areas filled in by the x-1 value but those areas would dim proportionately to the rest of the picture (like non-overloaded pixels do now) and wouldn't leave you with a grayish blob.

In the case of RED cameras, the traffic lights would still light up as normal and you'd still have an accurate representation of when and where you're overloading the sensor, you just wouldn't be capturing the overloaded values.

What's the benefit to the user of allowing the sensor to overload in a manner you're going to want to correct eventually anyway (unless you're Steven Soderbergh)?

From a technical, logistical or design standpoint, is it easier for the sensor maker to let the sensor overload and then pretty it up versus limit the signal on the front end?

Is what I'm saying pie-in-the-sky nonsense?

Sweet mother of God, this is exactly what I'm trying to put across! I'm going to have to start paying you Brian to explain things for me!

Please, let's all try to get in depth of what Brian suggested here!

 

[Scott Crawley]

If the pixel you get just before you overload the sensor is captured white and the pixel you get when the sensor overloads is white (for a different reason) but runs the risk of producing artifacts, why would you ever go past the white that doesn't produce artifacts (x-1)...ever?

The short answer posed by most of the respondents is simply this: If you don't wish to overload the sensor, then don't overload the sensor. Control the amount of light reaching the sensor by whatever means you deem necessary and/or practical; reduce the aperture, reduce the exposure time, reduce the dynamic range of light in the scene... etc. What we do before the sensor is at least as significant (I propose more significant) than what the sensor does, or what happens after the sensor.

It is all about controlling the exposure. No?

Is it more complicated than that? Does it need to be? That can be difficult enough at times.

As to the question of why would you ever do this, one or two of us suggested that a lack of time or resources could be named as reasons, though not necessarily acceptable ones. Another might be the possession of a free and experimental spirit, as well as the confidence and understanding of how to prevent those artifacts. I think Brice, David, Phil and others have shown that artifacts are not an outcome written in stone... or rather the R3d. ;-)

Edit: Add Gunliek to that list, naturally.

 

[António Fagundes]

The short answer posed by most of the respondents is simply this: If you don't wish to overload the sensor, then don't overload the sensor. Control the amount of light reaching the sensor by whatever means you deem necessary and/or practical; reduce the aperture, reduce the exposure time, reduce the dynamic range of light in the scene... etc. What we do before the sensor is at least as significant (I propose more significant) than what the sensor does, or what happens after the sensor.

It is all about controlling the exposure. No?

Is it more complicated than that? Does it need to be? That can be difficult enough at times.

As to the question of why would you ever do this, one or two of us suggested that a lack of time or resources could be named as reasons, though not necessarily acceptable ones. Another might be the possession of a free and experimental spirit, as well as the confidence and understanding of how to prevent those artifacts. I think Brice, David, Phil and others have shown that artifacts are not an outcome written in stone... or rather the R3d. ;-)

I think you are missing the point. That is not what we suggest at all. What you talk about are solutions to the problem, and those solutions are expensive and time consuming. What Brian and me are suggesting is not having the problem at all. That's how this business works. Things evolve. Things get better. Things get more advanced. Dragon is the third generation sensor from Red, for a reason. They want to make their cameras better every time. If they shared your thought process, they would have stopped improving their products a long time ago.

But if you can't understand this and have nothing to contribute I suggest you step back. Because something tells me you are trying to sabotage this thread as well.

 

[Scott Crawley]

If what you have been saying is that Dragon is a welcome step toward a place where we have so much dynamic range available that we are free to spend more time being creative and less time and energy trying to create a proper exposure in the first place, then I say I am with you totally and always have been. If I still fail to understand what you are trying to communicate then it is not for a lack of trying. I am sorry if I was unable, in spite of your continued personal observations. Observations that I have tried to look past and avoid returning in kind in the spirit and interest of understanding and in the hopes that we all might better ourselves.

I think you are missing the point. That is not what we suggest at all. What you talk about are solutions to the problem, and those solutions are expensive and time consuming. What Brian and me are suggesting is not having the problem at all. That's how this business works. Things evolve. Things get better. Things get more advanced. Dragon is the third generation sensor from Red, for a reason. They want to make their cameras better every time. If they shared your thought process, they would have stopped improving their products a long time ago.

But if you can't understand this and have nothing to contribute I suggest you step back. Because something tells me you are trying to sabotage this thread as well.

 

[Brice Ansel]

I think you are missing the point. That is not what we suggest at all. What you talk about are solutions to the problem, and those solutions are expensive and time consuming. What Brian and me are suggesting is not having the problem at all. That's how this business works. Things evolve. Things get better. Things get more advanced. Dragon is the third generation sensor from Red, for a reason. They want to make their cameras better every time. If they shared your thought process, they would have stopped improving their products a long time ago.

But if you can't understand this and have nothing to contribute I suggest you step back. Because something tells me you are trying to sabotage this thread as well.

Maybe it will be perseve as a bit rude from my part but what ever.
I don't give a shit about the dragon, I don't give a shit about the futur.
I care about the present and the tool I have between my hands right now.
The title of the thread is EPIC MX HIGHLIGHT ROLLOFF THREAD
I hope the title is big enough for you Antonio.
Right now I'll continu to enjoy the doing with the tool who seems to be at your eyes not good enough.
See you when the dragon lauch then and chilled up and move on.

B.

 

[Scott Crawley]

I care about the present and the tool I have between my hands right now.

Indeed, this is the context in which I approached the whole discussion because it is practical, not because I lack vision.

Thanks

 

[António Fagundes]

Maybe it will be perseve as a bit rude from my part but what ever.
I don't give a shit about the dragon, I don't give a shit about the futur.
I care about the present and the tool I have between my hands right now.
The title of the thread is EPIC MX HIGHLIGHT ROLLOFF THREAD
I hope the title is big enough for you Antonio.
Right now I'll continu to enjoy the doing with the tool who seems to be at your eyes not good enough.
See you when the dragon lauch then and chilled up and move on.

B.

I'm sorry Brice, but I really don't understand what your post was all about. I know what the tittle of this thread is, but if don't agree, like or understand what is being discussed you can just abstain from posting. As I'm sure you have better things to do than reminding everyone what this thread is 'supposed' to be.

I'm trying to understand why Red sensors behave the way they do regarding overload highlight capture. I saw the same problem in Mark's Dragon footage. So, rather then trying to ignore this problem I'm trying to understand it. And no, I'm not asking what can be done to avoid it. I know how to avoid it. The problem and avoiding the problem are two separate issues. I'm trying to focus on the problem.

If anyone of you doesn't care or identify with this issue, you have plenty of other threads to visit. But if on the other hand you want to join the discussion and figure out if this is something that you might want to understand more about then by all means post your comments.

Brice, I think you done your bit. It would be great if you could walk off now.

Because I believe that since Brian exposed what the problem was so clearly, people can't ignore it anymore and will refuse the accept it, so I'm sure that from now on this will turn into a fight thread and is going to turn ugly and eventually die.

It's a shame because I think we are really onto something here.

 

[David Battistella]

I'm thinking this is a language thing.

One person seeing something as a problem, the other sees it as a characteristic of the technology that needs specific solutions to attain the desired result (if it is a film look).

Here we have all posted numerous examples of how the EPIC MX sensor is more than capable of dealing with highlight rolloff (in this instance being compared to film's rolloff).

There is no use arguing semantics or trying to make any digital sensor behave exactly as film would behave, it will always be its own thing. If you truly desire the look of 35mm film then convince every producer you work with to allow you to shoot film on every shoot (and take on all the appropriate costs and advantages as well).

This thread demonstrates many situations where the so called "problem" is easily resolved with this sensor/capture medium.

At this point it feels like the only thing being bashed is the EPIC MX, (why, because it is not film?) Of course it is not film. It is a new tool to which we are all trying to apply photographic principles and decipher the best practices to attain a "look" we want.

It's not about measuring a digital film pixel vs and EPIC pixel as compared to what it looks like on the screen.

Film itself is limiting in many ways.

I feel the EPIC technology is quite the contrary. We have tons of room to explore.

David