DR ..a good discussion!!

[Aaron Lochert]

^

the lightbulbs...



can anyone else beside arri make a camera that renders a pinpoint light source smoothly?

maybe not.

When you clip, you clip. Arri is no exception, they just roll it off really smoothly in their output transform:

The Sony LUT actually rolls out the highs pretty nicely, too. I think Rand was just trying to extract as much detail as he could below the clipping point, which is the trade-off with a smooth highlight transition. I believe this to be yet another reason why Arri only ever claimed 14 stops of dynamic range - the extra couple stops the camera actually sees are being used for an appealing shoulder/knee.

Here's the full f/4 clip ungraded with the Sony LUT, to give everyone an idea what you can expect "out of camera." I white-balanced on the grey square as well as set exposure so that middle grey was 50 IRE - which meant ISO 640, not 500 as the board states.

4K: https://farm2.staticflickr.com/1769/41182064060_9a5482ae9c_o.png

When it comes to RED, what I've noticed with the IPP2 output transforms is that the ones labeled "soft" actually make the transition more apparent because it brings highlight detail down in luminance, revealing the clip point. Where Medium and Hard roll-off might be brighter and appear more contrasty, that stronger curve means smoother highs as things that are just below clipping are pushed nearer to white. So I typically am using Low Contrast and Medium highlight on grades these days.

 

[rand thompson]

I think Rand was just trying to extract as much detail as he could below the clipping point, which is the trade-off with a smooth highlight transition

Aaron, you know me so well..Ha.Ha.Ha

 

[Björn Benckert]

I get really frustrated when people discuss raw cameras and claim that one has better rolloff than the other. These cameras clip at a certain point, its at a higher or lower point depending on the light sensitivity of the given camera. One camera clipping earlier than another does not mean the camera is better or worse then the comparing camera. It only means it clips earlier. Not until you mesure the other end of the stick, the low end. You get a full picutre of what the given sensor is capable of.

For example a camera that clip 2 stops before another camera but has 2 stops more in the low end, is to me a camera that is two stops better in light sensitivity. It simply got the same DR span but is more sensitive which is a good thing in my, and I think most DP´s book.

Then rolloff, on a C-mos sensor is nothing but how you choose to read the raw data. As they all clip linear. The dynamic rounding of the highlights is the colorists responsibility or who ever turns the raw data to a viewable image. As raw can be "developed" in many different ways it´s not linked to camera manufacturer.

Possibly you can say that the luts you use rolls the highlight of Alexa really nice but that has little to do with the performance of a Sony camera as luts are replaceable.

For sure Arri got a lot of head room in their ratings of their camera ISO´s but that says little or nothing about the actual performance.

 

[rand thompson]

I did a short video using sample ProRes 2K footage from Arri. The clips have been shot with the following settings (unless otherwise stated): 24fps, 800ASA, 3200K; shooting gamma for all clips is Log C

It contains;

1) Arri Alexa LF "Ungraded Footage"

2) Arri Alexa LF footage with the "Arri Alexa LogC to Rec 709 Lut" applied to it from the "Arri" Lut folder in Davinci Resolve.

3) Arri Alexa LF footage with a custom curve I made without the Arri Alexa LogC Lut being used.


2k 2048x1152 h.264 file of video at dropbox link below


https://www.dropbox.com/s/fsujliu5vlf7sz2/An Arri Alexa LF Test.mov?dl=0

Sample ProRes 2k (2048 x 1152) footage name "B001C002", shot with "SP" 40, I'm assuming that stands for "Signature Prime", @24fps

http://www.arri.com/camera/alexa/learn/alexa_sample_footage/

 

[rand thompson]

Here is a dropbox Link to 4 Ungraded .TIFF screengrabs of the Arri Alexa LF from Davinci Resolve of the Jpegs below for those who didn't want to download the video above. The .TIFFs are 2k 2048x1152 resolution, the same as the Prores Sample video above with the same shot settings as above.

Dropbox
https://www.dropbox.com/sh/9nr7bamaeqwggup/AAC6TNp10R1Q4zRCvs4JUOeZa?dl=0

[/url]Untitled1_1.1.1 by rand thompson, on Flickr[/IMG]

[/url]Untitled2_1.1.2 by rand thompson, on Flickr[/IMG]

[/url]Untitled 3_1.1.3 by rand thompson, on Flickr[/IMG]

[/url]Untitled4_1.1.4 by rand thompson, on Flickr[/IMG]

 

[Adrian Jebef]

Rand, may I ask what your "custom s-curve" is trying to accomplish? It feels extremely unnatural. The mids and high-mids are so jacked they render as a solitary tone making it all seem like plastc.

 

[Adrian Jebef]

When you clip, you clip. Arri is no exception, they just roll it off really smoothly in their output transform.

I get really frustrated when people discuss raw cameras and claim that one has better rolloff than the other. These cameras clip at a certain point, its at a higher or lower point depending on the light sensitivity of the given camera. One camera clipping earlier than another does not mean the camera is better or worse then the comparing camera. It only means it clips earlier. Not until you mesure the other end of the stick, the low end. You get a full picutre of what the given sensor is capable of.

For example a camera that clip 2 stops before another camera but has 2 stops more in the low end, is to me a camera that is two stops better in light sensitivity. It simply got the same DR span but is more sensitive which is a good thing in my, and I think most DP´s book.

Then rolloff, on a C-mos sensor is nothing but how you choose to read the raw data. As they all clip linear. The dynamic rounding of the highlights is the colorists responsibility or who ever turns the raw data to a viewable image. As raw can be "developed" in many different ways it´s not linked to camera manufacturer.

Possibly you can say that the luts you use rolls the highlight of Alexa really nice but that has little to do with the performance of a Sony camera as luts are replaceable.

For sure Arri got a lot of head room in their ratings of their camera ISO´s but that says little or nothing about the actual performance.

Cameras are different. They clip differently. They render images differently. Just because digital cameras see linear light does not mean they are all the same in their approach or most importantly in their output. The Alexa does not clip like Monstro which does not clip like Venice. It's not just the "clip point". It's not just the "highlight roll off". You can't make one camera look like the other. They are all unique. Tests are silly things as most are semi-helpful interior setups with charts and other ephemera. They usually tell us when the chip chart light grey looks lighter grey. So what? Who cares? Unless you're shooting a movie starring color checker charts I find it almost useless to the general public.

BTW here's how ALF and Venice render highlight clipping. Notice a difference?


ALEXA LF:

SONY VENICE

 

[rand thompson]

Rand, may I ask what your "custom s-curve" is trying to accomplish? It feels extremely unnatural. The mids and high-mids are so jacked they render as a solitary tone making it all seem like plastc.

Adrian,


Yeah, it kind of is. I was mainly "trying" to do two things. First trying to bring out more detail in the shadows than the Arri Alexa LogC to rec 709 lut did. Next, to try to negiate that "purple/magenta" halo around the lights by the sofa. This was prores 2048x1152, it was processed with "tetrahedral" interpolation and I only added the Arri log C to rec709 lut to the clip, nothing else was done to it. I managed to get rid of some of the purple/magenta halo around the lights but I am willing to admit that I may have pushed the image a little too far to do so.

[/url]test 1 with cirlce by rand thompson, on Flickr[/IMG]

[/url]test 2 with circles by rand thompson, on Flickr[/IMG]

 

[Adrian Jebef]

Ahso, makes sense.

Curves are hard as the more they are stretched to open up DR the more contrast is required to produce a proper image. Pushed too hard and you can get close to their breaking point when luminance solarizes. Usually in these kid of tricky situations I'll use multiple layers or nodes of correction to get the image to where I want it. Sometimes it's easier to use a simple general contrast curve first and then use more finite layers to target the blacks and mids for more specific control. But with this kind of test imagery it's all relevant.

Magenta/green halo is also tricky around pin-point light sources as a lot of the time it has to do with lenses and stop more so than camera and sensor.

 

[rand thompson]

Adrian,


You live you learn! I never really dealt with a lot of Arri footage before so still learning how to process it properly. You appear to have much more knowledge than I do about the matter, so in the future, I will count on you to keep me on the "straight narrow".


Rand

 

[David J. Buchanan]

You wouldn't want light bulbs to look like that anyway, it makes no sense to see the filament. The issue with having better darks than lights or +2 better in highlights vs darks is that most of us can add light but take away, esp outside in bright sunlight is thzt it's hard to do.

 

[Hrvoje Simic]

Anton,
I understand what you are saying. But why would I need to tell you something you can see and assess with your own eyes?

The key word in the thread name helps to answer that.

I get that you are more of a "talker" than I am. "Talk" is great but a picture is worth a thousand words and a video is worth even more for those who may not have the technical knowledge of some of the other members.

Yeah, but this is a forum and discussion gets impaired without moderation.

 

[rand thompson]

The key word in the thread name helps to answer that.



Yeah, but this is a forum and discussion gets impaired without moderation.

Understood, and I will try to do a better job of this in the future.