High ISO behavior

[Christoffer Glans]

I've been making some noise profiles for Komodo and came across some artifacts I can't wrap my head around. When I get up to 5000 and above, there are some banding lines going across a very smooth white surface (underexposed white).
I've also seen that the differences between different ISOs create some huge shifts in green tint. Maybe this is something for a new firmware to adjust?

 

[Fidel Ruiz-Healy]

Following.

Also, what are those four black marks on the top right seen in the lines.jpg?

Been downloading and messing around with Kamodo footage and have also been finding a lot of chroma noise, some blotching, and color fringing. Don't know how much had to do with the lens/ settings of the clip I had but curious if you're finding the same. From what I was seeing, tt was especially noticeable at the lower resolutions like 2k 120. But again, don;t know much about the circumstances of some of these clips as they are just random files I'm finding online,

 

[paulcurtis]

Are those examples JPG compressed? I wonder how the DCT handles noise compared to wavelet of DSMC2 but if the forum has recompressed those then it's difficult to tell - is it possible to post PNGs with no compression?

I imagine global shutter will bring it's own unique noise issues - horizontal perhaps! In the iso noise what is the lowest ISO and the highest ISO in that range?

cheers
Paul

 

[Christoffer Glans]

Following.

Also, what are those four black marks on the top right seen in the lines.jpg?

Been downloading and messing around with Kamodo footage and have also been finding a lot of chroma noise, some blotching, and color fringing. Don't know how much had to do with the lens/ settings of the clip I had but curious if you're finding the same. From what I was seeing, tt was especially noticeable at the lower resolutions like 2k 120. But again, don;t know much about the circumstances of some of these clips as they are just random files I'm finding online,

At 2K you are cropping and essentially zooming in on the image when working in RAW. Because it is RAW and that means all pixels are RAW data from each pxiel on the sensor. Any manipulation of resolution makes it not RAW anymore. Also, the camera needs light as any other camera. It seems that most newcomers to cinema camera systems think that the norm is a Sony A7S heavily in-camera noise reducted high ISO clear star sky lit nightvision. They forget that all of that is heavily manipulated and not really a RAW controllable image. If you go through heavy noise reduction and work on the footage you can essentially get clean 12800 images out of a Red, but you lose a lot of color data and quality, just as with the A7S at those insane ISO levels. ISO is just gain, it should be as close to the base level as possible for the best possible image and it's impossible to get good images if you don't light things for the camera to see. There's a reason that Hollywood movies shot on Red doesn't look like shit and that's because of the cinematographer actually lighting and getting a proper exposure as well as heavy post.

 

[Christoffer Glans]

Are those examples JPG compressed? I wonder how the DCT handles noise compared to wavelet of DSMC2 but if the forum has recompressed those then it's difficult to tell - is it possible to post PNGs with no compression?

I imagine global shutter will bring it's own unique noise issues - horizontal perhaps! In the iso noise what is the lowest ISO and the highest ISO in that range?

cheers
Paul

Here's a temp link to the R3D and a TIFF file. I don't think that the "lines" or banding is an issue for normal shooting, but it would be good to know what that banding is and if it can be replicated on others Komodos. If you have a Komodo, push the ISO to 12800 and film a slightly underexposed evenly lit grey/white flat surface at K5600, then check if it behaves the same.

https://we.tl/t-i30wpHAt06