Epic-W more chroma noise than expected?

[Josiah Duncan]

I've only been testing the new Epic-W for a small while, and some of the videos that people are shooting online are incredibly clean at 800-3200iso, but I think I'm running into more noise in the blacks than I expected, even at 800 iso. Looking at some of the footage I shot this summer with the Epic Dragon at equal iso's I would even venture to say the noise is worse, which shouldn't be the case. The grain is smaller though, which helps, but the chroma noise (very noisy in the red channel) is almost distracting on a 4k monitor. On standard olpf. DEB seems to help a little.

Luckily the grain seems nearly the same in the blacks going from 800-3200iso, which leads me to believe it might be a black-shading issue? I've run multiple auto calibrations, and when I change shutter speed it takes a few seconds for the E to turn green like it's supposed to. However on the calibration menu when I press apply, no dialogue comes up that says "applying" like it did on epic dragon. Is this normal?

Maybe I'm being too critical and trying to push material that's too compressed, and perhaps none of it will matter when the new Helium CS is released. I'm just trying to get as comfortable with exposing properly on this thing as I was with dragon. If anyone has any insights that would be great, I've attached a still and an r3d, make sure to view in 4k or zoom into the bokeh in the shot.


Jpeg: http://img.pixady.com/2016/11/301647_a001c0341114v7s000.0000047_copy.jpg

R3D: http://s000.tinyupload.com/download.php?file_id=62906838898841120840&t=6290683889884112084006016

 

[John Brennick]

Hey Josiah, which OLPF were you using?

 

[Milano Questa]

I've only been testing the new Epic-W for a small while, and some of the videos that people are shooting online are incredibly clean at 800-3200iso, but I think I'm running into more noise in the blacks than I expected, even at 800 iso. Looking at some of the footage I shot this summer with the Epic Dragon at equal iso's I would even venture to say the noise is worse, which shouldn't be the case. The grain is smaller though, which helps, but the chroma noise (very noisy in the red channel) is almost distracting on a 4k monitor. On standard olpf. DEB seems to help a little.

Luckily the grain seems nearly the same in the blacks going from 800-3200iso, which leads me to believe it might be a black-shading issue? I've run multiple auto calibrations, and when I change shutter speed it takes a few seconds for the E to turn green like it's supposed to. However on the calibration menu when I press apply, no dialogue comes up that says "applying" like it did on epic dragon. Is this normal?

Maybe I'm being too critical and trying to push material that's too compressed, and perhaps none of it will matter when the new Helium CS is released. I'm just trying to get as comfortable with exposing properly on this thing as I was with dragon. If anyone has any insights that would be great, I've attached a still and an r3d, make sure to view in 4k or zoom into the bokeh in the shot.


Jpeg: http://img.pixady.com/2016/11/301647_a001c0341114v7s000.0000047_copy.jpg

R3D: http://s000.tinyupload.com/download.php?file_id=62906838898841120840&t=6290683889884112084006016

Compression artifacts?

 

[David Battistella]

12:1 compression and slightly underexposed would get you about the sam amount of noise, but I don't really see where the issue is in that image? You shot at 2.8 1600ISO, how much ND were you using? COuld it be that the ND is adding to that noise?

When I graded it back processing with RWG and RG10 it looks pretty smooth. (though it looks like you were going with teal and orange LUT) It might be your LUT or the agressivness of the grade or you may be addin saturation PRE instead of POST in RCX.

David

 

[Josiah Duncan]

Using Standard olpf. It could definitely be compression artifacts, this was shot at 12:1 8k ws but im getting similar results with 7:1. It was shot at 1600iso through a variable ND, but developed at 800iso. It's really just the red channel that's standing out to me in the example. It looks super smooth at 1080p but the red grain in 4k full screen is what I'm noticing.

Thanks for helping, just trying to find that sweet spot for now. The texture is great for some things once I clean some of it up with DEB but not great for every project.

David - I am using a basic lut based off DC2>RLF>LUT, and getting the same result with or without nd (at the same iso of course). I am looking closely on a 31 inch 4k monitor though, could you post a fullres version of that? I'm no color expert but find that RG10 is reducing exposure by around 1 1/3 stop so I'm guessing the gamma is really different.

 

[Stacey Spears]

Can you point out in the image what is bothering you?

 

[Graeme Nattress]

Log3G10 puts mid grey at 1/3 (or 33.33...%) so it's not reducing exposure, just using a different target point for mid grey. Any development LUT you'd use on Log3G10 would take that into account. Because Log3G10 encodes a very wide dynamic range, the mid grey point for encoding had to shift a little compared to what is used on lower dynamic range curves like RLF.

I looked a tthe R3D here and it looks pretty good and normal. It seems that whatever LUT you're using (is it a straight 1D LUT or a cube?) is boosting saturation in such a way as to vastly amplify any noise. If you want to share the LUT I can take a look and see, or maybe you could grade to your end result in a different way that doesn't unduly exasserbate noise.

Graeme

 

[John Brennick]

Log3G10 puts mid grey at 1/3 (or 33.33...%) so it's not reducing exposure, just using a different target point for mid grey. Any development LUT you'd use on Log3G10 would take that into account. Because Log3G10 encodes a very wide dynamic range, the mid grey point for encoding had to shift a little compared to what is used on lower dynamic range curves like RLF.

I looked a tthe R3D here and it looks pretty good and normal. It seems that whatever LUT you're using (is it a straight 1D LUT or a cube?) is boosting saturation in such a way as to vastly amplify any noise. If you want to share the LUT I can take a look and see, or maybe you could grade to your end result in a different way that doesn't unduly exasserbate noise.

Graeme

Hey Graeme,

kind of off topic, but somewhat related to this issue here. Is DEB applicable for Helium footage? And if it's not right now, is that something that would be? Also, would DEB ever be something that the RED ROCKET-X could process? I've always enjoyed the look of RED footage with DEB applied and the red noise I'm seeing in his example, while I'm not sure if it's olpf related or not, looked similar to skintone highlight olpf noise. Just curious about the future of DEB. Thanks Graeme!

 

[Michael Hoffman]

I think they're referring to the heavier amount of chroma noise (though subtle) - in the lower light and soft focus areas of the wheat grain. You can see the red pixels.

I'm used to running a chroma reduction (I have a homemade filter that decreases JUST those areas I use on Blackmagic shoots)... but I get the point. This is a higher end cinema camera.

Curious to see if this pops up with my Epic W.

 

[Josiah Duncan]

I tried to take my eyes off of it for a little bit and come back to re-grade. This time I used RWG>Log3G10>Lut>Saturation+Exposure adjustments. This time I enabled DEB and it helped tremendously. I'm also liking the colors coming out of this more than the RC2>RLF>Lut, and a lot of that could have to also do with applying all the saturation post-raw development. Just to clarify, the preferred way to add saturation is by using the bottom controls in cine-x, not the top?

The lut I was using is for Log>Kodak2393 from the Osiris pack, I've attached both the full res image and lut if anyone is interested.
Image: http://img.pixady.com/2016/11/269843_fixed.jpg
Lut: http://s000.tinyupload.com/download.php?file_id=51032897035252265521&t=5103289703525226552161740