Welcome to our community

Be a part of something great, join today!

Log in
  • Hey all, just changed over the backend after 15 years I figured time to give it a bit of an update, its probably gonna be a bit weird for most of you and i am sure there is a few bugs to work out but it should kinda work the same as before... hopefully :)

Scarlet / Noise

S

Stephane Marino

Well-known member
Joined
Jan 22, 2012
Messages
67
Reaction score
0
Points
6
Location
Nice, France
Website
www.stephanemarino.com
Hi,

I would like your opinions on the noise I have with my SCARLET :
In the attached R3D snapshot, do you think the amount of noise is normal ?
A001_C003_0724JI.0000052F.R3D

I shot with a Canon mount, 50mm, F4, 1/48 shutter, ISO 800 and did a black calibration just before.

Thanks for your help.
 
Last edited:
Stephane Marino said:
Hi,

I would like your opinions on the noise I have with my SCARLET :
In the attached R3D snapshot, do you think the amount of noise is normal ?
A001_C003_0724JI.0000052F.R3D

I shot with a Canon mount, 50mm, F4, 1/48 shutter, ISO 800 and did a black calibration just before.

Thanks for your help.
Click to expand...

First off my opinion is not a professional one, I just mess around with my scarlet a lot. I would say after looking at the waveform you need to crush the blacks for the piano. You have that piano at 14% IRE. Thats above the 9-12% IRE red even defines as Dark Grey. I think Black (not True Black) is supposed to be around 3-5%IRE and when I crushed the blacks down It looked pretty clear to me. If you leave dark areas where the should be there will be no noise, but if you map them up into the midtones you get noise.

My experience is that with RAW you can push you midtones anywhere you want, but 5-7 stops under middle grey can move only about a stop before introducing noticeable noise into the image. The difference between 3-5%IRE and 14% IRE is more than a couple stops. If I got it wrong, guys please let me know.
 
Last edited:
The window makes the histogram look well exposed, but everything else is actually underexposed – especially the piano.
If you expose at 800 it is very important to expose to the right – or expose at 320-400 iso and watch your highlights.
Just my 2 cents
 
I just read this too :
http://www.red.com/learn/red-101/exp...th-red-cameras
It also says to raise ISO to protect highlights and decrease ISO in low-light... exactly the contrary of what I was thinking.
It says that with higher ISO the middle gray is mapped to a darker tone (more tonal range for highlights), but then it should give a darker image no ???

HELP... :)
 
When we're talking about mid grey as recorded in the R3D the whole concept of ISO is that the correct ISO will always map that recorded mid grey to an mid grey in the final image. If that mid grey is recorded at a low code value you have more code values above it, hence more highlight protection, but more noise as the mid grey is recorded darker. The amount of gain the ISO control applies is always the right amount to map that mid grey to the mid grey in the final image.

Increasing ISO just makes the image brighter. If you retain the same image brightness by compensating with ND or aperture, you now have the same brightness for mid grey in the final image, but the mid grey code value is recorded at a lower value. As it's a lower value, you now have more available code values above it for more highlight information.

Graeme
 
Graeme Nattress said:
When we're talking about mid grey as recorded in the R3D the whole concept of ISO is that the correct ISO will always map that recorded mid grey to an mid grey in the final image. If that mid grey is recorded at a low code value you have more code values above it, hence more highlight protection, but more noise as the mid grey is recorded darker. The amount of gain the ISO control applies is always the right amount to map that mid grey to the mid grey in the final image.

Increasing ISO just makes the image brighter. If you retain the same image brightness by compensating with ND or aperture, you now have the same brightness for mid grey in the final image, but the mid grey code value is recorded at a lower value. As it's a lower value, you now have more available code values above it for more highlight information.

Graeme
Click to expand...

Once again Graeme you have me pulling a face like I have just seen a David fincher movie in reverse... :)

I would have said ( in my simple way) It's simply under exposed and the more you pull out detail in the dark areas the noisier it gets. Hence why I shoot at 320 (but I protect my highlights) , When I shoot at 800 I also see noise. Not that it worries me, because I like a little noise. :)
 
The small amount of noise you get at ISO800 is generally very good for an image.

ETTR is a different exposure method to "ISO". It doesn't necessarily mean you'll end up with a differently exposed image, but it's another route to get to optimum exposure.

Graeme
 
Graeme Nattress said:
The small amount of noise you get at ISO800 is generally very good for an image.

ETTR is a different exposure method to "ISO". It doesn't necessarily mean you'll end up with a differently exposed image, but it's another route to get to optimum exposure.

Graeme
Click to expand...

I just finished a night shoot... its all grainy because I pushed the hell out of the R3D's to get some more out of the blacks... but it still looks very nice and sharp. I'm more than happy... :)
 
Graeme Nattress said:
The small amount of noise you get at ISO800 is generally very good for an image.

ETTR is a different exposure method to "ISO". It doesn't necessarily mean you'll end up with a differently exposed image, but it's another route to get to optimum exposure.

Graeme
Click to expand...

Could you expand a little on why the "small amount of noise is generally very good for an image"?

I also don't really understand your second statement. I thought these were two parts of the one approach?

Always really appreciate your posts on the boards. Though sometimes like now I can't fully understand them :)

Best

Lliam
 
Graeme Nattress said:
When we're talking about mid grey as recorded in the R3D the whole concept of ISO is that the correct ISO will always map that recorded mid grey to an mid grey in the final image. If that mid grey is recorded at a low code value you have more code values above it, hence more highlight protection, but more noise as the mid grey is recorded darker. The amount of gain the ISO control applies is always the right amount to map that mid grey to the mid grey in the final image.

Increasing ISO just makes the image brighter. If you retain the same image brightness by compensating with ND or aperture, you now have the same brightness for mid grey in the final image, but the mid grey code value is recorded at a lower value. As it's a lower value, you now have more available code values above it for more highlight information.

Graeme
Click to expand...

Ok, now I got it !!
Thank you Graeme.
 
A couple of things: 1.) do you know what post processing that image went through (maybe noise reduction) and 2.) it's outdoors and while it may be overcast/cloudy-dusk there is a buttload of ambient light. Plus, because it's overcast, it's flat light, meaning the exposure and dynamic range is much less than what your image has (trying to balance exposure for dark inside with bright window). Basically the same as everyone else is saying -- since you lowered exposure slightly to capture more highlight detail, the stuff inside is actually underexposed... The quickest way to test the cleanliness of your image is to lower the ISO to 320 and look for mid-grey (around the window) to see how noisy it is (it'll be as clean as that second screen you posted.)

That frame grab could have been captured and rendered at 320ISO through and through with barely any tweaks in post... I think the guy that made that short is from Toronto and he's on reduser, so maybe you could get him to chime in on his settings and compare to that.
 
Last edited:
Hey Stephane, I played with you still and yeah like others posted, it seems pretty good once colored and brought to normal levels.

While I say that, I have experienced a lot of noise in areas of my frame I would not expect to be that noisy. Sometimes, in a high contrast light setup, the fall off of light has to move naturally. But, with that being said, Red is not a format you can point and shoot; which is a love-hate relationship for me. It is nice because it allows you to think of lighting more and think of your values. Once you understand what your going to get from set, to the final you will be better off. I still am playing with values and scenarios, all based on my own experience of understanding it. I do wish the image was a bit cleaner sometimes. Keep at it, test some more and see what works and what doesn't.

Based on your sample image, I think your camera is fine. And based on the last image of the man on the bench, I agree with Mike. That could have been shot with more light than the final output of it, hard to tell what the image went through to get that. Best of luck, keep trying to see what works.
 
I think it's a really great shot and you have an awesome lens.

Like someone else said, I'm new to Scarlet but have done a ton of testing with the noise as well.

A shot like this would be great for me. However, there's some considerations that I have to keep in mind. This isn't video. It's 4k film. 4x more info than in a my Canon7D. If it was on a theater screen it would look marvelous and video wouldn't.

Beyond that, as the other techniques mentioned, you could easily crush the blacks if your personal asthetics said so. However, I probably wouldnt do much on this. Last thing I want to mention, in my research concerning noise I read Canon's has two onboard "Denoisers" that does the grading before you get the image and what you get with Red is completely untouched RAW image. If you want to grade, you have to do it yourself.

I tested a bunch of Denoisers for my Scarlet "Job's Joy" including Nuke, and various AfterEffects filters. What I have found is for $99 NeatVideo (pro version) is the ultimate solution for any denoise concerns. I mean I had at times unusable images at the opposing end of the spectrum compared to this image and it came about very doable. My research on noise was concluded after my purchase of NeatVideo. Period. Done.

Just my limited but relevant experience with noise.
AVT
 
Last edited:
Seems right to me
a001c0030724ji0000052f0.jpg
 
You must log in or register to reply here.
Back
Top