Noisy blacks

[Steve Sherrick]

What???
Scarlet also has low compression,
as in, low as 3:1 just like the EPIC-X.

???

Can only get as low as 6:1 on Scarlet at 24fps in 4K.
I suppose if you wanted to shoot less fps you could get down to 3:1.

 

[Chris McKechnie]

Well it seems most users agree that 800 ISO is the best to shoot at, even outside.

So when you guys are shooting with daylight balanced LED's, what are you doing to combat the green spike? I know most sill use minus green filters, but that didnt really seem to help all that much. I have never seen my Rosco LED's look so green before on any other camera.

I really do appreciate all the opinions and feedback. This is just one of the many reasons why I chose RED.

 

[Lliam Worthington]

Great thread. So this is effectively a bump to keep it rolling.

+ 1 on the Gunleik book btw.

Cheers
Lliam

 

[Thomas Koch]

Well it seems most users agree that 800 ISO is the best to shoot at, even outside.

So when you guys are shooting with daylight balanced LED's, what are you doing to combat the green spike? I know most sill use minus green filters, but that didnt really seem to help all that much. I have never seen my Rosco LED's look so green before on any other camera.

I really do appreciate all the opinions and feedback. This is just one of the many reasons why I chose RED.

If they are daylight balanced, they shouldn't have a green spike, right? i have a couple 6500 K lights (prototypes) and use a minus green to get them down to 5600 and they look right at that temp. The final versions, they got down to 6500 by more precisely binning the LEDs.

 

[Chris McKechnie]

Like fluorescence, LEDs are a discontinuous spectrum light source - which means they do not output the entire visible spectrum of light and have spikes in certain wavelengths (green being one of them). This green spike is more prominent with the RED camera system. I am not sure why, but it is something about the image sensor that makes some LEDs (and my LitePads) look even greener than they actually are.

 

[Martin Stevens]

Can only get as low as 6:1 on Scarlet at 24fps in 4K.
I suppose if you wanted to shoot less fps you could get down to 3:1.

Ahhh, yes. I forgot.

However, 6:1 is not so bad.

TGWDT was at 5:1, and lots of other major features.

 

[Evin Grant]

We haven't seen much diffference in noise character till 10:1 and even then nothing bad till 12:1. I tend to shoot at 6:1 on the Epic simply because it gives me the record time I'm comfortable with on the 128GB SSDs.

 

[Florian Stadler]

If you look at the blue channel only you can see an appreciable difference in noise characteristics between 5:1 and 8:1. I'm sure 6:1 is a solid size/benefit compromise...

 

[Gunleik Groven]

Gunleik,

When are you going to write a book?

Hm... LOL!!!!

When it's commishioned and I can ask for articles from those I would like to have with me, maybe... But I guess really: Never.

Which would mean that I would ask for articles by Graeme, Michael Most, Gary Adcock, Michael Cioni, David Mullen, Nick Shaw as a start, in addition to myself.

If I should write a book, I think I would prefer the general theme being Digital Cinema, and Digital cinema cameras in TV productions, rather than a RED specific theme.

BUT with specified subchapters for the most relevant cameras, being (methinks)
The RED lineup, Arris cameras, C300, and Sony F65/F3

I would not spend time on the DSLRs, except maybe for a small chapter explaining the differnce between digital cinema cameras, and the DSLRs...

So...

But seriously:

Currently trying to de-geekify myself a bit and get back to mainly shooting and directing.
That's where it started and is at for me, really...

Artiocles and tests take so muuch time and focus away from what it is really all about.

Good to have that knowledge when you shoot, though...


LOL

 

[Alexey Milokost]

Gunleik, great article!!!
So what is your recipe on shooting with MX? Do you prefer 320 or 800 both in 3200 and 5600 enviroment? Do you advice to use 80C on lens? Please, share your best practice.
Thanks,
Alex

 

[Gunleik Groven]

Back after a while.

800 is allways a safe bet.

If you want to go bellow, do extensive testing and feel comfortable, don't just lower the ISO because you're out of NDs...



If the image looks great at 800, it probably is...

 

[Will Keir]

"As for color temp," you are talking about the Kelvin? I was under the impression Kelvin was just meta data, 3k, 5k, 8k, you can change it anyway you want in post.

Watch the histogram and the goal posts. Expose as far to the right of the histogram without clipping and you can bring everything down in post. As for color temp, the lower you go the more noise your going to get unless you expose it perfectly.

Noise goes away if you crush the blacks.

 

[Gunleik Groven]

"As for color temp," you are talking about the Kelvin? I was under the impression Kelvin was just meta data, 3k, 5k, 8k, you can change it anyway you want in post.

Sure, but that does not mean it doesn't affect your image and noise. Last time I checked, the sensor stays the same, whatever metadata we use, and the sensor has some preferenses.

G

 

[Will Keir]

My thought was that since it is meta data, you can set this at any value and in post get what you want. So in that light, it would not effect the noise which obviously is incorrect. Changing the Kelvin in camera DOES change the noise floor?

Early on the R1, I forgot to white balance and noticed the footage, once brought back to target kelvin, had more noise then when I just WB correctly to start.

I've read you article on ISO Gunleik and appreciate what you do for the community, thanks for keeping the info alive. I could understand ISO much better. Looking for more clear education on the subject.

Sure, but that does not mean it doesn't affect your image and noise. Last time I checked, the sensor stays the same, whatever metadata we use, and the sensor has some preferenses.

G

 

[Scott Crawley]

My thought was that since it is meta data, you can set this at any value and in post get what you want. So in that light, it would not effect the noise which obviously is incorrect. Changing the Kelvin in camera DOES change the noise floor?

No. Setting K in the camera does not effect the noise floor directly. It could indirectly have an effect if, for example you set the camera too blue and as a result you were to warm your tungsten sources. It's the same kind of equation as ISO/aperture, but applied to color balance.

The sensor is hungry for blues. If you feed it a diet of tungsten AND under expose you are compounding your trouble. Logically the opposite should also be true; If you plan to be working close to the noise floor you will get better results if you light in daylight temps or blue filter the lens. For the most part the old guard here seems to prefer the former.

That is how I understand it.

 

[Scott Snare]

If you choose to shoot tungsten you need to protect the blue channel with enough light to keep noise low. Really only a problem when the scene is dark and the histogram starts to clip blacks. Shooting tungsten can be an issue in low light situations.

 

[Chris McKechnie]

Love this thread and the wealth Of knowledge in it.

 

[Matt Ryan]

Ditto to what everyone has said about kelvin and shooting in tungsten. Underexposure is easier to achieve in tungsten given the sensors need for blue channel exposure. This can cause excess amounts of noise in your image.

For example, a simple test is go outside at night and shoot a car under a street light at say 1600 and wide open. Take that image into RedcineX and switch the meta between 3200 and 5600. Other than the color shift look specifically for noise in the image. You will see less noise when you slide the kelvin to daylight. Ive done it several times and always the same result. Its not the camera's fault, its our fault for not giving it optimal light for the given shot.

The sensor is daylight rated and for tungsten you lose a couple stops in exposure (noise wise ) which results in more noise if you forget to compensate. Like said previously, when shooting tungsten it is more important to nail proper exposure and if possible throw some gels on the lights to improve the blue channel which is most responsible for noise in your images. When shooting tungsten i usually rate the camera at 500asa if i dont have the gels or daylight balanced fixtures available. This improves my images significantly. All in all, although kelvin is just metadata, it makes a difference what temp your lighting with and the exposure given that color temp.

 

[Will Keir]

Steven, in an ambient light / sundown situation with a F3 lens, how do you expose properly? Impossible? I should be able to film a dark scene without noise?

The simple answer - the shot isn't exposed properly. Also, be careful of color temps of your lights.

 

[Gunleik Groven]

"As for color temp," you are talking about the Kelvin? I was under the impression Kelvin was just meta data, 3k, 5k, 8k, you can change it anyway you want in post.

Ouch. Get a life Gunleik, don't pop in on all these threads...

BUT

WB from daylight to tungsten happens through compressing the RED signal and lifting the blue.
(Just watch your RGB wAveform in RCX-P while you slide and it becomes very clear to you what happens...)

And that is all good and fancy and is exactly how you want it to work. Nothing wrong with that at all!
AND the processing is done through float-curve operations, which is also very nice and handy as it does not ADD noise, like the old "gain/mult" operations did, but it makes the noise that is there more visible in the chanels which is underexposed.

Thus... As the sensor is daylight ballanced, this means that under tungsten the redchannel gets clamped and the blue gets attenuation. And the further towards low kelvin you ballance the camera the more significant these two operations are and affect noise and image carracteristics.

All fine. But this is where metadata all of a sudden affects your exposure. I have seen multiple examples where people have lowered their blues on set after WB'ing, because the image looked too blue, and then wanted to make a low exposed blue-looking scene and comming home in post with a terribly underexposed image, they had some confidence in on set....

so...

The other effect, in the higher end of the exposure, is that the RED channel can be clipped before it is really overexposed (in the RAW signal). Clipped/clamped red looks like... the yellowis/metallic video-overexposed skin we all know from TV and have seen very often and is a signature element of video-ugly.

The third effect I think I see in the RED images, is that the processing seems to change a bit at or around 4400 kelvin. Lower than that, it looks like there may be some gain operations involved in addition to the curves. At least I have found that around there, it seems to add noise to low exposures in the blue channel.

(derailing here, but I conciously call it a low exposure, NOT an underexposure. There's nothing bad in exposing LOW, if you are certain you want a dark image, it's when you try to lift it all kinds of stuff hits fans....)

So.... There are at least a couple of things to remember.

Kelvin is only metadata, but like ISO - the image metadata creates may affect your lighting and exposure choices in a bad direction, and lights and exposure is NOT metadata.

And:
In an RGB grade, where low noise is a priority (I know that is not allways the case for me) you may want to set the Kelvin at a point where you find the noise pleasant and grade from there in RGB.

And:
Developing a noisy image too dark, definitely is a good way to give up whatever info was left in the darks.
Thus: RLF is your friend....

So how do I get around this?

Actually I just don't bother most of the time.
But I usually set the cam flat and shoot from there. (5600/800)
That way I can actually see if the redchannel is clipping or if it is only a metadata clip, and I can light and know about what exposure I have in the blue channel. But honestly THIS IS NOT HOW I THINK WHILE ON SET. Then I am only considering how to get a nice image home. But setting the cam flat is a way to NOT get unexpected results home.

Others have their other ways (Like Mark Toia) to work with these parameters. It doesn't matter HOW you deal with these issues. What matters is that you get intimate with your camera and post and find YOUR ways to deal with it and be comfortable with the images you see in the different stages and can controll exposure and noise.

That's the cool thing. Because there isn't really any bad exposures or bad noise. It is about controlling exposure and noise. You ride the camera, not the oposite!

That feeling is what makes me really love the Epic!