Changing ISO in post for RAW

[Hickup]

Hello,

currently I'm preparing for a trip with my C500 Mark II that will include capturing footage at night, specifically polar lights. Second camera will be a C70 with the Canon Speedbooster, both cameras will use the Sigma 14mm F1.8 (the C500 will use the cinema version of that lens).

Right now I'm trying to find the best approach on reducing noise because obviously it will be pretty dark. Apart from watching a lot of YouTube tutorials, I'm doing my own experiments as well and I ran into a pretty massive contradiction, at least from my point of view. Eventually somebody with better insights can help me?

There are a lot of claims that when shooting RAW, ISO can be changed in post and in my preferred NLE DaVinci Resolve Studio, the parameters offered for manipulating RAW are indeed allowing me to change ISO as well.

So I made a test and maxed all capabilities of the camera, e.g. using shutter of 1/12 and shot a series with alternating ISO settings. Starting from 100, I went to 800, 6400, 12800 and finally 25600, each one minute only. Both the C500 and C70 were side by side. I left internat noise deduction at default settings, although I found a tutorial that recommended changing Frame Correlation.

Once I got the clips in DaVinci, I started tweaking the footage and just like I suspected, changing the 100 or 800 ISO to 12800 was not even close to the footage shot in 12800 in-camera. There was a lot more noise, the colors were worse and observing the waveform display, it looks like changing ISO was doing nothing else than raising levels with the information still lost in noise.

So my conclusion is that changing ISO in post is not possible and proper attention still needs to be paid on exposing correctly with an appropriate ISO setting. Which is quite different from what several talking heads claim, at least the way I understood their explanations.

Am I missing something in my experiment or are those claims about being able to change ISO in post wrong?

Kind regards,
Frank

P.S. The footage captured with the C500 in ISO 12800 looked quite good, unfortunately the C70 could not quite keep up. Experiments with in camera noise reduction will follow.

 

[morgan_moore]

Im sure you are right. There is some sort of internal amplification before writing the data to carc

in 2015 red cameras just changed the in camera lut with an iso change - so fuddling in post had the same effect as fiddling in camera

not so with a modern canon

some testing may prove them to be dual iso in camer

so 6400+post post might equal 12800

 

[AndreeOnline]

Am I missing something in my experiment or are those claims about being able to change ISO in post wrong?

It's camera dependant. With REDs and Blackmagic's Pocket cameras you can change the ISO in post. For dual ISO cameras, you can change ISO with the defined bracket. If you shot a Pocket camera in the "ISO 400 bracket" you could choose between 100-1250 for instance.

Once you are in post, changing ISO in raw will change how bright or dark your footage looks, but you are not affecting the underlying linear data.

A bit of general info that might be too basic for you, but might help someone else:

Cameras capture linear data. Your settings at the time of image capture—specifically aperture and shutter—will dictate your ability to saturate the image sensor. Completely saturating a photo site will lead to clipping. Choosing a higher ISO in camera will lead to a brighter image preview which is also reflected in color managed in camera tools such as False Color. The brighter image might lead you to stop down the camera and "starve" the sensor of light (but at the same time you are "protecting the highlights").

You can look at ISO as a gamma curve that distributes the stops of dynamic range around middle gray. Selecting a higher ISO, and stopping down aperture and/or shutter to allow for it, protects highlights but pushes the meat of the images down into an area of less signal to noise = more noise.

Selecting a lower ISO, and opening up aperture and/or shutter to compensate, will lift more of the image away from the noise floor and produce less noise. At the same time, if you have a contrasty scene, brighter areas are more likely to clip sooner.

Many refer to the interplay of selecting an ISO and changing aperture to compensate as "rating the camera for ISO XXX". That is, you know that setting ISO to 1600 and not clipping will lead to better protected highlights (the desired stop distribution) and a noise level you are comfortable with. This would be "rating for 1600".

Remember to always expose for the light: direct or reflected. When shooting fireworks at night, you don't go high ISO and open your camera up just because everything is dark. The actual subject—the fireworks—are bright shining lights.

Being in the countryside at night, in a field, you don't go to ISO 25600 and a slow shutter at f1.4 to allow the muddy ground to hit 18% gray on the sensor if there is no light on it. We don't shoot the absence of light. If you do, you better capture a black image.

Polar lights are fine to shoot of course, and I would count that as a low contrast scene (the whole subject is at an even light level), which would make me want to go as low as possible with ISO. Still, polar lights have a very low EV (exposure value) so I'm guessing you'll still need ISO 6400, a good lens at open aperture and a slow shutter (which should be fine).

My goto would likely be to try lowering ISO by dragging the shutter in slow moving lanscape scenes with low light. If the camera is stationary, you could go pretty slow with the shutter, right? Keep doing tests.

 

[cpreston]

I'm interested in seeing whether your experiments with in camera noise reduction have any effect on RAW data. I would guess that it doesn't as I don't think any actual noise reduction is taking place. It is interesting that the dual gain sensor on the C70 can't keep up with the C500MKII at high ISO.

I did a test on my C500MKII for ISO's below 800 in RAW. I did not see any benefit at shooting at lower ISO's other than for monitoring purposes. I was grading in FCPX which doesn't have an adjustment for ISO. The image contained the same data and noise whether I was shooting at 800 or 200 ISO, but it was just harder to grade since all of the information at 200 was compressed into the lower exposure levels when the import LUT was applied.

 

[Hickup]

Thanks for all the answers!

I wasn't aware that it is camera dependent, but I think it is confirmed that changing ISO in post in Canon RAW will not recover data from within the noise level. I still need to learn a little more about color space transformation in DaVinci, but at least I understand now that RAW Light settings are not appropriate for tweaking ISO - it is still fully dependent on the ISO setting while the clip was shot. And in normal light conditions, I would stay with the base ISO 800 as well and not use the extended range, but night is different.

In regards to in camera noise reduction, I'm actually quite sure there will be no effect in RAW, the manual is quite explicit about it: For Custom Picture Settings it says: "Custom picture settings have no effect on RAW recordings". I still want to give it a try.

I do a lot of photography of polar lights, but for some time now, I try to get decent quality 4k video which is a lot different. I'm sure that slow shutter speed mode with 1/12 of exposure will be mandatory to keep ISO as low as possible. I managed to get a few shots last season, but weather was pretty bad during my time on the north cape and when it cleared up, there was no lights.

I bought the C500 mainly for the large sensor because in my experience, nothing beats sensor size for polar lights. When experimenting, I used a Canon 14mm F2.8 on the C70 with the speedbooster to get to the same aperture as the 14mm on the C500, so with comparable lenses, the C70's sensor cannot keep up. But the Sigma 14mm is offering F1.8, with the speedbooster that is an equivalent to F1.4, which is by far faster than the lens used for the experiment So in real world conditions, the RF mount plus speedbooster will likely make a big difference, I expect the C70 to produce great shots as well. I'm just super annoyed that the rumors about a RF mount for the C500 never realized and a little envious that most recent firmware upgrades for the C70 include features that I would appreciate to have on the C500 as well.

Kind regards,
Frank