GH5 Keeping noise under control at higher FPS?

[filmguy123]

Any guidelines on Max clean ISO at higher frame rates?

I'd say the GH5 is pretty clean up to 3200 ISO

But that's at 24p... even 60p?

But at 120fps, ISO at 3200 seems a lot noisier?

Any range guidelines i.e.:
24-60 up to 3200 ISO
60-120 up to 1600 ISO
120-180 up to 800 ISO

... ? I haven't tested those just wondering what feedback you may have on clean ISO as it relates to FPS, and why this happens? (Is it just the codec bitrate encoding noise worse? Would external recording make high ISO at higher FPS cleaner?)

 

[TheDingo]

I think you need to factor in noise-reduction in post, because camera NR has to do everything in real-time where NR in post can take a long as it wants to clean up the image.

 

[Cary Knoop]

I think you need to factor in noise-reduction in post, because camera NR has to do everything in real-time where NR in post can take a long as it wants to clean up the image.

True, but I think we are comparing two different kinds of noise here.

The noise from the sensor can be much more efficiently reduced before it goes to the encoder as it is still raw data.

Once encoded, technically a lot of the noise disappears and only reappears as DCT pattern distortions.

As a test try to encode a white noise pattern and see what happens! Noise, after all is the enemy of effective compression.

I think the Panasonic engineers were wise in limiting the ability for users to take out too much noise reduction because the noise damage once going through the encoder cannot be recovered in post.

 

[TheDingo]

I think the Panasonic engineers were wise in limiting the ability for users to take out too much noise reduction because the noise damage once going through the encoder cannot be recovered in post.

The GH5 also has temporal noise-reduction, so I'm hesitant to turn in-camera NR down to -5. ( currently using it at -2 )

When I can find the time, I want to do a bunch of NR tests where I compare different camera NR settings when combined with NR from NeatVideo. With NeatVideo I have build a NR profile that is specific to the actual camera settings including ISO, so I am probably going to pick something like 3200 ISO for the first set of tests, and then see what camera NR setting works best with the NeatVideo profile. Once I figure out 3200 ISO settings, I'll try going down to 800 and see what works best there.

 

[filmguy123]

These NR tests are really important. I'm no good at these sorts of scientific tests myself, but hearing from some of you in the community regarding this would be super helpful!

So any guidelines or idea as to when higher ISOs should start being avoided at higher frame rates? I am guessing this is a function of total bitrate (ie an aprox estimation of how many bits is going to each frame, which I know can change depending on frame activity as its variable), but if we knew roughly how many bits per frame were needed before high ISO noise starts becoming probelmatic, that would be helpful in developing a rough guideline such as:

24-60 up to 3200 ISO
60-120 up to 1600 ISO normal-high activity, up to 2500 ISO low activity
120-180 up to 800 ISO, up to 1250 ISO high activity

^^ I pulled those numbers out of my butt... but see what I am getting at?