Lumix S5 video noise comparison

[robin0112358]

I believe that everyone should decide for themselves what settings optimise their photography/videography. This means shooting test footage. I recently compared Natural and V-Log profiles at four ISO settings on the Lumix S5. Discussion and embedded YouTube video at my blog

Please, Log in or Register to view URLs content!

.

 

[inspiredtimothy]

Isn't much of the noise in VLOG due to less aggressive noise removal, not necessarily that the camera does worse in that mode?

 

[robin0112358]

Isn't much of the noise in VLOG due to less aggressive noise removal, not necessarily that the camera does worse in that mode?

Aggressive noise removal always produces artifacts, unless it's a time-consuming "AI" mode like DXO Raw has. Any in-camera NR would produce a smoother image free of necessary detail. I simply don't see that in the footage.

 

[Michael_too]

When I hook up my S5 to a TV and play around with picture profiles and noise settings I can clearly see that the noise suppression on the default profiles is quite aggressive. Even when setting it to -5 it is stronger then say 0 with vlog. It shows most clearly on human skin as we humans are very well trained on seeing if skin looks correct, healthy etc. wheras we don't look quite as carefully at things like apples and such. In vlog, with NR set to zero, the image looks very life-like with accurate skin texture, despite some noise showing up but even increasing the NR a bit quickly gives a plastered-over look. In the standard profile you can't get the same sense of realistic skin texture even at the lowest NR setting.
These subtle differences become very hard to spot once you record a file, send it to youtube and then watch it again. Youtube acts as an aggressive noise filter. So for practical purposes it might not matter to everyone equally.
The point of vlog is that it is as close to RAW as possible without being RAW while the other profiles already have some preferences built in according to what the manufacturer thinks makes a good image. The natural profile has the least amount of stylization baked-in of them all.
It also means that vlog must be exposed a bit differently than the standard profiles. Vlog must be exposed in such a way that the dynamic range of the scene is captured as completely as possible which might mean you sometimes intentionally under- or overexpose as you can later easily shift the exposure in post and get the same result as-if you had shot the scene with that other exposure. The standard profiles are designed to expose as close as possible to how you want the end result to visually look. Shifting the exposure in post can give inaccurate results compared to shifting the exposure during recording.

 

[robin0112358]

Good info. I would be interested in reading any technical information or actual comparisons that demonstrate this NR. Certainly it cannot explain the huge differences I am seeing. I honestly was shocked to see the V-Log footage.

I recall a parallel conversation, likely on the Blackmagic Forum, about BRAW. Folk complained that it had aggressive NR compared with ProRes RAW. I could see a small difference if magnified but in realistic footage it was irrelevant. But that was a tiny difference.

I mention in either the article or the video that these shots would have benefited from over-exposing in camera when using V-Log. Admittedly I should do another shoot and try to do better by V-Log. But my time is limited for such things.

ETTR is a bad solution for me. My usual shooting is in more aggressive lighting situations where I cannot over-expose. The last concerts I shot (stills only) had blown out lighting spots everywhere... it cannot be helped with stage lighting. The last thing I would want to do is push the exposure to lift the musicians out of the noise regime. Half the frame would be blown out.

If anything, I want to to the opposite: pull back the exposure one stop in-camera, then boost shadows in post. In V-Log, that is going to do horrific things in terms of noise. In Natural it works just fine.

 

[Michael_too]

Vlog should offer the most dynamic range, so it should be ideal for stage lighting scenarios. It's just that Vlog spends a lot of it's available bandwith on the shadows where the noise lives and less so on highlights compared to the standard profiles. This has the advantage that it samples the shadows including its noise very accurately, making the noise very visible but also very suitable for cleanup in post. Standard profiles hide the noise by spending fewer bits on it to begin with and adds some NR in the mix. If you are willing to do your NR in post Vlog will work well but if you prefer to get it right in camera and use the algorithms built into the camera for which you already paid good money than I can't blame you for that.
I shot some stage performances in bog standard rec-709 using camcorders with their tiny sensors. That was always a struggle with exposure. The S5 by comparison is sheer luxury for me.
Don't forget, when a stage performance is shot professionally, it is not uncommon that the lighting is adjusted to match the limitations of the cameras used. Getting equally good results without having any control over the lighting is sometimes just impossible.

 

[robin0112358]

Though I sometimes know the lighting tech, I don't have enough pull to get them to alter anything.

I agree that we are spoiled now in comparison to even a few years ago. I first shot gigs in the 1980s on film and had no idea what I was getting!

 

[CharlesH]

I mention in either the article or the video that these shots would have benefited from over-exposing in camera when using V-Log. Admittedly I should do another shoot and try to do better by V-Log. But my time is limited for such things.

Lumix cameras have a waveform monitor, and I always use this to set exposure with V-Log, from way back when I first had a S1H. The terms over-exposure or ETTR don't mean much with log video, and a good practice is to learn to ignore any exposure meter readings. I expose so the brightest part of the scene doesn't over-expose on the waveform monitor. I usually shoot at the V-Log base ISO of 640, and up to the second base of 4000 without noise in the shadows.