A7s and 8bit banding

Those samples are procesed: there's a ton of contrast added to them. You'd be hard pressed to see the banding in the original footage. You can try, actually, he posted the link where you can find it!!
You have to add contrast when correcting the low contrast of SLOG2 so it's a valid test IMO. I've not seen banding in uncorrected SLOG2 footage - it appears when you process it.
 
Before you all go buying shoguns- see this thread- Premiere (have not tested the new 2015 update) has problems with XAVC- specifically the video engine does not access the complete color space of the codec. Actually Nate Weaver explains it best: http://www.dvxuser.com/V6/showthread.php?337008-Severe-Banding-in-S-amp-Q-mode-180FPS/page2
I'm seeing better quality from Resolve with regards to less banding but this doesn't fix the issue with the loss of resolution and other edge artefacts from the internal UHD-HD downsampling. Also I'm not seeing no banding in Resolve.
 
You have to add contrast when correcting the low contrast of SLOG2 so it's a valid test IMO. I've not seen banding in uncorrected SLOG2 footage - it appears when you process it.
Indeed. I never understood why people think 4:2:2 is not precious anymore... why? Because it is 8-bit??
 
Indeed. I never understood why people think 4:2:2 is not precious anymore... why? Because it is 8-bit??

422 which is an old school video term really doesn't mean 8 bit, you can have 10 bit 422. The 422 refers to how many samples (resolution) there is of Luma and Cr and Cb. So 4 means full resolution of Luma (1920 across) but the 22 means half of the colour difference channels. 420 is even lower resolution on colour channels. Our eyes are more sensitive to Luma changes so 'compressing' the colour channels is valid. However this is really a TV distribution format. When capturing it's in our interests to capture as much as possible. Both in terms of colour depth and resolution.

It's generally considered that 422 10bit is the minimum for TV capture but actually it's all a bit in the air as these days pretty much anything goes. Film capture would ideally be RAW or 444 (meaning full resolution in all channels)

The A7 series is artificially restricted really. If you compare a RAW still to a frame from the video there is a lot more tonality in the still.

There's no reason why Sony can't have 10 bit output for example or in fact some form of RAW. I don't buy the 'heat issue' because RAW would need less internal processing. But of course if they did that then it may cut into sales of the higher end cameras.

Perhaps the A7s II big party trick would be same sensor size but better quality off it and that would neatly let it sit next to the A7r II.

cheers
Paul
 
Here's a thought experiment.

The stills out of the camera are excellent in terms of skin tones (once you move away from the default Adobe settings). I have some ICC profiles that work really, really well.

But i wonder how i can match the video output to the stills.

These ICC profiles assume a RAW linear starting point, but video is in a Slog2 starting point.

If i could get a DNG of a colour chart (the kind we use to create massive LUTs with) and get that chart into lightroom i can apply the colour settings and create a LUT that goes from linear to good skin colour.

Then if i can turn Slog2 into linear and apply the same LUT perhaps i can match the RAW workflow. To get Slog2 into linear means reversing the colour matrix that the camera is using. So i would need to find this colour matrix.

So the crux here is i need to generate a DNG from a 32 bit linear image so that can be read by Lightroom and reverse the colour matrix of Slog2 back into RAW.

Anyone?

cheers
Paul
 
I'm seeing better quality from Resolve with regards to less banding but this doesn't fix the issue with the loss of resolution and other edge artefacts from the internal UHD-HD downsampling. Also I'm not seeing no banding in Resolve.

Actually one thing that Resolve appears to do is to dither the image a bit. I don't know whether this is still the case in the latest version.

I ended up moving to SpeedGrade for numerous reasons, one is the scopes are more accurate because you can see 8 bit banding in SpeedGrade but not in Resolve because of this dithering. Once you know that you're looking at accurate scopes it's quite the eye opener reviewing various sources of footage. The effects of linear RAW become very apparent.

It's pretty annoying actually when you're trying to test footage and workflows. It really ought to be an option not a set-in-stone default.

cheers
Paul
 
I agree that the test Emanuel posted is useful. And I agree that you have to add contrast to slog2 in post. But I have never added THAT MUCH contrast. Not even close. So it's a useful test, but just as an indication that you may be limited if you want to go for an ultra-gritty look. In normal scenarios, I have never seen banding in my a7S XAVC-S footage. I don't use LUTs, though, so maybe that's where the issue comes from.


Also, I think Emanuel is perfectly aware of what 422 and 8-bit mean and what they don't mean. I think what he said is that people seem to care a lot about 422 when the files are 10-bit, but they disregard 422 if the files are 8-bit.
 
Also, I think Emanuel is perfectly aware of what 422 and 8-bit mean and what they don't mean. I think what he said is that people seem to care a lot about 422 when the files are 10-bit, but they disregard 422 if the files are 8-bit.

Perhaps i misread that. Apologies Emanuel, i'm not trying to teach you to suck eggs (which is a phrase that probably doesn't translate very well worldwide)

cheers
Paul
 
Actually one thing that Resolve appears to do is to dither the image a bit. I don't know whether this is still the case in the latest version.

I ended up moving to SpeedGrade for numerous reasons, one is the scopes are more accurate because you can see 8 bit banding in SpeedGrade but not in Resolve because of this dithering. Once you know that you're looking at accurate scopes it's quite the eye opener reviewing various sources of footage. The effects of linear RAW become very apparent.

It's pretty annoying actually when you're trying to test footage and workflows. It really ought to be an option not a set-in-stone default.

cheers
Paul
I quite like speed grade but unless I'm missing something obvious I can't alter individual hue values in Speedgrade as I can in Resolve. For instance with S-Gamut I find the skies are too cyan and the hue needs nudging to a blue which is easy to do in Resolve in the hue vs hue control.
 
I agree that the test Emanuel posted is useful. And I agree that you have to add contrast to slog2 in post. But I have never added THAT MUCH contrast. Not even close. So it's a useful test, but just as an indication that you may be limited if you want to go for an ultra-gritty look. In normal scenarios, I have never seen banding in my a7S XAVC-S footage. I don't use LUTs, though, so maybe that's where the issue comes from.
Interesting viewpoint. My bet too, even though, I'm still placing my bets on 4:2:2... ;-) Hence to not discard the clean HDMI recording even 8-bit, one of the most important features in any capture device IMO if we care about professional acquisition, grading and so on. With or without internal 4K. The broadcast standard, in short.


Also, I think Emanuel is perfectly aware of what 422 and 8-bit mean and what they don't mean. I think what he said is that people seem to care a lot about 422 when the files are 10-bit, but they disregard 422 if the files are 8-bit.
100% correct on my dictionary book :-) 8-bit doesn't necessarily mean 4:2:0...
 
Perhaps i misread that. Apologies Emanuel, i'm not trying to teach you to suck eggs (which is a phrase that probably doesn't translate very well worldwide)

cheers
Paul
No worries, Paul. Your post is informative anyway, that's what counts. Samuel is already an old chap from our chats in these pages. It is very likely understandable we can already succeed to communicate through smoking signs or telepathy when happens LOL :-)
 
Unfortunately, people tend to underestimate 8-bit because it is very connected to non-professional codecs. This is a misconception. I always remind when HDCAM was 8-bit 3:1:1 (1440×1080) but who cares? :-D DVCPRO100 (4:2:2 praised so many times by Barry!) was the king over here despite to be 144Mb/s vs 40Mb/s if my memory serves...
 
I don't think anyone disagrees that the right person can shoot a masterpiece in 8-bit 4:2:0.
The real question is: why do it the hard way?
Besides, the world is moving forward, what was accepted as a necessary evil (or even as top end) a few years ago does not necessarily hold true anymore. Processing is much cheaper and much faster nowadays, so is storage.

Now, different people have different requirements and different perceptions on what is good enough. A7s internal s-log2, once processed, simply isn't if you care one bit about good image density. And external s-log2, even though much better, does have similar issues, albeit to a smaller degree. Yes, you can surely shoot beautiful images with it. But if I have to second guess myself and wonder if it is going to work out every time a patch of skin goes in the shade...well, no, thanks. Yes, I'm that picky.
 
I've just ordered the Shogun. Not just because I think it will magically solve all the issues but other advantages are;
Scopes to nail the exact exposure - using them more and more in grading and realizing how important and useful they are
1 touch 1:1 focus zoom - far too fiddly to do on the A series
XLR audio with Phantom power in so I can use my 416 in certain situations ( subject to a test against my FR2LE)
4k straight from the sensor so I can do my own resampling to 1080
Useful on the BMPCC as well to replace shooting with the awful screen.
Well that's how I talked my way into it anyway.
Will report back....
 
I don't think anyone disagrees that the right person can shoot a masterpiece in 8-bit 4:2:0.
The real question is: why do it the hard way?
Right :thumbsup: 8-bit 4:2:2 is closer to 10-bit 4:2:2 than to 8-bit 4:2:0, that's what makes the whole difference...
 
My problem with 422 is that it often requires an external recorder, and those are not good for the kind of fast shooting I do all the time. I need a small camera package that I can move from tripod to stabilizer to shouldermount in a few seconds. This are my very personal needs and for people with more time to shoot I completely understand why you'd want to get 422 prores.

What I see myself shooting in the near future is internal 4K 8-bit 420, which has the same color resolution of 1080p 444.
 
So far no problems with footage from the Shogun. That's a bit of an understatement though - the quality is quite simply amazing!
 
Even more improvements in Resolve! When grading and you hold down the alt key it only applies the tonal adjustments to the luminance channel which seems to allow quite significant tonal changes without banding!
 
Back
Top