Why recording LOG with an 8bit codec is most probably going to get you in trouble.

[Liam]

It depends - grading can be as simple or as complex as you like. 

thanks, ha, didn't realize that was a link at first. Very informative. Just was surprised that you could get anything at all from a log image with like 3 sliders considering the way people talk about it

[Ed_David]

that link to that music video shot on the fs7 and a7s blew my mind.  I need to reevaluate how good the fs7 can be with the right DP, lens, lighting, and colorist

so much to always learn

 

and Ebrahim is right - it's more than 8-bit codec.  8-bit just means less colors available and also less luminescence data so it can get more noise - but also the sensor itself affects the noisiness of it all.

All complicated -the codec, the native sensor and photo sites, and just everything else.

so find the best way to maximize a codec and I think slog on the a7s was very noisy

[hmcindie]

It becomes obvious if you read through the 'history' of the A7s colour issues - first the A7s was too green, then it was too yellow, now it's too plasticky....

It's WAY too green AND yellow. The A7s is frankly sickly (this with the cinegammas, with slog it's slightly better - And I love the highlights with slog).

[Ed_David]

gamma and matrix are two seperate things.

Log and CINE gammas control the dynamic range and contrast

matrix is the color - this I prefer PRO for more natural skin tones and use Kholi hick's settings for the best skintones.  Just experiment around.  See what combination works best.  I think generalizations like it's too green and yellow isn't give the camera a fair chance.  But then again, I'm a hypocrit because I make fun of the nx1 being too magenta.  SO who am I to judge?

[cpc]

With LOG comes great power and it is down to the user how good ends up looking. There's a lot of terrible looking LOG footage out there due to the individual grade rather than any codec weakness and I don't think 10bit ProRes LOG from the Blackmagic Cameras looks significantly different or better to 8bit LOG. 10bit is overrated for everything aside from keying. I have graded 10bit from the GH4 to Shogun and didn't see an advantage in the grade even when pushing it to extremes to see what was lurking in the murky depths!! In my view, it doesn't suddenly benefit skin tones. That is more down to the sensor and image processor than 8bit vs 10bit.

Take for example 14bit raw as one extreme in terms of colour / bit depth... only way you can go more extreme on the specs than that is 16bit on the Sony F65! To make use of 10bit let alone 14bit and 16bit the sensor has to deliver an extremely wide dynamic range AS WELL AS a massive colour gamut. Now, we already have experience of 14bit with Magic Lantern raw. Is it significantly better than 10bit raw from a Blackmagic in terms of codec or grading flexibility? No. The sensor makes more difference. The sensor in the 5D Mark III is very good, very clean, wide colour gamut. The one in the BMPC is noisy and has a harsh run off into the highlights. Great codec, yes, but 10bit vs 14bit with that same sensor performance would have made very little difference.

Now we have a shining example of how good 8bit LOG can look for colour... he is called Mr Canon 1D C. You saw how it compared to the NX1 which I previously sung the praises of for being actually very very nice indeed at capturing vivid, rich, satisfying rec.709 colour.

What LOG does for a camera, regardless of 8bit vs 10bit, is quite frankly magic and should be highly regarded.

There's 'light LOG' like on the 1D C by the way, which doesn't go as flat as S-LOG and doesn't dramatically change colour, leaving it very saturated. I think S-LOG does go to extremes in terms of dynamic range... but it is not 8bit which causes the weirder colour vs the 1D C is it? Because they are both 8bit codecs! A7S owners, indeed RX10 II, RX100 IV owners can do an experiment. Shoot 4K 8bit LOG video. Shoot a raw still. Grade to match. Compare. Difference will be smaller than you think!

As an a7s owner, with all due respect, I can say that you are wrong. There is huge difference between s-log2 and raw. Both color and image density are significantly worse with s-log2 compared to the raw stills. Also, a7s externally recorded 4k downscaled to 1080p is easily more gradable than the in-camera 1080p.

good examples here: http://blog.inventome.com/Blog/2015/2/a7s4Ktests/Sony-a7s-Exposure-and-Noise-Workflow-with-UHD-and-Odyssey-7Q-Plus

 

Sensor, noise, processing, etc surely matter a lot but downplaying the importance of the available tonal precision is pointless. Yes, you can have badly quantized and processed data encoded in 10 bits, and it will still be crap. And you can have well sampled, quantized and processed data encoded in 8 bits. And it will STILL be bad if the curve is too flat for the available space. Ideally, one needs well processed data in at least 9-10 bits for log encodings with extensive DR.

It is no coincidence that the father of all log curves Cineon log is 10-bit. Back then nothing was taken for granted, nor there was any prejudice about 8 bits, so this was extensively researched by Kodak. They concluded that 9 bits are likely enough but went for 10, since 10 fits better with computers; plus, a little bit of redundancy doesn't hurt.

And lets not forget chroma subsampling. After all 4:2:2 is twice the chroma info contained in 4:2:0, and 4:4:4 is four times the chroma info contained in 4:2:0. So not all 8-bit is created equal.

In general, wider color gamuts do help grading by having better color separation, but there is no point having a wide camera gamut if there isn't the color precision to encode it properly as is the case with S-Gamut and s-log2 in the a7s. Note that many digital cameras have native color gamuts wider than film leading to color precision smaller than scanned film when encoded with the same bitdepth.

[Don Kotlos]

good examples here: http://blog.inventome.com/Blog/2015/2/a7s4Ktests/Sony-a7s-Exposure-and-Noise-Workflow-with-UHD-and-Odyssey-7Q-Plus

Very informative link. Thanks

[Ed_David]

4:4:4:4 12 bit raw means nothing if the sensor's readout or photosite information is limited - it's the back end vs front end

I think a few people made that point in regards to the BMC 4k camera.

So many factors to take into account than 8-bit vs 12-bit, and log vs raw vs rec 709 - Ebrahim made this point really well in his post on all of this.

 

Just find a camera that you like and learn how to light and compose better.

By messing with instagram on my android galaxy s4, I have found myself getting better frames in my cinematography work.  Not endlessly speculating on what camera will give me what I need in codecs.  Just by shooting with a crappy camera and figuring out what I can squeeze out of it.

[TheRenaissanceMan]

4:4:4:4 12 bit raw means nothing if the sensor's readout or photosite information is limited - it's the back end vs front end

I think a few people made that point in regards to the BMC 4k camera.

So many factors to take into account than 8-bit vs 12-bit, and log vs raw vs rec 709 - Ebrahim made this point really well in his post on all of this.

 

Just find a camera that you like and learn how to light and compose better.

By messing with instagram on my android galaxy s4, I have found myself getting better frames in my cinematography work.  Not endlessly speculating on what camera will give me what I need in codecs.  Just by shooting with a crappy camera and figuring out what I can squeeze out of it.

It's fun to be black and white about these things, but there's a reason high-end cinematographers require Prores HQ at an absolute minimum. 

Also, why's everyone bashing the BMPC? It's not great in low light, but the highlight rolloff is actually really nice. 

Andrew: There's a slope of diminishing returns going on. 14 bit to 16 bit is a pretty small difference, 14 bit to 12 bit is noticeable but not huge, 12 bit to 10 bit is a drop in quality but acceptable if you want efficiency, and 10-bit to 8-bit is absolutely huge. 

If these distinctions didn't matter at all, why would professional cameras offer 12-16 bit recording? If they could get the same results with less back end, don't you think they would? How can you argue against the importance of color depth when you just spent an entire article railing against a 4:2:0 camera for not matching up against the colors of your 8-bit 4:2:2 camera? How can you argue against the importance of bit depth when you spend so much time talking about LOG profiles, which are designed for the sole purpose of imitating the same tonal precision as larger-bit footage in a smaller container? 

I'm not denying you your point of view. If you think 8-bit is awesome, okay. But it's important for a reviewer with your level of clout to be clear about their priorities and thought process, as that informs all your conclusions. Sometimes it feels like you've changed your mind about fundamental things without telling anyone, which makes your thought process as a critic a little inconsistent and tough to follow. 

Sorry if this comes across as rude or over-critical, because I don't mean to be. I've written as a film critic for a few years and I know it can be a struggle to get people on board with your train of thought. But there's an important difference between "10-bit doesn't matter" and "10-bit is nice, but I don't think we need it. Here's why."

Just a matter of clarity.

Cheers.

[Mattias Burling]

This is not an example of good grading or anything but its 8-Bit Log and if you ask me it holds up to be pushed around quite a bit.
I agree that there is a difference between Raw and Raw, 8-bit and 8-bit, and so on.
I would also say there is a difference between 12 stops of DR and 2 stops of DR, depending in the codec and sensor (how far one can push).

In the video I have pushed shadows and highlights quite a bit, applied a LUT (again, just a quick test) and done some keying and adjustments.

[Don Kotlos]

This is not an example of good grading or anything but its 8-Bit Log and if you ask me it holds up to be pushed around quite a bit.

Mattias, I don't disagree that you can push the image quite a bit with an 8 bit log file.  What I am saying (and seeing) is that it is hard to reproduce tonalities since color information is just not there. Especially with skin color that we have been trained to discriminate tiny differences i.e. in redness  (blushing), blueness  (blood oxygenation-sickness), etc. 

The test that Andrew proposed will be quite telling. Get a frame from a log 8 bit video and try to match it to a perfectly developed  raw still from the same camera. From my experience with 8 bit video in general, That will be hard. A log file will just make it harder. 

 

[TheRenaissanceMan]

Was anyone denying that a good sensor is the first step in achieving good image quality in video quality? No? Cool. 

Was anyone denying that some 8-bit codecs can be pushed more than other 8-bit codecs? No? Cool. 

Are people saying that higher bit depth is essential to getting peak image quality out of LOG files originally designed for 10-bit, like S-LOG? Yes. Posting examples of 8-bit C-Log footage proves nothing, because it's a much less steep faux-log designed for lower bit-depth cameras. So let's focus on some non-Canon examples. 

[sunyata]

lots of different issues in this thread. 1) does log in 8bit ruin colors? 2) does much power come with log and is it magic? 3) are wide gamuts the real problem? 4) does Kodak choosing 10bit for Cineon  mean that 10bit is the least you can get away with before you see artifacts? 5) is it really all about the camera as a package and the sensor, codec, even the lens for example? 6) should we be more concerned with chroma sub-sampling? 7) is this debate incredibly boring and useless?

4) trick question, Cineon was designed to re-print a film negative, not based on digital to digital tests, it was also R',G',B'. so unfortunately i don't think it can be used as a fair comparison, even though it was how this workflow started. 

7) not at all, i just wasted several minutes. 

i think log has the same advantages with 8bit that it does with 10bit or any other depth, even though i disagree that 8bit is not distinguishable from 10bit unless doing keys. when you combine 4:2:0 compression with 8bit you get a negative re-enforcing effect (the blocks get much larger in dark areas because they have fewer codes to use), which shows up when doing lifts in particular. but all that is slightly separate from the initial log color question, of which log is the easy part to untangle; figuring out all the other stuff is really the challenge. so in that sense i think the "other" things that affect color, such as everything in question 3 and 5, are really the problem.

the thread started with 8bit log and color, but sub-sampling and bit depth came up so i thought i'd re-post this old video i did, it's exaggerated but hopefully useful. hit spacebar (pause) when the description changes.

 

 

[cpc]

re: Cineon

I don't know what do you mean by R', G,' B' (gamma corrected values?), but Cineon encodes film densities which do represent scene light intensities logarithmically, the same way digital log curves do. It is indeed meant to be printed on print film. But the comparison is fair I think. If anything, one actually needs more precision for digital because print film grain, as fine as it is, still acts as dither.

 

Incidentally, a few months ago I was involved with tone mapping and grading this real-time rendered short:

It uses high resolution 8-bit textures, so the source image is effectively 8-bit RGB (4:4:4) and the engine works in high bitdepth precision and/or floating point (so negligible quality loss with lighting and processing). There is a (perfectly flat) log conversion operation from the engine linear space at the end of the pipeline for film out simulation purposes (simulated Kodak print). It is parameterized for around 12 stops. The grade is done over this perfectly flat curve. I found that somewhere around 12 stops is where you start to see issues if you are picky (and I am), even considering the pulled down whites in this grade and the overlayed grain.

[TheRenaissanceMan]

re: Cineon

I don't know what do you mean by R', G,' B' (gamma corrected values?), but Cineon encodes film densities which do represent scene light intensities logarithmically, the same way digital log curves do. It is indeed meant to be printed on print film. But the comparison is fair I think. If anything, one actually needs more precision for digital because print film grain, as fine as it is, still acts as dither.

 

Incidentally, a few months ago I was involved with tone mapping and grading this real-time rendered short:

It uses high resolution 8-bit textures, so the source image is effectively 8-bit RGB (4:4:4) and the engine works in high bitdepth precision and/or floating point (so negligible quality loss with lighting and processing). There is a (perfectly flat) log conversion operation from the engine linear space at the end of the pipeline for film out simulation purposes (simulated Kodak print). It is parameterized for around 12 stops. The grade is done over this perfectly flat curve. I found that somewhere around 12 stops is where you start to see issues if you are picky (and I am), even considering the pulled down whites in this grade and the overlayed grain.

Wow, that's really cool! 

...but wasn't this topic about 8-bit 4:20 and 4:2:2 camera footage shot on physical cameras? I think we all understand that 8-bit can work very well for digital creations and is still the default delivery method, but the issue is whether it's an adequate acquisition bit depth for high dynamic range LOG-encoded footage. 

[johnnymossville]

The more powerful the tool, the more chance people will screw it up.  What it does is makes us have to sift through a lot of crap to find the gems.     I doubt Steve Jobs would want a LOG profile on an Apple Camera.    He'd come up with a half dozen "Insanely Brilliant" color profiles and want you to stick to them.  and 99% would be happy to follow.    

[cpc]

Wow, that's really cool! 
...but wasn't this topic about 8-bit 4:20 and 4:2:2 camera footage shot on physical cameras? I think we all understand that 8-bit can work very well for digital creations and is still the default delivery method, but the issue is whether it's an adequate acquisition bit depth for high dynamic range LOG-encoded footage. 

Thanks.

Actually, my whole point with this paragraph was that even uncompressed 8-bit 4:4:4 log has its limits and they aren't particularly high.  : )

[TheRenaissanceMan]

Thanks.

Actually, my whole point with this paragraph was that even uncompressed 8-bit 4:4:4 log has its limits and they aren't particularly high.  : )

Ah...must've gone over my head. You're too smart for me.

So you're saying that no matter what compression scheme we use, right around 12 stops of dynamic range is where 8-bit starts to show its weakness? 

[cpc]

Ah...must've gone over my head. You're too smart for me.

So you're saying that no matter what compression scheme we use, right around 12 stops of dynamic range is where 8-bit starts to show its weakness? 

That's what I've observed in this synthetic case but I'm not gonna claim that it is universal law.  

[animan]

THIS!

 

Shoot log with your wb set up correctly and you dont need to do any grading at all! - just use your curves/levels to get to the contrast level which leaves the highlights unclipped, themids natural and the shadows not crushed.  then boost saturation if required.  

you guys are confusing colour grading and colour correction, the video may not need correction but grading is an aesthetic choice

[richg101]

 

you guys are confusing colour grading and colour correction, the video may not need correction but grading is an aesthetic choice

I'm not confusing the two, I'm saying that if you;re shooting log on 8bit cameras it should be in order to maintain the greatest dynamic range on your camera files, for post adjustment of contrast, Not colour.  

Log or not,  if your WB is too far away from the end result you're in trouble.  If i want a teal grade I'd set wb to a lower kelvin number and increase the bias towards green/blue.  if i want a warm grade I'll go the opposite way.