Sony vs Canon colour science - does this explain the difference?

[Mattias Burling]

I don't know. People still use PCs, despite them being less idiot-proof than Macs.

Maybe. All I know is that if I asign a guy at work to do a video and he asks to use an A7sii instead of a C100 but needs 3h extra in post. Well, I think we all know what I will say.

And time isnt the whole issue either. If he gets sick and someone else needs to finish it with him being the only one thats familiar with S-log.

[Michal Gajdoš]

The color is not a physical property of any object, it is only subjective to our mind. What we are used to. There is zero science to your claims that Canon has "better" colors, there is no "better" color. A certain color you might find pleasing, is disgusting to others. Just as stated here:

Funny, when I had the 7D I always had to tweak the colors. Same happened with 5D files. I find Canon's colors so unnatural, in fact I find them too saturated and too contrasty.

I enjoy Sony's color much better.

On the other there are also articles saying the opposite, like this one : http://web.mit.edu/abyrne/www/ColorRealism.html, but as they say it works only to a certain degree. 

However it is important to note the difficulty of creating a nice look with the sony's S Log compared to baked colors of Canon straight out of camera. 

Personally, I don't care if canon has almighty color, if those grandpas cannot give me simple stuff like focus peeking or articulated screen (what is the almighty color shot to me if it is not in focus and my back is crying in pain) to more sophisticated stuff such as slow mo, better DR, better sensor etc. etc. etc. If only we could combine best of both worlds

 

[austinchimp]

Funny, when I had the 7D I always had to tweak the colors. Same happened with 5D files. I find Canon's colors so unnatural, in fact I find them too saturated and too contrasty.

I enjoy Sony's color much better.

true, and after working extensively with the original A7s, Red raw and now the Canon 1DC I can confirm that they all need skilful tweaking straight out of the camera. 

The difference for me has been that with Canon and Red files the first hour or so of grading is correcting the colour (removing weird casts, exposure, getting skin in the right area, making it look natural) and then I can go onto working the general look and feel. Where as with Sony - in any colour profile you choose - the first stage of 'fixing' the image so that it looks natural never really ends. I find it very hard to ever reach a point of having a natural look I'm satisfied with, and then any additional change I make to the image tends to knock all my previous work out of whack. With the Canon files, once I've got it to a natural, neutral point, further creative grading seems to go much smoother and doesn't stray out of a natural getting image as much.

What I'm basically saying is that in my workflow Sony A7s footage requires me to constantly wrestle with the image to stop people looking like aliens or zombies, whereas in Canon footage I can push and pull the image much further while people tend to remain looking like people.

As a working pro those hours saved and reduced stress is hard to put a value on.

To answer the question in the article: No, not really. Canon's color space is just more idiot-proof than Sony's. 

as an idiot, I greatly appreciate this!

[Sekhar]

IMO a lot of the color issues we see with Sony footage is due to indiscriminate and improper use of log. Folks seem to use log like it's a badge of honor, for every single situation and as a substitute for proper lighting, and as we know log and 8 bit is not a happy combination. A good way to assess and compare say an a7sII color would be to shoot without log, and I think it will also behave well in post shot that way.

[ajay]

Is it not true that Canon's Bayer pattern has more red filters compared to Sony's, which has more green? This is what I have understood and the reason why Canon's sensors have always been better with flesh tones.

 

AJ

[johnnyd]

IMO a lot of the color issues we see with Sony footage is due to indiscriminate and improper use of log. Folks seem to use log like it's a badge of honor, for every single situation and as a substitute for proper lighting, and as we know log and 8 bit is not a happy combination. A good way to assess and compare say an a7sII color would be to shoot without log, and I think it will also behave well in post shot that way.

I agree - I have two sony cameras (fs700 and a7s), both capable of slog2, but I mostly use the cinegammas when using internal codecs. I think the cinegammas are great gammas and easier to work with in post, especially when shooting with weaker codecs.

[LaneMc]

Andrew - which Herdwhite track is in the Nx1 vs 1DC video?  I combed their soundcloud and cannot find it.  Phenomenal track.

[TheRenaissanceMan]

Maybe. All I know is that if I asign a guy at work to do a video and he asks to use an A7sii instead of a C100 but needs 3h extra in post. Well, I think we all know what I will say.

And time isnt the whole issue either. If he gets sick and someone else needs to finish it with him being the only one thats familiar with S-log.

It's harder to make a great steak than a great burger, but I know which I'd rather eat. 

When I see Canon footage, it always looks like good Canon footage. Somewhat with Nikon, too. But with other manufacturers, it can look tons of different ways. So it's easy to get good results with Canon, and difficult (but possible) to get great results from manufacturers with wider-gamut, higher-bit depth footage. 

[kidzrevil]

Wait, I thought MatthewP pretty much debunked the "1Dc sees better color than NX1" conclusion in the shootout thread with his grade comparisons (see Page 3)? That would discredit the premise of this new article about technical vs. artistic view of color and 1Dc being better than NX1 that way. I.e., if you have good color info (technically speaking), you should always be able to tweak the image to fit your artistic preferences.

I agree with the tweaking part but with 8 bit footage I personally hate wrestling footage to get what looks good to my eye. You end up fixing one thing and breaking another...this is the primary reason I gave up my a7s for a Samsung NX1. Straight out the camera the colors are extremely good. Now I don't need LUT's to get something that doesn't look vomit green like Sony's do in cine4

[guru]

Is it possible to make a picture profile for A7x to look like Canon ? Or we have to wait for 5DmkIV 4k ML RAW

[Mattias Burling]

It's harder to make a great steak than a great burger, but I know which I'd rather eat. 

That doesnt matter to me. If where not done in time there is no dinner. Simple as that.

But in this example we are in luck since the A7sii is no stake imo. At best its the small sallad but that for some reason takes forever

[TheRenaissanceMan]

That doesnt matter to me. If where not done in time there is no dinner. Simple as that.

But in this example we are in luck since the A7sii is no stake imo. At best its the small sallad but that for some reason takes forever

I'm talking about color science in general, not the A7S II specifically. 

My work is mainly creative features, docs, shorts, and music videos, where I have don't have lightning-fast deadlines to meet. In those conditions, I've gotten much nicer results from beefier color spaces like Blackmagic's than Canon's very saccharine, Crayola look. Even with weddings, you generally have a couple months to deliver, and between Filmconvert, Speedlooks, and Magic Bullet, finding a decent look is within anyone's reach. If not, I doubt the camera is the problem. 

[Mattias Burling]

I'm talking about color science in general, not the A7S II specifically. 
My work is mainly creative features, docs, shorts, and music videos, where I have don't have lightning-fast deadlines to meet. In those conditions, I've gotten much nicer results from beefier color spaces like Blackmagic's than Canon's very saccharine, Crayola look. Even with weddings, you generally have a couple months to deliver, and between Filmconvert, Speedlooks, and Magic Bullet, finding a decent look is within anyone's reach. If not, I doubt the camera is the problem. 

Same here. Almost. I have always gotten better results with Raw, BMD prores or Canon C-Log.

Also older Panasonic camcorders.

(never had a mont to finish a doc unless its personal work. 1 week tops if its for broadcast. And as always thats the point. We all have different needs. Even among wedding shooters I bet there are those with much tighter deadlines.)

[kidzrevil]

I'm talking about color science in general, not the A7S II specifically. 
My work is mainly creative features, docs, shorts, and music videos, where I have don't have lightning-fast deadlines to meet. In those conditions, I've gotten much nicer results from beefier color spaces like Blackmagic's than Canon's very saccharine, Crayola look. Even with weddings, you generally have a couple months to deliver, and between Filmconvert, Speedlooks, and Magic Bullet, finding a decent look is within anyone's reach. If not, I doubt the camera is the problem. 

agreed. I've never came across a camera that was impossible to grade especially with all the options like filmconvert etc. people are just real fussy about cameras these days and there work ain't saying much lol....divas

[cantsin]

Back to the original subject of the thread: Andrew has a point that Canon cameras produce more pleasing/aesthetically appealing color than Sony (and Panasonic and Samsung, for that matter). But I don't believe that this has to do anything with sensor tech respectively color gamuts. After all, Nikon produces pleasing color, too, but uses Sony sensors.

We shouldn't forget that 8bit video - i.e. the signal recorded by Sony A7 cameras, most Canon C-series cameras, Samsung NX1 and Panasonic GH4 internally - represents only a fraction of the sensor's color information. In the case of 14bit sensor signals, 8bit video only contains 1/64th or 1.5% of the original color information, in case of 12bit sensor signals, only 1/16th or 6.2%. (You might argue that debayering isn't factored in here for the 12/14 bit signal, but on the other hand, 4:2:0 chroma subsampling isn't factored in for the 8bit signal either.) So it's all about which choices the jpeg/mpeg engine of the camera makes: which 98.5%/93.8% of colors it will throw away and which it will keep.

Cameras by consumer electronic manufacturers seem to be biased towards keeping a lot of green channel information because it will result in images that the untrained eye will perceive as sharper/more detailed. (The human eye can see greens better than other colors, a product of evolution and the age where hunters and gatherers need to spot prey or enemies in the woods.) It's probably the color science equivalent of edge sharpening, chroma oversaturation and increased micro contrast that is commonly used in consumer/amateur cameras to make images "pop" and appeal to the average consumers.

We shouldn't forget that Sony, Panasonic and (still) Samsung produce affordable consumer 4k cameras to boost sales of their 4k flatscreen TVs. I wouldn't be surprised if their color sciences is optimized for those TVs rather than for professional or cinematic/photographic video production.

[kidzrevil]

 

We shouldn't forget that Sony, Panasonic and (still) Samsung produce affordable consumer 4k cameras to boost sales of their 4k flatscreen TVs. I wouldn't be surprised if their color sciences is optimized for those TVs rather than for professional or cinematic/photographic video production.

good point ! That would make sense as I have seen in an interview a samsung rep saying that you can play the nx1 h.265 files on their tv's. Could be why they chose h.265 & their vivid color science

[3D-Kraft.com]

I suggest distinguishing more between the gamma profile and the color gamut which can be choosen seperately in the PP-settings of Sony A7S / A7RII / A7SII. My personal recommendation is to use the "Movie" color setting together with cine gamma 2 oder cine gamma 4. Red is not purple anymore and green isn't dominating anymore.

Even though S-Gamma may preserve some more of the sensor's dynamic range, in a 8 bit corset you get banding too quickly after stretching in post processing. When recording to memory card inside the camera, it is even more compressed (using only values between 16 and 235). Grading from the HDMI output (full using of the values from 0-255 and 4:2:2 subsampling e.g. by recording with Atomos Shogun) may help a little bit.

[theSUBVERSIVE]

From a presentation I watched from Alister Chapman I got to understand - not fully - a few things about colour science, logs, sensor, etc. with my own take on it since he wasn't talking about this specifically. He basically explained the differences between broadcast standard, how to use the extra stops you get in these cameras and how log works.

I think that the sensor itself is less important to the discussion because that's just a workspace, two manufacturers or more can use the same sensor and have very different results in the end. Colour science does matter, I can't agree with those that say like "it doesn't, you just have to grade better", that makes zero sense.

Actually in the presentation Chapman said something similar to Andrew's theory, Sony is an engineer company and they make it so you can get the most color gamut possible, therefore, they get more green as well, which not necessarity means it will make better images. But that's just one step, the other one is how you use it. He also talked about how the green end up shifting to fit the workable color gamut and he even explained the difference between S-log 2 and S-log 3 in more details and why S-log 2 is more complicated to grade or why you record S-log exposing to the right, about 60% zebra.

When you create a log or a picture profile, cine gamma, etc. that's when you start to use the information you have from the sensor and in this case, simply showing all the color gamut is not really the best option and that's when colour science comes. I don't know if I would use the word artistic here, but more like perception, so making the data you get into something that looks more natural to our perception does make a lot of difference, even more because during color grading you will be doing something similar.

But let's say that after the data you get, you make something unbalanced because you wanted to show all the color gamut possible, then it will be much harder to get a more natural look since you will be getting more green and depending on several factors like color space, compression, etc. you might not be able to get what you would had you created a more balanced log with a better colour science.

So basically with a better colour science you make a better use of the data you can get from the sensor, it doesn't necessarily mean showing more colors or even scientifically accurate colors. Of course that also means it's subjective as well and there is no a right and worng exactly, but overall, when the difference is notable you can say that one colour science is indeed better than the other and when it's closer, then it becomes more of a personal preference.

That's probably also why you have a big difference between measurements of stops, color gamut, etc. and practical, real life aplication because measurements don't really tell how you use it and that's also why those DxO ratings don't mean one camera is better than the other.

-

As for why the S-log 2 is harder to grade, in simpler words it's something like it's more curved than S-log 3. When you grade usually you will do it based on that middle and the S-log 3 from that part and above is more like a line and less like a curve, so when you grade it up or down, it's more uniform. With the S-log 2 you have to basically grade the shadows, middle and highlights separedly because if you do based on any of these 3 parts, the value you add or substract will be different in the other section, so it's not uniform.

That also matches the info I got watching a Sony guys talking about S-log 2 S-log 3. He said that S-log 3 is easier to grade and that S-log 2 is better for highlights while S-log 3 is better for shadows and if you look at the curves you can understand why, S-log 2 has more values for highlights while the S-log 3 has more values for shadows and mid tones. I think that "better" is not really the word, it's more like more nuances. Chapman also talked that in 8-bit codecs if you really want more information, you better off with S-log 2, if you want something easier to grade you can go to the others gamma curves. He explained why in low light you DON'T USE S-log at all, which totally makes sense, unless you are at night but filming a lot of light, recording in S-log will only make it noisier because although you will be getting more data overall, in the areas that matter it's the opposite, making is much noisier and that's also why S-log is no good for chromakey as well.

[Ed_David]

I really liked this post, Andrew.  

I think at a certain point, what I like in this film I just saw shot on super 16mm kodak vision3 (I think) - Carol, are the rich reds and secondary color of red in the face.

I think slog will get you there with good grading - scientifically it has more color information than film according to the chart.

 

I think the c300 c log does make really lovely skintones and gets you there easier, but the s gamut especially on the f35 can be incredible - you just have to get there.

Secondary color correction, bringing out the saturation on the reds - on the lips, also doing an offset for secondary skintones, it's all interesting stuff to get where you need to.

color grading takes a lot of work but you can get there with tons of practice and you can make any camera look really nice.

Some definitely can be easier to use to get there, but it's also easy to mess up even alexa footage when graded poorly.

[SRV1981]

Very interesting dialogue, glad i found this!