Calling all Sony users - share your whitebalance settings!

[jase]

It was said here numerous times and everyone who nowadays uses a Sony cam coming in the flavour of A7* / RX10* / A6* knows the sooner or the latter that setting a custom whitebalance is crucial to getting good results. I would like to create some kind of compendium (maybe even a wiki?) to collect this valuable information that each of us might have found out individually. By doing so, we might find "the" white balance settings for different lighting scenarios

I will start with my experience with the RX100 IV:

Daylight: quite easy, i always use 5600k - I never have dialed in a higher number than that (is that a mistake?)

Lowlight / tungsten: this is getting a bit more trickier. On the RX100, the right spot is definitely the 3000k range. Sometimes I get good results with 3300k, sometimes it is much better with something like 3900k. I am aware that different types of artificial lights need different kelvin settings, yet I did not succeed in finding something general applicable with the RX100.

I yet have to mess with the finer whitebalance settings for tuning it in any of the 4 quadrants...

[BrorSvensson]

i use 5900k in daylight with one shift down to magenta

[jax_rox]

Most 'non-film' lights, and even cheaper 'film' lights don't generally have high CRI, so will have a colour cast - which means it's almost impossible to have a 'this settin for this' other than to know the relative white balances (5600 k daylight, 3200 k tungsten).

I sometimes like shooting daylight at 6500 k, but that can change depending on what I'm running.

[Oliver Daniel]

White balance on the A7 camera line doesn't work. No maters what you do, it's never right. Something in the image always has a compromise. 

That said, I setup a custom white balance for every scene with the A7SII as the presets don't work at all, and very scene is very different. Normally there is a shift down to magenta to take out the ghastly green cast on skin tones. In post, I find myself having to dial out a blueish purple out of the shadows. Just the way it is. 

[cantsin]

Here's another take on the issue: I don't think that anything is wrong with Sony's white balance per se. When shooting raw stills, I rarely have to override the camera's WB values in the raw developer software (Lightroom, DxO, Raw Therapee). Auto WB is quite good and reliable, even in mixed light situations. - The real problem seems to be the JPEG/MPEG engine, its abysmal color science and the dead/plastic colors it produces; which, of course, can't be bypassed when shooting video instead of stills.

[Liam]

I realized a while ago white balance that was way too blue cleaned up really well and easily in post, while way too warm fell apart immediately. And actually at the time, I had green dialed way up in my camera by accident, so at "correct" white balance it was also pretty ugly after trying to fix it. Not sure this translates to like every compressed video camera, but intentionally making it blue could maybe bypass and be better than a greenish or whatever tint in a Sony when trying to do it properly. Anyone else have experience with this phenomenon? Obviously not at all helpful for delivering right out of camera

[jgharding]

Yes, this is due to chroma subsampling.

in 420 footage the 0 is the red channel. It is encoded with (Im rusty here) one quarter the resolution of the luma channel and half that of the blue channel. Red is not prioritised.

my exact number details could be rusty, but this is what's going on. The edges of red out of focus lights will show you just how fuzzy things can get. 

the more red your balance, the lower the resolution of your image.

when I used to use Canon 600D I used Visioncolor, which is a profile that enhanced blue, and used to balance cool. I also used a hack that upped the bitrate a lot. If you shoot very warm with and 8bit 420 codec then yes things are more likely to fall apart if pushed around.

so if in doubt cool/green it a bit would be general advice.

HOWEVER I've got rid of all my Sonys because I hated the colour, and the endless grading just to fix problems, so perhaps trying to go cool and green will just ruin your footage: you can but try!

[kidzrevil]

have you ever encountered the blue cipping issue when setting your white balance ?

[jax_rox]

On 07/05/2016 at 3:58 PM, jgharding said:

Yes, this is due to chroma subsampling.

in 420 footage the 0 is the red channel. It is encoded with (Im rusty here) one quarter the resolution of the luma channel and half that of the blue channel. Red is not prioritised.

Actually, in 4:2:0, both Cb (B-Y) and Cr (R-Y) are captured equally, but on alternating lines. 

I would posit that the reason things can fall apart on any digital sensor when shooting in warm white balances is because most sensors are weakest in the blue channel. Digital sensors are most sensitive to green - hence the proliferation of green screen. This mimics the human eye, which is also most sensitive to green.

Sensors are least sensitive to blue, so when you're shooting under Tungsten light with a Tungsten white balance, the blue channel gets very noisy, especially when compared with shooting daylight balance, where the light is mostly blue, and therefore the blue channel gets the 'saturation' it needs to stay 'quiet'.

Therefore it makes sense that even shooting daylight balance (considering digital sensors are 'biased' towards daylight) to saturate the sensor with more blue, will clean up better as the blue channel will remain quieter/less noisy as compared to shooting under tungsten balance.

I can't imagine chroma subsampling post-sensor does wonders for the blue channel either, so you've got an already noisy blue channel having its resolution reduced and compressed - which is why it's much less of an issue on a raw, or 444 subsampled camera.

[jgharding]

18 hours ago, jax_rox said:

Actually, in 4:2:0, both Cb (B-Y) and Cr (R-Y) are captured equally, but on alternating lines. 

I would posit that the reason things can fall apart on any digital sensor when shooting in warm white balances is because most sensors are weakest in the blue channel. Digital sensors are most sensitive to green - hence the proliferation of green screen. This mimics the human eye, which is also most sensitive to green.

Sensors are least sensitive to blue, so when you're shooting under Tungsten light with a Tungsten white balance, the blue channel gets very noisy, especially when compared with shooting daylight balance, where the light is mostly blue, and therefore the blue channel gets the 'saturation' it needs to stay 'quiet'.

Therefore it makes sense that even shooting daylight balance (considering digital sensors are 'biased' towards daylight) to saturate the sensor with more blue, will clean up better as the blue channel will remain quieter/less noisy as compared to shooting under tungsten balance.

I can't imagine chroma subsampling post-sensor does wonders for the blue channel either, so you've got an already noisy blue channel having its resolution reduced and compressed - which is why it's much less of an issue on a raw, or 444 subsampled camera.

Thanks for clearing that up, I knew i was rusty.

Still, it remains a fact that red comes out far more blocky than blue, so I'm guessing all of it as at play at once: chroma subsampling, blue sensitivity, bitrate priority and so on...

Even using Redcode raw with a red camera things are nicer at daylight balance. I think most sensors and processing pipelines are just optimised for it.

Red's web site has a good description of what';s going on which I should have read before posting:

http://www.red.com/learn/red-101/video-chroma-subsampling