FS7: FS7 RAW woes & data distribution between: 12-bit RAW, 16-bit RAW, & 10-bit S-Log

starcentral

starcentral

Veteran
FS7 RAW woes & data distribution between: 12-bit RAW, 16-bit RAW, & 10-bit S-Log

In the RED Raven and FS7 thread something interesting has come up that I feel deserves its own thread and discussion.

There was some discussion on excessive banding occurring with the FS7 using the 12-bit RAW output to the Odyssey recorder. An odd behaviour when the same issue did not seem to exist on the FS700 with 12-bit RAW output.

In the thread, I had written that with the way linear RAW works and assuming Sony's 12-bit RAW is a pure linear gamma curve there would actually not be enough data bits left at the lower end of the sensor's overall dynamic range of 14 stops. Sony would have to apply some rounding or bending of the linear gamma to help get some data bits covering the lower end of the tonal range.

This table of mine helps to illustrate what I'm talking about. As everyone knows with 12-bit you have 4096 total levels of grey. The highest stop takes 2048 of these levels (so half what is available overall) and then the 2nd highest stop takes 1024 levels, the third stop takes 512 levels, and so on and so on. By the time you get to the bottom you do not have any bits left to even cover the entire dynamic range of the sensor (14 stops). You'll actually notice the table only covers the first 12 stops and doesn't even get to the 13th and 14th stops as there would be no data available for them.

Screen_Shot_2015-11-12_at_9.05.48_AM_zpskauljbdc.png



16-bit RAW from the F5/F55 on the other hand seems to cover the camera dynamic range somewhat better because you're starting with 65,536 levels at the highest stop that by the time you get down to the lowest stop you still have some data levels available for it.

To put everything in perspective when it comes to data distribution between 16-bit RAW, 12-bit RAW, and 10-bit S-Log3 I've put together a table to show some differences. With a scene limited to 7 stops for example, notice the difference in data levels between ETTR and exposing for 18% middle grey for both 12/16 bit RAW and for S-Log. If shooting RAW it's easy to see that ETTR gives you more bits whereas with S-Log whether you use ETTR or expose for middle grey you practically end up with the same amount of data levels ie. 479 vs 539 levels.

Screen_Shot_2015-11-12_at_10.50.43_PM_zpsw2k99tb.png
 
Last edited:
So, what's the solution?

Fractional values? Mitch just posted on the NEX thread that CD do that by taking the 8-bit and converting it into a 10-bit wrapper via a proprietary algorithm. But what happens if it's a 10-bit into a 12-bit?
 
It also appears that the FS7 RAW may well be 12 bits but the lowest bit is always 0 and depending on mode, the lowest few bits can be 0. This was the case at the beginning but i don't have an FS7 and was just relying on files sent to me at the time to look into.

The FS700 RAW fairs a bit better but usually the LSB is 0. There's a long article on my blog about inside DNG which goes into this in more detail, specifically for the FS700 RAW though.

This means that 12 bit is actually 11 bit or even 10 bit in the field. Again i can only say for sure what i've seen, i can't say it applies to current firmware or all cameras.

Another difference is that FS7 RAW has white balance baked into it, whereas FS700 doesn't. This actually matters because when you take pre white balanced data into a higher bit depth and do the colour matrixing there then the resulting numbers have higher precision than if you are dealing with baked white balance in 10/11bit, there isn't that same mixing step.

Finally the curve off the FS700 isn't totally linear, there's a slight S shape which does help eek out the range a little. If the curve was strictly linear then you would never see more than 12 stops, and even in testing you can eek out another half, maybe another stop.

IMHO for all the good things about FS7, RAW output wasn't one of them. I wonder whether this is why the FS5 appears to have gone back to the same system as the FS700 RAW (single output). It will be interesting to see.

When you are able to delve into actual data coming off these cameras things get interesting, don't think for a second this is a Sony thing, all cameras do similar things.

cheers
Paul
 
Have you guys ever had 14 usable (!) stops out of your FS7?

I don't. I work with 12 usable stops.
-6 and -7 are for lowest shadows without Details that are usually crushed in post anyway.

At least with me.
 
I'm not sure of exactly how this works, or if it applies to the FS7 12 bit raw, but the linear 12 bit file from my D16 is written from a 16 bit first stage AD processor and unpacks as 16 bit values in every raw processor I've used. There are discrete pixel gray scale values in the lower ranges that are not possible with 12 bit linear AD. In any case 12 bit linear places middle gray at 512 with 3 stops of headroom.
With REC709, which is HD display values standard, you have an absolute maximum theoretical DR of 9.5 stops regardless of bit depth. Anything outside of that range on either end is hard clipped. REC709 video digital code values are not linear, they are gamma 1 log encoded, equal code value steps per stop of light value.
Other camera log curves are compression schemes to bring sensor data into display value ranges.
 
Last edited:
Just to add, since the Sony's use gain and pixel-binning for its HFR mode, the situation is probably going to get worse when trying to use the camera's max dynamic range and recording via 12-bit RAW. But it'd be real great if someone can do some tests using the Sony interface and AXS-R5 to see if the same problem exists.
 
Again that comes down to being able to advertise '240fps raw to an external recorder' without it actually being massively usable.

Here's one quick still.

240fps from FS7 Raw to 7Q Pro Res. LC709a LUT applied in Premiere and and I've pulled the blacks down a touch to try and clean it up

Asahi Noise test_zpsrdc26qp9.jpg


Not sure you can make it out all that well but most of the front of her green top is a pixelated mush. As I say I've actually brought that down from being exposed higher to try and protect it too.

2 x 4ft Kinos and an Arri L7c through a diffusion frame.

EDIT: Photobucket has raised the overall brightness by about a stop...
 
Last edited:
I definitely see the blocky's there on the sweater.

For a high key shoot like this the only solutions here might have been:

• shoot it in XAVC (10-bit)
• shoot with ProRes codec via 3G-SDI outputs to ext. recorder (10-bit)
• maybe try HDMI to ext. recorder (8-bit)
• for 12-bit RAW place any dark tones only 3 stops MAX below middle grey (thus clipping highlights and/or requiring you lower scene contrast ratio)
• shoot with a different camera


(Tell you something you don't know, I hear ya…. )
 
Last edited:
starcentral said:
I definitely see the blocky's there on the sweater.

For a high key shoot like this the only solutions here might have been:

• shoot it in XAVC (10-bit Had to be 240fps so nope :(
• shoot with ProRes codec via 3G-SDI outputs to ext. recorder (10-bit) Had to be 240fps so nope :(
• maybe try HDMI to ext. recorder (8-bit) Had to be 240fps so nope :(
• for 12-bit RAW place any dark tones only 3 stops MAX below middle grey (thus clipping highlights and/or requiring you lower scene contrast ratio) Borderline impossible with a shoot like this with so many people/skintones/clothes etc roating in and out
• shoot with a different camera Sadly this is the only solution :(


(Tell you something you don't know, I hear ya…. )
Click to expand...

Replies above.

Just to add, I have shot tons of 150fps internal XVAC-I stuff and while it isn't as sexy as the 4k 60fps stuff it is plenty usable for online. For example, this shot at 32secs - https://youtu.be/O3ygO6fZaCo?t=32
 
in the mid 90s I was shooting on a 22mp Hassy. (16bit 5k linear raw with very wonky software)

One absolutely had to control the lighting ratios and ettr

Im at a loss about this sony raw being bad, Im just thinking that with the correct exposure tools and a full understanding of the DR limitations it should look good

To some extent you should get an increase in tonal accuracy as a trade off to DR reduction.

But not having actual experience.. I dont know.

I do think some formal 'braket tests' need to occur before full judgement can be made however.

S
 
People talk about ETTR as if it were one thing. And that might be true if the brightest thing in your scene is a 90% white. But what if there are multiple light sources in your scene, and in some cases you want detail (a lampshade) and in others you don’t. Or you have bright clouds in your scene, and you want them to be somewhat visible. Now ETTR can mean almost anything.
 
For me personally, I now know what breaks the camera/image and will no longer try to force a square peg into a round hole and will just enjoy using the FS7 for the things it excels at, and rent in something when the job calls for it.
 
Ben, maybe consider trading in the FS7 for an F5? Then you have 16-bit's to cover 14 stops by renting an R5, you also get RGB 444 (HD only) which his clean as a whistle. It might also lessen the need for the Raven if you were considering that in addition to the FS7.

Let's hope Sony can somehow address the output of the 12-bit RAW from the FS7 and make it more like the FS700 - but perhaps that's not feasible, they don't care, or it's just plain impossible.

Who the heck knows.
 
Jim Ritscher said:
People talk about ETTR as if it were one thing. And that might be true if the brightest thing in your scene is a 90% white. But what if there are multiple light sources in your scene, and in some cases you want detail (a lampshade) and in others you don’t. Or you have bright clouds in your scene, and you want them to be somewhat visible. Now ETTR can mean almost anything.
Click to expand...

I see it as purely a 'science experiment' simply ETTR with linear creates a better s/n ratio.

If you want to protect your clouds, protect em, if your foreground is then too dark buy a light.

That is flippant I know and 'buying a light' (which could actually equate to a full grip truck for a large scene) doesnt fit into most modern productions... in which case it would seem that the camera is better used in a different mode.

My old hasselbad made wonderful images if you controlled the DR.. but I canned it as (at 25ISO) it most certainly didnt fit into 'modern productions'

S
 
Last edited:
Ha! Nope I don't intend to give any more money to Sony and like I say (possible A7r II aside), the FS7 is great for most jobs I do. Also I have less than zero faith in them ever even looking into something like you suggest. Just thinking about the F5 4k debacle gets my blood boiling.

I probably won't buy a Raven but might be in the market for the new Scarlet Weapon when that arrives if it gives me just that bit more for my money.
 
You must log in or register to reply here.
Back
Top