F3: Trouble with 10-bit

Finn Yarbrough

Well-known member
I am having a really hard time seeing the difference that 10-bit is supposed to make for chroma-keying.
EDIT: I don't actually mean 10-bit, I mean 4:2:2.

Here are two screen-grabs at 400% from my NLE (Premiere) using Ultra Key's default settings for both. Can anyone else tell the difference between the 10-bit ProRes 4:2:2, recorded via SDI to Blackmagic Shuttle, and the 8-bit internal recording? Obviously, I can refine the edges, but when I do, they still look the same to me.

Am I missing something?
 

Attachments

  • Screen Shot 2014-02-27 at 7.50.10 PM.jpg
    Screen Shot 2014-02-27 at 7.50.10 PM.jpg
    12.5 KB · Views: 1
  • Screen Shot 2014-02-27 at 7.50.23 PM.jpg
    Screen Shot 2014-02-27 at 7.50.23 PM.jpg
    12.1 KB · Views: 1
Last edited:
I am having a really hard time seeing the difference that 10-bit is supposed to make for chroma-keying.

Here are two screen-grabs at 400% from my NLE (Premiere) using Ultra Key's default settings for both. Can anyone else tell the difference between the 10-bit ProRes 4:2:2, recorded via SDI to Blackmagic Shuttle, and the 8-bit internal recording? Obviously, I can refine the edges, but when I do, they still look the same to me.

Am I missing something?

The difference in 10 bit vs 8 bit is usually not apparent in the untouched footage. (Though the difference in compression between xdcam and say prores from an external recorder can sometimes be seen). Typically, the difference in 10 vs 8 bit is very obvious in gradients when heavy contrast is applied in post grading. So in your examples, add a bunch of contrast to both shots (pre-keyed) and you will likely see banding in the 8 bit and smooth gradients in the 10.

8 vs 10 doesn't really affect chroma, but color space (4:2:0 vs 4:2:2) does. You will notice a difference here if you have a tough key to perform.
 
I'd take the internal XDCAM recording from the F3 and save myself time, money, and hard drive space. Only for S-Log would I go to an external recorder, but this can be shot internal too. I really like the internal codec.

 
This seems to be a rather short-sighted, conclusive statement without much merit. Do you have any idea why one might shoot with SLOG or RAW?
What if shooting to ProRes or DNxHD is desired in edit? How 'bout you want more audio channels? How about shooting to faster/larger/cheaper media?
Greenscreen? Extensive Grading? Tracking/stabilizing, etc.

Yes, internal recording/codecs are great for some, but the rest of the world there's plenty of reasons for external recorders with higher bit-depth and color-space.

Sam do you have a Samurai? If you do, why not do a test and show your findings? Unless your doing serious grading, I believe the difference is negligible, and to be honest, external recorders are pointless, expensive, and a waste of time.

Here is another test http://mikemarriage.lunarfilm.co.uk/Blog/files/XDCAM 35Mb/s Vs ProRes Test.html
 
Hey folks, thanks for weighing in. Point well taken about 10-bit vs 4:2:2, my title was misleading. My problem, however, remains.

The two screenshots I posted ARE greenscreen, that's what I'm talking about, not bit depth as my title suggests. One grab is from the internal coded, the other is from ProRes HQ recorded via SDI on a Shuttle.
I cannot, for the life of me, spot any difference in the key between the two methods, and I am hoping that it's just because I'm doing something wrong. Has the necessity of 4:2:2 for chroma keying been oversold?
 
No I agree with you, S-Log needs to be shot on an external recorder, my point is that for most projects, the internal codec is very good. I have been looking at the Pix 240i, the Pro Res 4444 might be worth the hassle of going external. I assume that at 330mbs a second, this must be close to RAW?

I actually want someone to talk me in to buying an external recorder, because at the moment, I still feel that they are unnecessary for most projects. Which recorder do you use? What is the main benefit of filming Pro Res for you? Do you still record to SxS while going external?

The Pix 240i cost £3,060.00 here in the U.K, would it not be best to put this cash towards another camera?

Sam, the on-board S-Log looks great.
 
Try shooting something bright red with thin curved edges on a dark background. It's usually something like this where I can see the extra color resolution of 4:2:2 over 4:2:0.
 
Surely agree that recorders are a hassle and not sure why the camera manufacturers can't get it right. It appears that Arri has done so, and perhaps Red…sorta.
I have about every recorder known to man, but like the Pix240 better than most of the others. (other than size/weight). Recording 12-bit 4444 is nice too!
 
I'm no expert on green screen but when I gave a test to my client of internal codec on an F3 compared to 422 on Samurai , she decided to spend an extra $150 a day for the Pro Res - and she was very cost conscious. Somewhere on the web I saw some tests of green screen with people smoking and throwing their hair around and there was quite a noticeable improvement on the 422. The guy who's head you pictured here doesn't have frizzy hair and isn't in motion so its not a tough test of the codec.
 
Seriously...

I thought that I had finally found the difference, but it turns out that was just because I accidentally applied the grade before the key in one of my examples.
Here are the two, 4:2:0 and 4:2:2 chroma keys side-by-side, with some motion this time.

IQKr8R3.jpg
 
Last edited:
Try it again with someone who has very frizzy hair.
Also what camera are you using? The recording is only part of the equation. For example a Panasonic HVX200 recording 422 100Mb pulls poorer green screens than an EX-1 at 420 35Mb because the initial picture has much lower resolution.

For most imagery on an Ex-1 or an F3 though I agree that 422 10 bit is a very minor improvement and I almost always shoot to the internal cards unless its green screen or SLOG. As long as your shot doesn't stress the codec you can even have good results shooting those. Look some guys claim they get great keys with a 5D.
 
Last edited:
If you shoot a variety of material, especially shots with motion, you'll see a real difference between XDCAM EX and 10 bit 4:2:2. The XDCAM codec has done some great work at reducing the artifacts that come along with the large compression, but this is at the expense of some sharpness and you'll definitely see motion artifacts with something more dynamic than your guy at the desk. I agree some external recorders can be a pain - but remember that many are extremely useful as monitors. I have been pleasantly surprised at the quality and utility of the small Blade, and it runs most of the day on the two small batteries, so only one cable needed. Another exercise that might show the differences more is to go to DnX instead of ProRes, whose 4:2:2 codec is somewhat nosier than DnX. I've bought an F55 now, so I don't use the Blade for recording, but the monitoring tools and bright picture can be a great help, and the 4" screen keeps the camera package small. And as mentioned above several times, the color space difference can be a big factor in pulling good keys.
 
the key thing with green screen is the lighting , regardless of what cam or codec u use, if u cant light for green screen, nothing will help, not cam , not codec....
 
the key thing with green screen is the lighting , regardless of what cam or codec u use, if u cant light for green screen, nothing will help, not cam , not codec....

Guys, come on. Look at the green-screen example I posted. The key is not too bad, right? That's not the point. The point is that its IDENTICAL. One is 8-bit 4:2:0, the other is 10-bit 4:2:2.

Can nobody point me to any concrete, comparative examples that really show the real-world difference 4:2:2 makes for keying? I understand the theory, but based on my results I'm not convinced.

Look at the single hair sticking out of the top of his head, for heaven's sake.
 
Back
Top