F3: 444 vs. 422 Anyone had experience comparing the two?

Shoots56

Carbonite Member
I'm trying to decide what recorder would be best going to an F3. S-log, 10 bit out to a 422 system, or is it worth the trouble and expense to go 444 dual link? (this is assuming the F3 has the paid Slog update)

I know it's more of an issue when doing FX or keying, but what I'm wondering is, is there any "mojo" in the 444 files that aren't there in single sdi 422 setup? Can you really tell any difference when moderately grading either?

This isn't really a technical question so its harder to ask I guess.
 
I would get a 444 recorder more for the 1080 50p/60p than 444:) But to answer you're question you won't see a big difference between 444 and 422. The biggest difference that you'll see is the file size:)
 
Agreed with Sanjin. The sensor is not capable of delivering true 444 output. I don't know the math in terms of sensor size, or more to the point the number of "photo sites", to deliver real 444 but it is significantly larger than the F3's sensor. The F35 on the other hand is a true 444 camera. The output from this camera is STUNNING for a variety of reasons. The 444 delivery is part of it. Like Sanjin said the best function of a dual link recorder is being able to capture 1080p60 footage (or is it 1080i59.97) from an F3.
 
Last edited:
There's been lots of discussion here on dvx on F3 420 vs 422 vs 444 and even YCRBR vs RGB and I don't think there is huge difference between 422 and 444 from what I've seen. But over 420 you will have more room in grading but not as much as everyone will claim. 8-bit vs 10-bit yet another debate.

Regardless 10-bit 422 will not let you down.
 
Even the Alexa can't do true 4:4:4 despite the Prores files are stored in 4:4:4. I would also consider getting a recorder more for the 1080p60 than 4:4:4.
 
Sidenote: The only sensors which can do 100% "real" 444 are RGB stripe sensors such as the ones found on the F35, F9000, and Genesis.
 
Last edited:
Is the 1080-60p function really that much better than the 720-60P function that can go over single SDI? Or, is it worth the extra hassle when the 720p, when simply uprezzed to 1080, seems to look pretty darn good (better than what an AF100 in "1080" 60p or DSLR can do)?

Then what is the real benefit of that expensive paid Slog option that many people agreed to buy? The ability to look at a LUT? The new recorders can provide that and still record a log file with the free 422 Slog update.
 
Well 1080 is 1080 obviously, but I did quite a bit of 720p 60p shooting to internal cards on the F3 and blew up to 1080p in post and they matched quite well. It is a tad softer but not always that noticeable unless you're really looking for the differences. To do 1080p 60p externally you will need a dual link recorder incase you weren't aware of that.

The expensive paid SLOG option was actually two things: slog, and RGB output capability over SDI (versus YCbCr video signal). As of firmware version 1.4 SLOG became "free" via gamma curve accessible through the picture profile in camera. That meant that the only difference for the paid upgrade was really getting the RGB over 10-bit SDI. pix240i was updated earlier this year to be able to record the RGB signal which is a pretty cost effective solution versus some other recorders on the market.
 
Last edited:
What's the advantage of shooting RGB vs. YCbCr?
My feeling ( after having bought the S-Log upgrade a few months before it was free) was that it was a waste of money. The LUT's suck anyway IMHO.
 
In the chain of events from processing the video image from the sensor to digital file, RGB is slightly less "processed" over YCbCr but hard to say how much better it really is.

The F5/F55 are also offering 3 flavours of the SR Codec: 10-bit 422 YCbCr, 10-bit 444 YCbCr, and 10-bit RGB. Perhaps its time to do some pixel-peeping tests to see if any difference can be found there.
 
Better to have 75% of your red and blue information than just 50%.

Shooting 4:4:4 instead of 4:2:2 is simply getting you more colour information. 4:2:2 is great, 4:4:4 is simply better. It's like raw in some ways - more than you need for many things, but better, if you need it.
 
To be fair, I just graded a 7 minute long video that was shot to SxS in Slog and it held up relatively well. The F3 has a clean image so this really helps in keeping the image from breaking down. So 10bit 422 is all you really need in 99% of situations if you ask me. If you don't need 50/60p then really think again about buying a 3G recorder. Regarding 720 50p uprezed to 1080p...well the F3 has what? about 850 lpph? So if squeezing to 720 then you lose only about 100 lines of rez. So this is probably why there isn't such a big difference between 1080p and 720p. Also, the F3 apparently does a good job in downscaling the image. But still, there is probably a 15-20% loss in sharpness.
 
Thanks Star.

And I know the difference between 422-444 10-8 bit, etc. and their benefits on paper and with other systems. This was about the F3 specifically. No matter, its a hell of a system for the money and it looks like a 422 10bit recorder will handle all that it can offer, or at least 99% of it.
 
Better to have 75% of your red and blue information than just 50%.
... It's like raw in some ways - more than you need for many things, but better, if you need it.

Brilliant words, which are absolutely true. If you can do it, why not? File-size, blah, blah...for the price of storage these days, you should be able to build the cost of storage right into your job.
 
Regarding pushing the image in post (grading), it's not really that much about chroma resolution then it is about the compression and bit depth.
 
I know how to beat a dead horse, watch this trailer shot in slog to internal cards on the F3 which is 8-bit 420. I've also done a ton of work when I owned my F3 using the cinegammas in 8-bit 420 and no one was the wiser. I will admit however that now I'm working with 10-bit 422, native on the F55 and feel better knowing there is extra bit-depth, bit-rate and color space to hold up my images in post.


http://www.dvxuser.com/V6/showthread.php?313651-Feature-trailer&p=1986363261#post1986363261

A couple of months ago I shot some key scenes for a director and his feature which he’ll try to raise funds to make. It was a few hectic days to get it all done and we had a small team and few actors. It’s all in Swedish so to explain the story it’s about a guy struggling with his relation to love, life and work. Title translated from Swedish – “A few words about love”.


I used my Sony PMW-F3 shooting to the internal SxS cards and using S-log gamma. Audio was recorded mostly externally but in some scenes we used the audio on the F3. For the tracking shots I used my Cinevate Atlas 30 slider and lenses were Canon L lenses. Mostly the Canon 50/1.2L. Grading in DaVinci by a colorist.
 
So I have 2 questions.

1) I know RAW is very flexible in terms of changing parameters after the fact, but what about just image quality and gradability of 4:2:2 10-bit compare to RAW. Is the jump in quality that huge?
 
A lot of that would depend on the gamma curve used to acquire the footage in 422 10-bit.

By the way, you can't really just look at the color space and bit-depth specifications alone, you have to consider the gamma curve being used as well.

So I'd say its like this in terms of "quality":

(Now before anyone jumps down my throat everyone's definition of quality is different. For example I would tell anyone RAW is the highest quality, yet there are people shooting with custom profiles in-camera using hypergammas who's footage would blow you away. Doug Jensen has some amazing footage shot with nothing more than a customized picture profile). Therefore "quality" is extremely objectionable and questionable.).

- RAW: King of quality, more like shooting film and forgiving in many aspects. Exceptional highlight recovery as long as you don't clip. RAW uses linear gamma which gives you exponentially high data bits in the highlight areas to work with. RAW is probably the easiest way to get a more FILMIC look from digital. Most cameras shooting RAW video do it in 12 or 16-bit.

- Log: slog/slog2 (Sony), clog (Canon), logC (ARRI) all do the same thing and apply a very serious gamma curve to the image before being recorded. The goal here is to allocate the data bits per stop in such a way that you get good information all around the low, mid and high tones of the image. Pushing shadows in log can sometimes be better than RAW because there is more data there to work with. Pushing shadows in general on log or RAW will increase noise however. Because of log design, it results in very smooth looking highlights and therefore is also very good at getting a more filmic and less "video" looking images. I read somewhere in a discussion somewhere once that 10-bit log was about equal and on par with 12-bit linear gamma.

- Hyper or Cine gammas: these can let you shoot up to 109% illegal signal giving you an extra free stop at the high-end and these gammas also move the mid-tones around so you can get brighter mid-tones in darker situations if needed without having to push it in post. When I worked with the F3 I used cinegammas for most of my shoots. By choosing the right gamma and being selective at how your expose your image with it, it can also produce a more "filmic" look after grading but needs a meticulous approach to lighting/exposure.
 
Last edited:
Back
Top