F3: Any real benefit to 444 output?

There's absolutely a benefit to shooting 4:4:4 instead of 4:2:2!

Even though the F3's sensor isn't actually high-resolution enough to deliver true 4:4:4 colour information for green, red and blue (i.e. 1080p of red, 1080p of green and 1080p of blue) - it does capture significantly more colour information in it's 4:4:4 mode than it does in 4:2:2.

The only clear, empirical evidence of this published to date is from MXR Productions' blog post here: http://blog.mxr.at/422-vs-444/

LjRsc6E.jpg

3mbYUs8.png


Have a look at the difference between the colour keys pulled from the 4:2:2 and 4:4:4 versions of the same shot - the 4:2:2 key has macroblocking up the wahzoo! Whereas the 4:4:4 key has enough colour information for a MUCH smoother key.

So although we don't have the sensor specs from Sony to be able to pinpoint the precise ratio of red and blue colour information to green on the F3's chip - this test at least tells us that there's a LOT more colour information to be found in the camera's 4:4:4 output.

Will you see much of a difference in footage that only requires a gentle grade? No, not really.

But will you see a difference when pulling secondary colour corrections or green/blue screen keys? Absolutely! (pulling secondaries is where you start to 'break' the data that 4:2:2 footage provides)

So depending on your shooting needs and grading requirements, 4:4:4 may or may not be a deal-breaker for you. But it will undoubtably make a difference if you're pushing your images around in post.

As for the economics of purchasing a new F3 in the present market? Personally I think it would be unwise (especially with the prices used units are fetching at the moment). But a used F3 (with the 4:4:4 upgrade), wrapped in the Element Technica F3 Body Armour (to make it more functional), and paired with one of the new Convergent Design Odyssey recorders (which will shortly be able to record up to 1080p60 in 4:4:4 to two different compressed codecs) - is a package that will basically offer the best bang-for-buck image quality in the world right now.

Now if you're wanting to earn camera rentals - look elsewhere, the F3 isn't in fashion. But if you want a personal package with a lot of oomph - the combination I listed above could be had for under $10k - and could do 90% of what anyone needs a camera to do.
 
Great, thanks for all your opinions. Yes, I'm just looking for a used F3 as a personal camera, trying to stay under $8K for the whole package with an Odyssey 7 or 7Q so I don't feel pressure of having to rent out the camera during downtimes. I've used a Gemini 444 recorder with the Genesis and really liked it. Great product, small and it works rock solid. But hey, I'm getting old. I would greatly benefit from the 7.7" Odyssey monitor. Tired of squinting into 5" LCD's all the time!!
 
The only clear, empirical evidence of this published to date is from MXR Productions' blog post here: http://blog.mxr.at/422-vs-444/

Not so fast.

His workflow involved recording 444 uncompressed to a portable macbook pro, but recording the 422 using a Samurai. How do we know for his comparison he didn't use the uncompressed version of 444 to the ProRes version of 422? And there are 3 flavours of ProRes 422 (actually 4 if you include the proxy version) so really how much of a color space comparison was this when codec quality was probably also a factor?

I will recreate the same test using the SR Codec in 444 and 422 and see if it produces similar results. As soon as you add a codec everything changes. On the F3 I saw virtually no micro-jaggies when doing some uncompressed tests over 10-bit SDI, yet the ProRes codec immediately introduced micro artifacts that were comparable to what you'd see shooting XDCAM.
 
Last edited:
Not so fast.

His workflow involved recording 444 uncompressed to a portable macbook pro, but recording the 422 using a Samurai. How do we know for his comparison he didn't use the uncompressed version of 444 to the ProRes version of 422? And there are 3 flavours of ProRes 422 (actually 4 if you include the proxy version) so really how much of a color space comparison was this when codec quality was probably also a factor?

I will recreate the same test using the SR Codec in 444 and 422 and see if it produces similar results. As soon as you add a codec everything changes. On the F3 I saw virtually no micro-jaggies when doing some uncompressed tests over 10-bit SDI, yet the ProRes codec immediately introduced micro artifacts that were comparable to what you'd see shooting XDCAM.

Hmm... compression artifacts rather than colour artifacts is a possibility. Though given that the sky doesn't have an awful lot of detail for the compression to contend with, I still suspect the quality of the key will have more to do with higher colour information. I'd be very interested to see a similar test from SR 422 and 444 on the F55 Dennis - let us know how you go with it.

I'll try and pull up some of my 10-bit 4:4:4 ProRes footage from the Alexa (recorded with the PIX240), and compare it to the 10-bit 4:2:2 ProResHQ material I have from the F3. We'll see what happens with secondaries on skin (the real torture test for codecs as far as I'm concerned).
 
But a used F3 ... paired with one of the new Convergent Design Odyssey recorders (which will shortly be able to record up to 1080p60 in 4:4:4 to two different compressed codecs) - is a package that will basically offer the best bang-for-buck image quality in the world right now.

A big +1

The F3 is almost funny right now because no one seems to understand what can be done with it set up this way, and how much better it is than all the $20k or less options out there in many ways.

Another test I just found https://vimeo.com/61160165
 
Last edited:
Whatever the actual debayer chroma resolution is for the F3, recording in 4:2:2 throws half the data away. 4:4:4 would certainly give more refined control in the grade if nothing else.
 
Too bad that we have to be forced to buy a bunch of stuff we don't need (elecronics) just to have a decent EF mounting system.
However, Jacek is about to get some serious competition for a third of the price as a result.
This will be interesting for all EF owners.

what is that competition that u r talking about? Hope its not Birger, as i dont think anyone believes anymore in that story ,-
 
So 10-bit 1920x1080 will give "more color" than 10-bit 960x540?

If you encode the "960x540" debayer ( not reality but that is another discussion), at 4:4:4 you will get the full 960x540. If you encode it at 4:2:2 you effectively throw away half the 540 chroma resolution you got from the deBayer. Data reduction is data reduction.

Working with raw footage from the D16, I have found that a maximum quality 8-bit 4:4:4 JPEG encoded frame sequence has about the same tonal response range as uncompressed 4:2:2 10-bit broadcast AVI for video grading.
 
These both look very good. I will make sure to check out thesemovies. Some of my favorite actors. I remember the guy in "Marry me for Xmas" from the Osiris web series.
 
Wu,

I recently shot two features, 1 recording to HQ, 1 to 444. I can't say that there's any obvious visual differences, but I'd like to think that the 444 graded much better in terms of subtle skin tone color.

"Dempsey Sisters" (proresHQ)
UP Original Movie: The Dempsey Sisters - YouTube
"Marry Me For Christmas" (prores444)
UP Original Movie: Marry Me For Christmas - YouTube

I'm currently debating on whether to purchase a pix240 or a samurai blade.


Location
Dallas, USA
Posts
194



These both look very good. I will make sure to check out thesemovies. Some of my favorite actors. I remember the guy in "Marry me for Xmas" from the Osiris web series.
 
His workflow involved recording 444 uncompressed to a portable macbook pro, but recording the 422 using a Samurai. How do we know for his comparison he didn't use the uncompressed version of 444 to the ProRes version of 422?...

I will recreate the same test using the SR Codec in 444 and 422 and see if it produces similar results.

Grug's info is pretty compelling especially when you go over the material in the linked article. Very interested to see what you come up with Dennis.
 
I'm sure there are some small differences but feel it would be difficult to call to a huge benefit.

I will do the test shot soon, I'm just waiting for the overcast weather to pass here in Toronto so I can get some nice blue sky. Overnight we had some dreadful ice storms with several trees broken and power lines down.
 
There's absolutely a benefit to shooting 4:4:4 instead of 4:2:2!

Even though the F3's sensor isn't actually high-resolution enough to deliver true 4:4:4 colour information for green, red and blue (i.e. 1080p of red, 1080p of green and 1080p of blue) - it does capture significantly more colour information in it's 4:4:4 mode than it does in 4:2:2.

So although we don't have the sensor specs from Sony to be able to pinpoint the precise ratio of red and blue colour information to green on the F3's chip - this test at least tells us that there's a LOT more colour information to be found in the camera's 4:4:4 output.

I agree and having shot both uncompressed slog 422 and 444 with my F3 I can see a noticeable difference in quality, especially when grading or doing fx work. Given that the F3 output is essentially 10bit rec709 video I feel the need to capture every last bit of information that the format is capable of holding.
 
Yes I've read the thread on the F5/55 Sony form and ordered the XQD cards. However after talking to a Sony tech rep, he wasn't sure if there was a recoverable memory in the XQD cards like there is in the Pro+ cards. I believe there is something like a mirrored RAID on the Pro+ cards but he wasn't sure if the XQD cards offered the same back-up.

According to Sony; "the S-Series features a Wear Leveling Function, Error Correction Code, Data Refresh Function and other new features to enable secure and stable recording of 4K content, helping to meet the rigorous demands faced by professionals. The N-Series gives consumers an affordable entry-level option that supports consumer 4K shooting."

Original source: http://blog.sony.com/press/sony-exp...seamless-and-stable-professional-4k-shooting/

I also wrote a review on my site: http://cinetiny.com/m7pz2o2h
 
Back
Top