FS5: FS5 4K color and bit depth

[ChristopherA]

At this moment I'm seriously considering what other users said about the FS700 + external recorder. That is one camera that can also be found used under our budget and it will make more sense than the F3. We have to consider all the accessories we are going to use with it and see if it can be a valid option. Other than that, I have no information whatsoever about the FS700 but I guess that I'm in the right place to start searching.

The FS700 is not a fun camera to handle or work with, but when paired with the Odyssey 7Q, it produces a spectacular 4K image. Be aware that other external recorders may not pull the RAW out of the FS700, and you don’t want to use the non-RAW output from this camera.

Honestly, if your budget can't stretch to an FS7 and 4k XAVC, it won't stretch to covering the recording and storage of 4k ProRes over a three year project (the storage requirements are three times as high as 4k xavc, that's a LOT of extra harddrives).


So whatever you're looking into, I'd be wary of the data requirements of 4k without the efficiency of xavc.

This. This. This. If you save $2000 on the camera rig but you have to spend that on drives, you lose. Be sure to take ALL the costs into consideration. XAVC-I is a magical format.

The big difference is the Codec efficiency as you state. XAVC can store more media than even ProRes LT (if I did not messed up my calculations), only surpassed by ProRes Proxy. That's a big advantage but a quick internet search revealed some quirks about the editing workflow with Davinci Resolve, which is now our editing NLE.

I have been working almost exclusively with XAVC-I footage for just over a year. Never had any problems. Do as Doug Jensen has recommended, and get some footage to test in Resolve. If you can’t, let me know and I’ll upload some clips for you.

FS7 and FS5 owner here.


First of all, this seems a non-issue. 80% of the time the Slog3 8-bit 4:2:0 Long GOP from the FS7 holds up in grading. I do it ALL THE TIME. From what I've seen owning an FS5 for about two weeks, it's identical.

I hate to disagree, but I have to. Although I’ve seen some decent stuff with the 8-bit 4k Long-GOP codec, I’ve also seen some awful footage and my tests found the codec quite lacking. It’s extremely limiting in grading and completely falls apart in high-contrast settings.


Just testing the codec (XAVC-L on FS7) in my office with the LC709A LUT, there was immediately banding on the wall. I hadn’t even begun grading. I would advise against using this codec, but by all means do your own tests.

my understanding is that 8-bit 4:2:0 4k can be down converted to 10-bit 4:4:4 1080p. I'm not an expert, but these guys seem to be: http://www.eoshd.com/2014/02/discovery-4k-8bit-420-panasonic-gh4-converts-1080p-10bit-444/

I believe they are mistaken. It is possible to take a RAW 4k and get a 1080p 4:4:4, but I believe that needs to happen prior to compression. Mitch Gross could speak more to this than I can, though. I do not believe it is possible to start with a compressed 4K 4:2:0 and get a 1080p 4:4:4. However, If your goal is 1080p there are better options—including shooting 1080 on an FS7 in 4:2:2 10-bit.

That's a bit like saying it's actually a 2.2K camera instead of 4K. I purchased mine for the express purpose of shooting a feature in 4K, scaling and reframing as needed and projecting in 2K. Saw it projected on a 40 foot screen last month and never noticed any resolution issues. I'm sure I was cropping as much as 30% at times.

I have also used the FS7’s 4k to reframe quite extensively. It does lose a little sharpness, but it still looks great.

Ergonomics apart we never had any issues when grading ProRes 422 HQ from BMPCC. On the other hand, we remember the big improvement on that front when we migrated from Sony's EX3 XDCam EX 8-bit 420 to the BMD cameras.

I had a project just recently where we had a shot looking out of an open refrigerator. Because of space requirements, we had to shoot that with a BMCC. The entire rest of the shoot was done with the FS7. Matching the color on the BMCC was a nightmare. Every time I have to use the BMCC, I end up regretting it in post.


Also, be aware that you have no real off speed options with the BMCC. I’ve been shooting 30, 36, and 48 for 24p quite a bit.

 

[zebratape]

3 years of 4k footage? damn, that is a lot of storage

 

[Tom Roper]

But even if it wasn't, my understanding is that 8-bit 4:2:0 4k can be down converted to 10-bit 4:4:4 1080p. I'm not an expert, but these guys seem to be: http://www.eoshd.com/2014/02/discovery-4k-8bit-420-panasonic-gh4-converts-1080p-10bit-444/

David Newman can be trusted. But so is this; 4k (UHD) 4:2:0 has the same number of chroma samples as 4:4:4 1080p. You could downsample to 1080p by simply discarding 3 out of every 4 luma samples while keeping all the Cb and Cr, and still have 8 bit 4:4:4 1080p. But what David is saying is by working in uncompressed 10 bit RGB space, instead of discarding the luma samples you can add 4 together to make one 10 bit luma sample, effectively giving you 10 bit 4:4:4 1080p. There is one small catch, you don't have 4 chroma samples to sum together like luma, so your 10 bit 4:4:4 RGB values will only have 8 bit precision for the chroma samples that went into the conversion to RGB. The luma samples that went into the conversion to RBG would be genuine 10 bit however.
Also worth mentioning since you gave anecdotal experience to claim 4:2:0 is good enough in most cases; I can give you a case where even 4k 10 bit 4:2:2 is marginal and can show banding artifacts; HDR with WCG (High Dynamic Range and Wide Color Gamut). 4k has a problem, and so will 8k, that you can't see the benefit from increased resolution alone except with very large screen sizes or when viewing very close. David Newman also alluded to that by pointing out the benefit that 4k can hold more color information and more dynamic range information that can be translated down to 1080p as he demonstrated above.
The future for 4k has to include better color and dynamic range than what most of us are working with, 709 color and gamma, particularly in light of the OP's planning for 3 years ahead. That's going to be the expectation of the viewers and content purchasers, titles already showing up in Dolby Vision 4k laser projection cinemas, Amazon, Netflix, MGO and latest generation 4k displays from Sony, Samsung, LG, Panasonic, Visio, that all are on sale right now; support HDR or Dolby Vision, and practically none of the film makers outside of those under NDA from Fox, Warner, Sony, Disney and Universal have much of a clue about. If you are looking 3 years ahead, you have to consider that 8 bit 4:2:0 even in 4k will be at best, ordinary.

 

[Dan Abrams]

Just testing the codec (XAVC-L on FS7) in my office with the LC709A LUT, there was immediately banding on the wall. I hadn’t even begun grading. I would advise against using this codec, but by all means do your own tests.

I believe there is a specific problem with internal LC709A recording, regarding the banding. Try it shooting it log and applying the LUT in Resolve and it's better. This is an internal processing issue, not the codec, is my understanding (which could be wrong).

 

[ChristopherA]

I believe there is a specific problem with internal LC709A recording, regarding the banding. Try it shooting it log and applying the LUT in Resolve and it's better. This is an internal processing issue, not the codec, is my understanding (which could be wrong).

If this is the case, that would be great news. I was using the baked in LUT, as almost anytime I'm shooting XAVC-L it's for something that won't get color grading attention. That being said, adding a LUT after the fact is easy. I'll test.

Edit: Test complete. This is correct. If you bake the LUT in camera, you will get very bad banding. If you record the LOG footage and put the LUT in post it looks much better.

 

[Tom Roper]

If this is the case, that would be great news. I was using the baked in LUT, as almost anytime I'm shooting XAVC-L it's for something that won't get color grading attention. That being said, adding a LUT after the fact is easy. I'll test.

In your case with the FS7, I would compare it to XAVC-I, the 10 bit 4:2:2 codec, *and* add the LC709A lut in Resolve to the OUTPUT. (output lut) then work backwards to isolate the nature of the banding you saw. Don't use LED lights for the test either.

 

[naavt]

Would you like me to shoot something that you think will be a particular stress test for the 8-bit 4K codec, and send the footage to you? Let me know what you'd like me to test and I'll send you the file and you can have your colorist play with it and try different things.

Hello Dan. I thank you in advance for your availability towards this. I plan to run several tests once I can get my hands onto the camera. Some of the ones I can remember right now:

Skin test with a X-Rite Checker
Profiling my Sekonic L758 meter and conducting a DR test
Shoot an extensive DR scene and under/over expose stop by stop to verify the camera limits at shadows and highlights
Shoot some skies and other surfaces to test for banding
Shoot a green screen to test color space
Shoot moving objects to test rolling shutter

Doing all that at 1080p and 4K to test for differences regarding the internal Codec used between both resolutions. I'll appreciate inputs on which Log I must use for better results.

I'll appreciate any footage that you find useful and can kindly share with me to start running some tests.

David Newman can be trusted. But so is this; 4k (UHD) 4:2:0 has the same number of chroma samples as 4:4:4 1080p. You could downsample to 1080p by simply discarding 3 out of every 4 luma samples while keeping all the Cb and Cr, and still have 8 bit 4:4:4 1080p. But what David is saying is by working in uncompressed 10 bit RGB space, instead of discarding the luma samples you can add 4 together to make one 10 bit luma sample, effectively giving you 10 bit 4:4:4 1080p.

For the purpose of the test I'll want to run once I have the FS5 with me, can you give me some insight on how to properly convert FS5 4K footage into 1080p 10-bit 4:2:2 (or better - if it's possible - 2K 10-bit 4:2:2)?

 

[Tom Roper]

For the purpose of the test I'll want to run once I have the FS5 with me, can you give me some insight on how to properly convert FS5 4K footage into 1080p 10-bit 4:2:2 (or better - if it's possible - 2K 10-bit 4:2:2)?

Andrew Reid made it sound a bit mystical but all you have to do from your 4k (UHD) 8 bit 4:2:0 project in Resolve is render an intermediate to an uncompressed codec (RGB 10-16 bit or RGB 10 bit Cineform) with your choice of format, DPX/AVI/Quicktime. That's the extra step he refers to. It only works because of the down sampling to 2k/1080p. If you rendered to the same resolution you started with (4k/UHD), you'd have a nice but wasteful 10 bit 4:4:4 with only 8 bits of actual precision. It would still be 4:4:4 chroma sub sampling like all RGB but the chroma samples would have been averaged across 4 pixels.

 

[Grug]

For the purpose of the test I'll want to run once I have the FS5 with me, can you give me some insight on how to properly convert FS5 4K footage into 1080p 10-bit 4:2:2 (or better - if it's possible - 2K 10-bit 4:2:2)?

You won't be getting 10-bit tonality from downsampling 8-bit 4k into 1080p. Only improved chroma subsampling. Which is silly. If you're going to end up with 1080p footage anyway, shoot it at 10-bit to begin with (either internally or recorded externally), then you'll have enough bit depth to use SLOG properly and get a nice result.

 

[Tom Roper]

You won't be getting 10-bit tonality from downsampling 8-bit 4k into 1080p. Only improved chroma subsampling. Which is silly. If you're going to end up with 1080p footage anyway, shoot it at 10-bit to begin with (either internally or recorded externally), then you'll have enough bit depth to use SLOG properly and get a nice result.

This is wrong. David Newman is correct and explains how. Read his tweets here. http://www.eoshd.com/2014/02/discovery-4k-8bit-420-panasonic-gh4-converts-1080p-10bit-444/

 

[alohype]

First of all, this seems a non-issue. 80% of the time the Slog3 8-bit 4:2:0 Long GOP from the FS7 holds up in grading. I do it ALL THE TIME. From what I've seen owning an FS5 for about two weeks, it's identical.

I have to disagree as well. They couldn't be any different. 4K XAVC-I is wonderful, you can take your image in grading anywhere you want and it can save you from white balance and exposure issues. I don't think I ever had a footage from XAVC-I that was so bad that couldn't be saved. 4K in XAVC-L on the other hand is weak in that regards. Forget about exposure or white balance shifts (at least bigger ones). You can still grade it but it falls apart A LOT quicker. And it's very heavy on computing power to they point were it's unusable (at least on average machines). I still use 4K XAVC-L but only when I'm very short on memory cards space, for long interviews or when I have time to setup and light everything properly, there's no time for grading and it will be launched just the way it was shot.

 

[naavt]

(...)If you're going to end up with 1080p footage anyway, shoot it at 10-bit to begin with (either internally or recorded externally), then you'll have enough bit depth to use SLOG properly and get a nice result.

FS5 only shoots 10-bit at 1080p and the main reason to shoot 4K is to reframe in post. One option we will not have if shooting 1080p.

I have to disagree as well. They couldn't be any different. 4K XAVC-I is wonderful, you can take your image in grading anywhere you want and it can save you from white balance and exposure issues. I don't think I ever had a footage from XAVC-I that was so bad that couldn't be saved. 4K in XAVC-L on the other hand is weak in that regards. Forget about exposure or white balance shifts (at least bigger ones). You can still grade it but it falls apart A LOT quicker. And it's very heavy on computing power to they point were it's unusable (at least on average machines). I still use 4K XAVC-L but only when I'm very short on memory cards space, for long interviews or when I have time to setup and light everything properly, there's no time for grading and it will be launched just the way it was shot.

So you are saying that FS5 UHD files (which can't be recorded on XAVC-I like FS7), aren't really that great for grading contrary to what some members have stated. I'm contacting Sony tomorrow and one of the things I'll keep in mind is to test that downsampling theory from 4K 8-bit 4:2:0 to 2K 10-bit 4:2:2. Shoot the same footage with a native 4:2:2 codec like ProRes on another camera and compare both in the grading suite. Has any FS5 owner already tried that?

 

[alohype]

FS5 only shoots 10-bit at 1080p and the main reason to shoot 4K is to reframe in post. One option we will not have if shooting 1080p.




So you are saying that FS5 UHD files (which can't be recorded on XAVC-I like FS7), aren't really that great for grading contrary to what some members have stated. I'm contacting Sony tomorrow and one of the things I'll keep in mind is to test that downsampling theory from 4K 8-bit 4:2:0 to 2K 10-bit 4:2:2. Shoot the same footage with a native 4:2:2 codec like ProRes on another camera and compare both in the grading suite. Has any FS5 owner already tried that?

Exactly. I am everyday FS7 user using both formats so I know strong points of each (L - smaller file sizes, I - everything else) and telling you that "I" flavour is miles better.

 

[naavt]

Exactly. I am everyday FS7 user using both formats so I know strong points of each (L - smaller file sizes, I - everything else) and telling you that "I" flavour is miles better.

Shame that Sony hasn't shipped their FS5 with the same FS7 Codec

 

[alohype]

Shame that Sony hasn't shipped their FS5 with the same FS7 Codec

I think that was marketing decision to differentiate them more. Price difference of body itself is negligible. FS5 also lacks of CineEI mode, something that is probably the best feature of FS7.
I treat FS5 basically as a 1080p camera. 4K (without external recording) is there just for kicks.

 

[naavt]

I think that was marketing decision to differentiate them more. Price difference of body itself is negligible. FS5 also lacks of CineEI mode, something that is probably the best feature of FS7.
I treat FS5 basically as a 1080p camera. 4K (without external recording) is there just for kicks.

I didn't tried it yet but that's also my first impression about it. It has the advantages of size, weight and uses cheaper media though.

 

[nsoltz]

No. Not including XAVC-I with the FS5 is not a marketing decision. The SDXC cards cannot sustain the data rates needed for XAVC-I.

 

[naavt]

No. Not including XAVC-I with the FS5 is not a marketing decision. The SDXC cards cannot sustain the data rates needed for XAVC-I.

I remember that when BMD first introduced BMPCC there were many issues with RAW recording to SDXC cards due to their limited capacity to write the amount of data needed, but the world moves forward and we already know that "tomorrow" someone will going to present us a card capable of that and much more, so that's not an excuse.

If you ask me, Sony has the inside knowledge, technology and budget to develop a proprietary card for their own cameras like RED did.

 

[Grug]

This is wrong. David Newman is correct and explains how. Read his tweets here. http://www.eoshd.com/2014/02/discovery-4k-8bit-420-panasonic-gh4-converts-1080p-10bit-444/

David's explanation seems sound, but your interpretation of it is off.

10-bit colour depth offers 1024 independent of red, green and blue. Being able to yield 1024 shades of green, but only 256 shades of blue and red does beat having 256 shades for all three, but it still does not equal proper 10-bit colour depth; and therefore leaves you open the same limitations that lack of tonality in red and blue presents in regular 8-bit footage.

There's no free lunch with these things, and it really does no one any good to let hyperbole cloud the issue.

 

[DLD]

No. Not including XAVC-I with the FS5 is not a marketing decision. The SDXC cards cannot sustain the data rates needed for XAVC-I.

Plus, the cost of the interface for the QXD is purely in the hardware.

If you ask me, Sony has the inside knowledge, technology and budget to develop a proprietary card for their own cameras like RED did.

It has the cards but one needs to install the appropriate electronics inside the camera too. That would involve several steps, from taking the data off the sensor to buffering (if needed) to the processing power (if needed, for compression or various sampling algorithms) and the interface to a faster card.

Look at GH4. It needed that YAGH attachment to handle the data and data processing that was too much for the camera itself.