Other: XAVC-I vs. ProResHQ

[Doug Jensen]

In the same thread that disparaged the FX6's variable ND filter, some people said that XAVC-I is a terrible codec and not fit for LOG. XAVC-I @ 24 fps records at 240 Mbps while ProResHQ is over 700 Mbps. That means that (all other things being equal) a ProResHQ file will be 3x larger than the XAVC-I version. In other words, you can get 3x more footage on a memory card when you shoot XAVC-I. But do you pay a penalty in image quality when using XAVC-I? To answer that question, I decided to shoot a few test shots with my FX6 to see how the two codecs compared. I recorded XAVC-I internally onboard the camera while simultaneously recording the camera's 10-bit output via HDMI to a Blackmagic 5" Video Assist. In these split-screen tests, the XAVC-I version is on one side of the screen . . . and the ProResHQ version is on the other. Both were shot with the exact same camera settings (S-LOG3) and have had the same grading applied in post. Only very minor adjustments have been made in Resolve to fine-tune and match the Lift/Gamma/Gain. Can you tell which is which? And if you can tell a difference, is it worth 3x bigger file sizes?

I apologize in advance for not using the right camera, the right lens, the right subject matter, shooting on the wrong day of the week, and that it wasn't the correct phase of the lunar cycle for doing camera tests. I will try better next time.

 

[Bassman2003]

Yep. If Prores is better, it is only "lab coat better". I think it is wise to do a cost benefit analysis when it comes to overall acquisition quality. Are your clients paying you enough to go for molecular level quality? Because if they are not, then it is all just internet cannon fodder.

 

[morgan_moore]

I apologize in advance for not using the right camera, the right lens, the right subject matter, shooting on the wrong day of the week, and that it wasn't the correct phase of the lunar cycle for doing camera tests. I will try better next time.

Doug your tests consistently show that 'in general use' the cheap sony sollution is as good as or if not better than the 'thrid party solution. codec vs recorder, ND vs Internal etc.

This is valuable to see and thank you

Sony are clever people and what they design delivers extremely well 98% of the time.

Personally I agree that if you are an FX6 owner it is most likely to fine to keep life simple and shoot with the provided ND and onboard codec.

If the the Sony FX6 is not delivering to your requirements maybe it is time for a Venice (if you shoot $100m features) or a new hobby. (if you shoot ducks in the park).

But there are also little niches that the sony may not handle wall. May not. The 2% that is outside that 98%.

So folk mey have legitmate reasons to want to test those special applications.

I noted that shooting through perspex windows has given me problems. Absolutery a 2% problem. But still a problem 2,% 1% of the time. 50% of the time if you shoot ice hockey

In terms of stressing a codec the 2% would usually sit in pulling back serious under exposure, or playing with high saturation (cop cars Disco LED wedding LED)

If you are a wedding shooting maybe LED rendition matters or shooting a 'blue light' show - there are many hours of horrific blue on many broadcast channels
.
Your tests have value, folks should appreciate that but also understand that they may not cover that 2% scenario

 

[Doug Jensen]

Sam, I appreciate your input but I am skeptical of that hypothetical 2%. If I was the one who believed there was a noticable visual difference 2% of the time then I'd certainly take the time to demonstrate that fact. So with that in mind, I would challenge anyone to provide some real-world side-by-side examples of where VND or XAVC-I don't provide the same results as the alternative -- even if we agree that it would only happen rarely. Until someone shows there is a difference even 2% of the time, I will contend they are basicallly the same 100% of the time. I gotta see the proof.

 

[jmone]

Nice example of how we view stuff BTW (we don't watch the raw footage). They are for all practical purposes identical.

Doug, when you were grading the original footage could you even tell if there was a difference (post ICT)?

 

[Doug Jensen]

Doug, when you were grading the original footage could you even tell if there was a difference (post ICT)?

Good question. No, they were not completely identical. The midtones and shadows were just slightly darker on the ProRes files. That's why I said " . . . very minor adjustments have been made in Resolve to fine-tune and match the Lift/Gamma/Gain." I had to make those adjustments by matching the waveforms rather than visually looking at the images themselves. Is the difference due to the different codecs or something to do with one being recorded onboard the camera and the other on an external recorder? I couldn't say and it doesn't matter to me because the differences were so slight. I couldn't honestly say that one was better than the other, they were just different. In fact, I can't even recall if I matched the ProRes to the XAVC or vice versa without going back to Resolve and looking at the grade.

BTW, I didn't change anything that would affect color, noise, sharpness, etc.

 

[jmone]

Thanks - helps reinforce the KISS approach with a stock FX6 build. It's too easy to get seduced by marginal add ons and then proclaim a night and day self confirming bias.

 

[jmone]

...and FWIW, in the latest Resolve Beta, XAVC-I with SLOG3/SGamut3.cine colour mgt has been shifted to the RAW tab on the Color Page. So you can fiddle with setting there as well. Note: No other XAVC-I profiles have been included in "Sony Raw" as they are calling it.... only SLOG3/SGamut3.cine.

 

[cyvideo]

In 99.9% of cases, I see no reason to shoot ProRes over internal XAVC. The same goes for RAW. For me at least, there are many downsides in terms of added kit making a shooting outfit heavier, more cumbersome and less nimble. Add to that the extra steps in post, and I ask why shoot external. I would agree with Doug that there is no appreciable reason for an external ProRes, or Avid recording for that matter in most everyday shooting requirements when you have a very highly efficient internal codec available to you on most Sony cameras. The fact that ProRes when recorded externally does not follow the same video gamma levels and attributes as the internal XAVC is why you have to make adjustments to the ProRes capture to match the internal XAVC LOG capture. That's the prime reason why Atomos added the "Legalise" switch to their monitor/recorders to try to match the ProRes recordings to the internal XAVC LOG recordings. Canon's XF-AVC codec is almost identical to Sony's XAVC codec bar bit rates. If you are prepared to accept that, it might be worthwhile for those tossing up the pros and cons of whether to shoot external ProRes vs internal XAVC to take a look at Kurtenbach Film's quite in depth overview of RAW and ProRes vs XF-AVC. I 100% couldn't agree more with his findings. Including his final comments on Long GOP when that is available to you as 10-bit 422 as it is o the A7s. Sadly, only available as 8-bit on the FX6 & FX9. If one wants a clearer explanation of the differences on levels between the internal XAVC recordings and external ProRes recordings, good old Gerald gives a pretty good overview of what is going on there.

Chris Young

XF-AVC better than RAW?

UPDATE Your Atomos Ninja V RIGHT NOW!

 

[bobtronic]

I'm sure most people are aware of that but posting it anyway. When recording externally to Prores make sure to set Data Levels to Full in Resolve in the Clip Attribute dialog. With internal recorded footage there is meta data that Resolve can use to set appropriate data levels. S-Log is always full data levels.

 

[cyvideo]

I'm sure most people are aware of that but posting it anyway. When recording externally to Prores make sure to set Data Levels to Full in Resolve in the Clip Attribute dialog. With internal recorded footage there is meta data that Resolve can use to set appropriate data levels. S-Log is always full data levels.

Exactamundo! Worth refreshing this point on levels. It really is the best way to tackle the issue.

Chris Young

 

[Doug Jensen]

I'm sure most people are aware of that but posting it anyway. When recording externally to Prores make sure to set Data Levels to Full in Resolve in the Clip Attribute dialog. With internal recorded footage there is meta data that Resolve can use to set appropriate data levels. S-Log is always full data levels.

Thanks for this tip. That is exactly why the levels on the internal and external recordings didn't match up for my tests. I don't normally record ProRes so I had no idea that Resolve would apply video levels to the clips. When I manually set the ProRes levels to data -- then there is virtually no difference whatsoever between the two. Good to know.

 

[Doug Jensen]

I
XF-AVC better than RAW?

This was an interesting video and I like his information. I just wish he'd use Resolve for grading instead of Premiere. In my opinion, it taints the whole test by trying to grade RAW or LOG in Premiere.

I think his conclusions about Canon's XF-AVC would apply to Sony's XAVC because they are pretty much exactly the same thing, but fortunately Sony RAW is 16-bit so that part of his comparisons would not be applicable to Sony cameras. Nevertheless, I still agree with his conclusions that RAW is hardly ever worth the added hassle while shooting. In post however, there is no difference to the workflow between RAW, XAVC or X-OCN. It's all the same.

 

[bobtronic]

Thanks for this tip. That is exactly why the levels on the internal and external recordings didn't match up for my tests. I don't normally record ProRes so I had no idea that Resolve would apply video levels to the clips. When I manually set the ProRes levels to data -- then there is virtually no difference whatsoever between the two. Good to know.

Great. That was my finding too, virtually no difference if everything is set up correctly. Sony's internal codecs provide great quality.

 

[AmbiSonde]

In the same thread that disparaged the FX6's variable ND filter, some people said that XAVC-I is a terrible codec and not fit for LOG.

Why did you feel it necessary to misrepresent what was said? No one said XAVC-I is a terrible codec and not fit for log. No one. You made that up. You want to engage in strawman arguments because you are desperate to 'reassert' your position as lord high priest amongst your flock.

What was actually said was that there were good reasons for shooting ProRes over XAVC-I, this comment was specifically made in relation to FS era cameras. Any professional who has shot SLog3 on the FS7 era cameras and who cared about the image being offloaded knew how borked SLog3 + XAVC-I could be under certain conditions. Shooting ProRes even when limited by the 12 bit RAW output of the FS series cameras could offer significant image quality improvement over XAVC-I. Shooting SLog2 + XAVC-I could also offer significant image quality improvements to SLog3 + XAVC-I.

When I discovered the issues with SLog3 and XAVC-I I sent my camera back to be checked out by Sony service because I felt sure that artefacts such as those present were a fault not a feature. The camera came back with a clean bill of health but also the same serious compression artefacts were present. The FS7 represented the best value camera of its time so ultimately I decided to work around the issues but I did briefly consider getting shot of the camera.

I even gave an explanation why SLog3 and XAVC-I on the FS7 had issues but this was met with aggression and condescension. Note, I have said NOTHING about FX era Slog3 + XAVC-I because I have not shot a single frame myself....yet. I suspect Sony has altered the codec which is why I suggested XAVC-I was a moving target. Wouldn't it be hilarious if in fact XAVC-I and Slog3 continue to have significant issues? I shoot XAVC-HS and have completed a side by side test of S-Cinetone, HLG1, HLG3 and SLog3 with that codec and see no tetris like macroblocking artefacts in the image. I do see more noise, more banding and larger macroblocks in SLog3 footage over every other gamma curve. S-Cinetone and HLG1 being the cleanest, and HLG3 being a tiny bit noisier. Only SLog3 rated at two stops lower approached the clean image of the rest of the gamma curves but at a significant cost to the dynamic range of SLog3. SLog3 at Base ISO and one stop below are noticeably noisier than all other configurations tested. It might not matter to you but it does to me.

Before I post images supporting my position is there anyone who wants to add themselves to the Dunning-Kruger Brigade above who hasn't already? I'll check back in a few days...

If you aren't interested in FS7 SLog3+XAVC-I artefacts and why ProRes could be a better choice just say and I won't waste my time.

 

[Bassman2003]

I think it is best for you to go AmbiSonde. Not that I want to discourage any input on a forum but it is clear that you are looking for conflict. I tried to go re-read your responses in these recent threads and you have removed them. That is just odd behavior imho. It is just a codec or a video camera. Pretty black and white. So why the hesitation to post images that support your findings? If you are seeing artifacts then why not show them for others to see and maybe contact Sony for future improvement? It is not a competition.

What any internet forum does NOT need is a few people that are constantly bickering. This place is hobbling along as it is...

 

[Doug Jensen]

Why did you feel it necessary to misrepresent what was said? No one said XAVC-I is a terrible codec and not fit for log. No one. You made that up. You want to engage in strawman arguments because you are desperate to 'reassert' your position as lord high priest amongst your flock..

Blah, blah, blah.
Hey buddy, it isn't about YOU. I haven't even given you a moment's thought since you promised you were leaving because DVXuser is "sad old folks home inhabited with dribbling geriatric ignorati who are so clueless they don't realise how clueless they are." Did I mention you in any of my posts about VND or codecs this week? No, because it's not about YOU. Plus you said you were leaving. But now you are back. Who would have guessed that?

It's flat-out impossible to misrepresent what YOU said because (as you have admitted) you went back and deleted and/or re-wrote all your old posts. How can anyone have a serious discussion with someone who goes back and hides what they said or re-writes their comments? I can't quote you because you found out you were wrong and deleted all your comments when challenged. In addition, we don't even know your name. We've never seen a single frame of video you've shot. We don't know your work history or anything about you that might help give you credibility. We know you can spew the technical jargon, but that don't mean squat.

If you think you blow into town, make outlandish comments, and then expect the dribbling geriatric ignorati to just accept your dumb opinions at face value, you are wrong. Even us ignorati can smell bullshit. And I suspect that's exactly why you got run out of your social media circles and came here in the first place. But, then again, I guess the social media people were all ignorati too, right?

If you must know, my testing this week was spurred by comments made by Rocketfuel.

The FX6 looks soft because the internal codec sucks big time. Even though it is an all "I" codec, compared to the XAVC-SI in the A7SIII, the FX6 implementation of the codec renders soft glowing images. I know for a fact because I compared simultaneous recordings I've made with my Ninja V+ in ProRes HQ and ProRes RAW. The files recorded externally are sharper and with much better color and skin tones. I hate the internal codecs of the FX6 because of this. I think that the internal codecs of the FX6 were developed a long time ago for the older cameras, and since then Sony has improved the compression algorithm and that’s why it’s better on the A7SIII. I wish that the FX6 could record ProRes internally like many consumer cameras can today like Fuji and Nikon. One other fact, is that the internal codec combined with the dismal electronic variable ND creates even more degradation to the image. There's a reason the Venice doesn't have the electronic VND. It sucks.

There it is. So, do you still want to claim that no one said XAVC-I is a terrbile codec? I found at least one. Unfortunately, I can't quote you because you deleted all your posts. So are you going to go back and delete your post today now that I have confronted you with actual facts?

What was actually said was that there were good reasons for shooting ProRes over XAVC-I, this comment was specifically made in relation to FS era cameras. Any professional who has shot SLog3 on the FS7 era cameras and who cared about the image being offloaded knew how borked SLog3 + XAVC-I could be under certain conditions. .

Hmm, not me. I've had an FS7 since the first month they started shipping and I still have mine. I guess I just don't care about image quality. :-(

When I discovered the issues with SLog3 and XAVC-I I sent my camera back to be checked out by Sony service because I felt sure that artefacts such as those present were a fault not a feature. The camera came back with a clean bill of health but also the same serious compression artefacts were present.

Did they return the camera with a note that said, please learn how to shoot with S-LOG and stop bugging us with your nonsense. Out of all the thousands of FS7 users in the world, only you have the discerning eye to see theflaws. Thank god you have set the record straight. I'm going to have to go back and reshoot all my video from the last 10 years.

I suspect Sony has altered the codec which is why I suggested XAVC-I was a moving target.

Prove it. What evidence do you have that XAVC-I is different than what it was a few years ago? When did Sony secretly give us "XAVC-I Verison 2"? Which firmware update gave us that update, or if it wasn't a firmware update, which specific cameras have the Version 2 codec and which ones do not? I am dying to find out the answer to this because you are the first person I ever heard claim it. Let's see the evidence.

If you aren't interested in FS7 SLog3+XAVC-I artefacts and why ProRes could be a better choice just say and I won't waste my time.

Save your time. You have no credibility.

However, I do look forward to hearing your evidence about XAVC-I Version 2.

 

[AmbiSonde]

Dougie,

Just the reply I was hoping for...

You are obviously confident in your position and I'm confident in my position so how about if I prove you wrong YOU NEVER post on this forum again and if I fail I NEVER post on this forum again?

How about it? How confident are you? It should be easy for you because I lack credibility and you are overflowing with it, right?

 

[Doug Jensen]

Dougie,

Just the reply I was hoping for...

You are obviously confident in your position and I'm confident in my position so how about if I prove you wrong YOU NEVER post on this forum again and if I fail I NEVER post on this forum again?

How about it? How confident are you? It should be easy for you because I lack credibility and you are overflowing with it, right?

Please don't try to drag me into your childish silly games. What's next, camera's at 20 paces? Do whatever you want. I shall do the same without giving you a second thought.

 

[AmbiSonde]

Please don't try to drag me into your childish silly games. What's next, camera's at 20 paces? Do whatever you want. I shall do the same without giving you a second thought.

Of course, the thought of not having your massive ****ing ego massaged here is too much for you to risk.

Imagine not noticing macroblock artefacts like this.

FS7, XAVC-I SLog3 HD crop at 100%

A fairly common grading operation requested by directors is to add more saturation into a SLog3 sky and to add a Grad. A simple operation like this can expose XAVC-I's horrendous macroblocking. The image below is obviously pushed further than a typical grade to better show the macroblocks but even relatively minor contrast adjustment can make a flat area of colour become extremely noisy due to the underlying posterisation cause by compression. These Tetris like macroblocks change every frame but the strange horizontal artefacts tend to be the most visible when grading.

I'd say anyone missing these sorts of artefacts really doesn't care about image quality or the quality of the product they hand off to clients.