Deadcode

1 hour ago, KnightsFan said:

It turns out that A is ProRes and B is H.265. As @Deadcode points out, B (H.265) has noticeably more detail. However, the ProRes file retains a tiny bit more of the chroma noise from extreme shadows the original video. As you can see in the 300% crop below, all that sparkly noise is simply gone H.265, and while the ProRes has some obvious blockiness, you can still faintly see the noise. This is of course a VERY extreme grade (see the curve from Resolve).

If you want to look at the original files (before I turned them into 380 MB monsters!) One final note: I accidentally encoded both with AAC audio, so ~6kb of the file size of each is an empty audio track.

ProRes: https://drive.google.com/open?id=102Ivc9Xa1Z7mPzCK8TgTqwh2GExMPOkZ

H.265: https://drive.google.com/open?id=1NxMofYvrHcP6DHx5VECRwMNw8qA4KXOD

These results are shocking for me ? h265 > ProRes ?! Insane. I think i will redo this test with my own RAW footage.

2 hours ago, KnightsFan said:

I did a quick comparison between ProRes and H.265 encoding, and thought I'd share the results with everyone. I grabbed some of Blackmagic's 4.6k RAW samples and picked a 3 second clip. In Resolve, I applied a LUT and then exported as an uncompressed 1080p 10 bit RGB 444 file. This is my reference video. From this reference file, I encoded two clips. One clip is H.265, and the other is ProRes SQ, which I compared to each other and the reference video.

The reference video (which I did not upload) was 570 MB.

One of these files was created from the reference file using "ffmpeg -i Reference.mov -c:v libx265 -crf 20 -preset slower -pix_fmt yuv422p10le H265.mov". The file size is 7.81 MB (1.4% of the reference)

The other file was created using "ffmpeg -i Reference.mov -c:v prores_ks -profile:v 2 ProRes.mov". The file size was 44 MB (7.7% of the reference, or ~5x the size of H.265)

In order to keep the comparison blind, I then converted both the ProRes and H.265 files to uncompressed 10 bit 422. So you shouldn't be able to tell which is which from the metadata, file size, playback speed, etc. You can download these files (380 MB each) and do extreme color grades or whatever stress tests you wish, and compare the quality difference yourself.

https://drive.google.com/open?id=1Z5iuNkVUCM9BgygGkYRXizzr6XnSXDK7

https://drive.google.com/open?id=1JHkmDKZU4qdS7qO_C0NwYH2xkey0uz7C

 

I'd love to hear thoughts or if anyone would be interested in further tests. Keep in mind that there are different settings for H.265, so this test doesn't really have implications for camera quality. since we don't know what their encoder settings are. However, it could have implications for intermediate files or deliverables, especially to sites with file size limits.

With ridiculous amount of crop there is a difference in texture detail. "B" is ProRes (left) and "A" is h265. (right)

Thanks for the test!

And none of you can figure out why this is happening. Beautiful.

The AF is reliable till -3EV (with A73), and live view exposure counts. 

Regarding to backlight situation if you go below -3EV the AF can struggle even in broad daylight.

The test is simple, use SLOG2, dial down to -2EV exposure and try to use AF. It will be unreliable.

If you dial up to +2EV everything will work perfectly.

In backlit situation the auto exposure metering is unreliable, if you push the exposure up manually above +2EV the AF will work perfectly.

Side note: in backlit situation i recommend to shoot with zebras turned on and set 46 +-5. Set your zebras to your subject, and you can avoid noise and still have some room in the highlights. (46IRE means +1EV for SLOG2 middle grey) If you are using auto metering, and set it to +2EV, it will be still underexposed. 

40 minutes ago, thebrothersthre3 said:

Yes

Awesome, an expert! Please show examples, reason, and solution!

6 hours ago, kye said:

 

Yep.. that's right, sorry I should have said.  Specifically the line in Resolve.  Theres a bit of a debate about if it's the skin colour line or just an indicator, but the point is that skin tones actually vary wildly in hue depending on lighting conditions.  The idea that skin tones should be on the line are shot down pretty quickly on the grading forums.

The top row are from the ARRI video, and the bottom row are from other videos that got good reception.

Here are a bunch of vector scopes from the ARRI video - notice the highly, highly, highly controlled skin tones!

These are a couple from the C200 demo video (IIRC) and note the more spread out line, and also that the line is between the indicator line and the Red reference box, and even slightly beyond the red in certain areas. 

Actually it's easy to even out the skintones. Just use hue vs hue curve with 4-5 adjustment points between red and yellow.

Dave Dugdale created an awesome tutorial about color matching between cameras and it's worth every penny. He also talks about the thickness of the skin tone "blob" what you can see above, and how to match it

What will be the "usual" bitrate for 8K?

For 1080p there was 28Mbps AVCHD first, then 50Mbps with newer codec's. For 4K the usual was 100Mbps. For 8K  IPB h265 i say 200Mbps seems to be the favorable.  V30 SD cards could be enough.

Actually i dont think processors are there yet. If these sensor's will be in the A7 line, they will be limited to 4K 60

But the same sensor will probably record 8K in Sony's next high end cinema grade cameras.

4 hours ago, kye said:

What is your impression of how transferrable things like this are?

I read the ML forums for a while and saw some instances of where figuring something out on one camera helped to solve it on other models, but I'm not sure if these were isolated instances or if this was more the norm.

If so, perhaps these findings might apply to the 5D3 and maybe 5D4?

5D2 is using Digic 4 processor 5D3 is using Digic 5.Everything that is possible with the Mark II is possible with the Mark III.

5D4 is not hacked yet. Yes, A1ex was able to run his code on the 5D4 (just like with the EosR) but he is not owning one, so he cannot continue the progress.

Until one of the skilled hackers own at least one newer (Digic7+) cameras we can not expect any progress.

17 minutes ago, FilmMan said:

Reddeercity post today...

I should mention that the 5D2 HDMI signal is 8bit 4.2.2 Full range (0-255) , Check it on my AJA Kona LHi video capture card .
So by this theoretically  "UHD 8bit(24 bit/px) 3840x2160 30p 4.2.2" would be in the realm of possibility . 
Here's hoping   

Dont get too excited, reddercity is brute forcing the registers in the past couple of years, and technically never archieved anything. The other forum members developed the new features from reddercity's findings . And unfortunetaly less developer using 5D2 every year.

But let's hope for the best

12 minutes ago, Andrew Reid said:

 

Z-LOG is an in-camera LOG profile for video shooting on the Nikon Z7 and Z6 mirrorless cameras, and most Nikon DSLRs.

Available now

Gorgeous! Please show more samples! Maybe early bird discount for the first 10 buyer? ?

34 minutes ago, Mattias Burling said:

Why don't we ever see the same criticism against Panasonic and Sony excuses. I mean I think the GH5 and A7iii are great cameras, but Im not gonna ignore the severe crippling and lacking of certain features. Yet the positive reviews never get the same bashing or called apologists, etc. Its cool to decide to live with the short comings of some cameras but not others. Im wondering if its because they are so mainstream and there for trendy to bash? 

Sony for example are always flying reviewers to exotic and expensive events but don't get as much flak for it. 
One of the biggest youtube reviewers was saluted as unbiased and fair because he went down hard on the Canon and said the Sony was better.
The same youtube have been traveling on the Sony account and said, "usually I never review Sony because they don't invite me to things.. but now I got to go to..."... 

Just a bit weird imo.

Dont forget the rolling shutter / overheat at a6300 / a6500, overheat / sunspot / bad battery life / poor AF with A7S II, The crappy 4K at fullframe with A7R II, the aliasing heavy  1080p with A7 II. Sony was bashed a lot in a lot of reviews

But they overcome the major weaknesses with the third generation. Still not perfect, but much more value for the money and much less pain in the ass comparing to previous generations.

Actually i dont remember if there was major problems with GH3/GH4. Maybe they were perfect? ?

31 minutes ago, Nikkor said:

Shure man, you make no sense at all, what are you trying to argument? There has to be continuity between posts.

To get those skintones you have to purposely fuck up. Seems like we have another psycho.

I fixed it for you, now you can rest in peace.

Let's get back to the original purpose of this video: noise levels...

2 minutes ago, Danyyyel said:

Whaaaaat, do you think that 90% of people buying a 3000+ usd camera use it at default setting. If it was a point and shoot or one of those cheap 500 usd camera, I would agree, but apart from being some rich guy just for bragging rights, most would configure their camera to some level.

I'm sure most of the Z7 customers are photographers who rarely record video. And they use factory default settings.

Those who want video oriented camera, will not buy Z7, nor EosR. The customers buy these cameras for photography.

And very few of the photographers are even heard about Davinci Resolve for example.

I was in several Sony Mirrorless/A7/a6xxx Facebook groups, and even after several years the "why my slog2 is so noisy" question is STILL on the table.

My point is: we, here on EosHD forums are only the small percentage of the users who actually want to examine 100% of the capabilities from these cameras. Others just use them out of box.

41 minutes ago, Andrew Reid said:

 

Again, in the Pocket 4K shot the man with the Apple tampon in his right ear seems off. It's under exposed.

I have not watched the video, I will just get annoyed... But can anyone tell me if Max says ANYWHERE in it, what the camera picture profile and white balance settings were, what kind of lighting was used, what the workflow was in post, what the codec settings were or what he did to the Pocket 4K to make it look about a stop darker than the EOS R at ISO 1600?

If there is anywhere near that level of basic detail in the video I'll be astonished.

I don't care if Max is a nice guy or not, he is misleading as shit with these basic cobbled together test shots.

He simply does not deserve that many subscribers on YouTube, based on the quality of his info.

But hey, that's YouTube for you!

He is showing every camera at their default settings.

So technically he is using the cameras just like 90% of the users do.

A lot of people complaining sony have the worst skintones. Because they dont know how to grade.

I can make the image from any camera to look good. It's just 30 sec work to fix the Z7 image in post.

And it's just few minutes work to match all of the cameras above to look the same. But we dont want that. We want to argue which one is better...

2 hours ago, Andrew Reid said:

The GH5S and Pocket 4K don't do supersampling either. Crisply detailed enough for my liking.

Technically the pocket 4K sensor size is the same as the eos r in 4k mode. I would love to see sharpness comparison with the same lens from your hands.

4 hours ago, thebrothersthre3 said:

Andrew seems fair about it to me. He ripped on the camera a lot initially but it surprised him when he actually went to use it. 

Can wait to see what happens after he tried BMPCC4K. ? 

3 hours ago, TurboRat said:

Why is the contrast higher on the GH5? bec of exposure?

Black levels and exposure are roughly matched, contrast is more or less close to each other. Technically you compare noise levels and shadow detail here, the contrast difference does not matter. 

1 hour ago, Danyyyel said:

If the Nikon Flat is that clean compared to the other logs in the shadows, it corroborate some test that showed that it was very close to log. You could clearly underexpose by one stop to preserve the highlight and come very close to the other logs. For the Nikon Z7 log it is exact contrary, I have not been able to download the high bitrate vimeo file for now, but we can clearly see from the highlight in the bulb that is clearly underexposed by at least 1 stop compared to the other log file. It would have been very nice to match the highlight of at least the Sony to see how much detail is left in the shadows.

Probably there IS a difference between N-Log and Flat, just like with the Sonys. Cine 4 has 2 stops narrower dynamic range compared to SLOG2

I think middle grey should be exposed to the same level. If we expose for the highlights it's much harder match the shots in post.

A7 III has the widest dynamic range, however the darkest parts of the image tends to shift magenta, as usual with Sony cameras

X-T3 seems like losing about 2/3 stop in the highlights and 1/3 in the shadows compared to A7 III, and the dark parts slightly shifts to green. I would redo this test with the original h265 file, because Resolve dont set the data levels correctly (at least Resolve 14), so if you created this comparison in resolve, maybe you missed the superwhites/blacks in the transcoding

EosR is impressive, it has same information in the shadows as the X-T3 but losing about a stop in the highlights

GH5 has the same highlight information as the X-T3, but losing about a stop in the shadows

The Z7 Flat has the most shadow detail but it losing more then 1,5 stops in the highlight compared to X-T3

Interesting to see how well the 8 bit low bitrate codec's hold's up against 10 bit high bitrate. The reason is probably the fully used dynamic range, in low DR + underexposed scene the difference would be much more visible after color correction.

Edit: in my first post in the shadow nosie comparison the correct camera order from left to right:

GH5 - A7 III - X-T3 - EosR - Z7, just like above

9 minutes ago, Andrew Reid said:

Ah you found the Pink Floyd album Thanks, exactly the kind of grades I'd like to see from original file on Vimeo. Keep em coming!

(Click video title to go and download it)

I like the Kate Bush one!

Z7 has best Kate Bush! That's the N-LOG or Flat shot?

I think that one also brought the most out of the blacks too. Cooke lens clearly visible next to the Floyd LP.

On the sidebyside picture there is the Nikon FLAT. Nikon N-Log +2EV is actually missing from your comparison! The original N-Log is underexposed, i am unable to recover anything from the dark areas. 

EosR - A7 III - X-T3 - GH5s - Z7  (lifted) shadow noise at +2EV:

Thanks for doing this comparison!

3 hours ago, Andrew Reid said:

Pixel peeping is alive and well I see!

Nikon Z7 is fine. It is a miracle in full frame from 8.2K down to 4K. Very nice detail level. Very good dynamic range and colour, which Max's test doesn't even touch on.

Z7 in APS-C is similar but less aliasing / stair stepping - but we are talking 4K here, so at normal viewing distances you don't even see the aliasing in full frame mode, let alone in the oversampled Super 35 mode.

The EOS R is soft. So what. It's 4K. You have more detail than you'll ever need unless cropping 4x into the image.

You wanna know why I don't do tests like this very much anymore?

1. Nobody views your film or music video at 400% crop. The absolute sharpness level in 4K means JACK SHIT. What you want is a soft stable cinematic image - not hard digital sharpness. In fact it's an *advantage* to have a softer image for YouTube, when the player is scaling it down to fit any number of screen resolutions - especially a 1080p screen. It looks more natural when people view it downscaled or even on a 4K TV from normal viewing distances. In the first case the downscaling works badly with a digitally sharper, harder image vs a softer, more cinematic one. In the second case the natural downscaling from the human eye at a longer viewing distance makes a less hard 4K image at 1:1 look more natural and less fatiguing than a "harder" image which shows more emphasised detail. We have plenty enough detail in 4K as it is, even on the EOS R and to overemphasise it, like in Max's video, is a BAD THING.

2. The test by Max claims to be about image quality when he's only testing one small aspect of it and not even very well. He's actually looking 90% at the sharpening levels in the menus, rather than outright performance of the image. All the cameras apply a different level of sharpening to bring out extra detail. You can dial it down or up. So what? How natural and cinematic does that fine detail look to the real viewer? That is the real question.

3. Max's video tests just one aspect of the camera and seems to imply it's 90% of what makes a good image. A wide shot of a building with constantly shifting light at dusk so that not even the lighting conditions are matched on each comparison shot. It says nothing of colour, dynamic range, skin-tones, lenses, sensor size, rolling shutter, motion cadence, codec performance, macro blocking, mud, compression, grading, bit-depth, LOG profile performance and  how easy or not it is to grade. These are the things that determine the final result. These are the important things and not ONE in isolation but ALL together.

Go back and do a proper test Max that takes you longer than half an hour... But no, he's got subscriber numbers and viewers to chase so it must be done quick!

Why the Z7 is singled out for criticism because of Max's test is beyond my understanding.

It justifies the pricing over the A7R III because it is a flat out better camera in every aspect of image quality and handling.

It justifies the pricing over the D850 because it adds video AF which actually works, gets rid of the mirror, fixes the ergonomics in live-view, shaves the pounds off and at the same time maintains the incredible video quality.

He did this test with factory default settings. Low light and dynamic range comparison is coming soon.

If we compare cameras in their "best" settings and match them in post, it would give a lot of room to cheating... and favor one over another.

I dont understand why  you are defending the EosR or the Z7. In one post you hammering the EosR because it's shit, in another post you bless it's glorious image.

Thank you for doing this comparison! Actually GH5/GH5s is better than i expected. 

24 minutes ago, Yannick Willox said:

The positive I take from this vid then is even with awful grading, the P4K looks much better. Your two second grade does help, but it still makes the GH5 look quite bad IMO. She still needs to go and see a physician.

The image from the BMPCC4K is superior to GH5, but not this much. It's awesome if you just put a REC709 LUT to BMD Film how good the results are. But it does not require advanced grading skills to make the GH5 image look good. 

This is my problem with Wofcrow's Alexa comparison too. We can see REC709 LUT's side by side. But what happens if you want to match the GH5 and BM's image to Alexa? Is it possible? (Yes...) how hard is it? . Can the cheap cameras keep up? (in some cases, yes). 

13 minutes ago, Yannick Willox said:

This seems like quite an honest review. All the cons are mentioned, and then you have the image quality & price point.

We already discussed this comparison several pages earlier. This video favors the BMPCC4K in every scenario: poor grade on the GH5 in skintone example, dynamic range comparison while the GH5 is rolling in it's natural PP, BM on gimbal vs GH5 in handheld... It a joke.

15 minutes ago, Slothorp said:

Attached you can find some dng files taken from the test. And in the link below, you can download the same test but shot in prores. 
 

 

 

 

A002_10130258_C022_000225.dng

Blue 400 iso.dng

blue 3200 iso.dng

red 400 iso.dng

red 3200 iso.dng

Thank you very much.

It's a processing error just like with sony footage.

Switch to bmdfilm log and gamut on the raw tab, and add a gamut mapping node to convert it to rec709 with saturation and luma knee. The color will roll-off perfectly fine in all of the examples.

But.

With v3 colorscience (i have resolve 14) the hue is shifted from red to orange. Strange. Maybe v4 is more accurate.

9 minutes ago, Slothorp said:

Later in the evening, yes. 

You can also download a prores file from my previous test here :

 

The prores is already clipped. RAW is cannot be clipped, because gamut is chosen in the post process. And i think this is the same problem what we already saw with sony cameras.

Probably the raw is perfectly fine with proper post processing. It should be. I hope so.