Z9 on test - N-RAW no better than H.265?

[mercer]

On 2/10/2023 at 6:17 AM, Andrew Reid said:

500Mbit in HEVC is like having 1000Mbit in H.264, so that A6500 (which I still have btw) was 100Mbit and still stood up well, so the a1 truly has some thick files going on.

X-H2 ProRes would be an interesting comparison.

I love the idea of RAW but it never seems to compel me to use it that much.

Sigma Fp-L 12bit needs a test though...as does the 5D3 Magic Lantern as that truly was proper RAW!

High bitrate h.265 combined with the high dynamic range of sLog3 does sound like a winning combo.

I took the FP out for another spin yesterday, and I really do like the footage. It handles very similarly to ML Raw while also losing a step in my post process. Even a few years after its release, with the help of Sigma's continued devotion to FW updates, I think it's one of the best buys out there right now, regardless of its limitations.

But again, it is very similar to working with 5D3 files, minus some detail that the 14bit ML files offer and the pure, unencumbered joy of shooting with the 5D3... so I'm still on the fence.

I still prefer RAW to any other compressed codec. It really helps to fix both my mistakes and the inherent obstacles of perfect shooting conditions with run & gun/guerilla filmmaking. I've submitted to the idea that it's going to take a lot of storage space and a longer post-workflow getting the footage into FCPX.

I'd love to see 5D3 ML Raw mixed into your test, pleas do that.

If you do get a chance, I'd love to see some footage from the XH2 ProRes. There really isn't much online. Originally, I preferred the idea of the "s" model, but after more thinking, the XH2 offers a very unique feature set... plus I didn't realize that the XH2 also shoots internal ProRes... which is huge for me and would really delineate the purpose of a second camera.

[Caleb Genheimer]

4 hours ago, TomTheDP said:

I am not saying it's an issue, just a reality. 12 bit linear RAW gives a similar amount of information as 10 bit log. This is why I assume people aren't seeing a difference when shooting linear 12 bit over the internal 10 bit on the S1H. Bitrate could still be an issue as 150-200mbps these H265/H264 codecs are doing isn't a lot of data to work with. 

 The Fuji XH2s does 14 bit ADC when shooting 4k up to 30fps. I am not quite sure on Blackmagic cameras. The URSA 4.6K had a similar dual gain sensor as the Alexa but I don't know if the color depth was the same. 

Atomos claims this about their recorders with the A7S3/FX3
"The Ninja V captures tremendous detail with 12-bit log files with over 68 billion color graduations! This is converted from the camera’s 16-bit linear RAW output. This process has the advantage of being visually indistinguishable from 16-bit Linear recording, while still having practical data rates."

Obviously the good old 5D MK3 does 14 bit with magic lantern, if that counts. 

ARRI does this with their cameras, they package 14 bit into a 12 bit log as there is really no benefit to 14 bit linear over 12 bit log, at least according to them. 

 

Yep sounds like we’re on the same page. Not 100% on the Pocket cams, but if they’re 14-bit, they’re doing something wrong, because it performs just like 12-bit (pretty sure that’s what it is though). 
 

The XH2s is indeed one of the outliers. Hopefully that means 14-bit is coming to full frame soon. It makes sense Fuji would do it in their S35 cam, as the have a FF “gap” in their lineup (APSC straight to Medium Format). 

[MrSMW]

On 2/11/2023 at 6:09 AM, newfoundmass said:

With all the hubbub over raw I do think we're pretty close to it being irrelevant for all but major productions

This is my ‘issue’ with raw to date…

A: How much faff and storage and work would I need to do to use it?

B: Do I even ‘need’ to?

The answer is right now, not really.

But do I want a more ‘cinematic’ (whatever that is) image?

Yes. Not so much for my clients but more for my own craft pleasure.

So intrigued by raw, but will only use it if there is an actual benefit and it’s in camera and easy to use.

[Llaasseerr]

16 hours ago, Andrew Reid said:

That is because ProRes 422 HQ in LOG looks just as nice, with smaller file?

If you're just playing the footage back then it's hard to see the loss in chroma information in a typical compressed 4:2:2 file, except maybe on very saturated primary colours where you see the steps on edges. But if you're doing heavy post manipulation then it's much more obvious that it's lacking.

If you change the colour space to YUV  (YCbCr) and look at the chroma channels (Y being luminance) on something like internal XAVC 4:2:2 Sony footage, there is practically nothing there. It's so blocky and compressed and undetailed. 4:2:0 is obviously a lot worse. That's the difference. But for most people with a camera that costs a few thousand dollars, they aren't going to be bothered by that.

[gethin]

The sensor in the z9 is designed for speed. I've found Nikon flagship D4/5 cameras lacking in dynamic range.  It was the first thing I googled when it came out: what the video dynamic range was like, and from everything I could find it wasn't great. And yep the odd video sample with gnarly highlight rolloff. Say that itqs dynamic range is limited in a Nikon group and people go slightly nuts 😆.  I'm a Nikon shooter, im used to waiting: so I'm waiting for the z6iii or z8, and Nikon hopefully getting back to the 14 stops of the d8xx's

[TomTheDP]

12 hours ago, Caleb Genheimer said:

Yep sounds like we’re on the same page. Not 100% on the Pocket cams, but if they’re 14-bit, they’re doing something wrong, because it performs just like 12-bit (pretty sure that’s what it is though). 
 

The XH2s is indeed one of the outliers. Hopefully that means 14-bit is coming to full frame soon. It makes sense Fuji would do it in their S35 cam, as the have a FF “gap” in their lineup (APSC straight to Medium Format). 

Yes definitely. With Blackmagic I believe the Fairchild sensors were dual gain but not the newer pockets or the Ursa 12k. 

So much goes into cameras its hard to determine what is actually advantageous. Recently I used both the Sigma FP and ARRI Alexa Classic on a feature film. It was a good opportunity to put both cameras through their paces in a variety of scenarios. During production I began to question my decision to use the Alexa classic due to its size. However now that I just finished coloring the film I can say the Alexa was so much easier to work with in post. It just looks nice without any work needing to be done. It is also so easy to adjust. We only shot in prores 422 sometimes at 1600 iso for low light scenes. It still looks amazing. 

I worked on a short film where we shot on the Z-cam S6 in 4k prores 422 and Panasonic S1 in the 10 bit 4k codec.  We came back to reshoot a scene and used the Alexa. Again the Alexa felt so effortless in post. I am sure my post workflow for the Panasonic could be better. But I can't argue with my eyes. Probably not a fair to compare a once 80,000 Hollywood camera to a prosumer $2000 camera though. 

Even when shooting in 422 it's still doing a 14 bit readout. But I think the color science and processing is almost just as important. I know that with the right post workflows most cameras can look identical these days when shooting 10 bit log or RAW. But the simplicity of cinema cameras has really drawn me away from prosumer gear, seeing as you can get a lot of older RED's and Alexas for reasonable ish prices these days. 

[androidlad]

13 hours ago, Caleb Genheimer said:

Yep sounds like we’re on the same page. Not 100% on the Pocket cams, but if they’re 14-bit, they’re doing something wrong, because it performs just like 12-bit (pretty sure that’s what it is though). 
 

The XH2s is indeed one of the outliers. Hopefully that means 14-bit is coming to full frame soon. It makes sense Fuji would do it in their S35 cam, as the have a FF “gap” in their lineup (APSC straight to Medium Format). 

14bit readout on X-H2s absolutely does not offer any tangible benefits. The sensor design cannot take adavantage of higher bit-depth readout as the extra bits are really only there to quantise noise.

 

[Davide DB]

Couldn't it all be much simpler (Occam's razor)? Nikon's Raw implementation is crap.

The test would need the raw of a Red or Canon.

[TomTheDP]

On 2/12/2023 at 9:45 AM, androidlad said:

14bit readout on X-H2s absolutely does not offer any tangible benefits. The sensor design cannot take adavantage of higher bit-depth readout as the extra bits are really only there to quantise noise.

 

The 14 bit ADC mode seems to increase dynamic range over the 12 bit ADC. The entire patch range went up to 15 stops. 

[BabsDoProd]

That's really strange that the H.265 of the A1 is coming out ahead of Z9's NRAW! Now that internal recording has been tested, what about testing the external RAW of the A1 against NRAW and the external 8K of the Z9 to see how that measures up? Also, despite the issues present, how do the clips hold up in post? Perhaps there's some attribute to NRAW that can be fixed in post correction? Lastly, could there have been any kind of setting on the Z9 that could have altered these results?

[ND64]

You make it look more complicated than what its. Z9 sensor is noisier than A1 sensor, and using 12bit ADC to read out a noisy sensor doesn't give the best image quality. Nikon use the same raw codec for high efficiency raw still images, and photographers have no problem with that. BTW, raw recording benefit is not just correcting huge exposure mistakes in post. First, log profiled compressed images gonna suffer from banding. Its inevitable. Second, raw preserves the bayer pattern, so there will be freedom to use state of the art demosaicing algorithms in future, which are getting better and better with AI. Third, who doesn't hate baked-in white balance? 

[TomTheDP]

1 hour ago, Eric Calabros said:

You make it look more complicated than what its. Z9 sensor is noisier than A1 sensor, and using 12bit ADC to read out a noisy sensor doesn't give the best image quality. Nikon use the same raw codec for high efficiency raw still images, and photographers have no problem with that. BTW, raw recording benefit is not just correcting huge exposure mistakes in post. First, log profiled compressed images gonna suffer from banding. Its inevitable. Second, raw preserves the bayer pattern, so there will be freedom to use state of the art demosaicing algorithms in future, which are getting better and better with AI. Third, who doesn't hate baked-in white balance? 

I assume the stills are 14 bit which makes a big difference, at least from my experience comparing stills to video RAW on most cameras. 

But like you said it could be as simple as just an inferior sensor. I would say the A1 is likely doing NR but it sounds like the underexposed areas look very organic. 

[D Verco]

On 2/13/2023 at 1:45 AM, androidlad said:

14bit readout on X-H2s absolutely does not offer any tangible benefits. The sensor design cannot take adavantage of higher bit-depth readout as the extra bits are really only there to quantise noise.

 

Either way, the xh2s sensor is probably the best on the market under 8k

[deezid]

8 hours ago, D Verco said:

Either way, the xh2s sensor is probably the best on the market under 8k

While the image processing on this camera even falls behind Sony...
Also 16 stops of patch range are quite common, even with 12 bit ADC.

https://www.cined.com/panasonic-lumix-s5-lab-test-rolling-shutter-dynamic-range-and-latitude/
https://www.cined.com/blackmagic-pocket-cinema-camera-6k-lab-test-dynamic-range-latitude-rolling-shutter-more/
 

[Shooper]

I shot a feature last year on the S1 and S1H in ProRes Raw and absolutely loved the results.  Towards the end of last year, though, I considered jumping to the Nikon Z9, not immediately, but when I needed to as it wasn't clear the S line would have a future.  Luckily the S line does have a future and I have not been able to find an Nraw sample that makes it look good.  Curious to see the ProRes Raw from the cam, but bummed its not fully resolution.

Thanks for doing these tests.

[Kino]

All this proves is that NRAW may not be optimized yet in N-LOG gamma on Resolve, as other users have reported much greater DR using different gamma profiles:

It is a very new raw format and not user friendly just yet. Your article also fails to take into account the impressive color depth of NEV files, which Resolve lists as 14 bit RAW. Is Nikon secretly giving us 14 bit RAW? All I can say is that the NRAW files display rich color and tonality that I've rarely seen on other mirrorless cams shooting 12 bit RAW:

It will not replace a $30K RED V-RAPTOR, but it does amazingly well up against RED Helium:

If you follow this channel, you will notice that they switched from using the A1 to the Z9 when the latter received the RAW update.

 

 

 

 

 

[PannySVHS]

I have not seen anything convincing on the web regarding better image quality than a S1 or other high end mirrorless. Curious to see some other work which really tests this camera.

[Kino]

 

[Fairkid]

I tried the Z9 on a few jobs but didn't keep it in the end. The colours are beautiful and I loved the ergonomics and battery life but as you mentioned the workflow wasn't ideal. I just couldn't get decent dynamic range out of it using the supplied Nikon conversion LUTs. As it's a bit of a niche camera there wasn't much info out there on how to get the best out of it. 

[ND64]

20 hours ago, Kino said:

 

I liked the DR in this shot