Thank you for a very comprehensive insight. As an electronics person, I suspect the new partially stacked topology is introducing some electrical noise on the sensor output, if it is the case, it will be interesting to see how Nikon resolves this. Of course they can cheat and do some noise-correction in the FW, but this will impact other areas then. We will most likely see this issue disappear as they improve on the sensor design going forward, maybe play around with the application and location of the partial stacked layout. Just speculation on my side though, take it with a grain of salt. I had to chuckle when you mentioned the variance in the Z8 sensor in your last test, when I saw that, I immediately thought it was thermally induced, and then you mentioned it. Haha, funny moment right there. Anyway, much value your efforts in doing these tests, will probably guide peeps to hold off on investing as soon as new technology drops. I was considering grabbing a z6iii to partner with my Z9 and Zf bodies, but after careful consideration and due to a special deal I got on a new Z8, I got that instead. Truth be told though, my d850 still produces better stills than my z9. Not sure if it is my imagination, I might need to run some tests as well. But the difference isn’t big.
24 vs 45 MP, and 24 is flickering in +3 eV. 😢 If they can't solve this issue immediately, it will be a shame for Nikon. People have waited years for Z6iii with competent Af and tech update, finally when they released it with video focused features and this is happening.
Is it fair to guess this is electrical noise in either the sensor or the support circuits and that there won’t be a feasible adjustment Nikon can make via firmware because the noise is so ridiculously varied? But I suppose the more important question is whether this is a manufacturing defect which can at least be fixed at the factory level and better behaved cameras delivered in the near term or whether there’s a more fundamental design issue which will require many months to overcome. In the mean time, I can’t build my video work around such poorly performing cameras. Bummer. I was planning to work the Z6III really hard in the months ahead. Very frustrating.
...and i was about to switch from my z8 to z6iii because I considered the z8 video too noisy... I really expected z6iii would do better. Is the issue fixable with a firmware update? Thank you for your effort 👍
Fantastic work. Thanks. At 6:14, the black frame video of the Z8 seems to show vertical banding, while the Z6III doesn't have any discernable pattern. I think this matches what we see in the scatter plots. We'll see if there is some firmware fix from Nikon, and how they compare after that.
The banding you see in the blackframe video is just an artifact of the fast-rendering logic of the TicoRAW library - it doesn't show up in the full-quality rendered footage. If you watch Adobe ACR/LR closely you can sometimes see that same grid pattern during the interim renders of blackframe NEFs.
How did you push the exposure? In my experience with z8 the best results I can get best shadows (not green) when I set raw tab to linear gamma rec2020, and then convert that to davinci wide gamut and intermediate gamma then then back to rec709 gamma 2.4. Exposure and shadow boost works best when applied in raw tab in resolve. Exposure boost while in wide gammut mode (between two color transform nodes) does look inferior to my eye. Are you able to upload both clips?
Based on this, it seems when things are exposed correctly, this flickering isn’t an issue. Pushing things too far to recover shadow detail, seems like it will also end up making the entire footage unusable as the brightness levels will also be increased. While the shadow area is the point of focus, it would be nice to see the result of the entire image with these settings applied.
@@testcams Hmmm… seems like with this sensor we may have to do the Sony way of overexposing SLog3 by 1.3-2.0 stops during recording and reducing brightness in post, to reduce shadow noise, then apply noise reduction if needed. This new sensor seems like it will certainly require a different way of working. This reminds me of when the FX30 launched, videos were out saying Sony lied about Cine EI and having low shadow noise. Then Sony put out a video on how to properly expose when using Cine EI but, the controversy continued until people found the video, followed the directions, then it became a recommended camera for budding filmmakers. Like the FX30, the Z6 III may not be to everyone’s taste but, still performs well overall. I have the FX30 and use SLog3 and Cine EI at times where I feel I may encounter shadow noise, I’m curious to see how applying this method to the Z6 III works once I get mine. I think most people buying this camera will stick with SDR and those that want more control will chose NLog or shoot RAW. These test provide some things to keep in mind when shooting. Let’s see what Nikon has to say about this.
@@truthkeeperfilms That can be cleaned up in post by applying some noise reduction, if you end up there or you can shoot at 6400, stop the lens down or increase the shutter speed to compensate, either will do. I’ve used my FX30 set at ISO 2500 (not Cine EI mode) for nighttime street scenes and cleaned up the noise (granted Sony has some NR applied in camera) in post. I like capturing the scene close to what it is, I’ll have to see how the Z6 III does once I get my hand on it.
Hopefully this is able to be resolved through an update. I’m sure as long as you expose the image correctly you should be fine either way, until Nikon finds a solution of course.
Any preliminary thoughts as to real-time correction for the flickering issue in the z6iii? Wonder if it’s addressable via firmware, or if only in post?
It's a waiting game at this point. I know lots of Z6 III users from many forums and FB groups have contacted Nikon about this issue and Nikon will have to say/do something about it soon
I had multiple shots of a statue change focus after I used manual focus with focus peaking turned on. I watched the focus pulse out of focus in the viewfinder. I had the AF-assist illuminator turned off and the issue seemed to go away once I turned it back on. I had the same result if I tooggled the switch to manual on the 24-70 F4 lense.
Well it is the first partially stacked sensor i guess we are starting to see why? I wasn't going to buy one personallu i already own a Z8/ZF combo. Its a downer for videographers that were looking forward to this i guess or not if you have a highend PC and you dont need 5+ stops of exposure there is the declicker and AI stuff that can correct your footage.
A few months ago I started a thread on the DPR Science forum with an open question - why don't any of the stacked sensors on the market use their fast readouts in their video modes? The theory I proposed was that it was due to power or thermal considerations. Someone in that thread posted a reference to a seemingly-leaked engineering datasheet for the stacked sensor in the Sony A9, which documented that its stacked memory was disabled for all video modes due to overheating. Heat in sensors doesn't just cause overheating but has noise implications as well. Here's the link: www.dpreview.com/forums/post/67470284
@@testcams So if Nikon can do a firmware update that reduces the Z6III's faster readout speed in video modes it will reduce the noise across the board? I guess if that isn't baked into the sensor it makes sense.
@@MB-dq2gz - Slowing down readout negates the whole purpose of the PS-Sensor no? Look at the jello effect of the Sony A7IV, absolute nightmare BUT, the DR is insane picking up almost 15-stops of DR with a good median of 12.5-12.8 usable
@@truthkeeperfilms not exactly as it stands it is faster than most of the competition Z8/z9 and quite a few Sony and Canon cameras. Bringing it down the Z8 level should be more than acceptable.
I actually did an experiment this morning on the Z8, comparing the DR of 14-bit raw vs 12-bit NRAW vs JPG vs C60 JPG, the last of which uses the slower sensor readout path of the video mode. You can see my full Z8 sensor readout measurements here: horshack-dpreview.github.io/RollingShutter/#Nikon%20Z8 The C60 was no cleaner than the single-shot JPG, even though C60 uses a 1/90 readout speed vs 1/268 for the single-shot JPG. Here's an animated PNG I created which demonstrates this: photos.smugmug.com/photos/i-c7fb57v/0/LSdq8gCZ5n9BHnjDLC6dMtQ6zmVQv9Q5qsxMN6P4c/O/i-c7fb57v.png
This would make sense as I've seen this in my tests on my Z9 - when swapping ISO there's a brief moment of flashing and colour changes before it settles down and becomes really consistent.
Do you mean when Gerald said the Z8's noise on the waveform flashes for a second when switching to the dual gain ISO? If so that' unrelated to the luminance shift I posted about.
wow, shocking. I had such hopes for this camera, specifically for better dynamic range and nicer highlights, better low light noise because of lower res sensor, than the z8
Do you have a *source* for this? "at ISO800 which runs the sensor at ISO100." If HLG base ISO400 then wouldn't that be ISO64, again I'm just guessing here, but that would make NLOG Base ISO800 equal to ISO125.
The source is me :) HLG is +2EV, so ISO 100 as well. You can see my original DPR thread of the subject here: www.dpreview.com/forums/post/67101863 I'm working on a thorough N-Log deep-dive video that will demonstrate this and lots of other technical details not available elsewhere. In the meantime the easiest way to demonstrate this is to see which N-Log ISO the dual gain ISO kicks in, ie the shadows get markedly cleaner. For example, on the Z6 III N-Log shadows get cleaner at ISO 6400, which is +3EV above its sensor's actual dual gain ISO of 800, as seen in stills. For the Z8, the N-Log shadows clean up at ISO 4000, which is +3EV of its true dual gain ISO of 500. Another way to demonstrate is to watch the video waveform switching between N-Log ISO 800 vs SDR ISO 100, without changing the aperture/shutter-speed - the highlights don't move.
@@testcams Good call. They're long overdue for N-Log2 and hopefully they can take a page from Sony and favor the shadows at the native ISOs rather than the usual 50/50 split. Insane low-light performance is one of the biggest selling points for choosing a mirrorless camera and we know they're using the same Sony sensor.
I like the flickering. Its so cinematic
Thank you for a very comprehensive insight. As an electronics person, I suspect the new partially stacked topology is introducing some electrical noise on the sensor output, if it is the case, it will be interesting to see how Nikon resolves this. Of course they can cheat and do some noise-correction in the FW, but this will impact other areas then.
We will most likely see this issue disappear as they improve on the sensor design going forward, maybe play around with the application and location of the partial stacked layout. Just speculation on my side though, take it with a grain of salt.
I had to chuckle when you mentioned the variance in the Z8 sensor in your last test, when I saw that, I immediately thought it was thermally induced, and then you mentioned it. Haha, funny moment right there. Anyway, much value your efforts in doing these tests, will probably guide peeps to hold off on investing as soon as new technology drops.
I was considering grabbing a z6iii to partner with my Z9 and Zf bodies, but after careful consideration and due to a special deal I got on a new Z8, I got that instead. Truth be told though, my d850 still produces better stills than my z9. Not sure if it is my imagination, I might need to run some tests as well. But the difference isn’t big.
You are the man. Thanks for your work! That is illuminating 😂. You got me subscribed with the scatter graph in the end.
The Z8 distribution chart is so interesting, thanks for sharing
You’re a genius!!.. looking forward to see better work around for this issue
24 vs 45 MP, and 24 is flickering in +3 eV. 😢
If they can't solve this issue immediately, it will be a shame for Nikon. People have waited years for Z6iii with competent Af and tech update, finally when they released it with video focused features and this is happening.
Excellent work here man!
Is it fair to guess this is electrical noise in either the sensor or the support circuits and that there won’t be a feasible adjustment Nikon can make via firmware because the noise is so ridiculously varied? But I suppose the more important question is whether this is a manufacturing defect which can at least be fixed at the factory level and better behaved cameras delivered in the near term or whether there’s a more fundamental design issue which will require many months to overcome. In the mean time, I can’t build my video work around such poorly performing cameras. Bummer. I was planning to work the Z6III really hard in the months ahead. Very frustrating.
...and i was about to switch from my z8 to z6iii because I considered the z8 video too noisy...
I really expected z6iii would do better.
Is the issue fixable with a firmware update?
Thank you for your effort 👍
Fantastic work. Thanks. At 6:14, the black frame video of the Z8 seems to show vertical banding, while the Z6III doesn't have any discernable pattern. I think this matches what we see in the scatter plots.
We'll see if there is some firmware fix from Nikon, and how they compare after that.
The banding you see in the blackframe video is just an artifact of the fast-rendering logic of the TicoRAW library - it doesn't show up in the full-quality rendered footage. If you watch Adobe ACR/LR closely you can sometimes see that same grid pattern during the interim renders of blackframe NEFs.
@@testcams Thanks
How did you push the exposure? In my experience with z8 the best results I can get best shadows (not green) when I set raw tab to linear gamma rec2020, and then convert that to davinci wide gamut and intermediate gamma then then back to rec709 gamma 2.4. Exposure and shadow boost works best when applied in raw tab in resolve. Exposure boost while in wide gammut mode (between two color transform nodes) does look inferior to my eye.
Are you able to upload both clips?
Exposure was pushed in the raw tab. This is done in linear space before the CST.
Based on this, it seems when things are exposed correctly, this flickering isn’t an issue. Pushing things too far to recover shadow detail, seems like it will also end up making the entire footage unusable as the brightness levels will also be increased. While the shadow area is the point of focus, it would be nice to see the result of the entire image with these settings applied.
The scene was exposed to within 1/3EV of highlight clipping. I don't consider a +2EV shadow adjustment representative of pushing things too far.
@@testcams Hmmm… seems like with this sensor we may have to do the Sony way of overexposing SLog3 by 1.3-2.0 stops during recording and reducing brightness in post, to reduce shadow noise, then apply noise reduction if needed. This new sensor seems like it will certainly require a different way of working. This reminds me of when the FX30 launched, videos were out saying Sony lied about Cine EI and having low shadow noise. Then Sony put out a video on how to properly expose when using Cine EI but, the controversy continued until people found the video, followed the directions, then it became a recommended camera for budding filmmakers. Like the FX30, the Z6 III may not be to everyone’s taste but, still performs well overall.
I have the FX30 and use SLog3 and Cine EI at times where I feel I may encounter shadow noise, I’m curious to see how applying this method to the Z6 III works once I get mine. I think most people buying this camera will stick with SDR and those that want more control will chose NLog or shoot RAW. These test provide some things to keep in mind when shooting. Let’s see what Nikon has to say about this.
@@waynerm002 - The worst ISO on the Z6 III is above 3200-5000. You'll see it even if you properly expose. At 6400, it cleans up nicely
@@truthkeeperfilms That can be cleaned up in post by applying some noise reduction, if you end up there or you can shoot at 6400, stop the lens down or increase the shutter speed to compensate, either will do. I’ve used my FX30 set at ISO 2500 (not Cine EI mode) for nighttime street scenes and cleaned up the noise (granted Sony has some NR applied in camera) in post. I like capturing the scene close to what it is, I’ll have to see how the Z6 III does once I get my hand on it.
Hopefully this is able to be resolved through an update. I’m sure as long as you expose the image correctly you should be fine either way, until Nikon finds a solution of course.
Any preliminary thoughts as to real-time correction for the flickering issue in the z6iii? Wonder if it’s addressable via firmware, or if only in post?
It's a waiting game at this point. I know lots of Z6 III users from many forums and FB groups have contacted Nikon about this issue and Nikon will have to say/do something about it soon
I had multiple shots of a statue change focus after I used manual focus with focus peaking turned on. I watched the focus pulse out of focus in the viewfinder. I had the AF-assist illuminator turned off and the issue seemed to go away once I turned it back on. I had the same result if I tooggled the switch to manual on the 24-70 F4 lense.
Wow thank you!
Well it is the first partially stacked sensor i guess we are starting to see why? I wasn't going to buy one personallu i already own a Z8/ZF combo. Its a downer for videographers that were looking forward to this i guess or not if you have a highend PC and you dont need 5+ stops of exposure there is the declicker and AI stuff that can correct your footage.
A few months ago I started a thread on the DPR Science forum with an open question - why don't any of the stacked sensors on the market use their fast readouts in their video modes? The theory I proposed was that it was due to power or thermal considerations. Someone in that thread posted a reference to a seemingly-leaked engineering datasheet for the stacked sensor in the Sony A9, which documented that its stacked memory was disabled for all video modes due to overheating. Heat in sensors doesn't just cause overheating but has noise implications as well. Here's the link: www.dpreview.com/forums/post/67470284
@@testcams So if Nikon can do a firmware update that reduces the Z6III's faster readout speed in video modes it will reduce the noise across the board? I guess if that isn't baked into the sensor it makes sense.
@@MB-dq2gz - Slowing down readout negates the whole purpose of the PS-Sensor no? Look at the jello effect of the Sony A7IV, absolute nightmare BUT, the DR is insane picking up almost 15-stops of DR with a good median of 12.5-12.8 usable
@@truthkeeperfilms not exactly as it stands it is faster than most of the competition Z8/z9 and quite a few Sony and Canon cameras. Bringing it down the Z8 level should be more than acceptable.
I actually did an experiment this morning on the Z8, comparing the DR of 14-bit raw vs 12-bit NRAW vs JPG vs C60 JPG, the last of which uses the slower sensor readout path of the video mode. You can see my full Z8 sensor readout measurements here: horshack-dpreview.github.io/RollingShutter/#Nikon%20Z8 The C60 was no cleaner than the single-shot JPG, even though C60 uses a 1/90 readout speed vs 1/268 for the single-shot JPG. Here's an animated PNG I created which demonstrates this: photos.smugmug.com/photos/i-c7fb57v/0/LSdq8gCZ5n9BHnjDLC6dMtQ6zmVQv9Q5qsxMN6P4c/O/i-c7fb57v.png
8:35 do you think this is what Gerald Undone spoke about that the Z8 needs few seconds to clean up? He posted a video of something similar
This would make sense as I've seen this in my tests on my Z9 - when swapping ISO there's a brief moment of flashing and colour changes before it settles down and becomes really consistent.
Do you mean when Gerald said the Z8's noise on the waveform flashes for a second when switching to the dual gain ISO? If so that' unrelated to the luminance shift I posted about.
wow, shocking. I had such hopes for this camera, specifically for better dynamic range and nicer highlights, better low light noise because of lower res sensor, than the z8
That was already there on my first gen Z6 … disappointing :( but I want to get the Z8 anyway
Jesus. An entire matlab analyser! You need a few $$$ for your effort!
Do you have a *source* for this? "at ISO800 which runs the sensor at ISO100." If HLG base ISO400 then wouldn't that be ISO64, again I'm just guessing here, but that would make NLOG Base ISO800 equal to ISO125.
The source is me :) HLG is +2EV, so ISO 100 as well. You can see my original DPR thread of the subject here: www.dpreview.com/forums/post/67101863
I'm working on a thorough N-Log deep-dive video that will demonstrate this and lots of other technical details not available elsewhere. In the meantime the easiest way to demonstrate this is to see which N-Log ISO the dual gain ISO kicks in, ie the shadows get markedly cleaner. For example, on the Z6 III N-Log shadows get cleaner at ISO 6400, which is +3EV above its sensor's actual dual gain ISO of 800, as seen in stills. For the Z8, the N-Log shadows clean up at ISO 4000, which is +3EV of its true dual gain ISO of 500. Another way to demonstrate is to watch the video waveform switching between N-Log ISO 800 vs SDR ISO 100, without changing the aperture/shutter-speed - the highlights don't move.
@@testcams Good call. They're long overdue for N-Log2 and hopefully they can take a page from Sony and favor the shadows at the native ISOs rather than the usual 50/50 split. Insane low-light performance is one of the biggest selling points for choosing a mirrorless camera and we know they're using the same Sony sensor.
Very interesting - I would have expected the Z6 to do better ... 🦘