My Z6 III tech thread

Quick demonstration of PRNU for an upcoming video about CMOS image sensor noise sources I have planned.

Animation: Z6 III PRNU Demonstration

The shot-to-shot blacklevel variation issue (stills) I discussed last month was originally thought to be just a curiosity without real-world implications. Since that time it was discovered this same issue causes flickering in video mode - here are a few videos I've done about the video flickering issue:

• Nikon Z6 III Video Flickering issue (with workaround)
• Nikon Z6 III vs Z8: N-RAW Video Noise Comparison
• Nikon Z6 III Video Flickering: Not just extreme conditions

Today I did a deep dive comparison of two Z6 III blackframe raw images, one which I've labeled "typical" and another "bright" - these are just relative terms for these two specific raws that are examples of the shot-to-shot blacklevel variation issue.

Here is the analysis showing the raw histogram differences between the two raws. I identified the specific ADU for this specific instance that causes the "bright" raw to be rendered with noticeably higher luminance than the "typical" raw. To verify this I used my OctaveRawTools framework to remove the specific ADU in the bright raw (changed all pixels at ADU 1,011 to the blacklevel value of 1,008), after which its luminance of the "bright" raw normalized to the "typical" raw.

Composite Image: Z6 III Blackframe analysis of blacklevel issue: Two raws analyzed and compared

Does this mean that image bursts will have inconsistent luminance levels and applying the same adjustment in post-processing will yield varying results?

MJ_Photo38 said:

Counting all the camera I own (including compact cameras and 20yo DSLRs) I have 13 cameras...

... pretty sure my girlfriend would like to have a word or two on "how much is too many" !

Click to expand...

Well that sounds like a good start. I know I'm well past that, but I actively shoot, probably, 8 or 10 including film and digital.

Today I compared the variation of 20 NEF blackframes between the original Z6 and Z6 III. Here are the results:

Spreadhseet: Z6 vs Z6 III shot-to-shot variation comparison

Summary:

• Red channel: Z6 III has 3.5x more variation in mean vs Z6
• Green_1 channel: Z6 III has 52x more variation in mean vs Z6
• Green_2 channel: Z6 III has 95x more variation in mean vs Z6
• Blue channel: Z6 has 33x more variation in mean vs Z6

Here's what this variation looks like for the processed raws (ACR). A Gausian 5.0 blur was applied to isolate luminance shits in visualization.

Animation: Z6 vs Z6 III shot-to-shot blacklevel variation @ 20fps, Blur applied (5MB)

Looks like Z6 III may not be the best choice for astro, especially when stacking multiple frames. Also, pretty sure you meant 'isolate luminance shifts in visualization". :-D

briggart said:

Looks like Z6 III may not be the best choice for astro, especially when stacking multiple frames. Also, pretty sure you meant 'isolate luminance shifts in visualization". :-D

Click to expand...

I giggled at that typo as well :-D

Horshack said:

Today I compared the variation of 20 NEF blackframes between the original Z6 and Z6 III. Here are the results:

Spreadhseet: Z6 vs Z6 III shot-to-shot variation comparison

Summary:

• Red channel: Z6 III has 3.5x more variation in mean vs Z6
• Green_1 channel: Z6 III has 52x more variation in mean vs Z6
• Green_2 channel: Z6 III has 95x more variation in mean vs Z6
• Blue channel: Z6 has 33x more variation in mean vs Z6

Here's what this variation looks like for the processed raws (ACR). A Gausian 5.0 blur was applied to isolate luminance shits in visualization.

Animation: Z6 vs Z6 III shot-to-shot blacklevel variation @ 20fps, Blur applied (5MB)

Click to expand...

The spreadsheet has coefficient of variation (COV) calculations for the channel mean variation but those are N/A since COV can only be applied to positive values. The stdev for those don't need to be scaled to the mean for a COV anyway since they're all in the same units, so the stdev itself can be compared. Here's the corrected sheet and summary stats:

Spreadhseet: Z6 vs Z6 III shot-to-shot variation comparison

Summary:

• Red channel: Z6 III has 29.38x more variation in mean vs Z6
• Green_1 channel: Z6 III has 34.99x more variation in mean vs Z6
• Green_2 channel: Z6 III has 32.61x more variation in mean vs Z6
• Blue channel: Z6 has 27.31x more variation in mean vs Z6

Mr Bolton said:

briggart said:

Looks like Z6 III may not be the best choice for astro, especially when stacking multiple frames. Also, pretty sure you meant 'isolate luminance shifts in visualization". :-D

Click to expand...

I giggled at that typo as well :-D

Click to expand...

What typo?

It seems obvious by now that the Z6iii sensor has different strengths than the Z6 non-stacked sensor. How does the Z6iii sensor "stack" up against other stacked sensors. IOW, is this stuff better controlled or non-existent in other stacked sensors? Also, because of the resolution difference, can the results be directly compared- thinking of the issues when people try to compare the noise between the Z6 and Z7.

--
Ruby
(If you can't see my posts it's because I often say things that get them deleted!)

Horshack said:

Horshack said:

I think just releasing and re-engaging AF would be enough (At least it has been for me on the Zf and previous Z bodies when this happens). In fact, putting your hand over the lens might cause the camera to completely miss focus or focus on the wrong thing because it has to "start over" in some respects, whereas a quick release and press might not...

Click to expand...

I know you've written before you've seen this issue on other Z bodies (not just the Zf) but I have seen no evidence of that. I've tried reproducing it myself on several other models and it's only occurred on the Zf and now the Z6 III.

Click to expand...

So, based on a Reddit thread about the issue, I got to testing and I can confirm that the Z8 and Z9 also have this issue.

I don't think any EXPEED 6 camera has this issue though. The live view stream seems to always stay at 60 fps in AF-C regardless of the shutter speed or light level.

Horshack said:

Horshack said:

Does increasing the EVF refresh rate to say 120 fps help any?

Click to expand...

Haven't tried yet but I doubt it.

Click to expand...

Unfortunately it does not help.

Ruby Rod said:

It seems obvious by now that the Z6iii sensor has different strengths than the Z6 non-stacked sensor. How does the Z6iii sensor "stack" up against other stacked sensors. IOW, is this stuff better controlled or non-existent in other stacked sensors? Also, because of the resolution difference, can the results be directly compared- thinking of the issues when people try to compare the noise between the Z6 and Z7.

Click to expand...

Not great, primarily because of the "partially stacked" design, which I think is also FSI not BSI. Whereas I think I understand that the fully stacked designs in use today are generally BSI. Compare the DR and other metrics, read noise, etc, of the Z6iii vs R5 ii or Z8. Generally it's over a stop penalty at ISO 100, but more overall the noise seems generally high,

On a different topic, I would love to see an update from @Horshack regarding the flickering and changing black level issue on a post 1.10 FW camera. I was pretty shocked to see how bad this reflected in footage I captured the other day, both in HLG and N-LOG.

patrol_taking_9j said:

Ruby Rod said:

It seems obvious by now that the Z6iii sensor has different strengths than the Z6 non-stacked sensor. How does the Z6iii sensor "stack" up against other stacked sensors. IOW, is this stuff better controlled or non-existent in other stacked sensors? Also, because of the resolution difference, can the results be directly compared- thinking of the issues when people try to compare the noise between the Z6 and Z7.

Click to expand...

Not great, primarily because of the "partially stacked" design, which I think is also FSI not BSI. Whereas I think I understand that the fully stacked designs in use today are generally BSI. Compare the DR and other metrics, read noise, etc, of the Z6iii vs R5 ii or Z8. Generally it's over a stop penalty at ISO 100, but more overall the noise seems generally high,

On a different topic, I would love to see an update from @Horshack regarding the flickering and changing black level issue on a post 1.10 FW camera. I was pretty shocked to see how bad this reflected in footage I captured the other day, both in HLG and N-LOG.

Click to expand...

Here you go:

Horshack said:

patrol_taking_9j said:

Ruby Rod said:

It seems obvious by now that the Z6iii sensor has different strengths than the Z6 non-stacked sensor. How does the Z6iii sensor "stack" up against other stacked sensors. IOW, is this stuff better controlled or non-existent in other stacked sensors? Also, because of the resolution difference, can the results be directly compared- thinking of the issues when people try to compare the noise between the Z6 and Z7.

Click to expand...

Not great, primarily because of the "partially stacked" design, which I think is also FSI not BSI. Whereas I think I understand that the fully stacked designs in use today are generally BSI. Compare the DR and other metrics, read noise, etc, of the Z6iii vs R5 ii or Z8. Generally it's over a stop penalty at ISO 100, but more overall the noise seems generally high,

Click to expand...

Click to expand...

It wouldn't make sense to go to all the trouble of partial stacking, on a slow, old FSI sensor.

Horshack said:

Horshack said:

On a different topic, I would love to see an update from @Horshack regarding the flickering and changing black level issue on a post 1.10 FW camera. I was pretty shocked to see how bad this reflected in footage I captured the other day, both in HLG and N-LOG.

Click to expand...

Here you go:

https://www.dpreview.com/forums/post/68161481

Click to expand...

Yes, I stand corrected. The reporting online is inconsistent. Did also find an interesting tidbit below, posted I believe by Sony engineer in Taiwan.

Mr Bolton said:

IMX410E/IMX820 adopted by Z6iii & a7s4

Partially-Stacked CIS is actually misleading in name.

This CIS is completely different from the traditional Stacked CIS.

In fact, the simulation layer (pixel layer) is the same as IMX410.

Just slightly modify the logic circuit layer below.

Then use Cu-Cu Hybrid Bonding technology to connect

Back stacking AD made with advanced manufacturing process

The Stacked part is actually in the AD, not in the imaging area.

Basically, it can be regarded as a high-speed version of BSI.

The reading speed under full-pixel 14Bit can be about 3.5 times faster.

CIS's Cu-Cu Hybrid Bonding and

The so-called Partially-Stacked CIS is the exclusive technology of SSS.

But SSS did not classify it into Stacked CIS.

It is called high-speed BSI CIS.

Therefore, Partially-Stacked CIS

It is very likely that it is a word invented by Nikon MKT.

The purpose is to make it easier and more intuitive for consumers to understand this technology.

And it has greater market publicity value.

Click to expand...

patrol_taking_9j said:

Yes, I stand corrected. The reporting online is inconsistent. Did also find an interesting tidbit below, posted I believe by Sony engineer in Taiwan.

patrol_taking_9j said:

IMX410E/IMX820 adopted by Z6iii & a7s4

Partially-Stacked CIS is actually misleading in name.

This CIS is completely different from the traditional Stacked CIS.

In fact, the simulation layer (pixel layer) is the same as IMX410.

Just slightly modify the logic circuit layer below.

Then use Cu-Cu Hybrid Bonding technology to connect

Back stacking AD made with advanced manufacturing process

The Stacked part is actually in the AD, not in the imaging area.

Basically, it can be regarded as a high-speed version of BSI.

The reading speed under full-pixel 14Bit can be about 3.5 times faster.

CIS's Cu-Cu Hybrid Bonding and

The so-called Partially-Stacked CIS is the exclusive technology of SSS.

But SSS did not classify it into Stacked CIS.

It is called high-speed BSI CIS.

Therefore, Partially-Stacked CIS

It is very likely that it is a word invented by Nikon MKT.

The purpose is to make it easier and more intuitive for consumers to understand this technology.

And it has greater market publicity value.

Click to expand...

Click to expand...

Is this the first leak of the A7S4? lol

That would be the 3rd camera to use that sensor after the Z6III and Lumix S1II.

This is unusual for Sony to use a sensor that is already used by their competitors, when it comes to full frame cameras they usually get out their new tech first.

Only exception that would follow the A7S4 route is the IMX 571, which was first released in 2018, but Sony waited until 2022 to use in one of their cameras (the FX30)