Z 7 Fixed Pattern Noise (FPN) results at PhotonsToPhotos

[bclaff]

Fixed Pattern Noise (FPN) testing at PhotonsToPhotos shows a visible horizontal pattern:

[ATTACH alt="Best viewed "Original Size" "]2027426[/ATTACH]
Best viewed "Original Size"

The false colors aren't important; they are simply to help visualize patterns.
The ideal result is a random pattern.

Single is a single image while stacked are 16 "identical" stacked images.
Patterns that show in stacked images are "fixed".

These are 256x256 crops from the center of the sensor.

Without agonizing over the technical details it does look as if horizontal bands might be visible under certain conditions.

Given the much cleaner D850 results:

[ATTACH alt="Best viewed "Original Size" "]2027427[/ATTACH]
Best viewed "Original Size"

It would appear that this may have to do with the On Sensor Phase Detect Auto-Focus (OSPDAF). This effect has been seen on other OSPDAF sensors.

--
Bill ( Your trusted source for independent sensor data at PhotonsToPhotos )

 

[Dibyendu Majumdar]

Would be useful to see a comparison with Sony A7R3.

 

[Alex Permit]

Would be useful to see a comparison with Sony A7R3.

Here you go

Loading…

www.photonstophotos.net

 

[bhollis]

Fixed Pattern Noise (FPN) testing at PhotonsToPhotos shows a visible horizontal pattern:


Best viewed "Original Size"

The false colors aren't important; they are simply to help visualize patterns.
The ideal result is a random pattern.

Single is a single image while stacked are 16 "identical" stacked images.
Patterns that show in stacked images are "fixed".

These are 256x256 crops from the center of the sensor.

Without agonizing over the technical details it does look as if horizontal bands might be visible under certain conditions.

What conditions? Pushing shadows?

Given the much cleaner D850 results:


Best viewed "Original Size"

It would appear that this may have to do with the On Sensor Phase Detect Auto-Focus (OSPDAF). This effect has been seen on other OSPDAF sensors.

What other OSPDAF sensors? This effect appears to be more pronounced with the Z7 than any other camera listed on your site.

Trying to get a handle on how significant an issue this really is.

 

[RaLuC]

Fixed Pattern Noise (FPN) testing at PhotonsToPhotos shows a visible horizontal pattern:


Best viewed "Original Size"

The false colors aren't important; they are simply to help visualize patterns.
The ideal result is a random pattern.

Single is a single image while stacked are 16 "identical" stacked images.
Patterns that show in stacked images are "fixed".

These are 256x256 crops from the center of the sensor.

Without agonizing over the technical details it does look as if horizontal bands might be visible under certain conditions.

Given the much cleaner D850 results:


Best viewed "Original Size"

It would appear that this may have to do with the On Sensor Phase Detect Auto-Focus (OSPDAF). This effect has been seen on other OSPDAF sensors.

That looks quite ugly to me. However, can someone please clarify the real world relevance on picture quality here?

Thanks

 

[gurgeh]

If your photography involves taking 16 black frames, stacking them, and then amplifying them so that you see all the noise, this could have some serious consequences!

 

[Alex Permit]

That looks quite ugly to me. However, can someone please clarify the real world relevance on picture quality here?

Thanks

It theoretically effects your latitude to lift shadows in underexposed shots.

 

[beatboxa]

If your photography involves taking 16 black frames, stacking them, and then amplifying them so that you see all the noise, this could have some serious consequences!

LOL!

For the question: essentially, it means that certain rows of pixels (likely the PDAF pixels) have different sensitivity--and that when you take a picture--especially one with completely even lighting--you could see slight banding in the raw file.

And this is one thing to remember: this is the raw file. Other manufacturers may interpolate these pixels differently pre-raw, by estimating more from other nearby pixels. (I'm sure Nikon does some level of this as well). It's possible that Nikon's raw > JPEG software may additionally tweak this.

This banding will get more pronounced if you take many pictures and stack them, from the exact same position, for a still scene.

To simplify for general understanding:

ie. if most standard pixels register a value of "100" and PDAF pixels are half-masked, they may only register a value of "50"--which Nikon then multiplies by 2.

Unfortunately, not everything is this simple. The "noise level" also gets multiplied by 2. So when you take a dark frame (lens cap on), pixels may register near "10"--but then these PDAF pixels get multiplied by 2, so they register 20. And this is how we see them on this test.

And the masking or angles may not be perfect, and pixels may clip sooner. So in a regular scene, you may end up with "48" or "52" instead of "50"--and these little variations may add up.

Nikon presumably counters this by not just simply multiplying by 2--but rather estimating from other nearby pixels as well. But obviously this algorithm isn't perfect.

Sony probably does something similar. This could explain their hunger for stars.

And Canon uses dual pixel--two raw images, where it then merges the pdaf pixels. We could see something similar in Nikon over the next few years, since they've patented "quad pixel."

But this clearly explains a lot about the Z7--and the D850. Maybe their engineers weren't lying after all when they said they didn't include OSPDAF on the D850 to avoid IQ issues.

 

[bhollis]

That looks quite ugly to me. However, can someone please clarify the real world relevance on picture quality here?

Thanks

It theoretically effects your latitude to lift shadows in underexposed shots.

That's my concern--that banding will undercut the sensor's otherwise excellent dynamic range.

Could be the kind of first-generation teething issue that ordinarily keeps me from buying first-gen products. Hope not. We'll see.

 

[bclaff]

That looks quite ugly to me. However, can someone please clarify the real world relevance on picture quality here?

Thanks

It theoretically effects your latitude to lift shadows in underexposed shots.

That's my concern--that banding will undercut the sensor's otherwise excellent dynamic range.

No effect on dynamic range

 

[bhollis]

That looks quite ugly to me. However, can someone please clarify the real world relevance on picture quality here?

Thanks

It theoretically effects your latitude to lift shadows in underexposed shots.

That's my concern--that banding will undercut the sensor's otherwise excellent dynamic range.

No effect on dynamic range

I understand that there's no effect on DR per se. But the question is, will banding show up when lifting shadows (thereby affecting one's ability to make use of the sensor's DR)?

It would really help if you could provide some real-world insight on these findings. Is this a show stopper or a hiccup?

Despite my usual aversion to buying first-get products, I've got a Z7 on pre-order. This is the sort of thing that could cause me to decide to cancel and wait. Just need to get a better sense of how big a deal this really is.

 

[bclaff]

If your photography involves taking 16 black frames, stacking them, and then amplifying them so that you see all the noise, this could have some serious consequences!

Stacking is a diagnostic technique to make patterns that are present in single images more apparent.
The pattern is there is the single image, just a bit harder to see.
If you push a single image hard, you may very well see this pattern.

Nikon cameras (mostly Sony-based sensors) rarely show FPN (the Toshiba D7100 would be an exception).
It remains to be seen in this will show up in normal photos; I suspect it will under certain conditions.

 

[bclaff]

That looks quite ugly to me. However, can someone please clarify the real world relevance on picture quality here?

Thanks

It theoretically effects your latitude to lift shadows in underexposed shots.

That's my concern--that banding will undercut the sensor's otherwise excellent dynamic range.

No effect on dynamic range

I understand that there's no effect on DR per se. But the question is, will banding show up when lifting shadows (thereby affecting one's ability to make use of the sensor's DR)?

The technical test shows an effect both at the shadow end (black) and at the highlight end (illuminated).
I suspect that if this shows in "real" images it's more likely to show in very bright areas rather than in the shadows.

It would really help if you could provide some real-world insight on these findings. Is this a show stopper or a hiccup?

I suggest taking dpreview gallery or studio shots and seeing whether you can provoke any unwanted effects.

Despite my usual aversion to buying first-get products, I've got a Z7 on pre-order. This is the sort of thing that could cause me to decide to cancel and wait. Just need to get a better sense of how big a deal this really is.

I wouldn't "panic". If this were a wide-spread issue I think early reviewers would have already noted it.
I suspect it will show under some circumstances.

Mostly it's evidence of one of the ways that the Z 7 isn't simply a D850 sensor in a mirror-less body.

 

[bhollis]

That looks quite ugly to me. However, can someone please clarify the real world relevance on picture quality here?

Thanks

It theoretically effects your latitude to lift shadows in underexposed shots.

That's my concern--that banding will undercut the sensor's otherwise excellent dynamic range.

No effect on dynamic range

I understand that there's no effect on DR per se. But the question is, will banding show up when lifting shadows (thereby affecting one's ability to make use of the sensor's DR)?

The technical test shows an effect both at the shadow end (black) and at the highlight end (illuminated).
I suspect that if this shows in "real" images it's more likely to show in very bright areas rather than in the shadows.

It would really help if you could provide some real-world insight on these findings. Is this a show stopper or a hiccup?

I suggest taking dpreview gallery or studio shots and seeing whether you can provoke any unwanted effects.

Despite my usual aversion to buying first-get products, I've got a Z7 on pre-order. This is the sort of thing that could cause me to decide to cancel and wait. Just need to get a better sense of how big a deal this really is.

I wouldn't "panic". If this were a wide-spread issue I think early reviewers would have already noted it.
I suspect it will show under some circumstances.

Mostly it's evidence of one of the ways that the Z 7 isn't simply a D850 sensor in a mirror-less body.

Thanks. I appreciate the response.

 

[syberman7]

If your photography involves taking 16 black frames, stacking them, and then amplifying them so that you see all the noise, this could have some serious consequences!

I was planning a one man retrospective gallery exhibition on just this type of imagery.

 

[rhlpetrus]

That looks quite ugly to me. However, can someone please clarify the real world relevance on picture quality here?

Thanks

It theoretically effects your latitude to lift shadows in underexposed shots.

That's my concern--that banding will undercut the sensor's otherwise excellent dynamic range.

No effect on dynamic range

I understand that there's no effect on DR per se. But the question is, will banding show up when lifting shadows (thereby affecting one's ability to make use of the sensor's DR)?

The technical test shows an effect both at the shadow end (black) and at the highlight end (illuminated).
I suspect that if this shows in "real" images it's more likely to show in very bright areas rather than in the shadows.

It would really help if you could provide some real-world insight on these findings. Is this a show stopper or a hiccup?

I suggest taking dpreview gallery or studio shots and seeing whether you can provoke any unwanted effects.

Despite my usual aversion to buying first-get products, I've got a Z7 on pre-order. This is the sort of thing that could cause me to decide to cancel and wait. Just need to get a better sense of how big a deal this really is.

I wouldn't "panic". If this were a wide-spread issue I think early reviewers would have already noted it.
I suspect it will show under some circumstances.

Mostly it's evidence of one of the ways that the Z 7 isn't simply a D850 sensor in a mirror-less body.

It certainly is not, given the OSPDAF that was implemented, not existent in D850. That is a possible cause of that, no? Basic silicon and stills are likely same, with that "interference" added. Video is another story, wiring, reading rates, etc, are likely different.

 

[bhollis]

That looks quite ugly to me. However, can someone please clarify the real world relevance on picture quality here?

Thanks

It theoretically effects your latitude to lift shadows in underexposed shots.

That's my concern--that banding will undercut the sensor's otherwise excellent dynamic range.

No effect on dynamic range

I understand that there's no effect on DR per se. But the question is, will banding show up when lifting shadows (thereby affecting one's ability to make use of the sensor's DR)?

The technical test shows an effect both at the shadow end (black) and at the highlight end (illuminated).
I suspect that if this shows in "real" images it's more likely to show in very bright areas rather than in the shadows.

It would really help if you could provide some real-world insight on these findings. Is this a show stopper or a hiccup?

I suggest taking dpreview gallery or studio shots and seeing whether you can provoke any unwanted effects.

Despite my usual aversion to buying first-get products, I've got a Z7 on pre-order. This is the sort of thing that could cause me to decide to cancel and wait. Just need to get a better sense of how big a deal this really is.

I wouldn't "panic". If this were a wide-spread issue I think early reviewers would have already noted it.
I suspect it will show under some circumstances.

Mostly it's evidence of one of the ways that the Z 7 isn't simply a D850 sensor in a mirror-less body.

It certainly is not, given the OSPDAF that was implemented, not existent in D850. That is a possible cause of that, no? Basic silicon and stills are likely same, with that "interference" added. Video is another story, wiring, reading rates, etc, are likely different.

If, as has been posited, this FPN is an artifact of the Z7's OSPDAF, then I'm surprised to see that the Sony A7RIII appears to be free of this effect.

Any thoughts on why this might be so?

 

[beatboxa]

Mostly it's evidence of one of the ways that the Z 7 isn't simply a D850 sensor in a mirror-less body.

It certainly is not, given the OSPDAF that was implemented, not existent in D850. That is a possible cause of that, no? Basic silicon and stills are likely same, with that "interference" added. Video is another story, wiring, reading rates, etc, are likely different.

If, as has been posited, this FPN is an artifact of the Z7's OSPDAF, then I'm surprised to see that the Sony A7RIII appears to be free of this effect.

Any thoughts on why this might be so?

As I noted in my post above, there are different techniques to counteract this. Sony probably interpolates the values for these pixels using the surrounding pixels, pre-raw.

 

[bhollis]

Mostly it's evidence of one of the ways that the Z 7 isn't simply a D850 sensor in a mirror-less body.

It certainly is not, given the OSPDAF that was implemented, not existent in D850. That is a possible cause of that, no? Basic silicon and stills are likely same, with that "interference" added. Video is another story, wiring, reading rates, etc, are likely different.

If, as has been posited, this FPN is an artifact of the Z7's OSPDAF, then I'm surprised to see that the Sony A7RIII appears to be free of this effect.

Any thoughts on why this might be so?

As I noted in my post above, there are different techniques to counteract this. Sony probably interpolates the values for these pixels using the surrounding pixels, pre-raw.

Correct me if I'm wrong, but this seems to suggest that this effect is, at least to some extent, correctable through firmware.

Which raises the question whether the image samples that were used to generate these results were from production cameras with final firmware? (I'm guessing that bclaff is confident that these results are representative of production cameras/firmware.)

 

[JimKasson]

I performed a spectral analysis on the raw channels of dark frames captured by Rishi at ISO 64. fs is the sampling frequency.


At all but the lowest spatial frequencies, the spectrum is white


At all but the lowest spatial frequencies, the spectrum is white


Interpolation over the PDAF rows causes spikes in the blue channel at 1/3 and 1/6 the sampling frequency, in addition to a half-db lowpass rolloff.

Jim

--
Posted as a regular forum member.

the last word the last word - Photography meets digital computer technology. Photography wins -- most of the time.

Photography meets digital computer technology. Photography wins -- most of the time.

blog.kasson.com