E-M1 Mark II Fixed Pattern Noise (FPN) Results

Started Jan 23, 2017 | Discussions
bclaff Forum Pro • Posts: 13,093
E-M1 Mark II Fixed Pattern Noise (FPN) Results
32

Photo-Response Non-Uniformity (PRNU) runs about 0.5% which is typical and there is no discernible pattern.
2D Fourier Transform (FT) does indicate a small amount of signal processing (usually noise reduction).
This is why there is a small white circle in the center of the FT.
PRNU image on the left, 2D FT on the right:

Note the subtle grid pattern on the left, we'll get back to that!

Dark Signal Non-Uniformity (DSNU) is good. DSNU image:

It does exhibit an interesting pattern which is more apparent if we look at the 2D FT:

Those white dots indicate something happening, mostly with a frequency of 4 pixels in both the horizontal and vertical.

Let's look more closely at that DSNU image by tweaking contrast and zooming in:

That is a very distinctive pattern. It's the on chip Phase Detect Auto Focus (PDAF).

In fact, we can discern a 16x16 pattern that looks like this (idealized):

Pretty cool!

This pattern is very subtle and would only concern people stacking multiple low light frames such as astro-photographers.

-- hide signature --

Bill ( Your trusted source for independent sensor data at http://www.PhotonsToPhotos.net )

Olympus E-M1 II
If you believe there are incorrect tags, please send us this post using our feedback form.
whumber
whumber Veteran Member • Posts: 4,154
Re: E-M1 Mark II Fixed Pattern Noise (FPN) Results

bclaff wrote:

Pretty cool!

This pattern is very subtle and would only concern people stacking multiple low light frames such as astro-photographers.

Ah, great observation.  I had noticed this pattern when doing some median blending and wasn't quite sure what was happening.

 whumber's gear list:whumber's gear list
Fujifilm X-T1 Olympus OM-D E-M1X Olympus E-M1 III OM-1 Olympus M.Zuiko Digital ED 60mm F2.8 Macro +10 more
OP bclaff Forum Pro • Posts: 13,093
Re: E-M1 Mark II Fixed Pattern Noise (FPN) Results

whumber wrote:

bclaff wrote:

Pretty cool!

This pattern is very subtle and would only concern people stacking multiple low light frames such as astro-photographers.

Ah, great observation. I had noticed this pattern when doing some median blending and wasn't quite sure what was happening.

Yeah, I think you would need flat to remove it.
Generally only required for specialized applications.

-- hide signature --

Bill ( Your trusted source for independent sensor data at http://www.PhotonsToPhotos.net )

GBC Senior Member • Posts: 1,454
Re: E-M1 Mark II Fixed Pattern Noise (FPN) Results

Very interesting. Thanks.

knickerhawk Veteran Member • Posts: 7,363
Re: E-M1 Mark II Fixed Pattern Noise (FPN) Results

bclaff wrote:

Photo-Response Non-Uniformity (PRNU) runs about 0.5% which is typical and there is no discernible pattern.
2D Fourier Transform (FT) does indicate a small amount of signal processing (usually noise reduction).
This is why there is a small white circle in the center of the FT.
PRNU image on the left, 2D FT on the right:

Note the subtle grid pattern on the left, we'll get back to that!

Dark Signal Non-Uniformity (DSNU) is good. DSNU image:

It does exhibit an interesting pattern which is more apparent if we look at the 2D FT:

Those white dots indicate something happening, mostly with a frequency of 4 pixels in both the horizontal and vertical.

Let's look more closely at that DSNU image by tweaking contrast and zooming in:

That is a very distinctive pattern. It's the on chip Phase Detect Auto Focus (PDAF).

In fact, we can discern a 16x16 pattern that looks like this (idealized):

Pretty cool!

This pattern is very subtle and would only concern people stacking multiple low light frames such as astro-photographers.

Very interesting stuff. Thanks for sharing and I look forward to more of your analysis (especially with respect to what you can tell us about the noise reduction that might be going on and how this compares to other cameras). One thing I don't understand from your DSNU analysis above is how this pattern is attributable to the on-chip PDAF. Olympus reports that there are 121 PDAF points on the chip and obviously the 16x16 pattern you're seeing would imply far more dedicated PDAF sensels than what Oly claims. What am I missing here?

OP bclaff Forum Pro • Posts: 13,093
Re: E-M1 Mark II Fixed Pattern Noise (FPN) Results
1

knickerhawk wrote:

...

Very interesting stuff. Thanks for sharing and I look forward to more of your analysis (especially with respect to what you can tell us about the noise reduction that might be going on and how this compares to other cameras). One thing I don't understand from your DSNU analysis above is how this pattern is attributable to the on-chip PDAF. Olympus reports that there are 121 PDAF points on the chip and obviously the 16x16 pattern you're seeing would imply far more dedicated PDAF sensels than what Oly claims. What am I missing here?

Well ... many pixels comprise a single PDAF point.
Remember, the pixel is only about 3.34µ

Regards,

-- hide signature --

Bill ( Your trusted source for independent sensor data at http://www.PhotonsToPhotos.net )

tonyC1994 Regular Member • Posts: 333
Re: E-M1 Mark II Fixed Pattern Noise (FPN) Results
2

Bill,

Great observation and analysis! Thank you for sharing with us.

Tony

 tonyC1994's gear list:tonyC1994's gear list
Olympus E-M1 II Sony a7 III Sony FE 55mm F1.8 Sony FE 24-70mm F4 OSS Olympus 40-150mm F2.8 Pro +1 more
kenw
kenw Veteran Member • Posts: 6,931
Re: E-M1 Mark II Fixed Pattern Noise (FPN) Results

bclaff wrote:

Photo-Response Non-Uniformity (PRNU) runs about 0.5% which is typical and there is no discernible pattern.
2D Fourier Transform (FT) does indicate a small amount of signal processing (usually noise reduction).
This is why there is a small white circle in the center of the FT.
PRNU image on the left, 2D FT on the right:

Thanks Bill!

Quick question, are you saying above that there appears to be NR to account for the PRNU or that there is actually more generic RAW NR going on? Any thoughts on the magnitude of this NR compared to other m43 cameras?
--
Ken W
See profile for equipment list

 kenw's gear list:kenw's gear list
Panasonic Lumix DMC-GM1 Nikon Z7 Panasonic Leica DG Summilux 15mm F1.7 ASPH Nikon Z 14-30mm F4 Nikon Z 24-200mm F4-6.3 VR +42 more
OP bclaff Forum Pro • Posts: 13,093
Re: E-M1 Mark II Fixed Pattern Noise (FPN) Results
1

kenw wrote:

bclaff wrote:

Photo-Response Non-Uniformity (PRNU) runs about 0.5% which is typical and there is no discernible pattern.
2D Fourier Transform (FT) does indicate a small amount of signal processing (usually noise reduction).
This is why there is a small white circle in the center of the FT.
PRNU image on the left, 2D FT on the right:

Thanks Bill!

Quick question, are you saying above that there appears to be NR to account for the PRNU or that there is actually more generic RAW NR going on? Any thoughts on the magnitude of this NR compared to other m43 cameras?

The small white circle in the center of the 2D FT is an indication of some sort of "nearest neighbor" signal processing, probably Noise Reduction.
(This amount is much smaller than I often see although it's more than zero.)

PRNU doesn't show directly in these images, it's measured statistically.

Regards,

-- hide signature --

Bill ( Your trusted source for independent sensor data at http://www.PhotonsToPhotos.net )

kenw
kenw Veteran Member • Posts: 6,931
Re: E-M1 Mark II Fixed Pattern Noise (FPN) Results

bclaff wrote:

The small white circle in the center of the 2D FT is an indication of some sort of "nearest neighbor" signal processing, probably Noise Reduction.
(This amount is much smaller than I often see although it's more than zero.)

Thanks for the clarification. I'm pretty sure Oly RAWs are truly "lossless" so indeed likely some NR.

Be aware that Panasonic RAWs are "mostly" lossless in that they use a reference pixel followed by differences and depending on the pixel level there can be one or two bits of truncation. Typically this quantization would be swamped by shot noise at the levels this occurs but the reference + difference "compression" along with truncation could show up in a spectral analysis.

PRNU doesn't show directly in these images, it's measured statistically.

Ah, yes of course. I'm used to long wave IR detectors where NU is so poor you have to calibrate it out so you do know the exact NU pattern. Forgot for these analyses it is derived from a fit to the SNR curve.

Thanks for all your hard work!

BTW, I have GM1 and E-M5II if you need any sample files from those.

-- hide signature --

Ken W
See profile for equipment list

 kenw's gear list:kenw's gear list
Panasonic Lumix DMC-GM1 Nikon Z7 Panasonic Leica DG Summilux 15mm F1.7 ASPH Nikon Z 14-30mm F4 Nikon Z 24-200mm F4-6.3 VR +42 more
tonyC1994 Regular Member • Posts: 333
Re: E-M1 Mark II Fixed Pattern Noise (FPN) Results

bclaff wrote:

kenw wrote:

bclaff wrote:

Photo-Response Non-Uniformity (PRNU) runs about 0.5% which is typical and there is no discernible pattern.
2D Fourier Transform (FT) does indicate a small amount of signal processing (usually noise reduction).
This is why there is a small white circle in the center of the FT.
PRNU image on the left, 2D FT on the right:

Thanks Bill!

Quick question, are you saying above that there appears to be NR to account for the PRNU or that there is actually more generic RAW NR going on? Any thoughts on the magnitude of this NR compared to other m43 cameras?

The small white circle in the center of the 2D FT is an indication of some sort of "nearest neighbor" signal processing, probably Noise Reduction.
(This amount is much smaller than I often see although it's more than zero.)

PRNU doesn't show directly in these images, it's measured statistically.

Regards,

Bill, do you have 2D fft plots and analysis for other sensors? Would be interesting to read and compare!

 tonyC1994's gear list:tonyC1994's gear list
Olympus E-M1 II Sony a7 III Sony FE 55mm F1.8 Sony FE 24-70mm F4 OSS Olympus 40-150mm F2.8 Pro +1 more
OP bclaff Forum Pro • Posts: 13,093
Re: E-M1 Mark II Fixed Pattern Noise (FPN) Results

kenw wrote:

bclaff wrote:

The small white circle in the center of the 2D FT is an indication of some sort of "nearest neighbor" signal processing, probably Noise Reduction.
(This amount is much smaller than I often see although it's more than zero.)

Thanks for the clarification. I'm pretty sure Oly RAWs are truly "lossless" so indeed likely some NR.

Be aware that Panasonic RAWs are "mostly" lossless in that they use a reference pixel followed by differences and depending on the pixel level there can be one or two bits of truncation. Typically this quantization would be swamped by shot noise at the levels this occurs but the reference + difference "compression" along with truncation could show up in a spectral analysis.

All the effects you list, compression artifacts, quantization, would not probably show as signal processing.
Even though some compression schemes have a "nearest neighbor" behavior the effect is so slight that it goes undetected.
I know this because of my prior work with Sony 11+7 files.

PRNU doesn't show directly in these images, it's measured statistically.

Ah, yes of course. I'm used to long wave IR detectors where NU is so poor you have to calibrate it out so you do know the exact NU pattern. Forgot for these analyses it is derived from a fit to the SNR curve.

As it happens, I don't fit to the Photon Transfer Curve (PTC) to get my DSNU, gain, and PRNU values (!)

Thanks for all your hard work!

You are welcome.

BTW, I have GM1 and E-M5II if you need any sample files from those.

I'll PM you regarding the GM1. I don't have that yet.

Regards,

-- hide signature --

Bill ( Your trusted source for independent sensor data at http://www.PhotonsToPhotos.net )

OP bclaff Forum Pro • Posts: 13,093
Re: E-M1 Mark II Fixed Pattern Noise (FPN) Results
3

tonyC1994 wrote:

Bill, do you have 2D fft plots and analysis for other sensors? Would be interesting to read and compare!

I do have a lot of data and it is quite interesting; but it's not currently published.

I'm trying to get the numerical results such as DSNU, FWC, PRNU, QE into a table as well as the 2D FT into some accessible form.
These analyses aren't my full time work so I just haven't had enough bandwidth to get this accomplished (but it is in the works!).

Regards,

-- hide signature --

Bill ( Your trusted source for independent sensor data at http://www.PhotonsToPhotos.net )

Jude 101 Contributing Member • Posts: 523
Re: E-M1 Mark II Fixed Pattern Noise (FPN) Results

bclaff wrote:

tonyC1994 wrote:

Bill, do you have 2D fft plots and analysis for other sensors? Would be interesting to read and compare!

I do have a lot of data and it is quite interesting; but it's not currently published.

I'm trying to get the numerical results such as DSNU, FWC, PRNU, QE into a table as well as the 2D FT into some accessible form.
These analyses aren't my full time work so I just haven't had enough bandwidth to get this accomplished (but it is in the works!).

Regards,

Great work.  Have you done similar measurements in the High-Resolution mode?

OP bclaff Forum Pro • Posts: 13,093
Re: E-M1 Mark II Fixed Pattern Noise (FPN) Results

Jude 101 wrote:

bclaff wrote:

tonyC1994 wrote:

Bill, do you have 2D fft plots and analysis for other sensors? Would be interesting to read and compare!

I do have a lot of data and it is quite interesting; but it's not currently published.

I'm trying to get the numerical results such as DSNU, FWC, PRNU, QE into a table as well as the 2D FT into some accessible form.
These analyses aren't my full time work so I just haven't had enough bandwidth to get this accomplished (but it is in the works!).

Great work. Have you done similar measurements in the High-Resolution mode?

I do have unpublished Pixel Shift (PS) measurements for the E-M5 Mark II.
I consider this a form of High Dynamic Range photography and don't present it with my "normal" data for that reason (nor do I include any other similar cameras in my results).
I would have to figure out what's a reasonable presentation.
Honestly, this is below the DSNU, FWC, PRNU, QE table as well as the 2D FT visualizations in my priorities.
BTW, where to stop? 10-bit/12-bit-14-bit? Lossy/lossless? Mechanical shutter/electronic shutter? Etc.

Of these shutter mode is probably of the most practical interest.

Regards,

-- hide signature --

Bill ( Your trusted source for independent sensor data at http://www.PhotonsToPhotos.net )

SHood Veteran Member • Posts: 5,978
Re: E-M1 Mark II Fixed Pattern Noise (FPN) Results

bclaff wrote:

Jude 101 wrote:

bclaff wrote:

tonyC1994 wrote:

Bill, do you have 2D fft plots and analysis for other sensors? Would be interesting to read and compare!

I do have a lot of data and it is quite interesting; but it's not currently published.

I'm trying to get the numerical results such as DSNU, FWC, PRNU, QE into a table as well as the 2D FT into some accessible form.
These analyses aren't my full time work so I just haven't had enough bandwidth to get this accomplished (but it is in the works!).

Great work. Have you done similar measurements in the High-Resolution mode?

I do have unpublished Pixel Shift (PS) measurements for the E-M5 Mark II.
I consider this a form of High Dynamic Range photography and don't present it with my "normal" data for that reason (nor do I include any other similar cameras in my results).
I would have to figure out what's a reasonable presentation.
Honestly, this is below the DSNU, FWC, PRNU, QE table as well as the 2D FT visualizations in my priorities.
BTW, where to stop? 10-bit/12-bit-14-bit? Lossy/lossless? Mechanical shutter/electronic shutter? Etc.

Of these shutter mode is probably of the most practical interest.

Regards,

Yes, I would like to understand the impact of using the electronic shutter on the E-M1ii.

-- hide signature --
 SHood's gear list:SHood's gear list
Olympus E-M1 II OM-1 Olympus M.Zuiko 300mm F4 IS Pro Olympus 12-100mm F4.0 Olympus 17mm F1.2 Pro
Jude 101 Contributing Member • Posts: 523
Re: E-M1 Mark II Fixed Pattern Noise (FPN) Results

bclaff wrote:

Jude 101 wrote:

bclaff wrote:

tonyC1994 wrote:

Bill, do you have 2D fft plots and analysis for other sensors? Would be interesting to read and compare!

I do have a lot of data and it is quite interesting; but it's not currently published.

I'm trying to get the numerical results such as DSNU, FWC, PRNU, QE into a table as well as the 2D FT into some accessible form.
These analyses aren't my full time work so I just haven't had enough bandwidth to get this accomplished (but it is in the works!).

Great work. Have you done similar measurements in the High-Resolution mode?

I do have unpublished Pixel Shift (PS) measurements for the E-M5 Mark II.
I consider this a form of High Dynamic Range photography and don't present it with my "normal" data for that reason (nor do I include any other similar cameras in my results).
I would have to figure out what's a reasonable presentation.
Honestly, this is below the DSNU, FWC, PRNU, QE table as well as the 2D FT visualizations in my priorities.
BTW, where to stop? 10-bit/12-bit-14-bit? Lossy/lossless? Mechanical shutter/electronic shutter? Etc.

Of these shutter mode is probably of the most practical interest.

Regards,

Gotcha, thanks.

OP bclaff Forum Pro • Posts: 13,093
Re: E-M1 Mark II Fixed Pattern Noise (FPN) Results
1

SHood wrote:

bclaff wrote:

....
BTW, where to stop? 10-bit/12-bit-14-bit? Lossy/lossless? Mechanical shutter/electronic shutter? Etc.

Of these shutter mode is probably of the most practical interest.

...

Yes, I would like to understand the impact of using the electronic shutter on the E-M1ii.

So far I only have the raw files for mechanical shutter but I'd be happy to do the electronic shutter analysis if someone will provide the appropriate files.
(Prior contributors "know the drill" but I'm always happy to train new collaborators )

Regards,

-- hide signature --

Bill ( Your trusted source for independent sensor data at http://www.PhotonsToPhotos.net )

knickerhawk Veteran Member • Posts: 7,363
Re: E-M1 Mark II Fixed Pattern Noise (FPN) Results

bclaff wrote:

knickerhawk wrote:

...

Very interesting stuff. Thanks for sharing and I look forward to more of your analysis (especially with respect to what you can tell us about the noise reduction that might be going on and how this compares to other cameras). One thing I don't understand from your DSNU analysis above is how this pattern is attributable to the on-chip PDAF. Olympus reports that there are 121 PDAF points on the chip and obviously the 16x16 pattern you're seeing would imply far more dedicated PDAF sensels than what Oly claims. What am I missing here?

Well ... many pixels comprise a single PDAF point.
Remember, the pixel is only about 3.34µ

Sure, I understand that, but what I guess I was mistakenly assuming is that the pattern would look more cross-like. What you show looks more like a generalized matrix that isn't particularly weighted/structured. I find that interesting, but I'm not well-versed on how OSPDAF systems are architected.

Thanks again.

uMad Contributing Member • Posts: 551
Re: E-M1 Mark II Fixed Pattern Noise (FPN) Results
2

bclaff wrote:
2D Fourier Transform (FT) does indicate a small amount of signal processing (usually noise reduction).

very nice work!

are there any other current cameras that do that? (noise reduction)

It has often come up with Sony, Fujifilm or Nikon, but I have never ever seen proof like yours

Keyboard shortcuts:
FForum MMy threads