Fixed Pattern Noise (FPN) Fun

[bclaff]

So far I have analyzed various aspects of Fixed Pattern Noise (FPN) for 14 cameras:
Fujifilm Finepix F700, S5 Pro, X-A1
Nikon D40, D200, D300, D700, D5100, D5300, D7000, D7100
Pentax MX-1
Ricoh GX200
Sony A7S

I'm still in the data gathering phase and if you'd like to help just send me email (preferred) or a Private Message (PM).

In constructing my analysis I considered using a Region Of Interest (ROI) that is not exactly centered on the sensor because I know some sensors are actually bonded from more than one part.

Here's an example from the D700:


D700 stacked red channel 256x256 centered on the sensor

Although this is interesting to see, I guess I will be moving my ROI.


And here's another "interesting" image:




S2 Pro stacked all channels

This is an S2 Pro that was subject to recall by Fujifilm because of sensor issues.

There will be a separate posts in the future with specific FPN results such as Dark Signal Non-Uniformity (DSNU), Photo Response Non-Uniformity (PRNU), and well as some other data that naturally fall out of processing the images. Right now I'm still crunching data, hoping to get more, and trying to figure out the best way to present the information.

--
Bill (visit me at http://home.comcast.net/~NikonD70/ )

 

[Abe R Ration]

I've done some pattern noise measurements on Sony A7 - here is the classic global local warmng "hockey stick" (actually I don't know what causes it, I just speculate that it might be heat - anyone else?)


Read noise averages per row of Sony A7.

In addition to this larger read noise of the bottom edge of the sensor there is anotter peculiarity: the patter noise isn't fixed, thus it's not at all easy to get rid of. (Though, the "hockey stick" end reduces to more normal level by subtracting frames, at least on ISO 1600 test image - the above one is ISO 100 where the "hockey stick" is more visible.





--
Abe R. Ration - amateur photographer, amateur scientist, amaterur camera buff

Abe R. Ration's lens and camera blog

This blog is about camera and lens technology and related sciences

aberration43mm.wordpress.com

 

[John Sheehy]

I've done some pattern noise measurements on Sony A7 - here is the classic global local warmng "hockey stick" (actually I don't know what causes it, I just speculate that it might be heat - anyone else?)

In addition to this larger read noise of the bottom edge of the sensor there is anotter peculiarity: the patter noise isn't fixed, thus it's not at all easy to get rid of. (Though, the "hockey stick" end reduces to more normal level by subtracting frames, at least on ISO 1600 test image - the above one is ISO 100 where the "hockey stick" is more visible.

If you are measuring the sigma of a line of pixels, that sigma can be greatly affected by low-frequency "noise" such as amp glow. My Canon 7DmkII has that in the lower corners of the frame, with a slow gradient to the middle bottom of the frame. The gradient, while not appearing as a noise, measures as noise. What kind of results do you get if you do a high-pass filtering of the frame before measurements? That would isolate the cause a bit.
--
John
http://www.pbase.com/image/55384958.jpg

 

[Abe R Ration]

I've done some pattern noise measurements on Sony A7 - here is the classic global local warmng "hockey stick" (actually I don't know what causes it, I just speculate that it might be heat - anyone else?)

In addition to this larger read noise of the bottom edge of the sensor there is anotter peculiarity: the patter noise isn't fixed, thus it's not at all easy to get rid of. (Though, the "hockey stick" end reduces to more normal level by subtracting frames, at least on ISO 1600 test image - the above one is ISO 100 where the "hockey stick" is more visible.

If you are measuring the sigma of a line of pixels, that sigma can be greatly affected by low-frequency "noise" such as amp glow. My Canon 7DmkII has that in the lower corners of the frame, with a slow gradient to the middle bottom of the frame. The gradient, while not appearing as a noise, measures as noise. What kind of results do you get if you do a high-pass filtering of the frame before measurements? That would isolate the cause a bit.

I did some tests. Here's the chart of the line sums after high pass filtering. It's quite interesting as the "hockey stick" goes mostly away.




Hockey stick almost gone

I also decided to do a regular unsharp mask with large diameter (to unfiltered image) - this is a crop of 1000x1000 from the bottom edge with appropriate contrast to make anything visible:


Interesting

I don't think the amplifiers are the cause for this, as the redout it done horizontally (or at least I imaing it is due to the direction of the stripes) and amplification is done before the thousands of ADCs do their job.

But I imagine it's heat based regardless.



--
Abe R. Ration - amateur photographer, amateur scientist, amaterur camera buff

Abe R. Ration's lens and camera blog

This blog is about camera and lens technology and related sciences

aberration43mm.wordpress.com

 

[Abe R Ration]

I did some tests. Here's the chart of the line sums after high pass filtering. It's quite interesting as the "hockey stick" goes mostly away.

Correction - I don't mean sums, but standard deviations. I measured the standard deviation of the pixels on a row, not sum.

Hockey stick almost gone

I also decided to do a regular unsharp mask with large diameter (to unfiltered image) - this is a crop of 1000x1000 from the bottom edge with appropriate contrast to make anything visible:


Interesting

I don't think the amplifiers are the cause for this, as the redout it done horizontally (or at least I imaing it is due to the direction of the stripes) and amplification is done before the thousands of ADCs do their job.

But I imagine it's heat based regardless.

--
Abe R. Ration - amateur photographer, amateur scientist, amaterur camera buff
http://aberration43mm.wordpress.com/

--
Abe R. Ration - amateur photographer, amateur scientist, amaterur camera buff

Abe R. Ration's lens and camera blog

This blog is about camera and lens technology and related sciences

aberration43mm.wordpress.com

 

[John Sheehy]

I don't think the amplifiers are the cause for this, as the redout it done horizontally (or at least I imaing it is due to the direction of the stripes) and amplification is done before the thousands of ADCs do their job.

But I imagine it's heat based regardless.

Either that, or some kind of inconsistency in the silicon. Is it the same in every frame?

I've never seen that artifact before; it looks more like an exposed shot with residue on the sensor.

 

[Abe R Ration]

I don't think the amplifiers are the cause for this, as the redout it done horizontally (or at least I imaing it is due to the direction of the stripes) and amplification is done before the thousands of ADCs do their job.

But I imagine it's heat based regardless.

Either that, or some kind of inconsistency in the silicon. Is it the same in every frame?

I've never seen that artifact before; it looks more like an exposed shot with residue on the sensor.

Correction 2: The hockey still exists after differnet highpass filters - I blundered the original measurement.

Of the unsharp and the hockey stick the thing to realize is that these abnormalities are extremely minute.

Another observation: as the exposure time goes up, the "hockey stick" reduces, though the unsharpmask thingie remains. While the latten seems dramatic, it is really just about half a DN above the rest of the data in average. But an interesting phenomenon.