bclaff Loc: Sherborn, MA (18mi SW of Boston)

Since Signal is going to be linear Signal to Noise Ratio (SNR) is just a proxy for noise.
Your SNR values are quite low. This must be a very dark target.
You should expect 14-bit and 12-bit to match until you get down to a read noise that 12-bit can't measure.
You can get an idea of what ISO settings, if any, would be limited by 12-bit by looking at the PhotonsToPhotos Read Noise in DNs chart. If the 14-bit read noise is greater than about 4 (2 on the logarithmic y-axis) then 12-bit will be limited.

selmslie Loc: Fernandina Beach, FL, USA

That's where I have been getting my information for the average and S.D. that I used in the plots.

Since there is no Bayer array I use Monochrome2DNG to flag the file so that it skips the demosaicing step. That makes it easy to find. Note that RawDigger does not report the aperture because I used a Leica Summicron that cannot communicate with the camera.

For the Bayer array cameras there is more information so I have to be careful copying the numbers from the G or G2 row. The Voigtlander Ultron can talk to the camera.

For the monochrome camera


For the Df

selmslie Loc: Fernandina Beach, FL, USA

It's actually a very bright target. It's the white display of an empty folder on a computer screen calibrated to 120cd/m2.

To avoid any issues with the screen's pixels the lens is focused manually at infinity and held close to the screen.

selmslie Loc: Fernandina Beach, FL, USA

I only took measurements at base ISO. All three cameras are well below 2 in read noise at base ISO.

I was testing for linearity so I varied only the shutter speed.

If I had been varying the ISO I would have used a constant shutter speed. That's an entirely different investigation.

bclaff Loc: Sherborn, MA (18mi SW of Boston)

Very strange. By my calculation this is only 1/3 stop brighter but the values are much higher. Your SNR should be much higher for a bright target.

bclaff Loc: Sherborn, MA (18mi SW of Boston)

You might find this target helpful since it exposes red, green, and blue more evenly.

(Download)

selmslie Loc: Fernandina Beach, FL, USA

I’m looking only at the green channel because it’s always the leading histogram in a neutral target. Green captures 50% of the light. I’d rather not complicate my life.

I have used that target for setting up UniWB but I have never found any benefit from that concept, just plenty of reasons to avoid it.

plumbbob1

Can I borrow that quote?

jlg1000 Loc: Uruguay / South America

Yes there is.

Total noise is the sum of 1) Dark Noise, 2) Read Noise, 3) Photon Shot Noise and 4) Fixed Pattern Noise.

1) Is generated because of electrons conducting because of the temperature being higher than absolute zero.

The value is exponential with temperature and is fixed by quantum effects once the technology is choosen

It is heavily related to the type of material and gate geometry... CCD, CMOS, FinFET, etc.

Bad thing is that by reducing the size of the gates, dark noise increases, because of quantum tunnel effects... we live in a strange Universe.

This is the reason because big telescope sensors are cooled to cryogenic temperatures (about 4K = -450ºF). You don't want to do that with your camera.

2) Is related to normal electric signal noise an can be traced back to shielding and - ultimately - to the decay of radioactive particles in the surrounding of the sensor.

It has been *very* minimized, and that is the main reason because modern cameras have lower noise than older ones.

We are almost at the bottom of it, and radioactive decay can already be identified in the noise signal... no not much more improvement in that department. Again, it is a quantum effect...

3) It is simply related to the fact that light comes in discrete photons.

Because the photons hit the sensor in an *average* correct number, but at inherently random quantities it depends only on pixel size (geometry) and exposure time (time)

Fo example, for typical 4um x 4 um modern pixels, the well capacity is about 60,000 electrons, and with a quantum efficiency of 50% this amounts to 120,000 photons... and at low light (maybe 20 foot - candles), and short exposure time (1/1000 s), the maximum amount of photons per pixels will be about 200 or 300... so slight random variations in the amount of photons received pixel to pixel (even a single photon) will be noticed when amplified or recovered in post.

There is absolutely no way to avoid this effect, it is encoded in the nature of light itself (quantum, quantum, quantum...) Shot noise is a *fixed* amount based on geometry, lighting and exposure time, and is *unrelated* to the composition or technology of the sensor. Once the size of the pixel is set, everything shot noise is set in stone.

4) Can be disregarded by calibration... either by the camera itself (high end cameras have a calibration file in firmware) or by taking "lights" and calibrating in post.

Bottom line: YES there is a *very* definitive floor for sensor noise.

TriX Loc: Raleigh, NC

Excellent explanation. Thanks.

photon-collector Loc: Tampa Bay Area, Florida

So, what does all this bullshit mean in real practice?

TriX Loc: Raleigh, NC

I guess it means it’s beyond your comprehension

OldSchool-WI Loc: Brandon, Wisconsin 53919

_________________________(reply)
It means avoid noise as much as possible with the correct camera, AND THE CORRECT EXPOSURE slightly on the dense side if you have a choice.-----------------

OldSchool-WI Loc: Brandon, Wisconsin 53919

___________________________(reply)
From above:----[photon-collector wrote:
So, what does all this bullshit mean in real practice?


I guess it means it’s beyond your comprehension
Digital: Canon 5D IV, 70
___________________I guess we are off in the new year where we left off in the old?----Insults upon insults????-------------ew

plumbbob1

I think it was a great discussion, however, for someone else, not me.
My point is there are other ways to deal with noise that doesn't require so much science.
That's the best way for me. Like my camera is very forgiving in this area and others.
Just the same, if you understand the conversation it would have been very inlightening.

But mostly, I hope our conversations can remain civil and cordial.