No I really do get it but you are not seeing past what you think is correct.
Sorry, he is posting what is actually correct and you really do not get it.
I'll put it simply:
Noise is generated by the sensor, not the lens. It is dependant on the technology of the sensor to reduce/remove it. All the lens aperture does is control the amount of light hitting it.
This is where is your central mistake. Most of the noise
is not generated by the sensor. It is inherent in the structure of the light hitting the sensor and its SNR is dependent on the amount of that light, so the aperture controlling the amount of light controls the noise.
--
Tinkety tonk old fruit, & down with the Nazis!
Bob
It is inherent in the nature of light that you cannot make that assumption. The whole structure of QED is based on removing all assumptions. We do not fully understand light but can predict it's behaviour and unfortunately the model only works when we don't make assumptions that we can't measure. You cannot say noise is inherent in the structure of light, you can only say that we can measure noise in the structure of light when we place a sensor in front of it. This is the reality we live with because it is the only one that allows us to predict the behaviour of light with any accuracy.
It's a little beside the point because noise reduction will be a part of sensor and software development. If it does become a lens technology then I would guess that it will be more in the way of lens coatings, maybe active ones, and not effective aperture diameter.
It's kind of a circular argument that you present. There will be random interactions when the light hits the sensor, this is well predicted by the science available. It is therefore best if you can maximise the amount of light hitting the sensor. So you then have to maximise exposure, but there is a maximum you can't exceed. So you therefore have to measure and control point light sources because lenses also focus highlight and shadow points onto your sensor and does not spread your total light evenly. It's impossible to maximise your exposure without being able to control intensity of point sources, which shutter speed and aperture do. But you want to remove this and replace it with something that reflects two properties that aperture cannot fully control?
Sorry, you are talking complete nonsense. Shot noise is a well understood phenomenon, the theory has been done, the measurements support the theory completely. Really, it isn't up for argument, except for those who want to argue with all the physics to support their pet theory.
No, I'm well aware of it, (it is not the only source of noise though, other sources are
produced at the sensor). As I said it is in the nature of light that you can't make assumptions, only make observations. We do not fully understand the nature of light but do have a model that predicts how it will behave, but you can only confirm it by making a measurement.
This leads to an interesting development. We cannot predict the course or outcome for single photons as to whether they will result in image noise, simply because we can't measure it without altering it, changing what it will do. But we can accurately predict the amount of noise that will result when we have a large enough group of photons. As you can't predict single photons but only product the proportion you can only say that for
x number of photons you will get
y amount of noise. You are forced to calculate it from total noise.
But, as it happens this is exactly the way light seems to behave, as shown by observation. Shot noise is a proportion of total light and it's distribution is not dependant on intensity at the sensor. You see it in the shadow areas of an image only because the signal to noise ratio is less there. The more total light you have the greater the signal to noise ratio, so larger sensors have less noise and it's not dependant on intensity at point sources. But there is an upper limit to exposure for individual pixels, if this limit did not exist then
f-stop would indeed be redundant, exposure in digital would be exactly as you say, redundant and it would simply be noise control, a light gathering exercise. This is how digital behaves when you are well below the limit, i.e. you have sufficient headroom to increase exposure without exceeding the limit, but not how it behaves when you are at the limit.
Because in digital exposure your best noise control is close to this limit then control of highlights becomes important. Digital exposure is an exercise in highlight and noise control to control highlight exposure you have to control point intensities across formats and focal lengths which is what the
f-stop system does.
It does not fully control shot noise because to produce an image you have to respect this limit, and the limit is defined entirely by the sensor not the lens aperture.
So in the case of noise aperture influences noise up to the point of "over-exposure" only and so cannot control noise through the full range of it's adjustment.
The only thing it can control through its full range of adjustment is the amount of light passing through it.
Are we not heading for a state of confusion again? ;-)
I don't think that you ever got out of one.
Also I think you may be assuming that although you change the numbers and what they represent on the lens barrel that they will maintain the same calibration, i.e. each click will vary the amount of light in a completely proportional way, always halving or doubling it (you still need to control highlights). This may not be convenient or even make the best sense as if you adopt a new set of numbers to label lens apertures that the actual calibration may alter to suit the new numbers. In fact calibrating a lens to one system and labelling it with another to me may cause a state of confusion. ;-)
