Just so's you know, in some cameras the raw data values are decreased, proportionally to: (base ISO/selected ISO).
The "net effect" on the great majority of cameras is sensor under-exposure ... for example, ISO 1600 under-exposes the sensor to 1/16 of it's capacity, in other words -4 EV.
Nice wriggle ...
Is there a credible reference that supports the statement?
There is nothing wrong with my word choice. Shooting "at" an ISO, with no further qualifying words or context, means using both the setting and the intended exposure that the setting was created for, to any reasonable person, IMO.
I think that a lot of people who feel that they know better than most like to get involved in dogmatic "over-reaction" to show how smart they are, and how wrong others are, as an end in itself.
I don't think that it is wise to insist that "ISO" always means one thing; namely the setting with that number. I have proposed many, many times that we adopt the habit of writing/saying "ISO setting" and "ISO exposure index" when they are not assumed to be the same number, but it seems that some prefer to use the term monolithically, because it can help them feel smarter when they get to say that higher ISO doesn't cause more noise, despite the fact that a higher ISO exposure index, does increase noise relative to signal, whether or not you crank the setting up, too. IMO, the exposure index is actually the more substantial of the two "ISO" meanings. When my camera reaches the top ISO setting allowed by auto-ISO and the JPEG is dark, the ISO I am shooting at, in my mind, is the exposure index; not the setting.
What's gain?
The term "gain", which is output divided by input, without explication, is ambiguous. You can have gain where the input and output units are the same: current gain, voltage gain, power gain, etc. You can have AC gain and DC gain. At higher frequency, you can have voltage gain margin at phase crossover. Those gains are dimensionless.
You can have gain with different units: conversion gain is an example, where the numerator is voltage and the denominator is electrons. The dimensions of conversion gain is voltage over charge. One gain of a transconductance amplifier has the units of current over voltage.
Circuits can have several different relevant gains. The voltage gain of a source follower is less than one, but the current gain is more than one, and so is the power gain.
I have designed many systems with mixed analog and digital processing. I referred to manipulations of signal level as gain, whether that gain was achieved through analog amplification or digital computation. But maybe I and the folks I worked with were outliers. I know we referred to raw integer digital representations as "counts", and I don't see many using that terminology when describing cameras.
And, by the way, when one says that a circuit has some kind of gain, that doesn't mean it amplifies. Gains can be less than unity.
Strictly incorrect, I'd say - "gain":is yet another word that, used by itself, means anything that Chuck Norris says it means.
Having worked for sixteen years as an electronic controls engineer, I needed to be more specific, i.e. voltage gain, current gain.
I never encountered "charge gain" in my work, but have seen source follower FETs described as "charge amplifers". I never understood that term because charge=capacitance x voltage and there is rarely a big fat capacity on the output - and a voltage by itself has no charge at all.
Though these are often called 'gain' I would argue it's a misnomer, because if the units are different where is the gain? Suppose apples a kg of apples costs $1. Then the price chnages and the cost $1.50. Is that gain? I would say that 'gain' without qualification has to be dimensionless.
I think that one has to have cognisance of the function you're talking about. When it's a case of a function between exposure measurements and points in a colour space, I can't see that would naturally be described in terms of 'gain'. It's bound to be a mathematical function, and therefore mathematical operators are the natural terminology, whether or not those operations are the result of digital or operational circuitry.
Which is at the root of my object to the use of 'gain' in this context, it naturally engenders an idea of light (or an equivalent) being 'gained' when there is 'not enough' rather than the essentials of the rendering process. I like to describe that to people unfamiliar as akin to painting by numbers - and indeed it can be performed by look-up tables and often is. I find it really unnatural to talk about LUTs in terms of 'gain' and I think that most people would,
In my book, the term "gain" without a qualifier is an increase of something in size, amount, or value ... thank you, Bard ;-).
I use 'charge gain' because Eric Fossum once corrected me thus, and who am I to argue?
Considering also that charge = current x time, then in some contexts it can be said that source follower (current amplifier) "amplifies" charge too. Input current here can be for example decharging current of some capacitor, like pixel well or something similar.
Which is exactly the point. Bothe the pixel SF and the input to the column amp are MOSFETs, with at least in theory infinite input impedance. So what we are interested in is that the charge on the input of the column amp is larger than the charge on the input of the pixel SF. How long it takes has some importance, but secondary.
Unlike the old days, eh?
Today I forgot to check my camera settings and shot at 400 ISO by mistake:
I was 17 when i passed my full ham licence, back then i could have designed and built a transceiver from memory . not any more "use it or loose it" 😁
...or you can set shutter speed and f/stop and let the camera worry about ISO.. Often useful when working with long lenses...
My everyday 'snapshot' mode is A mode with auto ISO, minimum shutter speed set to Auto (which means it's FL related) with a suitable scale factor. All I need to worry about is setting teh DOF I want and then the ISO display to see if I might need to compromise that. If I need to control the shutter speed for more than stopping shake, I just switch to M and leave auto ISO on, then I have shutter control as well.
The Source Follower in a 4-transistor Pixel implementation is necessary because otherwise reading the e- charge off the photodiode's floating diffusion sense node would effectively destroy it. It is a necessary evil because its gain is typically less than one and it introduces noise that dominates what we normally call read noise around here.
Some references for the above at the bottom of this article
www.strollswithmydog.com/photographic-sensor-simulation/
