CatMarley Loc: North Carolina

Please tell me if my understanding is correct. The ISO is an analog amplification. The sensor is a light sensitive cell that reacts to light by generating an electrical signal. Photons are stimulating the flow of electrons. This is basically and analog process. Each cell has a limited capacity to convert light to electrical energy. Low amounts of energy creates a low signal, and at maximum capacity creates the top highlight signal. This signal is amplified depending on the ISO input. This energy is then digitized and processed into the raw file. So the ISO is actually an arbitrary counter more or less standardized to correspond with exposure values based on aperture and shutter speed. Each increment of EV corresponds to a doubling of the ISO.

R.G. Loc: Scotland

Until selmslie compiles his answer, I'll address these points.

Strictly speaking, photons release electrons, which generates charge. If you put a specific amount of charge into a capacitor with a known value of capacitance it will generate a specific voltage, and that is the signal that the sensor generates.

I believe that ISO is used to make the exposure produced by specific aperture and shutter speed settings to line up with what a standard exposure would be for those settings. I believe that in theory all cameras using similar shutter speed, aperture and ISO settings should produce a similar exposure, and the manufacturer has control over ISO to make that so. In practice I'm not sure how similar such exposures would be, but they must be broadly similar, or Sunny 16 and the like wouldn't work for any cameras that were an exception to that standard.

a6k Loc: Detroit & Sanibel

The sensor itself has exactly the same sensitivity regardless of the ISO.
Yes. I was not disputing that if you mean "regardless of the ISO setting if above base". However, there are some cameras which state their base ISO as, for example, 50. There are others (ex: Fuji) which state the base value as 200. Most are using a value of 100 (ex: Sony and others).

On the one hand I think (layman's understanding) that the sensor material is pretty much the same in all of the CMOS sensors. If so, then they are merely varying the starting point for analog amplification. But perhaps due to design tweaks the difference is not due to amplification but to the sensitivity of the sensor itself. I don't know the answer but it seems logical to me that the question is at the root of this minor issue.

I'm really not addressing amplification at all. Or at least I'm trying not to do that.

..snip..

My earlier statement about a sensor that can use a lower ISO was not quite what I meant. What I meant to convey was that it has a greater capacity - deeper "buckets".

So I'm still thinking that a more sensitive sensor will produce the same output voltage pre-amplificaiton with less input in terms of photons. But a more sensitive sensor will thus have a higher ISO base number than in a camera with a different number. Yes? No? Other?

blackest Loc: Ireland

I think part of it is sensor area per cell, the iPhone has a very small sensor and base iso is around iso 25 iso 36, whats interesting is that at iso 200 the iq has degraded to the point where it's noticeable i'd rate the iq as similar in terms of noise as around iso 1600 or 3200 iso of my k5. Current technology seems to be around a stop better for noise but looking at similar generation sensors the larger the cell size the better the signal.

R.G. Loc: Scotland

Sounds about right (if you're comparing like with like).

selmslie Loc: Fernandina Beach, FL, USA

No, ISO is a digital amplification.

Some cameras (like Canon at low ISO) need to adjust the analog signal from the sensor as it is converted to a binary value for the raw file because the analog process is not sufficiently linear - directly proportional to the number of photons that struck the sensor. ISO invariant sensors have pretty much solved this problem but it is inherent in all analog to digital conversions and it is exacerbated by heat and the difficulty of manufacturing a perfectly uniform sensor.

Whatever signal is being measured at base ISO (100, 200 or whatever your camera supports) is multiplied by a constant representing the ratio of (new ISO)/(base ISO). A full stop increment would give you a multiplier of 2x but a third stop increment would use a multiplier of about 1.2599 (the cube root of 2). That is available with your camera's ISO dial.

But look at the Exposure slider in whatever program you are using. You can make even finer adjustments than that. Not just the cube root of 2 but as fine as the hundredth root, a multiplier of 1.00696. So although you can make a 1/3 step adjustment in your camera by changing the ISO, you can make a virtually identical adjustment on your computer with a 0.33 adjustment with the Exposure slider.

Analog amplification is just not good enough to deal with ratios that are so precise. The ten inch slide rules that we used in engineering half a century ago were a form of analog computer. They were capable of about 3 decimal digits of precision. To reach 4 digits of precision we would have needed a slide rule that was 100 inches long. Analog electronic computers faced a similar challenge with precision. Every extra decimal digit of precision might easily require a ten-fold increase in investment.

But digital 16-bit math easily produces 5 digits of decimal precision. Digital computers doing math with 32-bit numbers can deal with ten decimal digits of precision and 64-bit numbers can do integer math to twenty digits of precision.

Digital computers can easily add and subtract integers or floating point numbers but multiplication and division is not so direct. However, they can convert the numbers into logarithms, add or subtract them and then convert the logarithm back into integer or floating point numbers.

But the bottom line is that, if you multiply an integer by a floating point number and then truncate or round it back to an integer, you introduce rounding errors. It's the rounding errors that show up as banding in the blue or gray sky of an image and as noise in the shadows.

selmslie Loc: Fernandina Beach, FL, USA

Not all sensors contain exactly the same materials so there may be somewhat different voltages to be read. But the voltage may not be the only difference. One interpretation of Heisenberg's uncertainty principle is that you cannot accurately measure something without affecting its value, so a larger pixel or one with more capacity can be read more accurately.

So stating that one sensor is more sensitive than another is probably a gross oversimplification. Let's just say that, if the manufacturer provides a wider range of ISO settings, they are implying that they might have a more versatile product. Caveat emptor.

JD750 Loc: SoCal

Yes that is essentialy correct. Semisile is not wrong, but he is giving way more technical detail than is needed, nobody cares about those implementation details.

If I may, I would add that increasing the gain also (ISO) results in more noise in the image. Which is something we care about.

Camera makers and electronics engineers have worked hard to make cameras that photographers can relate to based on prior technology. After all many of us were very familiar with SLR cameras. And manufacturers gave us the D - SLR camera. Very similar, almost the same, as the beloved SLR, except for the sensor. And they gave is ISO to replace ASA, as a measure of sensor sensitivity. But ISO and ASA work in the same way as far as the user is concerned.

A very important rule of Product Development is "Never introduce a product that exceeds the understanding of your customers". It is only now that we are just starting to see the possibilities of abandoning the past and looking toward what the new technology can do by itself.

selmslie Loc: Fernandina Beach, FL, USA

ISO as a standard for film speed has been around since 1974, before digital. It's just a new name for ASA from a different organization. DIN is still with us.

selmslie Loc: Fernandina Beach, FL, USA

The relationship of ISO and noise is a chicken and egg situation. If you raise the ISO you can reduce the exposure but less light is more noisy because there are fewer photons being captured.

Close your eyes in a dark room and all you will see is noise. Open them and you might see dark colorless shapes but the noise is still interfering. That's analog noise.

But if you make two identical exposures at two different ISO settings in the camera, both images should contain the same level of analog noise. Increasing the brightness of the darker image on your computer will reveal digital noise as I showed here and here. That seemed counter-intuitive at first because the image taken with the lower ISO was noisier. But the noise in this case was the result of increasing the numeric values of small integers and revealing rounding errors.

But saying that the source of noise is analog or digital is an unwinnable argument. Both photons and electrons are integer quanta and they arrive at the sensor in a random nature. From that point on both analog and digital processes can drop the ball and cause the noise that we see.

But in the context of this thread, ISO invariance, the real limitation on invariance is the recovery of tonality from a raw file that was recorded with too little exposure for the chosen ISO or too low and ISO for the chosen exposure - the chicken or the egg. The recovery process on the computer is purely digital and it does a better job with tonality than with noise.

radiojohn

If I read the original post correctly, this technique depends on true manual exposure, not aperture-priority automatic.