There are good technical reasons, but there still is a design choice to be made. It has to do with how the sensor works and how "ISO" is implemented in a digital imager.
A quick summary of the path taken from photons to "bits":
- Photons hit the sensor and are converted to electrons (this is a photodiode).
- The electrons are stored in the pixel (this is typically a capacitor partially formed by the photodiode and any readout transistor). The stored charge of electrons in this capacitance produces a voltage in the pixel. Just think of a capacitor as a bucket that holds electrons just like water - the voltage on the capacitor will just be the level of the "water".
- At the end of the exposure this voltage/charge is read out by an amplifier - the readout method and amplification depend on whether it is a CCD or CMOS implementation.
- The amplifier gain combined with the characteristics of the pixel photodiode/capacitor determine the ISO. The amplifier gain is adjustable, the characteristics of the photodiode/capacitor are not. Adjustable ISO is implemented by changing the amplifier gain.
- This amplified voltage is passed to a analog-to-digital converter which changes the voltage into a number that the camera stores - typically a 12-bit number which means it has 4096 possible levels. Imagine a measuring cup with 4096 measuring lines on it.
A key point here is that the pixel can only store so much charge before it "overflows" - literally starts leaking charge into adjacent pixels. Prior to this occurring other troublesome things happen such as non-linear response. Bottom line is only so much charge, and equivalently so many photons can be shoved into a pixel before it starts to not work properly - just like a measuring cup in your kitchen can only hold so much water. A second key point is that besides the electrons converted from photons in the well there are also additional electrons from a variety of sources that we'd consider noise. Similarly a certain sized measuring cup in your kitchen is only accurate to a certain scale, you wouldn't want to try to use a 1 quart bottle to measure a teaspoon of fluid accurately.
So, how do we adjust ISO? For higher ISO settings the well fills with fewer electrons and we turn up the amplifier on readout. The result is we can use shorter exposures (need less photons) but as we amplify up the resulting voltage we also amplify all those noise electrons. In film terms this is very much like push processing film, you underexpose and overdevelop. It tends to work out much better in the digital world though - but the concept is the same. The limit is really just a question of how much noise you are willing to tolerate.
Well what about lower ISO settings? Well, remember we said that eventually a pixel can get to full and will stop functioning properly. Thus we can't arbitrarily lower the ISO by turning down our amplifier - given enough exposure, and thus photons, the pixel will just stop working properly. We could make the pixel less sensitive to begin with (a lower base ISO) but then our high ISO results would suffer.
So that's how things are limited, a pixel can only hold so much charge which sets the lower ISO limit and it contains a certain amount of noise which sets the upper limit.
There is one final trick possible, you could put a ND filter in front of the imager. Most folks put this on the outside of the lens, but a few cameras include one in between the lens and the imager. I believe the Canon G9/10/11 do this.
Hopefully that was vaguely understandable.
--
Ken W
Rebel XT, XTi, Pany LX-3, FZ-28, Fuji F30, and a lot of 35mm and 4x5 sitting in the closet...
