1

My understanding is that many cameras have ISO settings which are beyond the "true" ISO capabilities of the sensor, and that to achieve higher values, the camera itself performs "back-end digital amplification".

Photons To Photos publishes a table of sensor characteristics, including a value of the "ISO invariant point".

I have a number of (possibly silly) questions:

  1. As an amateur photographer, how much do I need to worry about this?
  2. Does the value of the invariant point allow sensible comparison of the degree of ISO invariance between different cameras (of the same brand or different)?
  3. If so, is better ISO invariance associated with a higher, or a lower, point?
  4. Is there a way I can measure this for myself?

I'd also appreciate any links to tutorials on the subject.

5

You probably don't need to worry too much about it for many pictures. If you're using a boosted ISO level, it's presumably because you need it - and you're happy with the quality you get at that ISO setting. Think of it as an added bonus that lets you get shots you otherwise might not.

At digitally boosted ISOs, you have fewer possible brightness values in the output image (since it's digitally stretched, the number of distinct values is limited to the input range you're stretching from) , and if the boost was big enough to significantly reduce the possible values you might start seeing a posterisation effect - but you'd need a very big boost for that to be obvious. (And interpolation when debayering to get the final RGB result might smooth it over somewhat).

So for most applications it's unlikely to be particularly noticeable.

Where it would come into play, though, is for applications like long exposure astrophotography, where multiple images are combined ("stacked") to reduce noise and then stretched and postprocessed to produce the final image.

In this case, since the image is normally stretched anyway in post processing, there is no benefit in digitally stretching things when capturing the source images - and actually a penalty, since stretching it means it reduces the dynamic range since input values at the brighter end of the input range will all map to the maximum value.

For normal photography, what you get by digitally boosting the ISO is convenience. You get a usable result with convenient brightness levels without the hassle of messing around with stretching a dark raw image, and most of the time, that's all you need - especially if you just want a usable image that you're not going to post process much. You don't normally care too much that you may have blown out some potential bright highlights - and in many cases, you may not have any significant areas that would be stretched to the maximum value. It also means that you can review your pictures easily on the camera LCD, instead of seeing a dark result (even if you could achieve the same end result after postprocessing).

On the other hand, if you wanted to do some fancy post processing (basically anything involving a significant stretch to part of the brightness range) for images that contain both bright and shadow detail, then running at a non-boosted ISO gives you more distinct values to work with. In this case, the non-boosted ISO image (with more distinct values than the boosted version) will let you end up with smoother tonal gradation.

Since stretching happens when you start digitially boosting the result, a higher invariant point is better (from the point of view of avoiding the stretching). It doesn't, by itself, determine other quality factors. You might, for example, have a sensor with a lower invariant point that has better quality / lower noise or whatever, than one with a higher invariant point but with worse quality/ higher noise.

With that caveat - that it's only one factor affecting overall quality - you should be able to compare values directly. But remember it's only affecting the point at which you start getting fewer distinct values, which may not be your primary quality driver. If you're into low light photography - where this is more likely to be relevant - than things like noise levels may be as - or more - significant.

2

A lot of worry over nothing! The software built-in your camera adjusts the gain (amplification) as needed to deliver a good exposure. This is like turning up the volume of a radio. As you up the amplification, some good signal and some bad signal gets amplified. In radio, we call this static, In imaging we call this noise. As the signal to noise ratio increases, at some point noise is visualized as a granularity not unlike gain in a chemical (conventional) film photograph. The bottom line is, no amplification, no image in feeble light. If the amplification is not to be employed then you must bust the light with flash or bring in additional continuous lights.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.