if the shot is bright then the signal to noise ratio is inherently low, if the shot is dark then the signal to noise ratio is high.
Other way round.It depends on the camera, some modern cameras (Nikon and Pentax particularly, less so Canon) allow you to push the shadows with little ill effect, have a Google for "iso invariant cameras".
But otherwise it is basically to do with the initial brightness of the shot - if the shot is bright then the signal to noise ratio is inherently low, if the shot is dark then the signal to noise ratio is high.
I'm curious: why is it exactly that I can shoot at, say, ISO 400, underexposing the scene, and get a large amount of noise in the photograph, as much or more noise than shooting at 800-1000 ISO and exposing correctly?
The slightly earlier Sony sensors, such as on the NEX7 and A77, are pretty close to ISO invariant. It was that generation of sensors which started all the discussion of ISO invariance. Rather like focal length equivalences there are endless debates about how many factors have to be invariant, over what ranges, and how close to exact invariance they have to be, to count a sensor as near enough ISO invariant. When I upgraded to an A77 from an earlier generation sensor I found the difference in dynamic range and shadow liftability quite startling. I only heard about ISO invariance years later.You get more noise when under-exposing and brightening the image in post processing, than you do by shooting at a higher ISO in the first place. This is because the gain applied before the analogue-to-digital conversion stage, ie by raising ISO, is much more efficient than brightening in PP.
The new breed of so-called 'ISO-less' or 'ISO-invariant' sensors are much better in this respect and shadow areas can be brightened in post with with little or no noise penalty. You still get noise of course, but no more than if you'd raised ISO first. There is only a few cameras around like this, notably those with the latest Sony full-frame sensors (Nikon D810, Nikon D750, Sony A7Rii although the Canon 5Div is also very good) and this has major implications for exposure control.
Yup. When the amount of signal is low relative to the amount of noise, then the SNR is low... although the difference between levels may be greater. (SNR is a measure of how much *more* signal there is than noise. when noise is greater than signal the SNR is negative)I think you may have that the wrong way round![]()
It's all about how much light the sensor actually receives, and ISO has no affect on that. There is a common misconception online that ISO is the sensor's "sensitivity" to light. It is not, it's amplification after the fact...like a volume/brightness adjustment on your computer.I'm curious: why is it exactly that I can shoot at, say, ISO 400, underexposing the scene, and get a large amount of noise in the photograph, as much or more noise than shooting at 800-1000 ISO and exposing correctly?
You get more noise when under-exposing and brightening the image in post processing, than you do by shooting at a higher ISO in the first place. This is because the gain applied before the analogue-to-digital conversion stage, ie by raising ISO, is much more efficient than brightening in PP.
It's all about how much light the sensor actually receives, and ISO has no affect on that. There is a common misconception online that ISO is the sensor's "sensitivity" to light. It is not, it's amplification after the fact...like a volume/brightness adjustment on your computer.
Camera dependent... but the best answer is always to record as much actual light as possible w/o clipping (i.e. don't use ISO to push the exposure).Or is it better to overexpose slightly and reduce it slightly later in PS