Digital Noise From the Sensor - some clarification

[RAKAMRAK]

Are "High ISO" noise and "long exposure" (at low ISO) noise the same qualitatively? Or is one slightly more managegable than the other? (Practical Reference: When photographing a night scene of a city is it preferable to choose high ISO and faster shutter speed, or is it preferable to choose low ISO slower shutter speed, assuming I do not want to consider the motion blur of moving subjects resulting from slower shutter speed as a factor)

[privatebydesign]

I would guess long exposure noise, it can be replicated and subtracted.  High iso noise is much more random.

[rs]

The noise is different. Unlike high ISO's, noise from long exposures is much more predictable, and easy to mostly remove with subtraction. Take a read about long exposure noise reduction, found on page 4 of this article: http://www.learn.usa.canon.com/resources/articles/2012/long_exposure_landscapes.shtml

[Pi]

Low exposure noise is photon noise, cannot be avoided. Long exposure noise (with enough exposure) is sensor noise, depending on the particular sensor. With extremely long exposures it can be really bad but with "normal" ones, it is less problematic than photon noise.

Subtracting a black frame (LENR) can cancel repeatable noise but it will increase the random noise. You lose even more DR, which is not great with Canon to begin with, and is really needed with night images.

In short, shoot at ISO 100 with LENR off.

[neuroanatomist]

Take multiple images at low ISO and a shutter speed of (desired SS / number of shots), i.e., if you want 4 seconds, take 8 shots at 0.5 s, then positively merge the resulting images.  Canon has implemented something like this as Multishot NR.  'Adding' the shots reduces the noise.

[Pi]

Take multiple images at low ISO and a shutter speed of (desired SS / number of shots), i.e., if you want 4 seconds, take 8 shots at 0.5 s, then positively merge the resulting images.  Canon has implemented something like this as Multishot NR.  'Adding' the shots reduces the noise.

What you probably want to say is to take 8 shots at 1/2 sec at ISO 100 instead of one at 1/2 sec at ISO 100, if the light is enough. Adding does increase noise (not different than subtracting) but averaging reduces it. When you add similar images, which differ by noise level only, you actually average them in the end.

If, instead, you take, say, 8 images at 1/2 sec each (and ISO 800), instead of one at ISO 100 and 4 sec, you can decrease the read noise (with Canon) but the shot noise is the same.

[RAKAMRAK]

I have not been chiming in after my first post, but I have been following your replies. Thank you all for your helps. So, now my understanding is low ISO (and the different techniques mentioned by Neuro and Pi) is the better choice.

[Kernuak]

But how about if the exposure time is much longer? For example comparison between ISO 100 for 40 minutes and ISO 400 for 10 minutes. I haven't tried the ISO 400 option, but I do know that noise from a 30-40 minute exposure is really high and I've come across information that it's better to use a higher ISO and lower exposure time, because of the heating up of the electronics.

[Pi]

But how about if the exposure time is much longer? For example comparison between ISO 100 for 40 minutes and ISO 400 for 10 minutes. I haven't tried the ISO 400 option, but I do know that noise from a 30-40 minute exposure is really high and I've come across information that it's better to use a higher ISO and lower exposure time, because of the heating up of the electronics.

As I said, with extremely long exposures, the sensor noise can be really bad. I would split it into several shots.

BTW, what exactly are you shooting at 40 minutes? For star trails, people usually stack 30 sec shots, but that is a lot of frames to stack. There is also the problem of pausing between the shots "long" enough to allow the sensor to cool down but not too long to avoid getting gaps in the trails.