Your tolerance to noise.

Exercise: Your tolerance to noise.

I used the  Kingsdown again for this exercise.  The camera was my Canon G1X.

With the camera mounted on a tripod and set at f16 I took eight images starting at ISO100 and working through to ISO12800.  The shutter speed spread was from 1/25sec to 1/3200sec.

The ISO 100 image was as good as the lens and sensor could gather with each pixel, as seen at 400%, showing its true colour and tone.  There were no lost pixels.

At ISO 200 the colours were less saturated.  No lost pixels.  No great change.

At ISO 400 the green areas were showing some graininess.  The image is starting to degrade but this is only visible at 400 %.

At ISO 800 the pixels are losing the identity at any change of colour or tone.   This is causing a softening of the image.  At normal viewing size this image is acceptable but at 400% the changes are obvious.

At ISO 1600 the pixels are clumping in groups of up to four which is leading to real loss of detail.  This is now visible at even normal viewing size.

At ISO 3200 the loss of detail is becoming obvious.  At 400% the reason is clear.  The pixels are forming large groups and the grain is masking any fine detail.

At 6400 the image is taking on the look of a Seurat painting.  All fine detail has been lost and edges are now blurred.

At 12800 the image is now no more than impression of the original.  At 400 % the Seurat look is complete because the image is so grainy.  There are now lost pixels that show a colour or tone that has little to do with what should be there.

The progressive increase in noise was linear and predictable.  The lesson is that one should use the lowest ISO that is practically possible.  There are times, however, when one is forced to use a higher ISO.  It is better to get an image, even if grainy, than none at all.  I  normally try to restrict myself to no more than ISO 400 on the Canon G1X and ISO 800 on the Olympus E3.

Below I produce two images, one at ISO 100 and one at ISO 12800.  They are of a detail in the image and are at 400%.  No playing with the noise filters would improve the second image.