Best amount of images for stacking

Ferguson wrote:

<snip>

Generally speaking for light frames the more data the less noise, the noise goes down with the square root of the number of images roughly, so 4 times the images is half the noise, 100 times is 1/10th the noise.

<snip>

Linwood

Part of the reason that people suggest little improvement after 25 lights is that the perceived reward drops rapidly because of the square root factor that Linwood points out.

If you take one light frame and compare it to a 4 image composite, you are delighted. The noise is reduced by a factor if two. And, wow! But to get the next factor of two you need to take 16 images (4X4). Wow again (maybe). And the next factor of 2 requires 64 images (4X4X4). By this time your reaction may be "ouch", not "wow".

Or more likely, after taking 16 images you only take another 16 images and look at the results. They are not impressive because it isn't a factor of 2 better. And you get discouraged.

We feel the pain linearly with the number of images, but the final product only gets better as the square root of the (total) number of images.

So more is always better, but you have to realize that each increment requires a lot more work.

There are other factors, such as read noise, which does not improve with the number of images, so the above explanation is a simplified one. But it captures what you need to know to answer your question.