Duplicating in the dark...

[Sterrenland]

As I tried to get to sleep after an evening of high cloud and no astronomy, I began to wonder about various things.

One nagging question was why don't we take a handful of darks, maybe 5 or 6, and then just duplicate them to create a batch of however many you need, rather than spending maybe hours taking them outside when we could be gathering more lights?

And dreaming of lights...what would be wrong with picking out one or two outstanding shots as determined in SubFrameSelector in PI, for example, and then duplicating a few of these? Would these not add to the quality of the stack?

Having thought these thoughts, I finally fell asleep.

[MarkAR]

Duplicating exposures just doesn't work.

The noise picked up in darks is not consistent in each frame which is what you want to cancel out noise. 

With lights you will be doubling the signal but also doubling the noise.

I'm no expert but it's something along those lines.

[Sterrenland]

7 hours ago, MarkAR said:

The noise picked up in darks is not consistent in each frame which is what you want to cancel out noise. 

I assumed the 'noise' in darks was pretty consistent between frames which is why it removes dust motes, etc.

7 hours ago, MarkAR said:

With lights you will be doubling the signal but also doubling the noise.

But isn't that the case with every extra light you add to an evening's stack? Choosing to duplicate one or two of your best lights that have captured excellent seeing, etc. must surely be a good thing? 🤔

Just wondering.

[MarkAR]

5 hours ago, Sterrenland said:

I assumed the 'noise' in darks was pretty consistent between frames which is why it removes dust motes, etc.

It's the flats that remove dust motes

5 hours ago, Sterrenland said:

But isn't that the case with every extra light you add to an evening's stack?

No, using a copied image has exactly the same noise pattern. Using different images has differing noise patterns.  If they didn't then you would only ever have to take one image of any single object.

Try an experiment, use one single image, copy it 10 times, the same for darks flats and bias. Stack them and see what comes out.

Edited May 22, 2020 by MarkAR

[tomato]

Just to add to the other posts, on a light frame the signal is pretty much constant, but the noise Is random, but is always present, so stacking original frames increases the signal but doesn’t increase the noise at the same rate so there is a benefit.

I have tried the copying the best images approach, it doesn’t work. It can help to not include your poorest images however.

Pixinsight can make a ‘Superbias’ frame from a limited number of actual bias frames, but that is a very specific type of calibration frame, a world away from light frames.

[CloudMagnet]

23 hours ago, Sterrenland said:

And dreaming of lights...what would be wrong with picking out one or two outstanding shots as determined in SubFrameSelector in PI, for example, and then duplicating a few of these? Would these not add to the quality of the stack?

Imaging and stacking is all about increasing the signal to noise ratio. You want to maximise the useful signal and lower the noise as much as possible.

For given single image lets say it has a signal of 10 and a noise of 2 (giving an SNR of 5)

If you just mulitply the frames by 10, you now have a signal of 100 but noise of 20. You still have the same SNR so no improvement is made. The maths behind stacking is that you will average out the noise but still add together the signal. So you would still have a signal of 100 but might have noise of 5, so your SNR is now 20 giving a better quality image. Its probably more complex than this, but it gives you an idea of the theory behind it.

 

[Sterrenland]

Thanks for everyone's input...I was just in a kind of wondering frame of mind. What people have said makes sense.

[MarkAR]

I actually did wonder the same thing a while ago, then realised no one was doing it so there must have been a really good reason.