When can you say "Enough subs!"

[namreg]

Hi guys

Wonder if anyone can answer this question.

I have a basic DSLR setup and, being unguided, am restricted to sub 2 minute exposures. As such, it is obviously necessary for me to take a large number of subs to get a decent SN ratio. I assume though that there must come a time when the SN ratio is as good as it's going to get and that adding more subs is not going to improve the image any further i.e. (and using unrealistic numbers to make the point) is 2000x60sec subs any better than 1000x60 sec subs?

If you know the answer to that you may also know what the cut-off point is, although that probably depends on the target I suppose.

Thanks guys.

[Demonperformer]

I think improvement is proportional to the square of the number taken, so

10 = 3.16x better

100 = 10x better

1000 = 31.6x better

2000 = 44.7x better

[namreg]

Thanks for that. That would suggest an improvement, relative to one sub, ad infinitum, which may be the case.

Probably a better measure would be the reduction in noise relative to the number of subs. There must come a time when the noise level is virtually zero (or at an acceptable level), and that would be the time to stop adding subs. Does anyone know how to work that out?

Thanks

[Demonperformer]

Sorry, Doug, obviously didn't express myself very well.

If you stack 100 subs you would have 10% of the noise, 400 subs would give you 5% of the noise, 1000 subs would give you 3.16% of the noise and 2000 would give you 2.23% of the noise. All compared to a single sub. Basically for each increment you have to take 4x as many subs to half the noise.

HTH

[RobH]

In practical terms, over 100 subs won't really improve things for you.

I go for between 50 and 100 (for luminance), depending on the brightness of the target, and no less then 25 each per colour channel.

What I also do, as I image over several nights, is do a quick stretch of any data I've gathered, and see if I start to run into noise issues. If I do, I shoot more subs until I can stretch the image as far as I want without noise.

Cheers

Rob

[namreg]

Cheers guys

What I also do, as I image over several nights...

That'd take several months, at the current rate of one clear night per month!

[RobH]

Cheers guys

That'd take several months, at the current rate of one clear night per month!

I'm actually working on a few projects over several years, and keep adding data each season....it pays to be very patient

[namreg]

So it seems!

Think I'll take the "patience of a saint" out of my sig!

[PortableAstronomer]

Stacking lots of short expsures will give you a nice clean, low-noise image, and get you a little bit deeper into the fainter stuff than just a few sub frames, but to get into the really faint parts you need longer subs. Noise increases by the square root of the number of frames, but the quality is proportional. So after stacking the noise becomes less and less in comparison to the 'light'/good signal.

[kookoo_gr]

you might also want to read these

Starizona's Guide to CCD Imaging

Starizona's Guide to CCD Imaging

Starizona's Guide to CCD Imaging

[namreg]

Thanks guys

[Atacamallama]

I'm actually working on a few projects over several years, and keep adding data each season....it pays to be very patient

I've been meaning to ask about this, when adding subs from different sessions do you just put them all together or process each night separately and then combine them.

What about darks?

Is it possible to combine subs of different exposure times? - and other settings for that matter?

And forgive me if this sounds naive, but will something like DSS still combine frames for different nights if the exact orientation is not the same on each night?

TIA

[Demonperformer]

You can put the lights & darks from each night in separate tabs in DSS, just putting you bias master frame in the main tab

[RobH]

I've been meaning to ask about this, when adding subs from different sessions do you just put them all together or process each night separately and then combine them.

What about darks?

Is it possible to combine subs of different exposure times? - and other settings for that matter?

And forgive me if this sounds naive, but will something like DSS still combine frames for different nights if the exact orientation is not the same on each night?

TIA

I put all of my subs from all the sessions together and stack the lot, but I've already calibrated them before I do it, as obviously the flats will certainly be different.

I put my calibration frames from different sessions in the folders containing the lights from those sessions, so they are all there when I need them.

Don't know about DSS, but maxim will align different orientations if they aren't too far apart. Otherwise, Registar has no trouble aligning different subs and I then just put them back into maxim for stacking.

[dph1nm]

There must come a time when the noise level is virtually zero (or at an acceptable level), and that would be the time to stop adding subs. Does anyone know how to work that out?

This is like how long is a piece of string .There is always something fainter however long you image. So the answer is to stop when you have the image looking as you want it.

BTW I read on the DSS yahoo group the other day that the "auto adaptive weighted average" is the correct way to combine completed images taken at different times.

NigelM

[ollypenrice]

It is certainly nice working with an unkillable amount of data. I combined two 5.5 hour images of M42 and could give it a nice old extra stretch in the faint areas once done.

But longer subs, hence guiidng, really are the way forward.

I am sometimes guilty of not restacking but combining one stack with a later stack. I haven't thought through the maths but I don't get the feeling there is much of a price to pay for this shortcut. I may be quite wrong.

Olly

[roundycat]

I have combined frames in batches in the past and also combined them as one single operation deliberately to see what the difference was. I could not see a difference. I think the maths would support this as well. The first thing is to increase the exposure length to the sky limit; if you have guiding problems - solve them. Then combine at least ten frames and you should have a noise free background. As measured by visual assessment.

Dennis

[themos]

One thing to watch out for is that sometimes taking OUT a few bad subs can make a better image. Keep an eye on any "quality" metrics your software supports and make sure you are stacking similar subs of good quality.

[namreg]

I tend to take more subs than I'm after anyway, allowing for those I'm going to have to bin. If I want 70 I'll take 100, have a quick look to see what the first 20 or so are like and if there aren't too many dud ones I'll tell DSS to use 70%

[ollypenrice]

I have combined frames in batches in the past and also combined them as one single operation deliberately to see what the difference was. I could not see a difference. I think the maths would support this as well. The first thing is to increase the exposure length to the sky limit; if you have guiding problems - solve them. Then combine at least ten frames and you should have a noise free background. As measured by visual assessment.

Dennis

Good info Dennis. If you have done the test I will stick with batches because it is much faster in Registar unless I'm missng something.

Olly

[FraserClarke]

combined frames in batches in the past and also combined them as one single operation deliberately to see what the difference was. I could not see a difference. I think the maths would support this as well

It depends on the maths you use to do the combining For a weighted mean (which is a pretty standard way of combining) it's mathematically identical. If your batches have the same number of subs in them (of the same length), you can even get away without the weighting.

For more complex routines, where you try to reject certain pixels 'cleverly' (i.e. to get rid of cosmic rays), then the results will be a bit different (mathematically at least -- they may not look any different to the eye of course).

[roundycat]

I find average to be completely useless. I always combine using SD Mask as the output.

Dennis

[RobH]

I find average to be completely useless. I always combine using SD Mask as the output.

Dennis

Me too Dennis, and I try to only combine subs of the same length.

Rob

[rfdesigner]

Hi guys

Wonder if anyone can answer this question.

I have a basic DSLR setup and, being unguided, am restricted to sub 2 minute exposures. As such, it is obviously necessary for me to take a large number of subs to get a decent SN ratio. I assume though that there must come a time when the SN ratio is as good as it's going to get and that adding more subs is not going to improve the image any further i.e. (and using unrealistic numbers to make the point) is 2000x60sec subs any better than 1000x60 sec subs?

If you know the answer to that you may also know what the cut-off point is, although that probably depends on the target I suppose.

Thanks guys.

One comment I haven't seen made is about readout noise.

As readout noise is fixed it can mean you get disproportional benefit from slightly longer frames at the short end of the scale.

IMHO I'd want the subs to be long enough that the dark current noise is significant compared to the readout noise. Have a play with darks and lengthen them until the RMS noise (not the absolute level, the variation in the level) is significantly above the level of the shortest frame

Derek

[namreg]

Thanks Derek.

And in English....?