Whats causing this!?

[Rustang]

I captured some Ha on the western veil last night and my stack straight out of DSS isant great and I havent changed any settings! Below is a crop to show the issue/noise!? Any ideas? my subs dont look that bad!

QHY 9 mono CCD

 

Edited July 18, 2021 by Rustang

[vlaiv]

What exactly is bothering you about that image?

I can see few things that I consider bad in image:

1. Image is over sampled, stars are not pin points, but rather "balls" of light

2. There is black point clipping evident - there is no smooth transition between faint regions and background but rather steep one. Data has been pushed more than it can handle and histogram has been clipped

3. Noise grain is too large - that is just consequence of DSS and the way it aligns subs - it uses bilinear interpolation. If you want to get better looking background noise (finer grain) - use Siril and Lanczos interpolation for frame alignment

[Rustang]

37 minutes ago, vlaiv said:

What exactly is bothering you about that image?

I can see few things that I consider bad in image:

1. Image is over sampled, stars are not pin points, but rather "balls" of light

2. There is black point clipping evident - there is no smooth transition between faint regions and background but rather steep one. Data has been pushed more than it can handle and histogram has been clipped

3. Noise grain is too large - that is just consequence of DSS and the way it aligns subs - it uses bilinear interpolation. If you want to get better looking background noise (finer grain) - use Siril and Lanczos interpolation for frame alignment

I havent done anything to it Viaiv, this is just the Stack straight out of DSS no processing what so ever and my issue is why is the Stack so bad to start with!?

Edited July 18, 2021 by Rustang

[vlaiv]

7 minutes ago, Rustang said:

I havent done anything to it Viaiv, this is just the Stack straight out of DSS no processing what so ever and my issue is why is the Stack so bad to start with!?

Not sure if I understand.

Can you post linear data in 32bit FITS format for inspection? Don't apply any curve changes from DSS to it.

[Rustang]

5 minutes ago, vlaiv said:

Not sure if I understand.

Can you post linear data in 32bit FITS format for inspection? Don't apply any curve changes from DSS to it.

So the above crop was from the 32 bit stack simply opened in photoshop and this is what it looks like, no processing what so ever! Stack attached

ShortWestVeilHaStack.TIF

Edited July 18, 2021 by Rustang

[vlaiv]

Same region as opened in ImageJ without any stretching:

Same region in Gimp after light stretch:

Btw, data is looking rather noisy if you push it a bit more, how many subs did you stack and what stacking parameters did you use?

[Rustang]

Its 2hrs 40mins worth so not long but not short! I have tried different stacking parameters and it makes no difference! I'm guessing that's the data worthless then!? Ive sort of noticed in other images that dark areas (sky) are noisy-er but this gets hidden better in processing on nebulas with lots of nebulosity such as the last image of NGC7000 so maybe because there isant much nebulosity and more dark sky the noise issue is showing up alot more than usual with more! I wounder if its just how it is with the camera or something else then!?

[Rustang]

This is what I mean above - the noise issue is still present in my NGC7000 stack with a little stretching! Look at the darker areas, they are even worse! yet the nebulosity is fine!!!?? Ive been putting up with it pretty much since using the camera as i kinda end up losing the noise but now with an image with so much sky its reallt not good so would be good to understand where the issue is coming from? Camera? intergaration time? camera settings? stacking issues? etc etc

Edited July 18, 2021 by Rustang

[vlaiv]

If you are concerned with stacking parameters - try setting everything to simple average (no rejection, don't use median - put everything on average) - and see how much improvement there is.

In the end - it could be that nebulosity is faint and you need more integration time - but I'd check with above average approach, just to make sure.

[Rustang]

8 minutes ago, vlaiv said:

If you are concerned with stacking parameters - try setting everything to simple average (no rejection, don't use median - put everything on average) - and see how much improvement there is.

In the end - it could be that nebulosity is faint and you need more integration time - but I'd check with above average approach, just to make sure.

Ok, il give that ago, from what Ive said above then do you think its just either integration time or some settings in DSS nothing else of concern?

[vlaiv]

3 minutes ago, Rustang said:

Ok, il give that ago, from what Ive said above then do you think its just either integration time or some settings in DSS nothing else of concern?

Yes, I'd like to rule out any strange setting in DSS that might cause problems (like setting rejection too high so you end up with just one or few stacked subs and similar) - but Veil is really faint and delicate structure. Not much signal there and signal that is there is rather faint - or rather "in strands / filaments" not in bulk so you can clearly display it in image.

[Rustang]

18 minutes ago, vlaiv said:

Yes, I'd like to rule out any strange setting in DSS that might cause problems (like setting rejection too high so you end up with just one or few stacked subs and similar) - but Veil is really faint and delicate structure. Not much signal there and signal that is there is rather faint - or rather "in strands / filaments" not in bulk so you can clearly display it in image.

Average does seem better!

AveragetestStack.TIF

[vlaiv]

6 minutes ago, Rustang said:

Average does seem better!

Indeed it looks a bit better.

Same region as before - but I did apply a bit of noise reduction this time.

I think it is mix of both - faint target and some setting in DSS.

[Rustang]

1 minute ago, vlaiv said:

Indeed it looks a bit better.

Same region as before - but I did apply a bit of noise reduction this time.

I think it is mix of both - faint target and some setting in DSS.

So going forward then, what ever the target, just use Average!?

[vlaiv]

26 minutes ago, Rustang said:

So going forward then, what ever the target, just use Average!?

Sigma reject is good algorithm that for most part should behave like average.

There are two important things to understand with sigma reject:

1. Meaning of sigma

If we have normal distribution of data, then 68.27% of all data samples falls in one sigma interval around mean value. 95.45% falls into 2s interval and 99.73% falls into 3s interval.

If you for example put 1 sigma as your parameter and have perfect data without any satellite trails or whatever - it will only stack about 70% of your subs. For regular data you want to have sigma between 2 and 3 because you want to stack between 95% and 99% of your subs.

2. For above to work properly - you need to have normalized data. Way you normalize your subs is very important

As target moves across the sky - it changes amount of atmosphere lights get thru and hence brightness of the target. If one frame is taken in zenith and another frame is taken when target is at 45° altitude, then second one will be just 70% signal strength compared to first one

Transparency changes during the night and so does light pollution (target moves and people turn lights on/off). This thing with light pollution changes background value.

This makes data incompatible for sigma reject, and in order to make it compatible - well you need to normalize it.

I'm afraid that I never had much faith in DSS frame normalization process - primarily because I don't understand how DSS does it. I know how it should be done - and options present in DSS don't really fit into my understanding of it:

In any case - you should fiddle with these settings - probably use Rational instead of linear and use middle for calibration method or something like that.

Don't use median methods (like in image above) - use regular average methods (simple Kappa Sigma, and keep number of iterations to say 3, so 3 and 3 is good value for Kappa-Sigma clipping).

[Rustang]

3 minutes ago, vlaiv said:

Sigma reject is good algorithm that for most part should behave like average.

There are two important things to understand with sigma reject:

1. Meaning of sigma

If we have normal distribution of data, then 68.27% of all data samples falls in one sigma interval around mean value. 95.45% falls into 2s interval and 99.73% falls into 3s interval.

If you for example put 1 sigma as your parameter and have perfect data without any satellite trails or whatever - it will only stack about 70% of your subs. For regular data you want to have sigma between 2 and 3 because you want to stack between 95% and 99% of your subs.

2. For above to work properly - you need to have normalized data. Way you normalize your subs is very important

As target moves across the sky - it changes amount of atmosphere lights get thru and hence brightness of the target. If one frame is taken in zenith and another frame is taken when target is at 45° altitude, then second one will be just 70% signal strength compared to first one

Transparency changes during the night and so does light pollution (target moves and people turn lights on/off). This thing with light pollution changes background value.

This makes data incompatible for sigma reject, and in order to make it compatible - well you need to normalize it.

I'm afraid that I never had much faith in DSS frame normalization process - primarily because I don't understand how DSS does it. I know how it should be done - and options present in DSS don't really fit into my understanding of it:

In any case - you should fiddle with these settings - probably use Rational instead of linear and use middle for calibration method or something like that.

Don't use median methods (like in image above) - use regular average methods (simple Kappa Sigma, and keep number of iterations to say 3, so 3 and 3 is good value for Kappa-Sigma clipping).

Ok, il try and get my head around all of that, it all makes some sense etc but I never realised that DSS could throw out some subs by itself! Thanks again for all your help, I owe you yet another beer! 👍

[david_taurus83]

2.5 hours of integration isn't enough imo. My Atik 460, which is more sensitive, doesn't yield acceptable noise reduction until I get around 6 hours or so of data. Even when I had my ASI1600 I was never impressed with anything less than 4 or 5 hours per channel. The reason the dark areas look noisy is because their low signal. That's the reason we keep going for more data, to smooth out those faint areas.

[Rustang]

I will hold my hands up and say I've been lazy with my integration time, I'm still fairly new to Narrow band imaging and probably will in time settle to longer projects, i.e longer integration time but for now its been fun trying lots of targets!  

[vlaiv]

38 minutes ago, Rustang said:

I will hold my hands up and say I've been lazy with my integration time, I'm still fairly new to Narrow band imaging and probably will in time settle to longer projects, i.e longer integration time but for now its been fun trying lots of targets!  

This was with taken with narrowband filters and QHY 9 camera?

What sort of exposure length did you use?

[Rustang]

1 hour ago, vlaiv said:

This was with taken with narrowband filters and QHY 9 camera?

What sort of exposure length did you use?

Yeah that's right, it's Adams old QHY 9 mono CCD with Baader NB filters. I've been doing 10min exposures for all of my projects. 

This short intergration time was only just to see what the data was like to start with, I do this with all of my images, stack the first couple of hours or so just to see before then gaining some more data. It was because I've never had it turn out like the above before which is what lead me to ask for help, my first initial test stack is normally noisy but never what was happening above!. I do normally try and get around 4-5hrs on each wavelength and know I should still be going for more. 

Edited July 19, 2021 by Rustang

[vlaiv]

20 minutes ago, Rustang said:

Yeah that's right, it's Adams old QHY 9 mono CCD with Baader NB filters. I've been doing 10min exposures for all of my projects. 

This short intergration time was only just to see what the data was like to start with, I do this with all of my images, stack the first couple of hours or so just to see before then gaining some more data. It was because I've never had it turn out like the above before which is what lead me to ask for help, my first initial test stack is normally noisy but never what was happening above!. I do normally try and get around 4-5hrs on each wavelength and know I should still be going for more. 

See if you can get longer exposures as well. That camera has ~8e of read noise. Ideally for NB you want something like 20-30 minutes of exposure if your gear is up for it.

[Rustang]

35 minutes ago, vlaiv said:

See if you can get longer exposures as well. That camera has ~8e of read noise. Ideally for NB you want something like 20-30 minutes of exposure if your gear is up for it.

Blimmey! what about light pollution, I'm in bortle 6!? would that limit my exposure time? obviously I can just try and see what happens. That will be a long process to get set of darks completed though! 😩