M31 - Comparing 18.5 hour integration to 10 hour integration

[Notty]

Really nice image(s) I'd take either or both of them & mount them on my wall with pride. TBH I seem to hit diminishing returns after 4-5 hrs anyway as the spread of quality of subs (according to DSS anyway) just widen downwards the more I take so I end up stacking the same old 60% from that flukey clear night anyway!  I'd love to get those red coals I've seen in others as well as your excellet image - I assume they're from the Ha input? I use Star Tools and if I use Ha as a Lum layer it just borks all my colours up!

[eshy76]

On 18/09/2021 at 15:04, Mr Spock said:

Beautiful image!

Thank you!

[eshy76]

On 18/09/2021 at 21:17, Notty said:

Really nice image(s) I'd take either or both of them & mount them on my wall with pride. TBH I seem to hit diminishing returns after 4-5 hrs anyway as the spread of quality of subs (according to DSS anyway) just widen downwards the more I take so I end up stacking the same old 60% from that flukey clear night anyway!  I'd love to get those red coals I've seen in others as well as your excellet image - I assume they're from the Ha input? I use Star Tools and if I use Ha as a Lum layer it just borks all my colours up!

Thank you very much! Yes the reds come from the Ha data...without the Ha those areas are a fainter pink. My R channel is 50% Ha and 50% Red data (I added the Ha data to the Red after masking the stars).

[Xilman]

6 hours ago, eshy76 said:

Thank you very much! Yes the reds come from the Ha data...without the Ha those areas are a fainter pink. My R channel is 50% Ha and 50% Red data (I added the Ha data to the Red after masking the stars).

Ah, perhaps that's why the image is not as detailed as I expected. I had ascribed it to the relatively small aperture of your refractor.

I urge you to produce another stack, this time, without damaging the stellar detail.

[eshy76]

6 hours ago, Xilman said:

Ah, perhaps that's why the image is not as detailed as I expected. I had ascribed it to the relatively small aperture of your refractor.

I urge you to produce another stack, this time, without damaging the stellar detail.

Thank you, but I don't understand - I masked the stars before adding the Ha to the red channel - so no Ha data touched the RGB stars...how does that damage the stellar detail, or have I missed something?

[Xilman]

44 minutes ago, eshy76 said:

Thank you, but I don't understand - I masked the stars before adding the Ha to the red channel - so no Ha data touched the RGB stars...how does that damage the stellar detail, or have I missed something?

Perhaps I misunderstand too.  If you mask stars you destroy the data for them which resided in the red channel, right? You then put in the predominantly diffuse emission regions on top of the stars. By doing so, it seems to me, that you reduced the contrast between the stars and their background by lightening the latter. If this is wrong, please educate me!

What I am suggesting is that processing each channel in its entirity and, perhaps, mixing them in different proportions you will very likely enhance the visibility of some features at the expense of others.  You might find it amusing and/or worthwhile to process your data to show the faintest possible objects visible in the galaxy as sharply as possible.  The global view might not be so aesthetically pleasing in some respects but is rather likely to be more so in others. There is a fair chance, IMO, that you could produce images which rival those taken by Hubble on the Mount Wilson 100-inch Hooker telescope a century ago.

I think I am arguing for a scientifically interesting image to be produced alongside a pretty picture.

 

[eshy76]

15 minutes ago, Xilman said:

What I am suggesting is that processing each channel in its entirity and, perhaps, mixing them in different proportions you will very likely enhance the visibility of some features at the expense of others.

 

Thanks for the clarification - on the above part of your post....I actually prepared the red channel right at the start of the process. Masking off the stars in the red data means trying to protect them before bringing in the Ha data as a 50:50 blend...the idea being to only blend the "non-star" parts of the red channel. Not doing this and bringing the Ha data straight into the red has, in the past, led to overly red stars for me. After doing that I then combined the R+ Ha, G and B data into an RGB image in a non-linear state (with only a background extraction on each stack beforehand) and after having aligned the brightness of each stack.

Basically, what I am trying to say is I definitely did not process each channel before combining, more the other way around - combining first and then processing. Of course, it's probably easier to get to the most realistic colour mix using an OSC instead of filters, but probably at the expense of details and noise in my skies.

I appreciate your interest in the scientific side of the images - for sure that is not my primary imaging goal at this stage of my astrophotography!

And actually, in the images above, I have performed...star reduction....so these images are unlikely to be the best for someone interested in the actual stars themselves, alas! I do have the original stacks with no star reduction performed and, if I get some time this weekend, will try to process a version with the original stars intact and post it here.

[Rodd]

I see the difference immediately.  Both are great, but the dynamic range in the 18 hour image is greater.  The darks more dark (dust lanes).  If you look carefully at the dust lanes near the core, they are much easier seen in the 18 hour image.  Yes, perhaps this is nit picking at little details.  But you know who inhabits those little spaces

[eshy76]

On 20/09/2021 at 16:45, Xilman said:

 

I think I am arguing for a scientifically interesting image to be produced alongside a pretty picture.

 

Hi Xilman - please find below a version of the 18 hour integration with no star reduction performed. I hope this yields more details!

[eshy76]

10 hours ago, Rodd said:

I see the difference immediately.  Both are great, but the dynamic range in the 18 hour image is greater.  The darks more dark (dust lanes).  If you look carefully at the dust lanes near the core, they are much easier seen in the 18 hour image.  Yes, perhaps this is nit picking at little details.  But you know who inhabits those little spaces

Thank you! Yes, the more I look at them, the more I see the subtle differences. The background also looks a bit less noisy and less affected by gradients to me in addition to your observations on the dust lanes in the longer integration.

I'm sure if I took the time to take in the image, there's a lot more in there, but it's a clear night tonight, so...!

 

 

[ollypenrice]

On 17/09/2021 at 22:49, CloudMagnet said:

There was a thread on there a few months ago looking at a comparison on the Iris nebula on integration time and it seemed to be that once you reached 10 hrs, the level of improvement afterwards really tailed off and you couldn't see much difference.

I think that once you get about 10hrs, you should really look at adding only the best quality data rather than just pure amount of time to get the best results. Otherwise, adding in data that is lower quality won’t improve the final image as your noise level will already be so low anyway.

This would depend on several things. I would agree that ten hours per filter may take you to somewhere fairly close to what's possible but that might mean ten Hrs in Ha, another ten in O111, ten in Luminance and, say, 2 Hrs per colour. It also depends on how faint the target is and, crucially, on your ratio of pixel area to area of aperture. (This is what really matters rather than F ratio.) And then there's light pollution to consider. The worse it is, the more exposure you need. We cannot, therefore, easily quantify the benefits of long integration.

Olly

Edited September 23, 2021 by ollypenrice
Typo

[CloudMagnet]

8 hours ago, ollypenrice said:

This would depend on several things. I would agree that ten hours per filter may take you to somewhere fairly close to what's possible but that might mean ten Hrs in Ha, another ten in O111, ten in Luminance and, say, 2 Hrs per colour. It also depends on how faint the target is and, crucially, on your ratio of pixel area to area of aperture. (This is what really matters rather than F ratio.) And then there's light pollution to consider. The worse it is, the more exposure you need. We cannot, therefore, easily quantify the benefits of long integration.

Olly

Agreed, there is a much deeper discussion to be had about the benefits of integration time and when you enter diminishing returns. It's not a simple black and white answer to the "best" way.

It often comes down to an individual's equipment as you say and also their opinion on when they feel it is not worth pursuing an image any further. I could put 20 hrs into a single picture, but it would take me 6 months with the UK skies. I personally find it much better to instead take 4 hours on 5 different targets, even though those images might be lower quality. Someone with 100 clear nights a year would probably put 20hrs on a target quite easily.

[Lee_P]

17 minutes ago, CloudMagnet said:

I could put 20 hrs into a single picture, but it would take me 6 months with the UK skies. I personally find it much better to instead take 4 hours on 5 different targets, even though those images might be lower quality. Someone with 100 clear nights a year would probably put 20hrs on a target quite easily.

I image under UK skies and aim for around 20 hours per image. I produce an image approx. every one month with winter skies, two months in the summer.

[CloudMagnet]

8 minutes ago, Lee_P said:

I image under UK skies and aim for around 20 hours per image. I produce an image approx. every one month with winter skies, two months in the summer.

Feel like trading places?

My total imaging time since the start of July this year comes to about 15hrs I think. Last winter, I went from the start of November to the end of February without a single clear night, I think I might be cursed in the North East.

[Lee_P]

21 minutes ago, CloudMagnet said:

Feel like trading places?

My total imaging time since the start of July this year comes to about 15hrs I think. Last winter, I went from the start of November to the end of February without a single clear night, I think I might be cursed in the North East.

Well you are the CloudMagnet!

[Adam J]

I think that main thing i would do here is instead of stacking all the images i would use the extra images to enable stacking based on star sharpness that will likely yeald an increase in detail, if SNR is good adding more will not make things much better.

[Xilman]

21 hours ago, eshy76 said:

Hi Xilman - please find below a version of the 18 hour integration with no star reduction performed. I hope this yields more details!

Thanks.  I will try to analyse your image asap but I have other things on my mind right now ...

An all-sky camera at Tacnade obnservatory reports a light ash fall, 3.1 km north of the eruption.  The INT and NOT, at least, up on El Roque, are not opening tonight because of the ash.  They are around 12 -- 15km away to the north.

Edited September 23, 2021 by Xilman
Add 2nd para

[eshy76]

11 minutes ago, Xilman said:

Thanks.  I will try to analyse your image asap but I have other things on my mind right now ...

An all-sky camera at Tacnade obnservatory reports a light ash fall, 3.1 km north of the eruption.  The INT and NOT, at least, up on El Roque, are not opening tonight because of the ash.  They are around 12 -- 15km away to the north.

Whoa - yes of course...

[eshy76]

1 hour ago, Adam J said:

I think that main thing i would do here is instead of stacking all the images i would use the extra images to enable stacking based on star sharpness that will likely yeald an increase in detail, if SNR is good adding more will not make things much better.

Thank you for this - what I would say is that I stacked using APP's quality weighting approach, which hopefully emphasised the better subs in the pack, but it is still a bit of an automatic approach...

...there is enough good data here to do something I haven't done since buying APP - use some criterion to pick subs to stack. I think I might give that a go at some point! Thanks!