Hi folks,
A slightly philosophical topic that I was curious about and hoping that the experienced folk on this forum might have a better view on?
Overall I would like to identify the best way of combining high SNR in deep sky images with the highest resolution (smallest and best looking stars). I enjoy looking at the close in detail of many DSOs. Sharpcap Brain solves optimization of SNR. But getting the best resolution at the same time as best SNR seems to be a challenge?
My best resolution final pictures (nominally about 4 arcsec so nothing brilliant but I think that Thames Valley seeing is probably habitually poor) seem to have been had by combining SC FWHM-filtered autostacks of very short exposures (e.g 3s) – short as of necessity using a (10”) Dobsonian reflector. Of course this set up is poor for capturing faint details – nebula extensions and galaxy arms. Nevertheless it consistently seems to have produced sharper resolution than longer exposure guided stacks for certain types of object - the cores of brighter galaxies, globular star clusters and bright planetary nebulae.
The general impression I get from SC and DSS FWHM values also supports the notion of better sharpness at short exposures. Despite guiding (e.g. an 8 inch Newtonian) at < 1-1.2 arcsec RMS (so the effect added in quadrature shouldn’t be that great?) the trend still seems to be that shorter individual exposures exhibited smaller FWHM values (by about 0.5 or so) allowing the FWHM filter to be set at a lower value. Stacking always results in a further increase in FWHM over single frames - but the final stacks of very short exposures remain sharper than stacks of longer exposures.
So is this a kind of ‘lucky’ imaging effect or is 3s much too long a time for such to apply (has to be < than 0.2s?) ? Or is it more likely that my longer exposure guiding or focussing or something is somehow faulty ? I may just be doing something wrong?
If, ‘lucky’ imaging by combining 2-3s exposures really is a thing ? -- then is judicious use of the brightness and FWHM filters in SC and stacking on the fly into autostacks an adequate method to exploit it and select the best frames? Operationally so much easier than dealing with thousands of frames in postprocessing selection.
I was thinking of maybe testing the notion more carefully on something bright and difficult like the double double (2.6 and 2.8 arcsec) in Lyra next Spring and seeing how short I can go at (F4.2)and possibly detect the split? I also wondered whether there is an easy way to combine the best of both worlds and to superimpose very short exposure stacks of the sharp bright centres of brighter DSO objects upon SNR-optimized longer exposure images that detect the fainter outer regions?
Anyway I’d be grateful for any comments/ thoughts on this - even if just to say it’s nonsense or to say that others see no diffeence in long and short exposure FWHMs - in which case I will know not to chase this.
Best wishes
TimH
Optimizing sharpness with SNR. DSO "lucky" imaging ?
Re: Optimizing sharpness with SNR. DSO "lucky" imaging ?
Tim
Research HDR techniques. A good land based explanation of HDR here https://astrotuna.com/category/hdr/ , courtesy of Brian (@oopfan).
The following are 3 stacks generated using Astro Pixel Processor - 60x10s, 60x20s and (60x10s + 60x20s). For the 3rd image all 120 frames were loaded, calibrated and stacked. The APP software dealt with the detail. Images captured with ZWO ASI 2600 and Sigma f/1.4 ART lens (stopped down to f/2).
I have also used the HDR facility in FOTOXX (which runs on Linux) to combine stacks. See https://www.astrobin.com/full/p63izu/0/ .
Dave
Research HDR techniques. A good land based explanation of HDR here https://astrotuna.com/category/hdr/ , courtesy of Brian (@oopfan).
The following are 3 stacks generated using Astro Pixel Processor - 60x10s, 60x20s and (60x10s + 60x20s). For the 3rd image all 120 frames were loaded, calibrated and stacked. The APP software dealt with the detail. Images captured with ZWO ASI 2600 and Sigma f/1.4 ART lens (stopped down to f/2).
I have also used the HDR facility in FOTOXX (which runs on Linux) to combine stacks. See https://www.astrobin.com/full/p63izu/0/ .
- comparison.JPG (123.2 KiB) Viewed 1563 times
Dave
-
admin
- Site Admin
- Posts: 13330
- Joined: Sat Feb 11, 2017 3:52 pm
- Location: Vale of the White Horse, UK
- Contact:
Re: Optimizing sharpness with SNR. DSO "lucky" imaging ?
Hi,
I think that there are a number of reasons why the short frames may give sharper stars:
* Less seeing variation over a shorter timescale. This is effectively the concept of lucky imaging that you are mentioning. I suspect though that the exposure length that you need to go below to get this effect to have much significance is very short – a few seconds or less. In general by the time you get into this range you are probably in the zone where you are not going to be able to get read noise optimised exposures.
* Less guiding variation over a shorter timescale. This is the same sort of effect as the above, but with the start position wobbling around not because of the atmosphere but because the telescopes tracking/guiding isn't quite perfect. This one will show an effect out to much longer exposures than the seeing related one. For instance 30s exposures may seem sharper than 60s exposures because of this.
* Bloating of stars due to oversaturation on longer exposures. Once the centre of a star has become saturated, it's brightness cannot increase any more as you further increase the exposure. However the surrounding areas will still increase in brightness until they to become saturated. SharpCap tries to avoid using saturated stars for FWHM measurements, but I think that DSS is happy to use them (it is looking for alignment points rather than sharpness measurements primarily). This is definitely going to be a real effect, but it should be potentially curable using the sort of HDR techniques that Dave mentions. I have the idea for HDR in live stacking as a long-term goal for a future version of SharpCap.
Cheers, Robin
I think that there are a number of reasons why the short frames may give sharper stars:
* Less seeing variation over a shorter timescale. This is effectively the concept of lucky imaging that you are mentioning. I suspect though that the exposure length that you need to go below to get this effect to have much significance is very short – a few seconds or less. In general by the time you get into this range you are probably in the zone where you are not going to be able to get read noise optimised exposures.
* Less guiding variation over a shorter timescale. This is the same sort of effect as the above, but with the start position wobbling around not because of the atmosphere but because the telescopes tracking/guiding isn't quite perfect. This one will show an effect out to much longer exposures than the seeing related one. For instance 30s exposures may seem sharper than 60s exposures because of this.
* Bloating of stars due to oversaturation on longer exposures. Once the centre of a star has become saturated, it's brightness cannot increase any more as you further increase the exposure. However the surrounding areas will still increase in brightness until they to become saturated. SharpCap tries to avoid using saturated stars for FWHM measurements, but I think that DSS is happy to use them (it is looking for alignment points rather than sharpness measurements primarily). This is definitely going to be a real effect, but it should be potentially curable using the sort of HDR techniques that Dave mentions. I have the idea for HDR in live stacking as a long-term goal for a future version of SharpCap.
Cheers, Robin
Re: Optimizing sharpness with SNR. DSO "lucky" imaging ?
Thank you both for your thoughts on this.
I had read something before about combining stacks of different length exposures - but didn't really understand how it works and they combine for HDR. That does seem to be the way to go then - and that will be something that I will explore. Thanks Dave for the links and names of some software to try this out with. It would be interesting if SC could one day incorporate it as a feature.
Anyway when I get the opportunity will try and look a bit more systematically at different exposure lengths on a single bright object using a single instrument on a single night plus pay serious attention to the guiding with PHD2 logs.
Incidentally - specifically on the double double question - do you know if UK-based imagers have been able to split them ?
Tim
I had read something before about combining stacks of different length exposures - but didn't really understand how it works and they combine for HDR. That does seem to be the way to go then - and that will be something that I will explore. Thanks Dave for the links and names of some software to try this out with. It would be interesting if SC could one day incorporate it as a feature.
Anyway when I get the opportunity will try and look a bit more systematically at different exposure lengths on a single bright object using a single instrument on a single night plus pay serious attention to the guiding with PHD2 logs.
Incidentally - specifically on the double double question - do you know if UK-based imagers have been able to split them ?
Tim
Re: Optimizing sharpness with SNR. DSO "lucky" imaging ?
Tim
An Astrobin search for 'double lyra' yields https://www.astrobin.com/search/?q=double+lyra. One of them was Belgium.
For my M42 HDR attempt, these are the FITS Liberator views (unstretched) for single frames at varying exposures. Trying to retain the Trapezium, I probably needed exposures less than 10s as well. After evaluating the histograms I took a 'best guess' as I believe that trying things is the best way to progress (or not).
Dave
An Astrobin search for 'double lyra' yields https://www.astrobin.com/search/?q=double+lyra. One of them was Belgium.
For my M42 HDR attempt, these are the FITS Liberator views (unstretched) for single frames at varying exposures. Trying to retain the Trapezium, I probably needed exposures less than 10s as well. After evaluating the histograms I took a 'best guess' as I believe that trying things is the best way to progress (or not).
Dave
Re: Optimizing sharpness with SNR. DSO "lucky" imaging ?
Tim,
I've held off from replying due to the word "DSO" in the title, but now that I see that the topic has shifted towards imaging stars, I can add something to the discussion.
Feel free to use sub one-second exposures for the Lyra double-double. They are bright at 6th magnitude. If you have a CMOS camera with adjustable gain, you might want to jack it up. I do exactly that when I use SharpCap to measure periodic error. With a 71mm aperture, I often use 350ms exposures with the gain cranked up to about 2000 on my Altair 290M. With my seeing conditions, a 350ms exposure is not short enough to "freeze" the atmosphere, but if I go down to 250ms then I begin to see the turbulence. The "threshold" exposure will be different for you, so you need to experiment. Set the frame rate to one per second, and then slowly shorten the exposure until you see the turbulence.
With regards to the output format, I output to PNG because my IronPython script needs to open it up to perform star detection. In your case, you might want to write to a SER file, and then break out the individual frames later on. Dave is the expert with that. Now, with regards to processing, you may be able to use the same software and techniques used for lunar and planetary. I haven't got much experience with that. If instead you want to process the frames in a conventional manner with software like DSS, then my preference is to visually inspect every frame for sharpness, and discard those that are contorted. I am a little suspicious of allowing software to do that, but that's just me.
At the AstroBin link that Dave offered, there is one image that I really like:
https://www.astrobin.com/357219/?nc=user
That is a stack of 31x one-second exposures of the Lyra double-double captured with an 8-inch Newtonian taken from his backyard in Austria. I have no idea what his seeing conditions but it looks quite good.
Brian
I've held off from replying due to the word "DSO" in the title, but now that I see that the topic has shifted towards imaging stars, I can add something to the discussion.
Feel free to use sub one-second exposures for the Lyra double-double. They are bright at 6th magnitude. If you have a CMOS camera with adjustable gain, you might want to jack it up. I do exactly that when I use SharpCap to measure periodic error. With a 71mm aperture, I often use 350ms exposures with the gain cranked up to about 2000 on my Altair 290M. With my seeing conditions, a 350ms exposure is not short enough to "freeze" the atmosphere, but if I go down to 250ms then I begin to see the turbulence. The "threshold" exposure will be different for you, so you need to experiment. Set the frame rate to one per second, and then slowly shorten the exposure until you see the turbulence.
With regards to the output format, I output to PNG because my IronPython script needs to open it up to perform star detection. In your case, you might want to write to a SER file, and then break out the individual frames later on. Dave is the expert with that. Now, with regards to processing, you may be able to use the same software and techniques used for lunar and planetary. I haven't got much experience with that. If instead you want to process the frames in a conventional manner with software like DSS, then my preference is to visually inspect every frame for sharpness, and discard those that are contorted. I am a little suspicious of allowing software to do that, but that's just me.
At the AstroBin link that Dave offered, there is one image that I really like:
https://www.astrobin.com/357219/?nc=user
That is a stack of 31x one-second exposures of the Lyra double-double captured with an 8-inch Newtonian taken from his backyard in Austria. I have no idea what his seeing conditions but it looks quite good.
Brian
Re: Optimizing sharpness with SNR. DSO "lucky" imaging ?
Many thanks once again Dave and Brian
It's amazing just how much you can learn on this forum!
And there was I was reflecting that with seeing limited to about 3 arcsec or so (for me - usually) image-stacking type astrophotography is inherently inferior to old-fashioned visual astronomy in it's ability to resolve double stars or other bright close details - and so back to a sketch book and pencil
. Ruminating on that was what was behind my original question.
Thanks for the link to the double/double photos - that answers that question then. If the objects are bright enough just turn up the gain and shorten the exposure enough and it works....the resolution is there. Yes I saw the shot from Austria -probably was good seeing for it to work even at 1s - maybe from up in the Alps somewhere ? Anyway the double double will be fun to try over a range of decreasing exposures -- trying to resolve it at a time beneath the seeing will also be a useful check that my optics and focusing are up to scratch.
I suppose that the " high dynamic range" benefits of processing short and long exposure stacks together and the 'seeing' resolution aspect of very short exposures are really two separate discussions related only by the fact that there are benefits as well as (SNR) downsides to short exposures.
I am beginning to learn that so much depends on exactly how you do things both in the initial imaging set up and in post processing -- and that this must be guided by which particular aspect of an object you are interested in. So with M87 for example -- with short exposures and a low stretch I got some images that clearly showed the bluish jet. However stretching the image enough to take in the halo and just possibly detect globular clusters obscured the jet completely. So it was really one thing or the other but would require complex post-processing HDR to get both aspects.
TimH
It's amazing just how much you can learn on this forum!
And there was I was reflecting that with seeing limited to about 3 arcsec or so (for me - usually) image-stacking type astrophotography is inherently inferior to old-fashioned visual astronomy in it's ability to resolve double stars or other bright close details - and so back to a sketch book and pencil
. Ruminating on that was what was behind my original question.Thanks for the link to the double/double photos - that answers that question then. If the objects are bright enough just turn up the gain and shorten the exposure enough and it works....the resolution is there. Yes I saw the shot from Austria -probably was good seeing for it to work even at 1s - maybe from up in the Alps somewhere ? Anyway the double double will be fun to try over a range of decreasing exposures -- trying to resolve it at a time beneath the seeing will also be a useful check that my optics and focusing are up to scratch.
I suppose that the " high dynamic range" benefits of processing short and long exposure stacks together and the 'seeing' resolution aspect of very short exposures are really two separate discussions related only by the fact that there are benefits as well as (SNR) downsides to short exposures.
I am beginning to learn that so much depends on exactly how you do things both in the initial imaging set up and in post processing -- and that this must be guided by which particular aspect of an object you are interested in. So with M87 for example -- with short exposures and a low stretch I got some images that clearly showed the bluish jet. However stretching the image enough to take in the halo and just possibly detect globular clusters obscured the jet completely. So it was really one thing or the other but would require complex post-processing HDR to get both aspects.
TimH