Jump to content

  •  

CNers have asked about a donation box for Cloudy Nights over the years, so here you go. Donation is not required by any means, so please enjoy your stay.

Photo

Sharper subs with shorter sub exposures?

  • Please log in to reply
39 replies to this topic

#1 Shiraz

Shiraz

    Viking 1

  • -----
  • topic starter
  • Posts: 605
  • Joined: 15 Oct 2010
  • Loc: South Australia

Posted 23 September 2019 - 07:00 AM

Hi

 

The chart below shows how measured FWHM decreased with shorter lum subs on a night of consistent average seeing and no excessive wind.

 

To counteract the effects of any seeing drift, a sequence of subs of various lengths (1x5 min sub and multiple subs of shorter lengths) was repeated 4 times. Focusing was redone regularly in SGpro, but not synchronised with the sub length sequence (ie, there was no focus bias in favour of any sub length). The process did not involve any sub selection at all - all subs were used and the FWHM from the PI SubFrameSelector script was averaged over all of the subs of any given length (from all 4 sequences). Fairly aggressive phd guiding was running at 1 second update rate on an average EQ8.

 

Got the same result on a couple of nights - the shorter the sub, the sharper the result (I also hand checked a few subs to make sure that star saturation was not blowing out the FWHM on longer subs - it wasn't).

 

I guess that this is due to slow seeing or tracking wander, but most of the exposures are well outside of the normal "lucky imaging" region and in any case, there is no image selection to prune off subs with poor seeing. I was surprised that the effect was so significant and also fairly consistent over the full range of sub lengths from 1 second to five minutes. Has anyone else ever seen anything similar - prior threads or experience? Does anyone have any opinions on whether this effect is real and/or what the mechanism might be?

 

Thanks for looking, Cheers, Ray

 

seeing4.jpg


Edited by Shiraz, 23 September 2019 - 07:48 AM.

  • elmiko, rathijit, LauraMS and 2 others like this

#2 happylimpet

happylimpet

    Soyuz

  • *****
  • Posts: 3995
  • Joined: 29 Sep 2013
  • Loc: Southampton, UK

Posted 23 September 2019 - 07:56 AM

This is why ive been using 10sec subs for years now. And 2sec subs for high-res stuff on occasion.

 

Thanks for taking the trouble to share these data - great to see what ive been  confident was the truth for a long time!



#3 ciraxis

ciraxis

    Mariner 2

  • -----
  • Posts: 231
  • Joined: 12 Feb 2014
  • Loc: Syracuse, NY

Posted 23 September 2019 - 08:00 AM

I have yet to really play with shooting shorter subs, how does the quality compare to say 1 hour of 2 minute shots to one hour worth of 10 second shots?

 

I'd be interested to see a comparison.   I've seen really cool shots with shorter subs but don't recall seeing a side by side with same camera, scope etc



#4 Jared

Jared

    Fly Me to the Moon

  • *****
  • Posts: 6068
  • Joined: 11 Oct 2005
  • Loc: Piedmont, California, U.S.

Posted 23 September 2019 - 08:44 AM

Hi

 

The chart below shows how measured FWHM decreased with shorter lum subs on a night of consistent average seeing and no excessive wind.

 

To counteract the effects of any seeing drift, a sequence of subs of various lengths (1x5 min sub and multiple subs of shorter lengths) was repeated 4 times. Focusing was redone regularly in SGpro, but not synchronised with the sub length sequence (ie, there was no focus bias in favour of any sub length). The process did not involve any sub selection at all - all subs were used and the FWHM from the PI SubFrameSelector script was averaged over all of the subs of any given length (from all 4 sequences). Fairly aggressive phd guiding was running at 1 second update rate on an average EQ8.

 

Got the same result on a couple of nights - the shorter the sub, the sharper the result (I also hand checked a few subs to make sure that star saturation was not blowing out the FWHM on longer subs - it wasn't).

 

I guess that this is due to slow seeing or tracking wander, but most of the exposures are well outside of the normal "lucky imaging" region and in any case, there is no image selection to prune off subs with poor seeing. I was surprised that the effect was so significant and also fairly consistent over the full range of sub lengths from 1 second to five minutes. Has anyone else ever seen anything similar - prior threads or experience? Does anyone have any opinions on whether this effect is real and/or what the mechanism might be?

 

Thanks for looking, Cheers, Ray

 

attachicon.gif seeing4.jpg

 

I’m unclear on something in your post.  Are these measurements after aligning and stacking so the integration time is comparable?  Or is this an average of the individual sub exposures alone?  I would think you would need to compare FWHM after aligning and stacking with similar integration times for this to be very informative.  Knowing that the individual subs may be sharper is not necessarily beneficial if the stacked frames are similar.  FWHM generally decreases when you stack, and it is possible that it would decrease more with the larger number of frames required with the shorter subs.  Worth checking if you haven’t already.

 

Next, a comment.  The question of optimal exposure length for highest resolution is likely to be moot if your exposures aren’t long enough for shot noise to dwarf read noise and for the details you wish to capture to be far enough off the noise floor that quantization error isn’t a problem.  In practice, there are relatively few deep sky objects that would benefit from, for example, 1s sub exposures given the current cameras available.  It is, however, quite possible that from a light polluted location in particular, where sky glow quickly swamps read noise, one could benefit from shorter subs than most of us are currently using.  I suppose we should all check our own equipment to see if your experiences are typical.



#5 mic1970

mic1970

    Viking 1

  • -----
  • Posts: 809
  • Joined: 09 Aug 2018
  • Loc: St. Louis Metro Area

Posted 23 September 2019 - 08:48 AM

I don't want to go to far off subject here, but...  What is your setup (scope, camera, ISO, etc)?  I'm new and trying this with little success.  The noise gets me.  How do you deal with that?  

This is why ive been using 10sec subs for years now. And 2sec subs for high-res stuff on occasion.

 

Thanks for taking the trouble to share these data - great to see what ive been  confident was the truth for a long time!



#6 UniversalMaster

UniversalMaster

    Viking 1

  • -----
  • Posts: 599
  • Joined: 20 Nov 2008

Posted 23 September 2019 - 09:16 AM

I'm a bit surprised to see that the effect is so insignificant. You only get about 30% better FWHM by decreasing the exposure time 300x. Guess this is why high resolution planetary imaging is done with very short exposures and high frame rates. That is most be what you need to beat the seeing.

Interesting comparison, though.

#7 Shiraz

Shiraz

    Viking 1

  • -----
  • topic starter
  • Posts: 605
  • Joined: 15 Oct 2010
  • Loc: South Australia

Posted 23 September 2019 - 04:40 PM

thanks for the thoughtful responses.

 

HappyLimpet, given that you find the same effect, any ideas on what the mechanism might be?

 

ciraxis, I haven't done such a comparo, because seeing is so highly variable over periods of hours, but I might try it anyway

 

Jared, these are averages over the subs in each group.... no stacking. I deliberately excluded stacking to reduce the number of variables - all that this shows is that short subs are sharper. I am going to have to spend a lot longer and get more data to do meaningful stacking - haven't done so yet. I agree entirely that very short subs are not useful for most DSOs and wanted to find out if a shorter-sub benefit applied in the region of sub-lengths that we use for dim objects. I had an inkling that shorter subs were beneficial but to my surprise there was no "shortness threshold" - shorter was better even when going from 5 minutes to 2 minutes or from 2 minutes to 1 minute. 

 

UniversalMaster Do you think that this is a seeing effect? I had guiding going at 1 second update, so this should have removed most of the long-period motion due to local seeing or mount wander. I was surprised to see any change at all going from 5 minutes to 2 minutes. Hi res planetary freezes low frequency seeing wander, but at frame rates above a few Hertz and it is generally used with bad-frame rejection. I think that the measured effect might be something slightly different, but don't really know.

 

Overall, I am fairly confident that the measurements do indicate that there is a real effect with my system under the conditions on the night. What I don't yet understand is what is causing it, since guiding should have well and truly tidied up any long term wander due to seeing and mount, especially with the long subs. I am not too sure what is left - maybe fast large scale excursions that occur infrequently and that are more likely to damage longer subs?? Or maybe the seeing causes slow warping of the scene that is partially frozen even with quite long subs?? 

 

Cheers Ray 


Edited by Shiraz, 23 September 2019 - 05:06 PM.


#8 RedLionNJ

RedLionNJ

    Soyuz

  • *****
  • Posts: 3695
  • Joined: 29 Dec 2009
  • Loc: Red Lion, NJ, USA

Posted 23 September 2019 - 05:05 PM

Given you use the phrase "tracking wander", I'd say it's exactly that.

 

With a shorter exposure, you're less likely to cavort into the extreme reaches of tracking errors.  Could be RA (more likely) or could be Dec, or even a bit of both.

 

I wouldn't rule OUT seeing, but you'd expect seeing fluctuations to be less reproducible.



#9 freestar8n

freestar8n

    Vendor - MetaGuide

  • *****
  • Vendors
  • Posts: 8825
  • Joined: 12 Oct 2007

Posted 23 September 2019 - 05:29 PM

I haven't measured this explicitly but I think many of us take it as a given that shorter exposures will be sharper - even beyond the time scale of seeing effects.

 

I think the dominant factor may be that the centroid is a biased measurement that is based on the star spot itself - and not where the guider is trying to place the centroid.  By that I mean - while you are guiding there are always errors and the true centroid is drifting around the target location.  For shorter exposures the true centroid won't drift much and the guiding error will be less - because even though the star spent some time off target, you measure the centroid based on the actual photons and where they landed.

 

In some ways this is similar to Fried theory, where you compare very short exposures with long ones and only seeing is playing a role.  For very short exposures the tip/tilt term of seeing is frozen and the photons accumulate in one place, whereas for longer ones the tip/tilt term sprays them around.  I think for guiding the situation is similar except the guiding timescale is longer.

 

Short guide exposures and fast corrections should help - but there is always error in the centroid and that error will wander over a larger area over time.

 

There are probably higher order effects involved and you could look at how the eccentricity changes with exposure - and even the Moffat profile.

 

Frank



#10 RedLionNJ

RedLionNJ

    Soyuz

  • *****
  • Posts: 3695
  • Joined: 29 Dec 2009
  • Loc: Red Lion, NJ, USA

Posted 23 September 2019 - 07:24 PM

 

I think the dominant factor may be that the centroid is a biased measurement that is based on the star spot itself - and not where the guider is trying to place the centroid.  By that I mean - while you are guiding there are always errors and the true centroid is drifting around the target location.  For shorter exposures the true centroid won't drift much and the guiding error will be less - because even though the star spent some time off target, you measure the centroid based on the actual photons and where they landed.

 

In some ways this is similar to Fried theory, where you compare very short exposures with long ones and only seeing is playing a role.  For very short exposures the tip/tilt term of seeing is frozen and the photons accumulate in one place, whereas for longer ones the tip/tilt term sprays them around.  I think for guiding the situation is similar except the guiding timescale is longer.

 

 

Frank

Frank - there is absolutely NO mention of guiding in the original post (or subsequents by the same OP).

 

I don't see any indication guiding is involved here - just the mount's tracking mechanism.



#11 freestar8n

freestar8n

    Vendor - MetaGuide

  • *****
  • Vendors
  • Posts: 8825
  • Joined: 12 Oct 2007

Posted 23 September 2019 - 07:35 PM

Frank - there is absolutely NO mention of guiding in the original post (or subsequents by the same OP).

 

I don't see any indication guiding is involved here - just the mount's tracking mechanism.

He says:

 

 

Fairly aggressive phd guiding was running at 1 second update rate on an average EQ8.

If there were no guiding at all then longer exposures will be trailed and a worsening is guaranteed.  But since there is guiding involved you would expect everything to be randomized after some shortish minimum exposure time.

 

But to show the biasing effect I'm talking about - here is a plot of the std of 1000 trials of different sized samples of normally distributed random numbers (std=1, mean=0).  With smaller numbers of samples there is greater likelihood they will be bunched up and some distance from the true mean - and I think the same thing happens with the star spot relative to the guide centroid.

 

There is no guiding or optics or anything in this plot.  Just finding the mean standard deviation in many trials of a given sample size.  And the standard deviation is much smaller when the sample size is smaller.

 

MeanSTDSampleSize.png

 

Frank


  • mic1970 likes this

#12 Jon Rista

Jon Rista

    ISS

  • *****
  • Posts: 23721
  • Joined: 10 Jan 2014
  • Loc: Colorado

Posted 23 September 2019 - 08:32 PM

Hi

 

The chart below shows how measured FWHM decreased with shorter lum subs on a night of consistent average seeing and no excessive wind.

 

To counteract the effects of any seeing drift, a sequence of subs of various lengths (1x5 min sub and multiple subs of shorter lengths) was repeated 4 times. Focusing was redone regularly in SGpro, but not synchronised with the sub length sequence (ie, there was no focus bias in favour of any sub length). The process did not involve any sub selection at all - all subs were used and the FWHM from the PI SubFrameSelector script was averaged over all of the subs of any given length (from all 4 sequences). Fairly aggressive phd guiding was running at 1 second update rate on an average EQ8.

 

Got the same result on a couple of nights - the shorter the sub, the sharper the result (I also hand checked a few subs to make sure that star saturation was not blowing out the FWHM on longer subs - it wasn't).

 

I guess that this is due to slow seeing or tracking wander, but most of the exposures are well outside of the normal "lucky imaging" region and in any case, there is no image selection to prune off subs with poor seeing. I was surprised that the effect was so significant and also fairly consistent over the full range of sub lengths from 1 second to five minutes. Has anyone else ever seen anything similar - prior threads or experience? Does anyone have any opinions on whether this effect is real and/or what the mechanism might be?

 

Thanks for looking, Cheers, Ray

 

attachicon.gif seeing4.jpg

I think the effect is real. I haven't done an explicit test like this, but I have had my fair share of 5-10s imaging, 1-3 minute imaging, and 10+ minute imaging. With 10+ minutes my stars are usually larger...2.5" or so, sometimes 3". With 2-3 minute subs they are usually closer to 2", although on nights of very good seeing I've been as low as 1.6-1.7" (at least while seeing was good...often doesn't last very long). With subs of just a few seconds the potential is there for stars under 2" all the time...with a key caveat.

 

What I have observed is that unless you DO cull poorer quality subs with short exposures (i.e. employ "lucky imaging" techniques), then the benefits of imaging in the seconds range vs. the minutes range from a final integration standpoint diminish to roughly zero. Without the lucky aspect, very short exposures are not worth the time, even though their potential for high resolution is higher. The cost of lucky imaging, though, is lower SNR. If you cull enough to really benefit, which is often throwing away 60-80% or so of your data, the cost to SNR can be significant. So you would really have to NEED maximum sharpness to make imaging with sub-minute length subs worth while.

 

Even with short subs in the 1-3 minutes range, to get the better resolution I was often throwing away 20-30% of my subs, sometimes more, to get better resolution. There is always that percentage of subs that are just truly crappy (at least, so seems to be the case for me...be it changes in seeing, wind, something funky with gear, etc.) If I stacked all the data (excluding stuff that was star trailed at least), then the benefits of 2-3 minute subs were usually not enough to support the loss of SNR from not using longer subs (and since getting the Mach 1 this is even more true, the differences have diminished to almost nothing unless I get fairly aggressive with culling "softer" subs when using 2-3 minute exposures.) 



#13 jhayes_tucson

jhayes_tucson

    Fly Me to the Moon

  • *****
  • Posts: 7157
  • Joined: 26 Aug 2012
  • Loc: Bend, OR

Posted 23 September 2019 - 09:40 PM

 

Overall, I am fairly confident that the measurements do indicate that there is a real effect with my system under the conditions on the night. What I don't yet understand is what is causing it, since guiding should have well and truly tidied up any long term wander due to seeing and mount, especially with the long subs. I am not too sure what is left - maybe fast large scale excursions that occur infrequently and that are more likely to damage longer subs?? Or maybe the seeing causes slow warping of the scene that is partially frozen even with quite long subs?? 

 

Cheers Ray 

 

In general, this is caused by seeing, which introduces time varying aberration into the wavefront.  Your camera is an integrating device so the longer the exposure is relative to the correlation time of the wavefront variation, the more effect it will have on the size of the integrated point spread function.  The strength of the effect of seeing on FWHM will depend primarily on a couple of things:  1) the magnitude of the wavefront aberration, and 2) how quickly the wavefront varies, which is directly related to the correlation time.  The wavelength that you use will also have a small effect.  Longer wavelengths (red) will be a bit less affected than shorter wavelengths.(blue.)  Of course the way that you guide and establish (and hold) focus can also introduce errors in this type of measurement.

 

Distilling all of that down:  It's mostly due to seeing.

 

John



#14 freestar8n

freestar8n

    Vendor - MetaGuide

  • *****
  • Vendors
  • Posts: 8825
  • Joined: 12 Oct 2007

Posted 23 September 2019 - 09:53 PM

I think the effect is real. I haven't done an explicit test like this, but I have had my fair share of 5-10s imaging, 1-3 minute imaging, and 10+ minute imaging. With 10+ minutes my stars are usually larger...2.5" or so, sometimes 3". With 2-3 minute subs they are usually closer to 2", although on nights of very good seeing I've been as low as 1.6-1.7" (at least while seeing was good...often doesn't last very long). With subs of just a few seconds the potential is there for stars under 2" all the time...with a key caveat.

 

What I have observed is that unless you DO cull poorer quality subs with short exposures (i.e. employ "lucky imaging" techniques), then the benefits of imaging in the seconds range vs. the minutes range from a final integration standpoint diminish to roughly zero. Without the lucky aspect, very short exposures are not worth the time, even though their potential for high resolution is higher. The cost of lucky imaging, though, is lower SNR. If you cull enough to really benefit, which is often throwing away 60-80% or so of your data, the cost to SNR can be significant. So you would really have to NEED maximum sharpness to make imaging with sub-minute length subs worth while.

 

Even with short subs in the 1-3 minutes range, to get the better resolution I was often throwing away 20-30% of my subs, sometimes more, to get better resolution. There is always that percentage of subs that are just truly crappy (at least, so seems to be the case for me...be it changes in seeing, wind, something funky with gear, etc.) If I stacked all the data (excluding stuff that was star trailed at least), then the benefits of 2-3 minute subs were usually not enough to support the loss of SNR from not using longer subs (and since getting the Mach 1 this is even more true, the differences have diminished to almost nothing unless I get fairly aggressive with culling "softer" subs when using 2-3 minute exposures.) 

Jon I think the increased sharpening is just due to less wander of the guide error - so the tighter fwhm follows fairly consistently from the shorter exposure time and there is no need to cull as there would be if it were more of a seeing effect.  The timescale here is huge compared to seeing - but it is about right on the scale of repeated guide corrections - each with fairly independent error.

 

There is nothing magic in this because if the guidestar tends to hang out on one side of the guide target - that offset will be removed when you register and stack.  The rmsd guide error isn't what shows in the image.

 

Here is a plot comparing trials of 5 samples with trials of 60.  With more samples the standard deviation will become more consistent - but with fewer samples you have a high probability the resulting std will be smaller than the true std of the population (1 in this case).  Only a small fraction end up being worse than the higher count version - and a very large fraction end up better.

 

With lucky imaging you are more committed to a small fraction of rare and nearly perfect samples - and that requires culling.  But with shorter - but still long exposures - the statistics are very different and you can get a benefit with shorter exposures.  The main thing you are trading off is read noise.

 

STDHistograms.png

 

Frank


  • classpath likes this

#15 Jon Rista

Jon Rista

    ISS

  • *****
  • Posts: 23721
  • Joined: 10 Jan 2014
  • Loc: Colorado

Posted 23 September 2019 - 10:31 PM

Jon I think the increased sharpening is just due to less wander of the guide error - so the tighter fwhm follows fairly consistently from the shorter exposure time and there is no need to cull as there would be if it were more of a seeing effect.  The timescale here is huge compared to seeing - but it is about right on the scale of repeated guide corrections - each with fairly independent error.

 

There is nothing magic in this because if the guidestar tends to hang out on one side of the guide target - that offset will be removed when you register and stack.  The rmsd guide error isn't what shows in the image.

 

Here is a plot comparing trials of 5 samples with trials of 60.  With more samples the standard deviation will become more consistent - but with fewer samples you have a high probability the resulting std will be smaller than the true std of the population (1 in this case).  Only a small fraction end up being worse than the higher count version - and a very large fraction end up better.

 

With lucky imaging you are more committed to a small fraction of rare and nearly perfect samples - and that requires culling.  But with shorter - but still long exposures - the statistics are very different and you can get a benefit with shorter exposures.  The main thing you are trading off is read noise.

 

attachicon.gif STDHistograms.png

 

Frank

In my own practical experience, even with shorter subs you still have to endure those periods where seeing gets worse. I have seen seeing change on relatively small time scales...going from terrible to good in a matter of minutes, to very good, then suddenly back to terrible again. Perhaps it has to do with living near a massive mountain range and the changing jet stream, perhaps there are local phenomena. I don't know what causes it, but I've seen sudden and dramatic changes in seeing.

 

This is often quite visually obvious with lunar imaging. I have started nights that I thought were awesome, only to have things go from a "slight ripple in an almost glass-like pond" to "a roiling, boiling pot of hot water", right in the middle of my captures. I've seen large scale, slow (relatively speaking, it could take 10-15 seconds to cross the face of the moon) waves of differing turbulence move by, where in front of the wave things may be great, or terrible...and behind the front things are the opposite. Then another will pass through some time later and it was back to how it was before.

 

When I'm chasing 1.6-1.7" FWHMs these kinds of effects can quickly bloat FWHMs by half an arcsecond or more. They can vary in degree as well. I've gone from sub-2" FWHMs to over 3-4" FWHMs in a single imaging session, even:

 

aNeygDW.gif

 

Some of this (at 5" plus) is due to less than perfect focus as well, as seeing got so bad it had a tough time finding focus reliably. 



#16 freestar8n

freestar8n

    Vendor - MetaGuide

  • *****
  • Vendors
  • Posts: 8825
  • Joined: 12 Oct 2007

Posted 23 September 2019 - 10:55 PM

In my own practical experience, even with shorter subs you still have to endure those periods where seeing gets worse. I have seen seeing change on relatively small time scales...going from terrible to good in a matter of minutes, to very good, then suddenly back to terrible again. Perhaps it has to do with living near a massive mountain range and the changing jet stream, perhaps there are local phenomena. I don't know what causes it, but I've seen sudden and dramatic changes in seeing.

 

This is often quite visually obvious with lunar imaging. I have started nights that I thought were awesome, only to have things go from a "slight ripple in an almost glass-like pond" to "a roiling, boiling pot of hot water", right in the middle of my captures. I've seen large scale, slow (relatively speaking, it could take 10-15 seconds to cross the face of the moon) waves of differing turbulence move by, where in front of the wave things may be great, or terrible...and behind the front things are the opposite. Then another will pass through some time later and it was back to how it was before.

 

When I'm chasing 1.6-1.7" FWHMs these kinds of effects can quickly bloat FWHMs by half an arcsecond or more. They can vary in degree as well. I've gone from sub-2" FWHMs to over 3-4" FWHMs in a single imaging session, even:

 

 

 

Some of this (at 5" plus) is due to less than perfect focus as well, as seeing got so bad it had a tough time finding focus reliably. 

If you are only concerned about variable seeing - then stacking a bunch of short subs is still better than one long sub - because you have the benefit of additional alignment of each sub.  There is no corresponding penalty at all doing short or long subs if you aren't culling in either case.

 

But if you do allow culling - and you want the sharpest final result - you still don't really lose snr with shorter subs - except for read noise - because you would be throwing out some of the longer ones also.

 

I think it's important to realize that seeing involves very fast tip tilt motion on the scale of milliseconds - and you can remove its impact by very short exposures - and culling.  But on the longer time scales all you have is an overall undulation of the fwhm - but the star is round and fixed.  In that situation shorter exposures would just chop up that overall variation into a range of small and large fwhm over time - but the mean wouldn't be affected.  And if there are guiding errors then you would again benefit from the additional alignment of each shorter exposure.

 

Everything is completely different when the reality of the situation is - the scope is being smacked left/right/up/down every second with guide corrections.  And those corrections will rarely be dead on - and the star in the image will be moving around a bit.  This is the only thing that makes sense to me as the cause for shorter exposures being tighter - and I don't see anything too surprising about it.

 

People think the guider is making perfect corrections and the star in the image is basically fixed - but with PE and drift and finite time for the guide exposure - the star in the image will be doing its own thing - and as a result it will have tighter locality on short timescales than long.  This is shown in the thread I did on guiding simulation:

 

https://www.cloudyni...guide-exposure/

 

If you look at the plots of what the star is actually doing in the image - it has a mixture of fast oscillations on the timescale of the guide correction - and overall variation on the timescale of PE.  Any part of the graph where you see it hanging out on one side vs. the other means that exposures in those regions will have smaller fwhm than an overall long one - because the local standard deviation of motion is smaller than the overall std.

 

One way to rule out guiding as a cause is to image stars near the pole with tracking off - and do a range of exposures.  But even then other surprising factors may play a role.

Frank



#17 happylimpet

happylimpet

    Soyuz

  • *****
  • Posts: 3995
  • Joined: 29 Sep 2013
  • Loc: Southampton, UK

Posted 24 September 2019 - 03:59 AM

I don't want to go to far off subject here, but...  What is your setup (scope, camera, ISO, etc)?  I'm new and trying this with little success.  The noise gets me.  How do you deal with that?  

EQ8, 12" newtonian @f3.7, ASI1600MM, lots of light pollution. Im sky limited in 2-8 seconds depending on (broadband) filter. So theres no penalty for me in using short subs.

 

thanks for the thoughtful responses.

 

HappyLimpet, given that you find the same effect, any ideas on what the mechanism might be?

I think its a general wandering of the scope pointing. The thing that reinforced my desire to do short subs initially was the observation that while watching repeated 10sec subs (or 5, or 2) come in, they never 'overlapped' perfectly with the preceding one, which means that if they had been integrated as one long 1min (or whatever) sub, the FWHM of the PSF would have been broader. Certainly sharper subs lead to sharper stacks, and its my feeling (though unconfirmed) that the PSF of the subs corresponds pretty well with the PSF of the stack.

 

Jon I think the increased sharpening is just due to less wander of the guide error - so the tighter fwhm follows fairly consistently from the shorter exposure time and there is no need to cull as there would be if it were more of a seeing effect.  The timescale here is huge compared to seeing - but it is about right on the scale of repeated guide corrections - each with fairly independent error.

 

There is nothing magic in this because if the guidestar tends to hang out on one side of the guide target - that offset will be removed when you register and stack.  The rmsd guide error isn't what shows in the image.

I agree with the above.

 

Basically shorter subs means less time for things to go wrong!!!!

 

I've long observed that my images of any given object tend to be amongst the sharpest of those online, which, given my unremarkable seeing and location makes me think my short subs method is working.


  • Shiraz likes this

#18 freestar8n

freestar8n

    Vendor - MetaGuide

  • *****
  • Vendors
  • Posts: 8825
  • Joined: 12 Oct 2007

Posted 24 September 2019 - 04:06 AM

 

I think its a general wandering of the scope pointing. The thing that reinforced my desire to do short subs initially was the observation that while watching repeated 10sec subs (or 5, or 2) come in, they never 'overlapped' perfectly with the preceding one, which means that if they had been integrated as one long 1min (or whatever) sub, the FWHM of the PSF would have been broader. Certainly sharper subs lead to sharper stacks, and its my feeling (though unconfirmed) that the PSF of the subs corresponds pretty well with the PSF of the stack.

 

Yes that is the key here.  The effect isn't explained by some kind of dynamics of a stationary star and its profile due to seeing.  It has to do with random-ish wandering of a star in the image plane around a target location on the timescale of seconds.  The guide corrections cause jumps here and there - and over time the true star spot will linger locally over short time spans while exploring the overall landscape that represents the full fwhm in a long exposure.

 

As a result, in short periods of time it will tend to hang out in a small part of that landscape - and thereby end up with a smaller fwhm.

 

This would result if the jumps are completely random around the target - as my plots above allude to - and it would also happen if there is somewhat coherent motion of the star due to guide corrections - as my simulation shows.

 

Frank


  • happylimpet likes this

#19 mic1970

mic1970

    Viking 1

  • -----
  • Posts: 809
  • Joined: 09 Aug 2018
  • Loc: St. Louis Metro Area

Posted 24 September 2019 - 07:13 AM

Well with a 3.7, 12".... you are pulling a lot of light in ten seconds.  Any tricks in pre or post processing you can share?

EQ8, 12" newtonian @f3.7, ASI1600MM, lots of light pollution. Im sky limited in 2-8 seconds depending on (broadband) filter. So theres no penalty for me in using short subs.

 

 



#20 happylimpet

happylimpet

    Soyuz

  • *****
  • Posts: 3995
  • Joined: 29 Sep 2013
  • Loc: Southampton, UK

Posted 24 September 2019 - 09:45 AM

Well with a 3.7, 12".... you are pulling a lot of light in ten seconds.  Any tricks in pre or post processing you can share?

Nope! Just carry on as usual!

 

One thing that might be useful is the acquisition software. Until recently I used firecapture, as that could work with 2sec subs and very little latency. But this didnt allow for autofocusing etc.

 

SGP is way too slow.

 

Now NINA is here and can certainly work with 10sec subs - it loses about 0.5 second between subs on my machine.



#21 Shiraz

Shiraz

    Viking 1

  • -----
  • topic starter
  • Posts: 605
  • Joined: 15 Oct 2010
  • Loc: South Australia

Posted 24 September 2019 - 10:04 AM

thanks everyone for the very informative discussion.

 

I had been concerned that, since any slow drifts due to seeing and/or mount errors would be removed by guiding, there should have been little difference between short and long subs (at least for subs above maybe 30 seconds). What I had missed was that seeing/mount errors can introduce minor offsets in the guidance loop that will persist for a significant time - I guess since since they are below the "min motion" threshold (of phd and presumably other guiding algorithms). These persistent minor offsets are on a much longer timescale than seeing fluctuations (as I understand it), and could indeed account for the observed results. Thanks happylimpet for the observations and Frank for the detailed explanation and modelling.

 

If this is in fact the mechanism, it raises some questions:

1. is it really a good idea to have min motion thresholds to stop excessive corrections in a guiding loop 

2. if a guiding loop runs for significant periods without correcting minor offsets, would it be better to switch it off and run unguided with short subs

3. Jon has indicated that the extra resolution is not available after stacking unless culling is used - maybe that is worth further investigation, since this whole exercise has been a waste of time if there is no end-use resolution advantage to short subs.

 

cheers ray


Edited by Shiraz, 24 September 2019 - 04:52 PM.


#22 happylimpet

happylimpet

    Soyuz

  • *****
  • Posts: 3995
  • Joined: 29 Sep 2013
  • Loc: Southampton, UK

Posted 24 September 2019 - 10:17 AM

 

3. Jon has indicated that the extra resolution is not available after stacking unless culling is used - maybe that is worth further investigation, since this whole exercise has been a waste of time if there is no end-use advantage to short subs.

 

cheers ray

I disagree with this point. The subs all end up sharper and so does the final stack. There will always be a handful of pigs, which you might as well throw out, but it makes very little difference to the stack (at least for my setup and experience).

 

One of the odd things about this is that as you watch the subs come in, every 10 secs for example, most of them look beautifully sharp, but all distinctly offset from each other. One might think they would be trailed in a 'connect the dots' sort of way, but thats not what i see.  Every now and again a rough one shows up which I will usually throw out, unless Im just trying to go very deep and not so bothered about resolution, which is rare.


  • Shiraz likes this

#23 Jon Rista

Jon Rista

    ISS

  • *****
  • Posts: 23721
  • Joined: 10 Jan 2014
  • Loc: Colorado

Posted 24 September 2019 - 12:50 PM

I disagree with this point. The subs all end up sharper and so does the final stack. There will always be a handful of pigs, which you might as well throw out, but it makes very little difference to the stack (at least for my setup and experience).

 

One of the odd things about this is that as you watch the subs come in, every 10 secs for example, most of them look beautifully sharp, but all distinctly offset from each other. One might think they would be trailed in a 'connect the dots' sort of way, but thats not what i see.  Every now and again a rough one shows up which I will usually throw out, unless Im just trying to go very deep and not so bothered about resolution, which is rare.

I wouldn't say that all subs are all guaranteed to end up sharper. That is not my experience. Perhaps in some locales, where seeing is good enough on a consistent enough basis, then that could be true. My point is it is not guaranteed to be true. Consider what you would get if you stacked the three exemplars from my GIF above. Would you get a 1.7" FWHM or a 5" FWHM in the stack? You certainly would not get a 1.7" FWHM, and you are likely to get over a 3" FWHM, probably quite close to a 5" FWHM...despite the fact that it is just one sub.

 

Now if you have a lot of subs, and use a clipping algorithm, then the softest subs will usually have their halos rejected. Problem then is the profiles of good vs. ok vs. poor subs can vary quite a lot, and this affects the quality of the stars in the final stack. 

 

There aren't any guarantees here. Frank is correct that there are added benefits to registration, I agree with that. But it is more complicated than that, and it depends on the night(s) you image over and just how things may vary through the subs as to whether stacking everything with short subs delivers better results than using long subs.

 

I think there also needs to be consideration for basic psychology. How many of you are going to be willing to throw away even a single 10 minute or longer sub? Let alone a few, many? But 2 minute subs? How many of those could you throw away before you psychologically felt you were starting to lose something? What about 10 second subs? There is a major psychological factor when it comes to culling bad data. With long subs, there is a strong desire to keep as much as you can, as every time you throw out even one, you are tossing a "large" chunk of data, a meaningful portion of signal (whether this is actually true or not, it is still the psychological sense.) On the other hand, throwing away a bunch of 2 minute subs? Eh, who cares. It is easy to get 2 minute subs if you need more...right? 

 

So to Franks point about throwing away the same amount of data with long subs as with short? I don't really agree that is generally the case...in practice. I think people are far more willing to take the resolution hit to keep their SNR with long subs. On the flip side, I think people are far more willing to take an SNR hit to keep their resolution with shorter subs. I think this applies regardless of whether in the end, you might in fact actually end up with the same SNR (I don't actually believe this is necessarily the case either, at lest...it is not guaranteed.) 

 

In my experience, it is both easier to throw away short subs, and when you DO throw away the worst offenders, it does have the ability to help improve the FWHM of the short-sub stack over keeping them in the short-sub stack. This is often true regardless of whether stacking all subs in a short-sub stack is better than a long-sub stack or not. It is easy to optimize with shorter subs, and you can take the benefits of shorter exposures even farther than just stacking them all if you do cull the bad data, than if you just always stack it all. 


Edited by Jon Rista, 24 September 2019 - 12:52 PM.


#24 happylimpet

happylimpet

    Soyuz

  • *****
  • Posts: 3995
  • Joined: 29 Sep 2013
  • Loc: Southampton, UK

Posted 24 September 2019 - 04:25 PM

I wouldn't say that all subs are all guaranteed to end up sharper. That is not my experience. Perhaps in some locales, where seeing is good enough on a consistent enough basis, then that could be true. My point is it is not guaranteed to be true. Consider what you would get if you stacked the three exemplars from my GIF above. Would you get a 1.7" FWHM or a 5" FWHM in the stack? You certainly would not get a 1.7" FWHM, and you are likely to get over a 3" FWHM, probably quite close to a 5" FWHM...despite the fact that it is just one sub.

 

 

I suppose it might also reflect the specifics of my system and location. It could be that my optics are routinely delivering a sharper image than the mount is capable of guiding within (if you see what I mean) and so I end up with many sharp images with significant offsets. But it could be different for others with different seeing characteristics, and mounts.

 

I know that for me my FWHM generally doesnt vary much through, say, 4 hours of 10sec subs. There'll be a very small number of clearly trailed subs (wind?) which I'lll cheerfully throw out but usually a few % at most.

 

FWIW my subs are typically just over 2" fwhm.



#25 freestar8n

freestar8n

    Vendor - MetaGuide

  • *****
  • Vendors
  • Posts: 8825
  • Joined: 12 Oct 2007

Posted 24 September 2019 - 05:16 PM

thanks everyone for the very informative discussion.

 

I had been concerned that, since any slow drifts due to seeing and/or mount errors would be removed by guiding, there should have been little difference between short and long subs (at least for subs above maybe 30 seconds). What I had missed was that seeing/mount errors can introduce minor offsets in the guidance loop that will persist for a significant time - I guess since since they are below the "min motion" threshold (of phd and presumably other guiding algorithms). These persistent minor offsets are on a much longer timescale than seeing fluctuations (as I understand it), and could indeed account for the observed results. Thanks happylimpet for the observations and Frank for the detailed explanation and modelling.

 

If this is in fact the mechanism, it raises some questions:

1. is it really a good idea to have min motion thresholds to stop excessive corrections in a guiding loop 

2. if a guiding loop runs for significant periods without correcting minor offsets, would it be better to switch it off and run unguided with short subs

3. Jon has indicated that the extra resolution is not available after stacking unless culling is used - maybe that is worth further investigation, since this whole exercise has been a waste of time if there is no end-use resolution advantage to short subs.

 

cheers ray

Thanks Ray-

 

Regarding 1 and 2, if you are autoguiding then you are doing it because your tracking isn't perfect and the star is wandering off target as a result.  And if you make a correction it means you have detected some shift of the star and are kicking it back.  There is nothing bad about this - it's just what you have to do if tracking isn't perfect.  You need to have error grow enough to detect it and then correct it.

 

The only bad thing in all this is if your corrections don't match the error well.  The solution there is to tune the guiding better.

 

But there is a strong sense in autoguiding that corrections are bad and should be kept to a minimum - but I don't think that makes sense.  Certainly if your tracking is perfect you shouldn't make corrections - because it is already perfect.  But if your tracking isn't perfect, your corrections should help - and you should make them as often as you can to keep the errors small.

 

I guess this dependence on exposure time could be an indirect measure of guiding quality.  If everything is perfect then there shouldn't be much degredation in longer exposures.

 

As for culling - I assume the *mean* fwhm in a batch of images with short exposure will be lower than in a batch with longer exposure, but the distribution may have outliers with somewhat high fwhm.  But if it were just a seeing effect, the mean would be the same - which is why this points to guiding playing a role.

 

Frank


Edited by freestar8n, 24 September 2019 - 05:19 PM.



CNers have asked about a donation box for Cloudy Nights over the years, so here you go. Donation is not required by any means, so please enjoy your stay.


Recent Topics






Cloudy Nights LLC
Cloudy Nights Sponsor: Astronomics