Does stacking alignment accuracy ever limit sharpness?

[timh]

Hi Robin,

A rather basic question. I just wondered whether stacking is something digital that either works - or doesn't - and, if not, whether there is a grey area where frames can successfully stack but the alignment process itself be sufficiently imprecise as to contribute to an increase in the apparent size of stars?

In particular I wondered whether there is any significant image quality advantage to be expected from, say, picking 15 alignment stars rather than 10? Secondly - where the decision to BIN is marginal (say pixel FWHM is about 5-6) - I wondered whether stacking alignment is inherently less precise when over rather than undersampling? - so, on that basis it would be better to BIN? Or does it not really make any difference?

Grateful for any insights

best wishes
Tim

[admin]

Hi,

there certainly is an advantage to using a few extra stars for alignment – the calculated one positions aren't exact – they can be affected by noise in the image, so the more styles that are involved in the alignment the more likely it is that the noise effect cancelled out by the averaging.

The stack will always have a little bit more blurring than the individual frames – two things contribute towards this. The first is the noise in the image which can let the star positions and hence the alignment will be slightly out. The second is that the transformation to align the new image with the stack is not necessarily a whole number of pixels and may include some rotation. Both rotating an image and trying to translated by a non-integer number of pixels will lead to some additional blurring.

I'm not sure that I can see any good reasoning why live stacking would affect your choice of binning or not in the way that you are thinking – the only effect that really comes to mind is that using binning significantly reduces the amount of computation that needs to be done to stack each frame. If your FWHM is large then this may be a consideration.

Cheers, Robin

[timh]

Thanks yet again for your clear answer Robin,

I will up the number of alignment stars then

The question about binning did indeed come from 1) wondering about rotation and how to optimise for Dobsonians and then 2) about how SC alignment dealt with non integer pixel numbers. I assumed that SC would force alignment rounding up or down to the nearest pixel. I had reasoned (I hope correctly) that the inherent pixel 'blurring' of any given object during one exposure from rotation is ~ (sine theta X the radius of the object in pixels). Then I wondered whether it might be better to have lots of pixels (i.e. oversample) so that 1 pixel movement then constitutes a smaller fraction of the total FWHM or to reduce the number of pixels (BIN) so that sine theta x radius is a smaller number ? It seems then to be six of one to half a dozen of the other?

Anyway (weather permitting) I am going to have a go and see whether I can successfully use exposures up to 30s or so (i.e. field rotation ~ 0.12 deg?) on small objects like M27 - rather than the usual 10s. It will be interesting to see how closely the quality of an image from a dob can approach that of EQ mounted systems for that type of object.

best wishes
tim

[admin]

Hi Tim,

once you allow for rotation, you inevitably end up with fractional pixel movements, so you may as we mode as well allow it in translation adjustments too. The effects of blurring due to the stacking will be (relatively) smaller if you are oversampling I think.

cheers,
Robin

[RonAM]

Hi Tim and Robin,

Robin, correct me if I’m wrong please. I understand that using the FWHM limit helps keep star size consistent during the livestack process.

Once stacking begins successfully, say 3 or more frames, I then activate and carefully watch the FWHM tab for a few more frames to set the limit just above the average measured FWHM.

Once that is set up, I then double check the histogram tab limits I set, and then for the actual image acquisition I restart the stacking.

Of 50-200 frames over a half hour time, I may have 3-5 that fail due to not meeting FWHM limit. A turbulent atmosphere might cause more to fail.

I observe the effects of field rotation with my alt-as dob Mount over extended time as a slow creep of bad/dim image at the edge of the livestack view. I save my image with adjustments at several points in time, and finally stop the stacking when the edge effects from field rotation get big enough to creep into the part I want. This is usually 15-39 minutes of total exposure time at 1582mm focal length.

Cheers,

Ron

[admin]

Hi Ron,

the FWHM limit that you are using is really designed to keep rogue frames out of the stack. It sounds like it's doing its job for you. What might cause a road frame? The telescope mount might get nudged causing vibration. There might be a particularly bad patch of seeing. Some thin cloud might drift through causing star widths to increase, and possibly other causes.

This is slightly different to the topic that Tim and I were discussing – basically we were discussing the fact that the stack almost always has a slightly higher FWHM than the individual frames that go into it. For instance if all your frames are FWHM of about three, you might find that the stack has an FWHM of about 3.5. This is introduced by the requirement to transform the images being added to the stack in order to align them with the existing stars in the stack. While it's unavoidable that there will be some increase in FWHM due to this, obviously we try to keep it as small as possible.

Cheers, Robin

[RonAM]

Thanks for the explanation Robin.
I’m always eager to learn more.

Cheers,

Ron

[timh]

Hi Robin and Ron,

I am also eager to learn more! Just to maybe help wrap this thread up. The thread title is probably misleading because I asked Robin the question about optimizing alignment because I was interested in trying to optimize parameters for using a Dobsonian

In theory, if field rotation were the only limiting factor for a GOTO dob then, for small objects of interest with a small radius - preferably those not too near the Zenith and not due South or North - and successfully held in the centre of the field then it should be possible (r * sine Theta )to go right up to exposures of 30s or so without too much blurring. in connection with this I was wondering about how the alignment process itself worked and whether it was better to oversample or align using a half sized frame (since only the centre is of interest)?

So using the few clear sky opportunities that there have been I did some experiments targeting M52 which is in the NE (field rotation < 20 deg/h at time of observation) trying a variety of exposures for total times of 10 minutes.

To cut a long story short I came to understand that my quest was somewhat quixotic and that I had boldy gone where doubtless many more experienced Astronomers have been long before . Taking Robin's advice I payed attention to the Drift Graph and quickly learned that the real limiting problem with the GoTo dob is not so much rotation as the rate of translational drift. Even kept level with a spirit level and accurately initially aligned the field was drifting by 100 pixels or so in 10 minutes meaning 1.6 pixels movement every 10s which is about 3X more significant than the (about) 0.6 pixel field rotation effect in the central third of the field (of my As1294 camera frame with a 250 mm DoB at F 4.16).

Examining the images carefully and blowing them up to a scale where I could see the individual pixels it was obvious that 5.1s offered better resolution than 11s and better again than 20 or 30s. At 5.1s stars were visible above noise that could not be seen at 20s and above.

So probably just proving the obvious but all good learning. In future I will probably generally be using the Dob at exposure time about 5s, slightly oversampling, gain about 200 (reduces pattern noise a bit relative to unity gain) and probably with the reducer to keep the F number as low as possible in order to make the most of the necessarily short exposure time. I do now have a manual field rotator (by the way Ron --you could also try this? just pause stacking and rotate the camera every 15 degrees or so of rotation so as not to lose so much of the edges). I intend to now see whether can use SC to autostack 10 minute accumulations of 5s exposures (to save disk storage space) for up to an hour or more in order to get long total exposure times with the Dob. If I have understood Robin's Brain model correctly (quite possible that I haven't) then the low read noise of the CMOS camera and the maths of stacking should still mean that signal to noise can be high in such a 'superstack' notwithstanding the short subframe exposure time?

Maybe at some point alt/az DOBs with 3 or 4 fold better tracking will come available but until then it looks as though there is a very good reason why DOB exposures don't stray much above 10s?

best wishes
Tim

[RonAM]

Hi Tim,

What Dob are you using?

Perhaps I’m lucky with my Starmaster or my criteria aren’t as strict, as I’ve been successful with up to 90 seconds exposure per frame.

Qualifier: This is for objects from about 30-70 degrees altitude. Performance is poor towards the zenith or horizon so I don’t even try unless the object is in the more ideal range.

I’ve learned 2 things that help tracking performance a lot, beyond good balance of the upper truss tube on the rocker box:

1. Keep the tracking rail clean of Teflon residue- see attached photos that show how the residue can build up. Only when clean will I get the longer tracking performance. These residues can cause minor fluctuations more than steady drift.

2. When 2-star aligning my Skycommander, I use higher magnification to be sure the target star is truly centered. I used to use lower magnification to make it go faster, but even a small deviation from true center can influence tracking performance. Best is to use the camera with SharpCap with the reticle to be certain the target star is centered.

I’m guessing all these are well managed in your system but wanted to mention just in case.

Finally, at my light polluted location and 14.5” aperture I’ve not used more than about 10 seconds exposure per frame with SharpCap. I’ve used up to 90 seconds with a DSLR, unguided, in the past, at dark sky locations.

Some day I’ll get to the dark sky and try SharpCap at the longer exposures. Then I’ll probably see what you note.

All the best,

Ron

[timh]

Hi Ron

I am using a Skywatcher 250 flextube GOTO - about 7 or 8 years old now. Bortle 6 skies not too far from London and a local town centre.

The Skywatcher GOTO tracking is certainly fine for visual use (something in field of about a degree or so will still reliably be there an hour later).

Noticed that it does really help to get the base set dead level with a spirit level (as I am sure you have also found). Agree with you on the balance (currently have a steel hammer taped to the bottom of the tube to counterbalance all of the camera paraphernalia at the other end).

With the reducer (skywatcher 0.9X) f becomes about 1050 and the F number 4.16 so quite fast...so focus and collimation critical. I attached an electric focuser so as to be able to do that remotely.

It is all now married up with an AS1294 PRO camera having 4.63 uM pixels. So should be about 0.88 pixels per arc sec under OK seeing) - and with short enough exposures Sharpcap does indeed give best FWHM score of about 3.5 pixels (i.e about right I think for a seeing limit of about 4 arcsec?). As I increased the exposure I noticed that the SC FWHM score (presumably of the same stars that it uses for frame alignment?) would go up -- so at 20s for example I was seeing FWHM mainly above 5.

It was the Sharpcap Drift Graphs that put me onto the size of the problem with drift. The graph told me (sorry I didn't capture a screenshot) that in my case, in ten minutes the field was drifting about 80-100 pixels. Therefore up to 1.6 pixels within a 10s frame - and enough to cause some star shape distortion and spreading. For visual use this amount of drift is inconsequential of course.

I am hopeful of improving on things (found out that my scope wasn't really level on the last run out) and if I can halve the drift then exposures of 10-15s will look sensible.

I must emphasise though that I am just being a perfectionist here. I have had some really good pictures taken through the dob -- I really like the way it is possible to get such big bright and highly magnified images so quickly. Mostly I have been using 10s exposure and gain 200-285 and where I have managed to stack up images to 20, 30 min or so some good pictures of galaxies -- just experimenting to try to work out what the absolute optimum conditions are. So far the experiments are saying that sharpness and resolution are better at lower exposure times (5-7s or so) ..although of course such short times come with their own price in terms of increased contribution of read noise etc.

Anyway ..the tip I got from Robin was to pay attention to the drift graphs. The FWHM score variation with exposure might also be useful? With a 14 inch scope you must be able to get an awesome amount of light onto the camera even with quite short frames?

best wishes
Tim