Consistent star shape problems. Why fix the telescope when it's errors can be fixed in post processing?

Blaine Gibby:
I would assume that if your stars are misshapen due to tilt in your optical path, then the signal of your DSO would also be proportionally distorted as well, decreasing sharpness or signal to noise ratio. I imagine the more accurately aligned your optics are, the better the final image. Looking through your posted images I would never have been able to tell otherwise so it must be working out ok in post processing,

100% Blaine ! I think people often forget that a star’s profile is one of the ways we can analyze the quality and alignment of the optics. We are not after perfect stars per se, we are after perfect optics.

I guess that then leads to a question of why you are into astrophotography. Is it just for the photo or is it the pursuit of a perfect optical train ( or as close as you can get ) and the challenges that presents. Both are fair reasons to do it but each way defines how you go about it.

Ryan Jones:
I guess that then leads to a question of why you are into astrophotography. Is it just for the photo or is it the pursuit of a perfect optical train ( or as close as you can get ) and the challenges that presents. Both are fair reasons to do it but each way defines how you go about it.

Hi Ryan,  That is certainly true.   It is probably a bit of both with me .  I take satisfaction in having the instrument  properly tuned and working - but of course it is pointless without delivering images - and also taking an interest in what the images represent and some of the science behind it.

Deconvolution is interesting and I just wondered how widely folk use it?   It is no good for fixing the actual star shape problem itself because a star is not a continuous function and in applying deconvolution you would normally mask stars  off but it is a pretty powerful process for fixing the problem that is implicit in seeing odd shaped stars  -- i.e. to improve the resolution of the objects of interest such as  nebulae etc within the image .

(Have a new suspect for the scope star-shape problem now -- the F4 coma corrector has some play within the focuser  - so I expect I'll now go back to tinkering with it again :-) )

Tim, I have a 12" newtonian with a mirror cell and mirror from OOUK and from my nightmares with it I would say the best thing is to melt it down and make a beer mug with the glass and some key chains with the mirror cell. Personally I'm after perfection and if it cannot be achieved mechanically I want nothing to do with it. One thing I would be sure to check with your newtonian is if the collimation is the same on both sides of the mount. Likely the issue that happens here (as is for me) is the mirror tips in the cell.

Matthew Proulx:
Tim, I have a 12" newtonian with a mirror cell and mirror from OOUK and from my nightmares with it I would say the best thing is to melt it down and make a beer mug with the glass and some key chains with the mirror cell. Personally I'm after perfection and if it cannot be achieved mechanically I want nothing to do with it. One thing I would be sure to check with your newtonian is if the collimation is the same on both sides of the mount. Likely the issue that happens here (as is for me) is the mirror tips in the cell.

Hello Matthew,  that  must have been a very disappointing bit of kit to say the least!  -  - I'm lucky not to have experienced that bad.  Quality control must be poor because to be fair my particular VX12  isn't junk - or at least in my judgement  has produced decent images- probably my best in fact.  But there is always the "could it be even better?" question --and  then "how much effort is it worth  for 20% improvement (versus  just  'patch' during processing) ?"  - as well as "what is it even reasonable to expect? " etc.  

I value the wisdom of others here because I am just not experienced enough with big F4 Newtonians to be able to make comparisons or to know what to fairly expect at the price point.  When I was considering the purchase the obvious competitor was the Quattro 300p .  But the weight difference would have meant  a whole new mount - loss of transportability etc.   Also - as far as I could see -  the same criticisms are levelled at all of the big Newts  anyway?

(cf v. the comments below from a review of the Quattro...    https://www.amazon.com/product-reviews/B01066PM46  just the same really

1) you need a seriously capable mount. With my camera train attached this scope weighs 60 lbs. An EQ6 or EQ6-R will not cut it...
2) you need a quality set of collimation tools. I use the Farpoint 2" Super Collimation Kit
3) there was significant tilt in the focuser on delivery. But the focuser has built in tilt adjustments to fix this
4) as delivered the primary mirror was shifting in the cell which changed the collimation substantially as you pointed the scope in different directions. I fixed this by adding shims between the mirror retaining clips and the edge of the primary. Do not over tighten the clips as it will result in pinched optics.)

So I plumped for the OO VX12 hoping to pre-empt any potential issues by adding a Baader steeltrack focuser and also a long dovetail and top-plate for rigidity.

Anyway  your advice is correct of course.   I did exactly that and used a Hotech laser for alignment - firstly of the secondary - and did indeed notice  that if I aligned the spot  on the mirror centre and then rotated the scope by 180 degrees in DEC the spot moved  off centre to the edge of the ring.  So something had certainly shifted or flexed .  After tightening everything up - including the secondary mirror supporting spider - there was less but still not zero movement.  In practice therefore I set the scope up for best collimation at the sort of elevations I normally image at - normally not too far from the zenith because of surrounding buildings/ trees etc. But it is another compromise.   If you are right about the mirror cell movement - which seems likely with its weight - then maybe the clip shimming is the way to go?   

Tim

Just to wrap this thread up should anyone ever read down this far. 

Somewhat ironically - having started a thread about the utility of deconvolution etc in post processing as a way of compensating for difficult to pin down faults in telescope set up -   I did in fact manage to fix the telescope.  Following  Matthew's comment above about mirror flop in the OO (UK) VX12 I contacted OO and they advised me to tighten up the nylon screws in the mirror cell.  I did this and now all seems well with good collimation, alignment and no significant movement of the laser spot from the mirror centre – it will be interesting to see how much of an improvement this engenders.

Tim

Chris White:

Tim Hawkes:
Just to wrap this thread up should anyone ever read down this far. 

Somewhat ironically - having started a thread about the utility of deconvolution etc in post processing as a way of compensating for difficult to pin down faults in telescope set up -   I did in fact manage to fix the telescope.  Following  Matthew's comment above about mirror flop in the OO (UK) VX12 I contacted OO and they advised me to tighten up the nylon screws in the mirror cell.  I did this and now all seems well with good collimation, alignment and no significant movement of the laser spot from the mirror centre -- it will be interesting to see how much of an improvement this engenders.

Tim

Good. 

Your final image will only be as good as the data you collect.  You can "process" bad stars to look better, but you will never be able to match the quality if you capture good stars from the beginning.  Also, your star quality is an indication of the quality of your nebula, galaxy, dust, etc.... Poor looking stars, means everything else is also poor quality.  We agonize over stars because they are the canary in the coal mine for what our optics are capable of. 

In the end though, this is really a subjective thing.  If you are happy with your stars then thats all that matters... but dont confuse this subjective satisfaction as an ability to fix telescope defects in post processing.   To me, it might still just look like garbage stars.

Chris,

I agree that if the stars in an image field exhibit a poor shape then you won't be able to fix the star shape itself in post processing (or at least I don't know a way).  Also agree that whatever that distortion of star shape happens to be in terms of the PSF then the same distortion will have been visited on every other part of the image - thus a planetary nebula  or whatever will equally have been imaged  suboptimally and be distorted just as the stars were.  However my understanding is that if the stellar PSFs  are distorted in a sufficiently consistent way  right across the image (defined by e.g. a given aspect  ratio at a given angle theta ) then - it  is possible to define the consistent 'distorted' PSF  (e.g. by averaging a number of stars)  and  - by using deconvolution-  reconstruct the undistorted image of the nebula.  With objects like M57 and the catseye nebula deconvolution does seem to work very well to recover very good sharp (and true -judged by comparison with Hubble etc)  images starting  from images in which the the stars were  not round (e.g. r ~ 0.85  but consistently so).  As you will know deconvolution works well on objects (or parts of objects)  that exhibit  good SNR and  where features are characterised by continuous functions  from pixel to pixel--- but not on noisy parts of the image (these need to be masked out) or stars themselves (also need masking out).   

What I  really don't know - but it would be fun to now test  - is whether  or not the  image that is finally obtainable  after deconvolution is noticeably any better starting from images with round stars  than from images with some (say r=0.85) distortion of star shape - but consistently so, 

I suppose the argument is that consistency rather than shape per se may be the more important determinant of eventual image quality for certain types of object.  So my original argument was that if your set up produces a consistent distortion it might not matter for certain objects -- I don't know it is just conjecture

But of course I agree -- better to have the star shapes right in the first place -  makes for a simpler life and it is nice to see all working as it should :-)

Tim

Chris White:
The most common causes of deficiencies are tracking error, focus, spacing, image train tilt and flex.... so how do you generate a valid PSF to accommodate error that can be somewhat random in value in between frames (focus, tracking) or impact the image differently in different areas (tilt) and expect that it will achieve an optimal result?

Hi Chris, 

Well since you asked... How indeed ?   And also I'd turn the question around,  Given that all these deficiencies are indeed common and that most data produced flawed in some aspect or another then how do you make the best of it?

Deconvolution is a process that certainly does work - even with fairly grotty (but consistently so) stars - and sometimes surprisingly well.  The whole purpose of it is to deconvolute out the 'true' image from an imperfect image using a mathematical description of the nature of the imperfection.  The more perfect the image you start from then the less impact you might expect the process  to have 

It doesn't matter that frames vary.  Deconvolution is not applied to individual frames  but to the  integrated aligned  linear image  - which will anyway comprise only the  better and more consistent frames .  Most importantly the variation across all the selected  frames  is combined and averaged in the integrated image.  The then measured PSF subsumes  and mathematically summarises the net distorting effect of  all of the optical, seeing, mount factors etc  combined within any particular image (or part thereof).  THe PSF  is valid because it is  a measurement  -  and  the geographical scope of  its validity  can be assessed by how consistent it is  and where across the image-  i.e. if the measured PSFs of unsaturated stars across  the entire image exhibit quite consistent values then you are probably good to derive an average PSF and  apply it to  deconvoluting  target regions (e,g high SNR and smoothly varying)  right across the image - otherwise not.

Deconvolution is certainly more appropriate to apply to certain subjects than others.  Dark nebulae and faint dust clouds?  - indeed forget it - and yes it needs to be used with care to avoid artefacts and certainly judicious masking is essential.

In my limited experience thus far the PixInsight implementation of the Lucy Richardson algorithm is easy to use and works well.  There are three examples in my gallery (M57 , the core of the catseye and the core of M94)  where  the field star shapes aren't great (r values about 0.85 as I recall) but the detail and resolution  of all of these objects is really quite good -  particulalrly so for UK skies.  In all of these cases deconvolution provided an obvious and indeed quite dramatic improvement in the resolution of detail.  

Generally I do think that is useful to think about what sort of processing is useful to recover the best  possible from flawed images.  It is a useful discussion to have  ...because that represents most of what most of us get most of the time.  Most equipment (that is affordable) has some sort of inherent problem  and at least 50% - probably more -of imaging  sessions end up generating flawed data (depending how picky you are) for some  reason or another -- as you say focus, collimation slightly out , haze - dew etc.  etc.  So as one obvious example -  not very well resolved RGB images can be perfectly OK  for providing chrominance  to an image if combined with good luminance data.  I see deconvolution as just another of the tools that can be used for trying to makes silk purses out of sow's ears.

Tim