When trying to work at 0.5 arc sec / pixel and below the practical limits on sharpness become apparent. With good collimation and a sufficient number of short exposures that only the best need be retained for integration then what - aside from the seeing - is there left to improve upon?
Interesting recent learning for me is that the mount movement in RA can be a real limitation - see below.
For a long time I have seen ellipticity (r value 0.8-0.9) and especially so in unsaturated stars. There are so many possible causes of this that pinning it down can be quite a detective story - and probably particularly so in long focus Newtonians.
I had at first put this assymetry down to uneven DEC/RA guiding - but then noted that it didn't go away even in 1 or 2s short frames - either unguided or when PHD2 RA/DEC RMS guiding figures were near even. Then - for quite a long time - I assumed that I must have some alignment error or problem with persistent poor collimation or uncorrected coma. But using a laser, Cheshire eyepiece and out of focus star views I eventually became confident that collimation and alignment were good. Finally I noticed that when I analyzed past images by measuring the PSF of stars that the direction of elongation was consistent both within and across images - it was always quite close to the RA axis. It was also the same after swapping telescopes. So obviously it couldn't be anything to do with the scopes, the seeing or tube currents either.
So it was puzzling. At high resolution (not so noticeable at lower) stars were elliptical even in 1s frames when guiding was better than 0.5 arcsec RMS , similar in RA and DEC and the telescope was well aligned and collimated?
Finally did some experiments to see just how short exposures needed to be in order for the stars to stop looking elliptical. Used Sharpcap to collect video .SER files at 100ms, 200ms and 600ms, used PIPP to then extract the .FIT files and then finally compare images made by registering and integrating the different exposure frame sets. The camera was a ZWO AS1294 MM PRO, the telescope was a PDS200 F 5.0 Newtonian with Baader MkIII flattener mounted on an Ioptron CEM70.
The attached images - all at 0.45 arcsec/ pixel - show some of the results. The left hand image is an integration of 917 x 100ms (gain 324) frames and the right hand 202 x 600 ms (gain 124) frames of part of the M3 core. The difference is quite striking. The 100ms image is really quite good for a total exposure of only 91s - it may be the best resolved (FWHM ~ 1.5) picture of M3 that I have obtained so far? By contrast, all the stars in the 600ms image - taken just minutes after the 100ms image -are blurred by elongation - again in the RA direction.
So the conclusion in the end was quite clear. My mount clearly suffers some high frequency oscillation as it tracks - between 2 and 10 Hz - and mainly in RA and up to a couple of arcsec peak to peak. PHD2 doesn't indicate the problem because it acts as a low pass filter only sampling every couple of seconds or so. I further confirmed this oscillation more directly (not shown) by making AVI videos of the frames - which clearly show significant oscillatory sideways movement over and above the general non-directional seeing jitter. It could be gear meshing or something else but I'm throwing it over to Ioptron to look at - whatever else the RA resolution of the mount is not 0.07 arcsec. I'd be very interested to know if anyone else has run this sort of quite rigorous test on their mount movement and what they see ? - the effects become relatively subtle at resolutions much below 1 arc sec / pixel and/or in longer exposures with more saturated stars and/or poorer seeing. So not an obvious limitation under many conditions.
Anyway the above was interesting learning if a bit frustrating -- and it also made me better understand perhaps why deconvolution has been quite such a beneficial a process for a number of my images. Deconvolution works particularly well for cases where you have consistent motion blur - which is what this is - right across the field. With this in mind I dug out some of the images of M57 that I took this time last year when I first started dabbling with the use of semi-lucky short frames to improve resolution. i.e. viewtopic.php?t=4243
So the second picture is essentially the same as last year's image of M57 but with some additional short frames added and deconvolution applied. The deconvolution makes a major improvement - and I was pleased to see that the added detail is not artifactual (can see evidence of the same features in the HST image). So while motion blur is a nuisance deconvolution can provide something of a post facto cure in some cases (the one thing it hasn't done is sort out the triangular shapes of stars - this time last year I had the Newtonian mirror clamps too tight - yet more learning
).Tim
