Perhaps this topic has been discussed over and over, but I thought it would not hurt to have a look at optimal sampling yet again.
If we would like to get the most out of a telescopic imaging system in terms of resolution and real estate, perhaps slight under-sampling combined with dithering and drizzle integration might yield most optimal results.
Slight under-sampling usually means we can use a smaller telescope that is less demanding on the mount and is cheaper than a larger telescope of the same built quality. It also means we cover a somehow larger field in our images as opposed to imaging at less arc-seconds per pixel with the same camera. Drizzle integration when applied properly recovers some detail thus increases resolution in the final image.
This was my exact thinking when I recently bought a new telescope. For good or bad, I resisted a very strong temptation to get a larger telescope. My new imaging system will sample outer space at 1.21 arc-seconds per pixel. Drizzle x2 will increase the scale to about 0.6 arcseconds per pixel, but more realistically I feel that the true detail in final images will be probably comparable to what I would get when imaging at about 1 arc-second per pixel (maybe less), while the filed of view captured in final images will be larger.
So in theory, if I permanently incorporate drizzle integration in my data processing, I should be getting seeing-limited detail as long as I will be shooting from a near coastal location in Australia. Of course I have not considered light pollution, pixel well depth etc, but that I feel is a different story.
Any thoughts, and also pointing errors in my reasoning, will be highly appreciated.
My setup is similarly based around a small scope (Esprit 100) and a cooled CMOS with small-ish pixels (ASI1600). I've been quite amazed at the detail the ensemble produces.
No doubt I'm undersampled at 1.4"/px but it's still possible to spot seeing limited nights - I have a couple of targets where the subs from one night are noticeably better/worse than the night before, according to PixInsight.
I'll have to try drizzling during integration since taking lots of subs with the ASI1600 is de rigeur...
Thanks Dunk, I hope you meant interesting thoughts in a positive sense. Just kidding
With my previous telescope I was sampling the night sky at 1.33 asp, and with 3nm filters I was most of the time undersampling by a small amount in Brisbane. In particular tiny stars looked squarish. Drizzle integration made them nice and round, but I never actually measured whether I got any more detail in extended objects, and if the only improvement was better looking small stars at the expense of a 4 times larger image.
I look forward to reading about your experiments with drizzle integration. It will take a while for PI to process them with drizzle integration if you have a lot of subs, but with drizzle the more subs the merrier (read: better)
As you probably know, I'm a fan of drizzle integration. My local imaging/drizzle experience has been at image scales of around 2 arcsec/pixel or more. At SRO we are imaging at 1.26 arcsec/pixel and drizzle works well there too, although the fantastic seeing probably helps.
I think your theory is sound and you should get good results.
I do a fair bit of drizzling and I am imaging at 1.158"/pixel. Depending on wind direction (amongst many other factors) I actually get some quite reasonable seeing in Melb some nights. The best I've seen has been 1.6" seeing but that didn't last long, routinely closer to 2-2.5".
I personally have found that even if you're at a FWHM of 2 it is beneficial drizzling as it makes processing considerably easier, more computer intensive but easier to draw out information.
Coming back to this now I have access to my data again...a nice bright example where the signal is good...just around and below the Keyhole of Eta Carinae Nebula.
Here's the same data represented in native (upsampled x2) and drizzled (x2) form, with no additional processing besides the STF being applied (post-crop).
Anyone care to comment on what my resolution might be now?
Stars look a little smoother and you're getting the noise reduction (really noise hiding) effect of Drizzle. I don't see much increase in resolution, Dunk. Would be worth comparing with SubframeSelector. It can measure stuff that's much harder to see.
I'm having a lot of fun tonight combining drizzled data from a refractor (TEC-160FL) with undrizzled data from a RCOS 14.5". It's working, but not as well as I'd hoped... so far.
I've been playing with some data from a fainter target (the Rosette) and observing the effect of changing the drop shrink variable. The data was collected on a different night where the seeing appears to be not so great (~3" FWHM) but the noise hiding seems to be favoured by higher drop shrink values.
I've been playing with some data from a fainter target (the Rosette) and observing the effect of changing the drop shrink variable. The data was collected on a different night where the seeing appears to be not so great (~3" FWHM) but the noise hiding seems to be favoured by higher drop shrink values.
Smaller drop shrink values trade off resolution for SNR, as you'd expect. TANSTAAFL
