Upsizing your resolution

[rogerg]

I probably should know the answer to this, but feel it's worth a discussion to broaden my knowledge.

The golden rule once was supposedly "image at 2 arc sec per pixel".

However, if everyone was imaging at no higher resolution than 2 arc seconds per pixel there wouldn't be a resolution advantage in having a larger telescope yet it appears in practice there is a resolution advantage. Some people produce higher resolution images of small objects than others who are at about the 2" scale.

While I never subscdribed to the 2" rule I can see where it comes from - limitations of viewing conditions primarily, as I understand it. I find 0.8" is my limit.

Are large professional telescopes limited to similar resoultions? Or if not then is it only because they are at the top of a mountain or is it a factor of mirror diameter?

In practice is a 24" RC limited to the same resolution as a 12" RC?

So, how does someone increase their resolution per pixel while retaining a sharp image?

Does it come down to just adaptive optics and moving to a mountain top or is there a factor such as mirror size which does in practice have a material impact?

Resolution is a factor of focal length and pixel size, to my knowledge. While possible to choose a combination of long focal length and small pixel size (perhaps to reach 0.5" for example) this would result in a horribly poor quality image. But in practice is there a way for 0.5" to become usable?

Obviously quality of optics, tracking and equipment stability has a big impact but I'm choosing to ignore these for this example and they are not variables I am intending to discuss.

I should also clarify I'm considering deep space CCD imaging. Not visual and not planetary. Although, it's interesting how planetary imaging compares, where the resolution is pushed to the physical limits of the telescope and perhaps only compensated for by increased number of frames hence increased SNR.

Regards,
Roger.

[PeterM]

Hi Roger,
I always thought 2 arc sec per pixel was the absolute upper and should be somewhere between 1-2 arc sec per pixel.
I am probably ignorant to this (Bock is the guru). My LX200, SXVH9 with Optec F7 reducer and binning at 2x2 produces 1.23 arc sec per pixels for me.
CCD Calc helps (you probably have it, if not link is below)
http://www.newastro.com/book_new/camera_app.php
Look forward to more informed replies here.

[Paul Haese]

Currently imaging withh the RC 12 and STL at 0.77". Seeing has to be good to get sharp detail but my recent image of NGC 253 (http://paulhaese.net/NGC253.html) looks pretty good. Adapative optics would help more.

[Shiraz]

Hi Roger

did a spreadsheet on this recently http://www.iceinspace.com.au/forum/showthread.php?t=99472.
Taking into account the atmosphere, scope and tracking system, best seems to be about 1 arc sec in good seeing (2 arcsec FWHM). In practice anything from about 10 inches and upwards (including an 8m scope without adaptive optics) will have the same resolution in 2 arc sec seeing. Bigger scopes may be better in exceptional seeing - eg if they are on a mountain top. Tracking mount performance is particularly important in good seeing.

Of course the other issue is that you need enough chip area to cover the object of interest, so you may choose to depart from the optimum resolution for other reasons.

regards Ray

[cfranks]

My DSI RC10C + QSI683wsg-8 mono, give me 0.6 as/pix. A major downside is the 33x25 arc min FOV so that the larger DSO's, M42, Rosette, Omega Cen etc., have to be built as mosaics, taking that much longer to acquire. The Rosette for example is a huge 16 frame so 16 times longer to get sufficient data. At 75 I might not live that long! :sadeyes:

Charles

[rogerg]

Thanks everyone.

Based on replies so far it seems my understanding was unfortunately correct that the only way people achieve higher resoultion is with nights of better viewing conditions, and that the limiting resolution is otherwise is the calculated 0.8-2 arc second region.

I suppose though if you had a big enough aperture telescope you're going to be able to take lmore exposures in the same time, so increase your signal to noise ratio, so hence make better results from poorer viewing hence smaller arc second scale. Perhaps that's the biggest advantage of a larger aperture in this respect. Taking it to the extreme I suppose if we could image a mag 16 galaxy at the frame rate used for Jupiter, hence acquiring 2000 frames in minutes, there's a good chance images of < 0.5 arc sec/pixel might become more usable and hence resolution improved.




Thanks for chiming in Peter. Yes, have used CCD Calc for years but I just had to wonder if there was more to it than the basic calculation we and CCDCalc use.



Thanks for adding an example Paul. Your 0.77" is very close to my preferred 0.8" of my 12" SCT. Although of course in theory the aperture is irrelevant but at least we're on the same page.



Thanks Ray, I missed that post. Your calculations match my understanding.

You are right about chip area, although that's not an issue for me, as I'm usually photographing galaxies smaller than 10 arc minutes. My goal is to have more resolution on galaxies smaller than 5 arc minutes.



33x25 is enormous :lol: for me anyway. You're lucky to have 33x25! :) My 0.86 gave me 10' x 7' with my ST7, and currently I have 16' x 11' with my ST8 :) It's OK for me, hunting down small galaxies :) The main disadvantage for me is reduced plate solve accuracy due to fewer stars in the FOV size.

[LightningNZ]

I'm amazed that in all of this, no one has mentioned Nyquist sampling - http://en.wikipedia.org/wiki/Nyquist%E2%80%93Shannon_sampling_th eorem

In a nutshell, unless you want to lose information you'll need to sample at least twice has high a resolution than the smallest features you want to resolve. A good practical value we used to use in microscopy is 2.3x to account for losses.

If you have a small star that that's 3 arc-seconds across and you've got square boxes to put it's light in then what if it's contained in 1 box? You'll get a square star! So you have to put it in more boxes, enough to avoid the aliasing issues mentioned in the rather mathematically heavy wikipedia article I linked above. Enough to ensure a rather roundish star anyway.

Hope this helps,
Cam

[Shiraz]

I'm amazed that in all of this, no one has mentioned Nyquist sampling - http://en.wikipedia.org/wiki/Nyquist%E2%80%93Shannon_sampling_th eorem

Hi Cam. Nyquist sampling was discussed in the thread below, which was linked to earlier

http://www.iceinspace.com.au/forum/showthread.php?t=99472

"the spreadsheet is nothing particularly special, but it does include the main variables and gives some idea of what is important. I wanted to look at system issues and used the Nyquist sampling requirement for a Gaussian - 2.35 pixel per FWHM - but of course, seeing cannot be predicted at anything like that precision, so 2 or 3 are probably just as useful."

regards Ray

[Peter Ward]

Interesting question.

I've found larger apertures do have intrinsically higher resolutions, plus exhibit less positional wander (due air-cell induced tip-tilt) in the location of stars at the focal plane.

This comes at a cost, stars look like fuzzballs most nights... eg big fuzzies on bad nights, and teeny fuzzies on good ones.

I've also found guiding adaptively makes a *big* difference on nights where the seeing is good.

FWHM's can easily be reduced by 30% or more with AO's.

The big edge professional observatories have is their high-altitude locations almost always have excellent seeing and transparency...

Even non-professionals have been taking this altitude advantage of late...look no further than Martin's world beating results taken from Sierra Remote.

[SkyViking]

As Ray and Cam mentioned, sampling is the key. Consider that the details of the object you are imaging cannot be accurately recorded if they are sampled 1:1, but that at least 2x2 and preferably 3x3 pixels should cover the smallest discernible features in the object, given the FWHM on the night. So if FWHM was say 2", then the ideal pixel resolution is 2/3 = 0.66".

Personally I image with 0.87" per pixel. I've found that on 70% of the nights I'm out it seems just right as my FWHM is usually hovering just above 2". On 10% of the nights I wish I had better resolution, and on the last 20% the seeing is worse than 3" and I'd clearly be oversampling (but on those nights I just shoot the RGB, always binned 2x2 :)).
Others may have completely different experiences depending on local conditions.

I really think the most important thing is to find a good balance between resolution and integration time for the typical objects you tend to image.

[rogerg]

Sorry for my tardy replies, I'm tired from work and it's Christmas so my IIS time keeps getting interrupted by other things ;)



Thanks for the info. I feel this has less relevance to my direct desire to image at greater resolutions because by virtue of doing so I am oversampling already, so square single pixel stars are never going to occur, the challenge is the opposite, to improve sharpness when oversampling.



Well I'm glad because I thought I had seen this demonstrated in photos by people with larger aperture telescopes which is one factor leading to the question in the first place.

Is it possible that just like large aperture telescope can "see through" thin canopy trees (because of the great overall mirror surface area vs the few patchy leaves in the way resulting in a usable but dimmed image) larger aperture telescopes can better average out the effects of atmospheric effects to arrive at a sharper image at higher magnifications because they are sampling through a larger area of the atmosphere, I wonder? I guess that depends how large pockets of atmospheric turbulence is.

My understanding is that for adaptive optics to be effective a bright guide star is required which permits high frequency reactions. Is that correct? Reason I ask is that a challenge I face at long focal lengths is guide star availability. This means it's kind of a catch 22 - wanting higher resolution (longer focal length) yet needing larger field of view (easiest achieved through short focal length) for bright guide star acquisition.

[Peter Ward]

Yes, that said attenuation by filters is no longer an issue with newer self guiding cameras. I've found ~10Hz is pretty easy in most areas of the sky.

[naskies]

Roger, thanks - this is an interesting thread! Another couple of titbits for thought...

Very few astro images have enough signal-to-noise to even remotely approach the diffraction limit of the sampling scale. For example, many people target their imaging around completing one LRGB image per evening or month, rather than aiming for a predefined SNR threshold. If all else were equal, larger apertures would equal better SNR.

Larger apertures usually also mean longer focal lengths, and thus larger chips (with correspondingly larger pixel sizes) are needed to maintain the same field of view and pixel scale. Since the performance characteristics of small and big chips can differ quite substantially, you can get other benefits such as increased dynamic range and lower limiting magnitudes with chips that have deeper wells.

[Logieberra]

I found this to be a very good read for choosing the best pixel size for your site and setup:

http://www.starrywonders.com/ccdcameraconsiderations.html

[rogerg]

Interesting aspect to consider. How would someone work calculate the SNR of their images and the desired SNR to aim for?



Interesting thought. I hadn't considered aspects of different pixel sizes such as deeper wells, only having considered the pixel size for the sake of calculating resolution.



Just to clarify, what I am interested here is not so much calculating the "best practice" best pixel size, more so investigating how the resolution can be pushed beyond that best practice to achieve higher resoultion, what factors can contribute to higher resolution being possible.

[Terry B]

Dear Roger.
I agree with most of what the others have said but am just commenting on the point above. Theoretical resolution for a scope is proportional to the scopes diameter and the wavelength of the light. Focal length has nothing to do with it. Otherwise a tiny scope with a very long focal length would give hubble level resolution.
Longer wavelengths have lower resolution as well but for light this isn't terrible important (For radio astromony it is very important)
Practically of course other factors limit the resolution as you have discussed. I image a 0.78arcsec/pixel but often will bin 2x2 expecially for photometry of dim objects just to keep the file size down. FWHM for me is rarely better than 2arcsec and I live at 1100m altitude. I would be supprised if anywhere in Australia has much better. The AAO at a similar altitude to me is has similar seeing.
Cheers

Terry

[rogerg]

Thanks Terry for your reply, you added a few interesting points there. I'll be honest I don't understand the impact of wavelength.

Your comment regarding Hubble is one factor which got me thinking along this line. Hubble isa bit of a special case because it's outside the atmosphere, but I was thinking more along the lines of a comparison between large aperture ground based telescopes vs standard amateur telescope size, and how it is that they end up with a resolution advantage with the larger telescopes.

With your site and AAO having FWHM not better than 2, I wonder what resolution AAO image at with their larger aperture compared to your 0.78 arcsec/pixel.

[Terry B]

The formula for resolution is
R = D/lamda
where D = diameter
lamda= wavelength.
Wavelength is essentially constant for light (ie 400- 700nm) and is a very small number conpared to the diameter of the scope. For radio frequencies the wavelength is much longer (21cm is a common frequency used) and even for a dish 20m in diameter it is a significant amount. This is why radio dishes need to be so large and why arrays of dishes are used to simulate a telescope that is many km in diameter to overcome this problem.
The AAO isn't a great example as they are not known for terribly high resolution. That is why other places in the world are favoured like high in the Andes or at Dome C in antarctica.

[naskies]

Since the aim in amateur imaging is usually to create "pretty pictures", there's probably not much point in going beyond "it looks nice/smooth enough" for most of us.

Although I have no direct experience or knowledge of it, I'd imagine that SNR for deep space imaging would be similar in principle to any other imaging:

http://en.wikipedia.org/wiki/Signal_to_noise_ratio_(imaging)

The signal would be the photon flux from the targeted deep space object. Noise would be everything else such as CCD sources (read noise, dark noise, non-linearity, quantisation), optical system losses and imperfections, sky background (light pollution, light from other objects), and so on.

I'm sure the theory for this would be well established in the professional astronomy literature by now. Perhaps those with a bit of expert knowledge in the area might be able to help us out?