Have I got this right regarding imaging, seeing and resolution?

Hi Piers,

I found this article most helpful for deciding which camera to match to my optics.
Matching Your Camera to Your Optics - Astronomy & Scientific Imaging Solutions (diffractionlimited.com) .

Cheers
Gernot

If your seeing is genuinely 1" (which is really excellent) then 0.4"/px sampling is not a bad place to be.   In general you would sample at 2x to 3x seeing to get the most out of your resolution.  This does require excellent tracking as well.   If you are not able to track to the image scale of your system, it will blur your images similarly to that of degrading seeing. 

The benefit of tiny pixels is that you can produce very high resolution images with modest focal lengths.  With a larger aperture and longer focal length your system will be more efficient and resolving power higher, but there are more challenges with longer focal lengths to consider. 

I'm using the same pixel size as you.  I imaged a ton with an Edge 925 as well as my 130mm refractor and compared the data.  My FWHM was identical for both setups.  While the Edge could produce an image with less integration time, I ultimately decided that for my skies (average or worse seeing) that it would just be more convenient to use the refractor only.

Something else to keep in mind is that you dont need a larger pixel camera with a long scope.  These 3.76um sony chips are the kings for astroimaging with any focal length.  You can always bin the camera (for smaller files) or just downsample in post processing if you are too oversampled.  The low noise and well behaved data from these cameras makes them an excellent choice for any focal length scope.  If you get an Edge and a large pixel CCD you might need to take 20 or 30 minute exposures to clear the noise floor.  With modern CMOS you could still get away with your 5 or 10 minute subs. 

My $.02.

Piers Palmer:
particularly if, as you say, 3.76um camera are some sort of sweet spot for astrophotography.

Just to clarify, it's not really the pixel size that makes the current generation of astrocameras the "sweet spot" choice.  These cameras are great because the noise is low, no amp glow, clean noise profile, etc... they also happen to have a pixel size that is ideal for wide-field imaging and can be used comfortably with long focal length as well.  Unlike CCD there is really not much benefit to binning in camera, so you can image at full resolution and just downsample in post processing to effectively improve your SNR and for display/viewing purposes. 

800mm focal length and IMX571 is a good combo for imaging galaxies.  You can crop to tighten up the view and if you collect enough (well dithered) data you can get a resolution benefit by drizzling.  Also with new tools like BlurX you can eek out even more detail from your image.

You might also watch this video (or just view the slides):  https://www.advancedimagingconference.com/articles/secrets-long-focal-length-imaging-john-hayes

AIC is now free to join so this is available to everyone.

John

Just a bit on what I have experienced so far imaging galaxies.   I have found that what you can do depends an awful  lot on conditions and also very much upon which particular aspect of of galaxy imaging that you wish to prioritise.   Go for deep detail in the brighter core or go for capturing the fainter parts?  If you are at the stage of making purchasing decisions it might be worth  looking into the crystal ball a bit on that and also consider what best fits your conditions.

So here is my example…

Under the best seeing conditions here in the UK Thames Valley for example I have captured  (linear) integrated images at as low as 1.9 arc sec  using a 200mm telescope and maybe 1.7 arc sec using a 300 mm telescope  - so the actual sky FWHM must have been a bit lower than that - at least transiently for the period of many of the short (3s) frames.

However these (relatively) good seeing conditions often coincided with poor transparency and vice versa.  So  good conditions for trying to bring out fine detail in the galaxy core – but not so good  for trying to capture the whole extent and faint outer reaches of galaxies.  

For broadband objects like galaxies filters don't help a lot and sky darkness is critical for seeing the outer reaches and fainter parts.  I live at a Bortle6 site so to capture frames that add information about the fainter parts I travel to a darker site – at least Bortle 4  or better to achieve 10 fold increase in SNR or more.

But just back to resolution it is possible to beat the seeing to a certain extent and gain more resolution than the seeing should allow - (e.g in some images I estimate being down to ~ 1.2 arcsec) - and particularly in the brighter regions -  by using a combination of 'semi' lucky imaging  and image deconvolution (or now BlurXterminator) .  These approaches work quite well on galaxies -   but only really on detailing the core regions and the brighter galaxies.  It is an aspect of galaxy imaging that I enjoy and that - to a certain extent - local conditions constrain me to.

In case you didn't happen to know (sorry if this redundant info)  - deep sky (so-called) lucky imaging involves stacking many short frames (~ 3-5s)  rather than using the normal longer exposure times.  Part of the power lies in selection  – many of the frames are rejected - so the ones you choose to stack are brighter and have lower FWHMs.  Particularly combined with deconvolution it is one route to getting a sharper image if you don't happen to live where seeing is good most of the time.  However because the frames are necessarily so short  SNR is a real problem - hence the focus on bright objects and brighter parts of objects.   For this approach therefore it really also helps to have a bigger D as well as a longer f and reasonable speed as well -  so ~ 10 inch Newtonians at F 4-5 etc.   The other aspect is deconvolution  for which it actually helps to oversample somewhat   – i.e.  if you want to end up sharpening  to an effective image resolution of 1.5 arcsec  or better then - to sample sufficiently -  would require an image scale of < 0.75 arcsec/ pixel.

To be honest though if I'd have lived in the Californian mountains and had 1 arcsec seeing then I wouldn't have bothered with lucky imaging at all probably….

I only mention the above - which may be a sidetrack in many ways - because if your skies and conditions did a resemble mine then it could impact your choice of equipment for galaxy imaging down the road.   So for the particular approach above I'd go  for a 10 or 12 inch scope,  f 1000-1200  and look to at least have the facility to  sample at 0.5 arcsec/ pixel   (you can always bin when it is not needed).

Tim

John Hayes:
You might also watch this video (or just view the slides):  https://www.advancedimagingconference.com/articles/secrets-long-focal-length-imaging-john-hayes

AIC is now free to join so this is available to everyone.

John

Thanks John. I’ve just signed up and will Take a good look!

Piers Palmer:

John Hayes:
You might also watch this video (or just view the slides):  https://www.advancedimagingconference.com/articles/secrets-long-focal-length-imaging-john-hayes

AIC is now free to join so this is available to everyone.

John

Thanks John. I’ve just signed up and will Take a good look!

Definitely do, it is very good and from some perspectives should be highly encouraged content for all imagers to consume. 

I specifically want to call out a few sections of the content (not in order):

• Slides 42-44. Love these graphs, as they can really help one to see how a system will work given different seeing conditions. For example, I will be deploying a 14" CDK with a 3.76um pixel camera (pretty common pixel size these days). In 0.5"-1" seeing, this combination will produce great results at bin 1x1. As the seeing shifts to 1.5"-2" that is when I should consider binning the data - which can be done nicely in post-processing.
• Speaking of Binning in Post, if you continue through the presentation, you will hit Slide 50. At the bottom of the slide there is some Pixel Math included by John that can be used to perform binning in post processing. Of course, there is also the Integer Resample process in PixInsight as well. Be sure to review all of the content here (and not just the math) as John gives some good insight (as always).
• Slide 18. john talks about optimal spacing for sensors here, and the real-world experience I have had matches what John has stated here. The larger the sensor you are working with, the more delicate those space adjustment can be to get things just right. I was surprised a little to hear that that C14 at F11 would have tolerances less than 100 microns, but it goes to show that for big chips one needs to be prepared to put the work in to improve image quality.

Of course, there is a lot more in this presentation than just these bits. The Seven Keys to Success he calls out are all very important and can help one step up their game immensely.

My guide camera is a 290mm mini. Even something as humble as that?!

andrea tasselli:
All this talk of lucky imaging with galaxies leaves me rather cold. It simply doesn't work. You'd need small angular field, very bright subjects and large, fast apertures. Else, you'd need very good seeing. And I know because I did it when nobody was doing it (but in my case were the handful of bright PNs). Now you can much better results but the recipe is still the same. And good seeing, as an average. At least for broadband subjects.

I tend to avoid the lucky imaging topics as it's not really something that interests me, and the results I've seen that were very good were basically all of the same very bright galaxies. Which isn't to say that it's bad or anything, my interests lie in the deeper integrations, fainter objects, and even wider field Galaxy Clusters.

For people interested in max detail there is no reason not to use very small pixels - for reasons that may not be obvious at first, but should make sense once it is explained.

Typical discussions of "pixel matching" to seeing only refer to a single exposure, and they ignore what happens under the covers when you align and stack.  This is where it helps actually to write code that does this - so the details are evident - but it is easy to understand.  When you align and stack, the pixels don't just stack on top of each other.  They are all being shifted and mapped into the destination image - and if the destination image has the same size pixels - it means each pixel will tend to contribute its info to a small tile of 4 pixels.  And that has a blurring effect.

There is no way around this and it happening every time you align and stack - but people don't tend to be aware of it because they don't "see" it happening directly.  But if you take a given exposure and try to map it into an image with a shift in x, y a fraction of a pixel - the result will be smoother and softer.  *That* soft image is the one used in the stack - not the original sharp one to which people apply Nyquist.

The key thing is that this blurring happens *on the scale of the pixels* - which means that smaller pixels will result in less blurring.

As for applying Nyquist here - it doesn't apply directly because the requirements of the Nyquist theorem are not met in the first place.  But even if you apply it as a rule of thumb and say "pixels should be 1/3 the fwhm" or something - you would be better applying an additional factor of 2 so it is 1/6 fwhm or so.  The thing is - smaller will always have benefit because - again - *the blurring happens on the scale of the pixels*.  Smaller pixels = less blurring = smaller realized fwhm.

For people who don't hope for better than 2.0" fwhm and end up using 0.5" pixels - they will likely get their wish and never do much better than that.  But I think everyone should aim for at least 1.5" fwhm - if they do in fact want max detail.  I haven't imaged from Umbria, but I did spend a few years in Dublin, Ireland and despite the horrible conditions there, I think 1.5" was possible on at least some nights, and pixels around 0.3" would be good for that.

In my case I use EdgeHD11 at f/10 with 3.78um pixels or 0.28" per pixel - and I often get fwhm's in the low 1" range.  But it does require good collimation, focus, and guiding.

There is always the issue of possibly more noise with smaller pixels and no significant gain if the seeing is poor - but in those cases you can bin or smooth the final result.  The key is that the small pixels minimize the blurring that happens during stacking - so you still benefit.  This is in contrast to having large pixels and not even realizing that you could have done much better on a given night.

But this is only relevant for people who do want max detail.  For wide field images where 2.5" or so fwhm is fine, there is no need to go below 0.5" or so pixels.  But there are lots of interesting and tiny objects up there that are rarely imaged - so it's a rich subject for study - with small pixels.

Frank

There is no way around this and it happening every time you align and stack - but people don't tend to be aware of it because they don't "see" it happening directly.  But if you take a given exposure and try to map it into an image with a shift in x, y a fraction of a pixel - the result will be smoother and softer.  *That* soft image is the one used in the stack - not the original sharp one to which people apply Nyquist.

The key thing is that this blurring happens *on the scale of the pixels* - which means that smaller pixels will result in less blurring.

Inter alia -- that is a very good point.  (A very dull thing to do I know) but in PI you can watch the error numbers go by as the frames get aligned.  I agree also about aiming for better resolution and taking the pixel size down which also matters if you eventually want to apply deconvolution.  FWHM ~ 1.5 or so is occasionally  possible  even in the Atlantic climate of Britain and Ireland.

And you would need to use OAG if you are striving for max detail. Switching from f/4 with guidescope (I assume you use guidescope?) to f/10 with OAG is a big step in complexity - but it should have benefit if you want max detail in small objects.

That isn't exactly true.