Binning...

[szymon]

So I've been starting every process I do with binning 2x2, because all of the tutorials show that this is "the right thing to do at the beginning", because "all modern data is oversampled". But then I recently started doing some maths, and I'm not so sure that my images are in fact oversampled, even considering the tiny pixels in my Hypercam 183M.

I use the following hardware:

Altair Hypercam 183M (5440x3648 2.4u pixels)
Celestron 80ED (600mm f/7.5)
IKHAROS Flattener/Reducer (0.8x)

Running the numbers in a random astronomy calculator found on the Web shows the following:

Resolving Limit (Dawes): 1.45" arcseconds
Ideal Resolution : 0.67" - 2" arcseconds
Resolution 1.03" arcseconds per pixel
Sampling Good
Camera Chip FOV 1.56° x 1.05°

https://www.bintel.com.au/tools/astrono ... 42e4af0916

1.03" arcseconds per pixels seems just about right, no? If so, it looks like my data isn't actually "oversampled" as such, and therefore by binning I am actually losing data -- no?

Or am I missing something here with the calculations?

[admin]

2 arcsec/pixel is a probably a decent guideline, but it's going to be dependent on seeing(!), focus and guiding (and anything else in your optical train that may diffract the light).

Unless you're based in the Atacama desert, atmospheric conditions are usually the limiting factor. 1.03" is... optimistic

From Wikipedia;

A 1.0″ seeing is a good one for average astronomical sites. The seeing of an urban environment is usually much worse. Good seeing nights tend to be clear, cold nights without wind gusts. Warm air rises (convection), degrading the seeing, as do wind and clouds. At the best high-altitude mountaintop observatories, the wind brings in stable air which has not previously been in contact with the ground, sometimes providing seeing as good as 0.4".

It is certainly not the case that "all modern data is oversampled". It is often the case however that data acquired with a modern DSLR is oversampled at even moderate magnifications when using the prime focus method (e.g mounting your DSLR without lens to your scope).

It is fairly easy to see whether your real-world data is oversampled. As a rule of thumb, if non-overexposing ("small") stars only occupy at most a 3x3 patch of pixels your data is not oversampled. If they occupy more than that your dataset is oversampled.

[szymon]

It is certainly not the case that "all modern data is oversampled". It is often the case however that data acquired with a modern DSLR is oversampled at even moderate magnifications when using the prime focus method (e.g mounting your DSLR without lens to your scope).

I assume that also goes for modern CMOS cameras like my Hypercam 183M (I guess that's pretty much the same as a modern DSLR really, with the Sony IMX sensor).

It is fairly easy to see whether your real-world data is oversampled. As a rule of thumb, if non-overexposing ("small") stars only occupy at most a 3x3 patch of pixels your data is not oversampled. If they occupy more than that your dataset is oversampled.

Ok, that's a useful guideline. So the following shows that my data is oversampled and can be safely binned without losing anything, right? (This is just the raw data after a wipe, no real processing yet, zoomed in in StarTools).

oversampled.jpg (444.37 KiB) Viewed 7278 times

[admin]

That - indeed - looks oversampled to me, with the smallest stars taking up a fair few pixels to describe what is essentially a point light.
Nice job on the Elephant by the way!

[szymon]

That - indeed - looks oversampled to me, with the smallest stars taking up a fair few pixels to describe what is essentially a point light.
Nice job on the Elephant by the way!

Thank you. I am learning so much about processing with StarTools, and enjoying every bit of it . I've actually been learning a lot, especially about how to recognise processing errors and work out how to correct them, from the NINA developers on their Discord. I use NINA for all my capturing now, but folks there will gladly discuss processing too! They do seem to be rather fixated on PI and dislike StarTools, although they've been impressed with what I've shown them of what it can do...

[admin]

That - indeed - looks oversampled to me, with the smallest stars taking up a fair few pixels to describe what is essentially a point light.
Nice job on the Elephant by the way!

Thank you. I am learning so much about processing with StarTools, and enjoying every bit of it . I've actually been learning a lot, especially about how to recognise processing errors and work out how to correct them, from the NINA developers on their Discord. I use NINA for all my capturing now, but folks there will gladly discuss processing too! They do seem to be rather fixated on PI and dislike StarTools, although they've been impressed with what I've shown them of what it can do...

It's just great to see people progressing so quickly, through asking all the right questions. Kudos!

Ah yes, the PI hive mind.
(slight PI rant coming up)

StarTools quickly earned a reputation as software "for beginners", because it's easier to get decent results, quick. Yet in terms of signal fidelity it also runs rings around the old sequential processing engines like PI.

"Surely you can't have your cake and it it too, right?" I've found that that is pretty much the train of thought of people who have never tried ST, or - more often - have never taken the time to understand the underlying tech that accomplishes just that. In the past I've seen different, odd or nebulous justifications of PI's "superiority" ("more control" with PI, ST is "black box", ST is "automated", ST is for "lazy people", etc.). As time passed, fortunately ST started drawing more and more advanced users away, mostly from PI (hence this long standing sticky for people curious about the difference and moving between the two).

It also probably doesn't help perceptions that I am keen to keep ST very affordable. Cheap things can't be just as good (let alone better!) than more expensive things, right?

PI was actually one of the main reasons why I started ST. Some (not all!) - frankly elitist - attitudes around PI rubbed me the wrong way. "Just get a better scope", "just get more data", "go to a dark site". No. How about I, as a developer, improve my software and signal path so that you don't have to spend more time and money than is necessary.

AP is not about "paying your dues", nor in time, nor in money. I also don't care how long you have been doing this or how long you took to process an image, or how much money you sunk into your setup, or even what software you used. It is immaterial to how good your work is, it is also immaterial to the weight of your opinion of the work of others. What matters is how well you use what you have at your disposal (gear, time, circumstances).

You, yourself, are a great example of someone who virtually came out of nowhere (from my perspective) and, in a short amount of time, is producing increasingly impressive images from a highly light polluted site. It's just fantastic to see and really makes my day! Meanwhile I can think of a number of prolific PI users, producing lesser work with better gear, more time and more favourable circumstances...

Cheers!

[almcl]

Beautifully put, as always, Ivo.

Bravo!

[matthew_carlos]

In other forums its mentioned that perfect binning is 3x3 star. What does that mean and how do I achieve this? In other words what do i need to do to get perfect or as close to possible binning and what should I see when I'm just there and just beyond it.

Thank you!

[admin]

In other forums its mentioned that perfect binning is 3x3 star. What does that mean and how do I achieve this?

StarTools - uniquely - allows for fractional binning, which means you do not have to choose integer numbers. This allows you to choose the perfect amount. The "perfect" amount is the amount where one unit of detail is perfectly described by one pixel. In practice this means that a non-overexposing ("small") star will take up roughly a 3x3 patch, but not more and not less.
For more information,see here.
Hope this helps!