That old chesnut 'Undersampling'..

[900SL]

For various reasons I'm pairing a 533MC Pro with a Samyang 135 and ASI Air Pro for a Greece trip end of September.

There's a lot of positives with this set up, including no guiding required, shorter subs (3 min Ha/Oiii NB at f2.8 v 9 min on Redcat at f4.9). more forgiving of tracking, polar alignment errors and easier to find targets (platesolving wide FOV)

I'm hoping to get 8+ hrs on some of the larger targets and it's a bortle 2 region, so should be whoopee data

 

However, there's always a however in AP:

1: Pixel scale is around 5.7"/px:  Am I likely to see a significant softening in images at this pixel scale at full screen size due to the lower resolution? I can't drizzle as I am not dithering on the tracker.

I'm not too bothered about blocky stars at 100% zoom (i'm not a pixel peeper) but I do like sharp nebulosity images on my monitor 

[vlaiv]

Undersampling does not produce blocky stars - that is just a myth (and very persistent one - like many other myths in amateur astronomy).

At 5.7"/px - you will certainly not be under sampling with lens - you might be even over sampling.

Just to give you some idea - at F/2.8 and with 135mm of focal length - your aperture will be opened to ~48mm. Airy disk diameter alone is 5.77" at 550nm. Add to that tracking and seeing and you'll be close to optimal sampling even with perfect aperture.

However - look at this:

At 30 lpmm we already have ~95% attenuation of signal. 1000um in a mm so 1000um / 30 = 33.33um - even for 16.66um pixel size we will see some sharpness loss, let alone for 3.75um pixel size. This is solely due to sharpness of the optics and not related to other things. Lens are simply not diffraction limited.

1 hour ago, 900SL said:

Am I likely to see a significant softening in images at this pixel scale at full screen size due to the lower resolution?

Under sampling does not produce softening at any pixel scale - quite the opposite.

What under sampling does is to produce aliasing artifacts when conditions are right (and you are truly over sampling). For example:

this image as is already shows some effects of aliasing, which means that it is already under sampled. Look at the side of this building - if you look at pattern produced by windows as a whole - you will see strange curved "wave" appear formed by windows - although in reality we know that there are no curved wavy like features on the building.

Look what happens when I resize down this image without taking care of under sampling issues (there are algorithms that can deal with this to some extent by using anti aliasing filters):

pattern just got worse and more visible.

That is aliasing and is effect of under sampling. There are no other effects - we don't get "softening" as effect of under sampling - both images are very sharp - just look at windows on the face of the building - they are very nicely defined and edges are high contrast. Nothing soft about this image.

With astronomical images it is very very hard to see effects of under sampling because of the way astronomical images are composed and because of the way under sampling artifact - namely aliasing works. Even when you are grossly over sampled with astronomical images - you won't see any aliasing issues.

In a nut shell - you don't need to worry about blocky stars, don't need to worry about under sampling and if your image turns out to be slightly soft when viewed at full size - that is because you are probably over sampling for that lens with 3.75um pixel size, as lens itself is not diffraction limited.

[Elp]

That MTF graph above is at F8?

[ONIKKINEN]

Here is an astrophotography example of a severely undersampled image:

And then a closeup at 500% zoom:

The smallest stars, the ones that are actually only one pixel look somewhat blocky at 500% zoom but none of this was visible at 100% because they appear like point sources. None of the more than a single pixel sized stars appear blocky.

In short, dont worry about undersampling, which may not even occur.

[vlaiv]

32 minutes ago, Elp said:

That MTF graph above is at F8?

Yep and it is not much worse than fully opened.

You can see both here:

https://www.samyanglens.com/en/product/product-view.php?seq=323

under data tab. Fact that both look pretty much the same at F2 and F8 is just a sign of a good design.

You can't really have diffraction limited optics with so much elements and corrected for most aberrations including flat field over large sensor size. That of course does not mean that this lens can't be used for lower resolution work - and indeed it works excellent for that.

9 minutes ago, ONIKKINEN said:

The smallest stars, the ones that are actually only one pixel look somewhat blocky at 500% zoom but none of this was visible at 100% because they appear like point sources.

Blockiness of stars (I just love how we use non existent words - no spell check support, is it blockyness or blockiness? ) is function of resampling method used. If one uses nearest neighbor - they get square, other methods give different results.

The closer one gets to ideal filter (unfortunately, ideal filter does not exist as we would need it to be spatially infinite - because it needs to be finite in frequency domain) - more "airy disk" like enlarged star becomes.

This is single pixel enlarged by 2000% (x20 enlargement) - with different algorithms used. Top left is nearest neighbor - which simply produces square, is the most simple method and is the root of the blocky stars myth. Back in the day most software used this for enlarging images as it is by far easiest method to both implement and execute - does not require fancy math knowledge, so people concluded that under sampled stars look like squares because that is how they saw it in software.

Top right is bilinear interpolation - it creates sort of "star" pattern in that one "pixel". Bottom left is bicubic and bottom right is quintic b-spline. Note how "first diffraction ring" starts forming with advanced methods. This is because both telescope aperture and limited number of samples do the same thing - they limit maximum frequency in frequency domain which results in ripple kind of effect in spatial domain:

This animation is brilliant at explaining actual effect - if you use infinite number of sine waves - you can actually get square pattern by summing them all - but if stop at certain point - you can't produce sharp edge and there is some residual rippling left. This is why airy rings form - if we had aperture of infinite size and we could capture all frequency components - we would simply have star image as a single dot - but since we cut off higher frequencies - we get this residual rippling that has not been canceled out by higher frequencies.

[900SL]

Thanks Vlaiv, Oskari, much appreciated. I'll take the SY 135mm.

[ollypenrice]

When I process Samyang images I de-star them in Star Xterminator and when I put them back I pin them down to a size which seems commensurate with the very wide field in the image. They are too small to be blocky! I can also give them a slight blur, and usually do. You can't expect the same resolution of nebular detail as you'd get from a larger optic, just as you can't expect as wide a field of view from a larger optic as you'll get from the Samyang.  If you choose targets which need the wide FOV, and don't invite cropping or pixel peeping, you'll get a super result. My personal favourite from all the images I've ever done was from the Samyang.

Olly

 

[900SL]

11 minutes ago, ollypenrice said:

When I process Samyang images I de-star them in Star Xterminator and when I put them back I pin them down to a size which seems commensurate with the very wide field in the image. They are too small to be blocky! I can also give them a slight blur, and usually do. You can't expect the same resolution of nebular detail as you'd get from a larger optic, just as you can't expect as wide a field of view from a larger optic as you'll get from the Samyang.  If you choose targets which need the wide FOV, and don't invite cropping or pixel peeping, you'll get a super result. My personal favourite from all the images I've ever done was from the Samyang.

Olly

 

Cheers Ollie.. which was?   (Personal favorite?)

[ollypenrice]

1 hour ago, 900SL said:

Cheers Ollie.. which was?   (Personal favorite?)

Orion - with Paul Kummer and Peter Woods. One shot colour, no narrowband filters.

 

Olly

[Anne S]

That's gorgeous.

Anne

[900SL]

11 hours ago, ollypenrice said:

Orion - with Paul Kummer and Peter Woods. One shot colour, no narrowband filters.

 

Olly

Blimey that's something. Took me a moment to find the horseradish and flame. Incredible really what you managed to dig out

[tomato]

Superb. I wonder if most of the whole sky would have this level of nebulosity and 'empty' patches would be few and far between? If you go deep enough is IFN just about everywhere?

[Paul M]

As absorbed as I was in vlaiv's tutorial (as always!), I think Olly's collaborative image just hijacked the thread!

I can't not comment

I've see  similar images but to zoom into that is a sublime experience. To see the relationships between 2 of the most famous DSO' s and whole bunch more is just great!

[ollypenrice]

6 hours ago, tomato said:

Superb. I wonder if most of the whole sky would have this level of nebulosity and 'empty' patches would be few and far between? If you go deep enough is IFN just about everywhere?

The Samyang goes deep very quickly but not all the sky is as packed as Orion. There are still dark regions but, as you say, a very deep 'whole sky' image would be interesting. We have it in narrowband but not in broadband (to my knowledge.)

Olly

[900SL]

On 06/07/2023 at 20:28, ollypenrice said:

The Samyang goes deep very quickly but not all the sky is as packed as Orion. There are still dark regions but, as you say, a very deep 'whole sky' image would be interesting. We have it in narrowband but not in broadband (to my knowledge.)

Olly

So what would you recommend as a good target for Crete in September Ollie? 35 N, bortle 2, 533MC Pro and 135mm f2.8, NBZ Ha/Oiii, around new moon

Should be able to get three hours on target per night, reckon 2 - 3 nights max without the wife throwing something.

I was thinking of the Veil because it fits the 533 frame nicely, and at a good elevation. 3 hrs with the 135 is like 9 hours with my regular scopes at f4.8

Maybe also the Eagle/Omega, these are maximum 40 degrees up and south over the sea

I've never been that keen on the North American nebula, doesn't do much for me

I was thinking of the Bubble and Lobster Nebula, but I think I'll need a lot of hours on this, as the Lobster looks pretty dim?

[ollypenrice]

16 hours ago, 900SL said:

So what would you recommend as a good target for Crete in September Ollie? 35 N, bortle 2, 533MC Pro and 135mm f2.8, NBZ Ha/Oiii, around new moon

Should be able to get three hours on target per night, reckon 2 - 3 nights max without the wife throwing something.

I was thinking of the Veil because it fits the 533 frame nicely, and at a good elevation. 3 hrs with the 135 is like 9 hours with my regular scopes at f4.8

Maybe also the Eagle/Omega, these are maximum 40 degrees up and south over the sea

I've never been that keen on the North American nebula, doesn't do much for me

I was thinking of the Bubble and Lobster Nebula, but I think I'll need a lot of hours on this, as the Lobster looks pretty dim?

I wouldn't recommend the Bubble-Lobster in broadband. The whole region is a red fest with little colour contrast and only comes alive in narrowband. I started it once but got so bored by it that I never finished the thing!

You should get a decent Veil but my own preference would be lower Milky Way, still around in September, I think. We're about to post a Samyang 24 panel galactic centre that would give you an idea of what's in there.

Olly