MTF of a telescope

[andrew s]

9 hours ago, mikeDnight said:

Well, after looking at the four images of Jupiter posted above, I have to say that if my 4" refractor gave an image like the supposed 4" refractor above, I'd give up on planetary observing. The views of Jupiter when its high in the sky are much more like the bottom two images, but with less vivid colours.

Just imagine what you could see with the 12".😏

More seriously,  it highlights the for need objective measurement to compare telescopes. 

Maybe FLO can set up a measurement service with a Shackscope40x40.

Regards Andrew 

[vlaiv]

9 hours ago, mikeDnight said:

Well, after looking at the four images of Jupiter posted above, I have to say that if my 4" refractor gave an image like the supposed 4" refractor above, I'd give up on planetary observing. The views of Jupiter when its high in the sky are much more like the bottom two images, but with less vivid colours.

Let me try again then, maybe this will better approximation to what you see in 4":

What do you think, does it look more like image of Jupiter you see thru 4"?

[vlaiv]

Here is interesting thing

I concluded that @mikeDnight had issues with scale of Jupiter - once image gets too large for a given blur - it does not look as sharp, so I offered the same image only scaled down to be more like the size viewed at the eyepiece.

Then I said - why don't I try to calculate actual Planet size if viewed at "standard" settings behind computer screen. So there are findings that are valid for "standard" settings - if your viewing conditions are different - your results might be different as well.

I calculated that if I used magnification of x176 at the telescope - Jupiter at 47 arc seconds will have apparent diameter of 2.3°. If we are sitting half a meter - 50cm away from computer screen and we have computer screen that has 96dpi resolution, then 2.3° will be ~80px across for planetary disk. So standard settings for viewing following images are 96dpi display and 50cm viewing distance.

Left is 4" Jupiter and right one is 12" Jupiter. Interestingly - not much difference between the two.

It looks like one would need higher magnification in order to see the difference between what aperture can deliver. I can still detect some softness in left image versus right one - meaning that x176 is too much for 4" instrument - you don't need as much magnification to fully resolve what aperture can deliver - that goes to show that x50 per inch is just not needed to fully resolve everything that telescope can deliver.

[vlaiv]

10 hours ago, jetstream said:

The marble and set up. I used a Vixen 2.4mm and the VIP barlow. Not sure if the fs is at the shoulder but if it is the mag is around 393x. Scope is a 630mm 90mm f7, APO triplet that star tests well. Distance is 54 yards on the Zeiss rangefinder. I was shocked at how well this scope did, I could see the 2 separations of colors- yellow and blue very well. The Nagler 3-6 was first tried but it was too easy.

I was hoping for few snaps thru the eyepiece as well

So you spotted two main belts?

[alex_stars]

As my contribution of comparison to reality, I placed a "edge target" (just a piece of paper half white, half black) at my 170 m distance test site and took some photos of that through my Skymax 180.

If you look at my previous post on the "edge spread function", well one can use that method to estimate the actual MTF of a scope. Here is my result:

From top left to bottom right: Actual BW image of the edge target (the rain softened the paper, but that is not problem). Data measured with my CCD of the Edge Spread (blue dots) and a data fit (red line). The reconstructed Line Spread Function. The theoretical Point Spread Function of the Skymax 180. And finally the MTF comparison.

Now what do you think? Opinions welcome. Two ideas to take along:

• I consider the edge well focused on the "Image" subplot. At least that is as good as I can manage with the mirror focus of the Skymax
• The small overshoot of the red curve above the green curve in the MTF plot is due to the data fitting in the Edge Spread Function. So consider the two curves (green and red) to be equal until the red one falls below the green one.

 

[andrew s]

1 hour ago, vlaiv said:

 

Left is 4" Jupiter and right one is 12" Jupiter. Interestingly - not much difference between the two.

 

To me, on my tablet, the right image has much higher contrast than the left and slightly more detail. A subjective opinion. 

Regards Andrew 

Edited February 4, 2021 by andrew s

[vlaiv]

25 minutes ago, alex_stars said:

Now what do you think? Opinions welcome.

You did fitting of the curve to the data. What model did you use and why? MTF looks like gaussian, so my guess is that exponential of sorts was used for fitting.

Did you use single wavelength of light? What you have here is averaged MTF for different wavelengths weighted according to QE response of your system.

Remember - 400nm has almost twice the cut of frequency as 700nm (x1.75 precisely). MTF will look much more "hanging" if you combine different wavelengths.

Try using raw data and doing row to row difference

V(x,y) = V(x-1,y) - V(x,y)

That is similar to differential filter (it will give you slope at constant intervals) - don't do any fitting but that should look roughly like profile of Airy disk.

Stacking is good option to reduce noise.

[vlaiv]

26 minutes ago, andrew s said:

To me, on my tablet, the right image has much higher contrast than the left and slightly more detail. A subjective opinion. 

Regards Andrew 

That higher contrast might be due to resampling used or it might be genuine feature.

There is another way to conduct experiment - we should just look at original images from greater distance. In my case - 1.5m produces similar effect as reduced images (since smaller images are 34% of original size). At over 2m away from screen - I can virtually detect no difference between original 4 images.

[alex_stars]

8 minutes ago, vlaiv said:

You did fitting of the curve to the data. What model did you use and why? MTF looks like gaussian, so my guess is that exponential of sorts was used for fitting.

Yes, as my sampling is not that great, due to the CCD I have. So I fitted a sigmoid function into the data. The MFT is a Gauss curve (half of it) as the LSF is a Gauss, which is the differential of the sigmoid.

10 minutes ago, vlaiv said:

Did you use single wavelength of light? What you have here is averaged MTF for different wavelengths weighted according to QE response of your system.

I wish, nope, just pure sun-light 😉. Exactly my point. because this is what happens when we image. So the theoretical MTFs (mine for 530 nm) do not compare to our images. Also we do not observe with our eyes on a single wavelength, so we always get a "average" MTF to see.

20 minutes ago, vlaiv said:

Try using raw data and doing row to row difference

V(x,y) = V(x-1,y) - V(x,y)

That is similar to differential filter (it will give you slope at constant intervals) - don't do any fitting but that should look roughly like profile of Airy disk.

Thanks, for the reminder, I do know finite difference methods 😉. However the data does not not justify a piecewise differentiation. Give me a super high-resolution CCD and we get going.

21 minutes ago, vlaiv said:

Stacking is good option to reduce noise.

That is true, but the issue here is not the noise of the camera in the white area or in the black area (look at the second subplot, first row). The issue is that my CCD does not have small enough pixels to resolve the ESF better.

However, as I image with that CCD, that is all I get and it nicely demonstrates the capability of my system to actually resolve the theoretical MTF of my scope.

[vlaiv]

7 minutes ago, alex_stars said:

I wish, nope, just pure sun-light 😉. Exactly my point. because this is what happens when we image. So the theoretical MTFs (mine for 530 nm) do not compare to our images. Also we do not observe with our eyes on a single wavelength, so we always get a "average" MTF to see.

Indeed we do - but our vision is such that majority of brightness that we perceive as sharpness comes from green part of spectrum - specifically around 500nm.

For that reason - ~500nm is often used as baseline for telescope performance assessment.

or rather 500nm for night time use and 550nm for daytime use. I think that Baader Solar Continuum filter is very good filter for assessing visual telescope performance by means of camera analysis.

13 minutes ago, alex_stars said:

That is true, but the issue here is not the noise of the camera in the white area or in the black area (look at the second subplot, first row). The issue is that my CCD does not have small enough pixels to resolve the ESF better.

However, as I image with that CCD, that is all I get and it nicely demonstrates the capability of my system to actually resolve the theoretical MTF of my scope.

Actually I think that your CCD is just sufficient to completely resolve image from Mak.

If I'm not mistaken, at around 500nm - critical sampling rate for 3.75µm pixel size is F/15. Mak is F/15 or slower (due to slightly lower aperture than 180mm), so you should be able to sample image properly. There will be only slight pixel blur, but I think it is rather small in this context.

However, I think there is difference between PSF obtained as differential from this edge softening and actual PSF.  Edge gradient is result of 2D convolution yet we here treat it as 1D case.

Let me show you what I mean. I generated airy pattern and made perfect edge in the image (just rectangle with value 1). Then I convolved image with Airy pattern and did following kernel on it:

0 0 0

1 -1 0

0 0 0

to extract the edge. Then I measured the edge (plot profile), and this is what I got:

Notice that ripples are not at zero like linear cross section of Airy pattern is:

This is because of 2d convolution so our values are not affected by values only to the left and right - but also those in "upper rows" and "lower rows" - which add to average and somewhat "smooth things out".

Doing simple edge contrast won't produce proper PSF

[mikeDnight]

2 hours ago, andrew s said:

Just imagine what you could see with the 12".😏

More seriously,  it highlights the for need objective measurement to compare telescopes. 

Maybe FLO can set up a measurement service with a Shackscope40x40.

Regards Andrew 

Objective measurement is really the way to go!  Of course I'm not really trying to say a 4" apo is a better telescope than a 12", but its images can be so pleasingly sharp, it can give seriously impressive views to anyone who carefully studies the detail on show. Hence the the often stated claim "packs a punch greater than its aperture class"!  But then there are a number of other variables - the studious nature, patience and visual acuity of the observer. 

[CraigT82]

I wonder if it's worth throwing in some real world images showing differences in resolving power, to perhaps complement and maybe compare to the simulations?

This set of images by Parisian imager Jean Luc Dauvergne (screenshot from Twitter) taken with the Takahashi Mewlon family: from left 180mm, 210mm 250mm and 300mm apertures. These are probably the closest to perfect optics you can get as an amateur, and with the atmospheric effects minimised by lucky imaging I think it would be interesting to compare to the simulated Jupiters for the same scopes?

Image credit: Jean Luc Dauvergne (@JLucDauvrergne)

[andrew s]

11 minutes ago, mikeDnight said:

Objective measurement is really the way to go!  Of course I'm not really trying to say a 4" apo is a better telescope than a 12", but its images can be so pleasingly sharp, it can give seriously impressive views to anyone who carefully studies the detail on show. Hence the the often stated claim "packs a punch greater than its aperture class"!  But then there are a number of other variables - the studious nature, patience and visual acuity of the observer. 

I am all for enjoying our hobby anyway you want to. We are all different and get our kicks in different ways.

I would love to have a Shackscope40x40 to  monitor seeing and test telescopes rather than a high end Takahashi refractor for visual but that's just me.

You I suspect would might differ on that.😉

Regards Andrew 

[alex_stars]

29 minutes ago, vlaiv said:

Indeed we do - but our vision is such that majority of brightness that we perceive as sharpness comes from green part of spectrum - specifically around 500nm.

Hence my choice of the 530 nm for the theoretical MTF.

30 minutes ago, vlaiv said:

Actually I think that your CCD is just sufficient to completely resolve image from Mak.

If I'm not mistaken, at around 500nm - critical sampling rate for 3.75µm pixel size is F/15. Mak is F/15 or slower (due to slightly lower aperture than 180mm), so you should be able to sample image properly. There will be only slight pixel blur, but I think it is rather small in this context.

You seem to misunderstand, I never said that my CCD is not capable of resolving images taken through the Mak. What I said is that it has not fine enough resolution to resolve the ESF for piece-wise finite differencing, which is a whole different thing altogether.

And yes I am able to sample images properly.

33 minutes ago, vlaiv said:

Doing simple edge contrast won't produce proper PSF

I wonder on what knowledge you base such a strong statement. Hmm. First off, I do not attempt to recover the PSF (point spread function), but the LSF (line spread function), which is the correct 1D equivalent of the 2D PSF. I never claimed to reconstruct a 2D PSF. And secondly, I urge you to read up on some scientific literature before you claim that a certain method is not valid, based on your quick assessment. For example pick any of:

• https://www.spiedigitallibrary.org/conference-proceedings-of-spie/5251/0000/Fast-MTF-measurement-of-CMOS-imagers-using-ISO-12333-slanted/10.1117/12.513320.short (so you see it even has an ISO standard to it)
• https://aapm.onlinelibrary.wiley.com/doi/abs/10.1118/1.1288682
• https://aapm.onlinelibrary.wiley.com/doi/abs/10.1118/1.597264
• https://www.spiedigitallibrary.org/conference-proceedings-of-spie/7109/710905/MTF-assessment-of-high-resolution-satellite-images-using-ISO-12233/10.1117/12.800055.short
• https://www.osapublishing.org/abstract.cfm?uri=oe-18-4-3531
• https://www.osapublishing.org/abstract.cfm?uri=oe-26-26-33625
• and soooo many more.

 

[vlaiv]

9 minutes ago, CraigT82 said:

I wonder if it's worth throwing in some real world images showing differences in resolving power, to perhaps complement and maybe compare to the simulations?

Actually - these images will always win over simulation as simulation tries to show what telescope can deliver visually - without processing and these images are processed.

In particular, there is part of processing that sort of negates MTF - that is sharpening. Point of sharpening is to do this:

Well, that second MTF line did not come out right - but you get the point - sharpening corrects the "sag" of MTF - all the way up to cut off point.

For that reason - image will look sharper after processing. In fact - look at this:

left is 5" apo image at eyepiece - and right is the same image with wavelet adjustments in registax. In reality 5in scope will not be able to deliver right image at eyepiece - but maybe it could deliver it when imaging.

[mikeDnight]

3 hours ago, vlaiv said:

Here is interesting thing

I concluded that @mikeDnight had issues with scale of Jupiter - once image gets too large for a given blur - it does not look as sharp, so I offered the same image only scaled down to be more like the size viewed at the eyepiece.

Then I said - why don't I try to calculate actual Planet size if viewed at "standard" settings behind computer screen. So there are findings that are valid for "standard" settings - if your viewing conditions are different - your results might be different as well.

I calculated that if I used magnification of x176 at the telescope - Jupiter at 47 arc seconds will have apparent diameter of 2.3°. If we are sitting half a meter - 50cm away from computer screen and we have computer screen that has 96dpi resolution, then 2.3° will be ~80px across for planetary disk. So standard settings for viewing following images are 96dpi display and 50cm viewing distance.

Left is 4" Jupiter and right one is 12" Jupiter. Interestingly - not much difference between the two.

It looks like one would need higher magnification in order to see the difference between what aperture can deliver. I can still detect some softness in left image versus right one - meaning that x176 is too much for 4" instrument - you don't need as much magnification to fully resolve what aperture can deliver - that goes to show that x50 per inch is just not needed to fully resolve everything that telescope can deliver.

I appreciate what youre saying, and that 176X may technically be higher than is needed, but I've always found around 180X works well on planets irrespective of aperture. I feel the left hand image to be a view on a mediocre night, while the sharper right hand image is much more in line with the 4" Apo view. Having said that, visually through the 4" the edges of the belts differ from those in an image, and are less colourful.

[vlaiv]

18 minutes ago, alex_stars said:

And secondly, I urge you to read up on some scientific literature before you claim that a certain method is not valid, based on your quick assessment.

My claim wan not that certain scientific method was not valid - my claim was that your method was not valid.

Did you use proposed algorithm for deriving of MTF from edge profile or did you just use 1D Fourier transform of 1D PSF profile that you got as differential of 1D profile of the edge?

If you did later - look at two PSF profiles that I got above - they are different and one is real PSF profile used in calculation and other is PSF derived using your method.

[Carl Au]

I am definitely confused now, but it is starting feel like the point of this thread is telescope bashing (refractors in particular) and rubbishing people's gear. 

I stopped using cloudy nights because of this type of carry on...pity.

Edited February 4, 2021 by Carl Au
Spelling mistake

[CraigT82]

I'm finding the thread informative and interesting.... granted there is one or two getting the hump but thats the internet for you. You can use the unfollow function at the bottom of the page (mobile version- top of page on desktop) if you need to? 

[alex_stars]

6 minutes ago, Carl Au said:

I am definitely confused now, but it is starting feel like the point of this thread is telescope bashing (refractors in particular) and rubbishing people's gear.

Just for full disclosure, I do not attempt to bash any type of scope. On the contrary I am in the position where I have decided to swap my 180 mm Maksutov to a 125 mm refractor, for may reasons, but mainly for the simple fact that the few exceptional nights where I can harness the power of my Mak are just too few and that a 125 mm refractor is a way better choice for me, especially considering the whole MTF discussion above.

9 minutes ago, Carl Au said:

I stopped using cloudy nights because of this type of carry on...pity.

same here. but it seems one has to be careful on any forum.

Clear Skies

[alex_stars]

24 minutes ago, vlaiv said:

My claim wan not that certain scientific method was not valid - my claim was that your method was not valid.

is that so....

24 minutes ago, vlaiv said:

Did you use proposed algorithm for deriving of MTF from edge profile or did you just use 1D Fourier transform of 1D PSF profile that you got as differential of 1D profile of the edge?

Obviously I did use a proper algorithm, published in a paper on the subject.... but you seem to take issue with what I have presented.

For me this discussion has gone too far, you seem to have the urge to prove yourself of being the "wise guy" on the forum and seem to take pleasure in criticizing valid contributions of other... which is not very scientific....

Some points of yours were valid, others not. I was hoping for a discussion among peers, but that does not seem to be possible.

Good day to you sir.

[vlaiv]

17 minutes ago, Carl Au said:

I am definitely confused now, but it is starting feel like the point of this thread is telescope bashing (refractors in particular) and rubbishing people's gear. 

I stopped using cloudy nights because of this type of carry on...pity.

Not sure if my posts are perceived like that - but my personal goal is finding the truth.

At this point - my major interest is discrepancy between what science and my personal (mostly related to imaging) - but limited experience suggest, and what other people report.

I would really like to know where I'm wrong with this if 4" APO can perform on par with 12" - or maybe it can't maybe my simulations are just off and everything is "shifted" (my 4" sim actually describes 2" telescope and 12" sim - 4" telescope because I got some quantity wrong or something). Maybe there is some observational bias at play here.

Also note - we are not talking of the performance of telescope under stars here - we are talking about best possible scenario - no atmospheric influence. With seeing in equation - I'm sure there are conditions when 4" APO performs on par with 12" Newtonian - however, I would expect the image in that case to be worse than both simulated best case images - again discrepancy.

 

[vlaiv]

7 minutes ago, alex_stars said:

Obviously I did use a proper algorithm, published in a paper on the subject.... but you seem to take issue with what I have presented.

For me this discussion has gone too far, you seem to have the urge to prove yourself of being the "wise guy" on the forum and seem to take pleasure in criticizing valid contributions of other... which is not very scientific....

Some points of yours were valid, others not. I was hoping for a discussion among peers, but that does not seem to be possible.

Good day to you sir.

I'm terribly sorry if I've offended you and I assure you that I don't have urge to prove myself to be a "wise guy". On plenty occasions I've been proven wrong and I have no problem what so ever accepting my mistakes and learning from them. If I strongly support a stand - it is not because it is my stand, but because stand is strongly supported by the fact that I have and logic. Either might be wrong - I might be operating under wrong assumptions or there might be a flaw in my logic. Again - happy to be corrected on either or both accounts.

In fact, I would be more than happy to explain exact steps in getting two graphs above so we can together explore them and try to find mistake in my method that lead to graphs being different if expected outcome is the same - PSF (as I would expect from your explanation).

In any case, please accept my apology if I've offended you in any way.

 

[jetstream]

4 hours ago, vlaiv said:

So you spotted two main belts?

If you mean those 2 belts in the marble then yes. The blue was easier than the yellow and I could see a very sharp top, bottom and middle contrast difference line. The yellow at 393x started going away but the blue could take more. I can see from this the colour makes a difference, at least to my eyes.

I rushed to set up and observe this as today its snowing and -35c is right after this for a week, so no more marble tests for a bit. Anyway this is why I didn't try to get an image through the eyepiece. I do have a TV Fonemate coming, kind of. We'll see if it shows up.

I'm going to set the 15" dob with the marble on the ice once I figure out the distance.

[jetstream]

1 hour ago, Carl Au said:

but it is starting feel like the point of this thread is telescope bashing (refractors in particular) and rubbishing people's gear

I'm confused why you think this-optical characteristics and their limitations are a fact of life. Before assuming that refractors are being bashed because of limitations in aperture, look through a good, largish-doesn't have to be expensive- reflector.

I'm learning a lot from this thread, in many ways and hope it continues on.