MTF of a telescope

[Captain Scarlet]

OK raw image attached here...

 

_S7A5565.CR2

[vlaiv]

2 hours ago, Captain Magenta said:

It's about 125m away, taken at prime focus with my Skymax180 at 2799mm FL and EOS 7Dmk2

Forgot to say. If you can also include following - that would be really helpful:

1. something of known size at the same distance - so we can determine pixel scale used - maybe ruler of sorts or just place something that you measured and know dimensions of at the foot of that pole.

2. precise measurement of distance to target (about 125m is "okayish" but 127.63m is better )

Problem is with Maksutov telescopes - they focus by shifting primary mirror - and their focal length depends on primary to secondary distance. There can be significant difference in declared and actual focal length - depending on focus distance and where in optical train camera sits (extension tubes and where is optimal focus position with respect to that).

Without above - we can produce "relative" curve, but with above info - we can produce absolute curve and compare it to theoretical curve for that scope.

[vlaiv]

18 minutes ago, Captain Magenta said:

OK raw image attached here...

Image is teensy-weensy noisy but here is the result:

this was green channel extracted (FitsWork and VNG debayering) and sample of 192x512 pixels taken with the edge. I did differential of edge and got this:

LSF is there but so is quite a bit of noise.

[Captain Scarlet]

I'll try to address the precise distance-to-pole thing tomorrow.

The FL of the scope as used for the photo, at 2799mm  +/- 17mm, with this combination of camera and adapters, I am highly confident of. In fact I've done quite a bit of measurement of this, ending up with the following chart. I had an extra approx 40mm of back-focus from the "B" point in the chart below. A longish thread (Im the OP on it )about this can be seen here.

 

 

[Captain Scarlet]

... just had brainwave (highly abberrated wavefront though) ... it's a parallel-sided pole, I'll measure its circumference tomorrow...

[alex_stars]

Hi all again,

I think I have found a very usable software that we all can use to play around with MTF estimations on our own. It is far more advanced than the scripts I wrote and works really well. I encourage everybody who is interested to try it out, here the link:

MTFmapper

It has a nice GUI and is well written. Was no problem to install on my Linux and does have a windows version, so we should be covered.

One can print out test sheets, also generate them and then place them at an adequate distance to cover the whole CMOS chip. The rest the software does for you.

Here are some examples from the homepage of MTFmapper:

Automatic edge detection and calculation of SFR (Spatial Frequency Response) / MTF charts for each edge:

Spatially resolved lens grid, as it resolves many edges over the sensor:

and lens profile as one is used from lens tests (this is contrast across the lens as a Meridional / Sagittal MTF curve

And it can do a lot more....

So I think we have our tool to use and compare

[alex_stars]

A yes, regarding the target chart distance for MTFmapper

One should place it at a distance so that it covers the whole sensor. As an example for my scope and setup (just if people are uncertain about the math):

• I soon gonna have a 975 mm FL scope
• I have a ASI 224 MC
• This results in a FOV for my sensor of 0.29° x 0.21° (just use astronomy tools in imaging mode 😀)
• Now if I print my test sheet on an A4 paper (297x210 mm) I can work out the distance
• Lets focus on the long edge of the sensor and the A4 sheet (so we have 0.29° for 0.297 m)
• The distance to target is distance = 0.297/tan(0.29°) which is 58.67 m (just make sure you either convert to radians or use a tan function suitable for degrees) I just enjoyed my old pocket calculator from high school
• This is easily do-able

The rest MFTmapper does for you.

As soon as I have my new scope in hands and some time I will try out MTFmapper and report back.

[alex_stars]

A yes, it gets better. The newest version of MTFmapper can do:

Lateral chromatic aberration

how nice is that for us refractor people 🥳

I can't wait to see all the comparison discussions on the forum

[JeremyS]

4 minutes ago, alex_stars said:

A yes, it gets better. The newest version of MTFmapper can do:

Lateral chromatic aberration

how nice is that for us refractor people 🥳

Did somebody call? 
🤣

[alex_stars]

Kind of don't want to let this thread die, so here is the long ago mentioned comparison between a C14 (planetary imaging favourite) and that 7 inch Mak.

Couple of things to take away from that graph. And I am sure if people have been reading until here, they can interpret the MTF graph as well. However one measure we did not mention in the discussion yet is the

MTF50 value: which is the spatial frequency at which the MTF of a given system decreased from 1 to 0.5 (horizontal dotted line) Below the MTF50 we definitely perceive a quality loss and this measure is often used for lenses when comparing.

• The advertised 7 inch Mak (actually 6.77 inch) is "equal" in transferring "contrast" with a 5 inch (actually 4.53 inch) unobstructed refractor. That's what we often read on the forum. You see it in the graph where the green and red line overlap perfectly. Note that the MTF50 value for both scopes is the same a 0.42/arcsec or at 2.38 arcsecs.
• A C14 is equal to a 248 mm (9.76 inch) refractor and the MTF50 value for both scopes is at 0.91/arcsec or at 1.09 arcsec.
• Hence we can argue that a C14 is really good until the average seeing limit (say at 1 arcsec) whereas the 7 inch Mak only is as good until say 2.4 arcsec. This is in line with the experience people have with such scopes.
• For us astronomers the question arises, how far to the right of the graph can we go, because that is the question on what scope to get.
• To spark a discussion, I plotted the 0.4 arcsec mark, which is the best seeing on the planet and I wonder if the C14 would be good enough to work to the full potential of such places.
• We could also turn the question and ask at which MTF value do we need to draw our line?

Looking forward to the next round of comments

[alex_stars]

I have added another two measures to the graph, the MTF20 and MTF10:

• At MTF10: the human eye can not resolve contrast anymore and camera noise even of good cameras is a serious issue. So we can use MTF10 as the real limit of what we can observe, both visually or with imaging. Note that the Dawes' limit is at higher frequencies but that is very theoretical. MTF10 is close to the Rayleigh Limit, which we should use to stay realistic.
• At MTF20: most human eyes can't perceive higher resolutions. If you think you can, well, go check if you can see the resolution limit of the screen on which you read those words. Most digital screens are designed to the MTF20 limit of the average human eye. I know I can't observe below MTF20. We could use MTF20 as a practical observation limit when working visually.

For me the MTF20 limit puts scopes into a nice perspective.

Here also some further reading links for the interested:

• https://www.imatest.com/docs/sharpness/

• How to read MTF curves (recommended!)

[Captain Scarlet]

On 21/02/2021 at 17:18, alex_stars said:

Kind of don't want to let this thread die ...

Have you seen Pulp Fiction? ... "Thread's dead, baby, Thread's dead..."

 

 

 

... actually i hope it isn't. I've been playing with MTFmapper and it is very good. I have a printout of one of the A3 test-chart versions and I'm awaiting the 80kph winds to die down before I can find a sufficiently convenient 50-odd metres target distance for my Skymax 180.

Edited February 23, 2021 by Captain Magenta

[vlaiv]

I would not call it quite dead yet. Maybe a bit stalled at the moment.

Got myself a phone adapter so I can do afocal imaging with phone. Guess that is probably simplest way for people to conduct any sort of test. It's not going to be very accurate test since we don't know:

1. eyepiece MTF

2. phone camera / camera lens MTF

At this point I'm not sure if I properly understand what sort of influence it might have on result. I also want to get c-mount lens to attach on my astro cameras so I can use that instead. Just a few moments ago, I realized there is this great thing called - step up / step down rings and there is also M52/T2 adapter to be purchased - which means I can connect proper lens to scope.

I'll need good laser range finder so I can accurately measure angular sizes. All of these don't cost much, but it does add up and I'm in situation where I have to really save up (my new house is going to be finished in few months and I'll need to move and purchase all the new stuff I won't be moving), so I'm not really sure when I'll get all the bits that I need for this.

To be honest, I'm sort of depressed by the fact that WinRoddier test using my synthetic data is not working properly - and I can't figure out why - is it my data or software itself. At the moment - there is no real benchmark that we can use to compare results.

I could write software that does similar analysis as WinRoddier - probably using different approach, since the way I see it - you don't have to have exact defocus for in/out images using my approach and you don't have to limit data set to just one set of in/out images. Not sure how fast it would work since it would do brute force search across Zernike polynomial space (well - not brute force - more like gradient descent along each coordinate since Zernike polynomials form orthonormal set). This means possibly thousands of FFTs per fitting. That can't be very fast .

But again - I don't have much time at the moment to be able to do that, and even if I did - how would we know if it provides good results? What would be our benchmark to test against?

 

[jetstream]

5 minutes ago, vlaiv said:

I don't have much time at the moment to be able to do that

Lots of time in the future Vlaiv, no rush.

 

6 minutes ago, vlaiv said:

What would be our benchmark to test against?

Shack Hartman or Zygo?

and... do we really want to know how bad or good are scope is anyway? one thing for sure- if the test results on a scope aren't that good- look out! the test or tester will take the blame...

Actually I eagerly wait for you to figure this out so I can try it.

[vlaiv]

5 minutes ago, jetstream said:

Shack Hartman or Zygo?

and... do we really want to know how bad or good are scope is anyway? one thing for sure- if the test results on a scope aren't that good- look out! the test or tester will take the blame...

Actually I eagerly wait for you to figure this out so I can try it.

By the way - I checked that link @andrew s provided and it is quite sensible option.

In fact - if we can get many tiny lenses - I think we would be able to build one . Not sure what the smallest available lenses are but if we could fit say about 100 over the span of 2" filter - well, one would need just a barlow and regular camera to their own test. Granted - profile of the barlow must be known for this to work

My greatest concern is collimation lens. Wavefront needs to be collimated properly and in configuration they are presenting on their website - where wavefront sensing is done close to focal plane - you have to match F/ratio of the scope used. How much aberration will collimation lens introduce?

 

 

[Captain Scarlet]

41 minutes ago, vlaiv said:

...

I'll need good laser range finder so I can accurately measure angular sizes. 

I take it you don’t have a mount with encoders? If so you could use the encoder output as a theodolite to measure small arcs? I do that with my ayo2 and dsc.

Or perhaps plate-solve the exact FL of one of your scopes and subsequently use the distance across the sensor between a pair of terrestrial features to derive angle from that?

Edited February 23, 2021 by Captain Magenta

[vlaiv]

1 minute ago, Captain Magenta said:

I take it you don’t have a mount with encoders? If so you could use the encoder output as a theodolite to measure small arcs? I do that with my ayo2 and dsc.

I have Heq5 and AzGti.

Both could in principle be used to slew mount just a bit and take two successive images. Then I would calculate pixel shift and divide that by angle that mounts tell me.

I wonder how accurate that would be given backlash and stepper resolution?

This is not a problem for prime focus setup - pixel scale is not that large and I can slew for example - 10-20 arc minutes and still have reference in the frame. I wonder if I use "very high magnification" - or rather high pixel scale needed to capture MTF?

Say I use 5.5mm eyepiece with my mobile phone. My phone is Xiaomi Mi A1 and it has 26mm (equivalent?) lens and 1.25µm pixel size. I say that it has equivalent lens since no way they packed 26mm FL lens in there. Sensor has size of ~1/2.9" so actual focal length of lens would be 3.8mm?

In any case, that would mean that 1300mm x 5.5 / 3.8 = ~1900mm at 1.25µm pixel size. That is 0.14"/px and for 4000px in width of sensor that is 560" or 9.333' - well, not bad.

If my calculation is ok - then yep, I could use mount to measure pixel scale.

Good point - I'll have to try that. I just need to tell mount to slew 5 arc minutes and see how much image is shifted in pixels.

 

[vlaiv]

I just realized I made mistake in above calculation.

5.5mm EP and 3.8mm camera lens will give smaller image by factor of ~1.44737 and not larger!

Eyepiece and camera lens here form relay lens pair. If both are equal focal length - they produce no magnification. If first is shorter FL than the second - then they produce magnification.

Unfortunately this means that phone camera is bad choice for testing telescopes as phone cameras have very short focal length lens (on the other hand - it is a good choice for EEVA as such combination will act as powerful focal reducer).

Above is still doable - but I need regular lens of about 50 or so mm and different attachment system. Back to the drawing board.

[festoon]

This may not sit properly in this thread...apologies if not I'm happy to move it.

I've been thinking about the MTF of optical systems and how to model the MTF of a lens from basic principles.

The minimum spatial resolution in lpmm for ideal lens with diffraction for 550nm light can be modelled as 1812.2/F where F is the focal ratio. From this we can calculate the Rayleigh Criteria as this number divided by 1.22 and MTF50 as half of the Rayleigh Criteria.

So fo an ideal lens the MTF50 at 550nm for a varying F number would be as below

Now in reality this is never the case. The actual MTF50 never reaches the theoretical due to the resolution of the sensor and optical abberations.

I've found this reference below showing the effect of the sensor pixel size on the MTF50 value (https://www.dpreview.com/forums/post/63709384)

Is anyone on here able to please help me to understand how to simulate this? i.e. how to go from the diffraction only to diffraction + pixel

[vlaiv]

3 hours ago, festoon said:

s anyone on here able to please help me to understand how to simulate this? i.e. how to go from the diffraction only to diffraction + pixel

Do pay attention that most lenses are far away from diffraction limited optics and while we can talk about theoretical MTF - most lenses will have much lower values.

As far as pixel size is concerned - in post on previous page of discussion I produced MTF of pixel. Pixel convolves point sampling - so FT of pixel will multiply FT of original signal. We just need to produce FT of pixel pattern and multiply with MTF of ideal aperture.

This is MTF of pixel where pixel is matched in resolution to lens (optimum sampling rate - or pixel size twice smaller than max wavelength produced by the lens).

In order to see complete MTF of pixel, here is case when pixel is much larger than optimum sampling rate:

First thing to note is that pixel FT does not have rotational symmetry - and it will impact MTF50 depending on direction.

This is profile of FT in X direction.

We can see from this graph - that MTF50 is equal to 0.8 of frequency related to the pixel size (frequency equal to wavelength of single pixel). Problem is however that such frequency is not recorded in our image as sampling records only those frequencies that have wavelength twice that of pixel size as pixel size is related to our sampling frequency.

We can talk of hypothetical MTF50 at focal plane dictated by pixel size - but never on the image recorded with those pixels.

Record image with large pixels and there will be some frequency that such pixels will reduce to 50% of their original value - but those frequencies will not be in recorded image as those frequencies are much larger than sampling rate will record.

Increase pixel size for any lens and you'll only get sharper image - never "softer due to large pixel size".

 

 

 

[Captain Scarlet]

My optical test bench in action ... Intes M603 shown here, having previously done the Skymsx 180 (with the target a good bit further up the lane!)

 

Edited March 7, 2021 by Captain Magenta
Pic loading problem

[alex_stars]

@Captain Magenta, did you run your photos through the MTFmapper? Just curious. I still did not manage to test my new APO, however I will post as soon as I get a chance to take some test images.

[Captain Scarlet]

Oops I just remembered I never answered this.

I did run them through MTFmapper, and the results demonstrate that both my skymax 180 and my M603 seem to be truly terrible. I was left scratching my head, as they certainly aren't terrible when I look through them either at night-sky objects or terrestrially. As you can see I photographed my A3 test target at about 35 metres, essentially the minimum focus distance for either scope. Various things could have conspired to render the sharp edges on the test sheet to become soft: missed focus, either because the cardboard target-back might have shifted in the wind; poorly printed; camera-shutter-shake. Or, the scopes ARE terrible, but I don't think so.

I'll have to do it again with better-controlled conditions.

Cheers, Magnus

[Deadlake]

34 minutes ago, Captain Magenta said:

Oops I just remembered I never answered this.

I did run them through MTFmapper, and the results demonstrate that both my skymax 180 and my M603 seem to be truly terrible. I was left scratching my head, as they certainly aren't terrible when I look through them either at night-sky objects or terrestrially. As you can see I photographed my A3 test target at about 35 metres, essentially the minimum focus distance for either scope. Various things could have conspired to render the sharp edges on the test sheet to become soft: missed focus, either because the cardboard target-back might have shifted in the wind; poorly printed; camera-shutter-shake. Or, the scopes ARE terrible, but I don't think so.

I'll have to do it again with better-controlled conditions.

Cheers, Magnus

What you want is this:

https://www.cloudynights.com/topic/757247-ai-based-wave-front-sensing-and-collimation/

If I follow correctly you want the wavefront..

 

[Captain Scarlet]

Thanks for the reminder… 😊