Mitutoyo M Plan APO HR 5X 0.21 Test Results

[rjlittlefield]

Usual math. Rayleygh criterion, sampling on the sensor, MTF, minimum sampling of the sensor at 2.5-3 pixels per LP ....

Ploum, maybe it will help if you try thinking of me as an extremely dense student who needs to have everything spelled out.

In your spreadsheet, I see lots of numbers. But I do not see any formulas, so I am having to guess where the numbers came from. You say "Usual math" etc. But I cannot match your numbers using any math that I know. So I need some help.

Let me give one example of my confusion. In your spreadsheet, line 210 is labeled "LP/mm". I assume this means "line pairs per mm". The numbers range from 752 to 520, so I assume this is line pairs per mm on the subject. I see that https://www.microscopyu.com/techniques/ ... microscopy says Rayleigh resolution is 0.61*lambda/NA, and I see your spreadsheet says "Ouverture numérique 0,21". But with lambda = 0.00055mm, 0.61*lambda/0.21 gives 0.001598 mm, corresponding to 626 line pairs per mm. No match with 752. Using lambda = 0.0005 gives 688 line pairs/mm, better but still no match. To match 752 line pairs/mm would need lambda = 0.000458. That's a very unusual value, so I am not confident (to say the least). Where does the LP/mm 752 come from?

Another example... Let me assume that for smallest FOV, you have somehow chosen LP/mm to match the Rayleigh criterion. Then the MTF of the objective for that LP/mm will be only about 9% = 0.09, as it always is for Rayleigh. From that, I conclude that the overall system MTF = objective MTF * sensor MTF, can be no larger than 0.09. But Line 212 of your table, "MTF globale", shows a value of 0.42 . Now I am totally confused. How is the MTF globale 0.42 computed?

These are only examples. In general, I cannot figure out how most of the numbers in your spreadsheet might be computed.

I understand there are some language difficulties.

If you would like to post your spreadsheet, or send it to me some other way, then I will try to make sense of the formulas that I find in it.

But right now, all I see are some numbers that I cannot make sense of. (Anybody who can, please jump in and explain them to me.)

--Rik

[Macrero]

Another comparison for those who (like me) trust their eyes more than graphs and charts.

Mitty 10 vs Mitty 7.5 @5X on sensor. 100% crops.

M7.5:



M10:



Again, the higher NA objective shows clear advantage.

I doubt the 5X HR would do much better than the 7.5X @5X, but it would be nice to see a comparison.

- Macrero

[Scarodactyl]

When my renovations are done I can take a better head to head with my 10x vs 10x hr. I am not sure where I stuck the source files and the edited one I did dig up is kind of ugly.

[ChrisR]

Interesting discussion.
I'm not a theory guy, but in my experience my Mitty 7.5/0.21 and Nikon S Fluor 4/0.20 outresolves/outperforms any 0.13, 0.14, 0.15 objective at any magnification on any camera I used them with: MFT (20MP), APS (26MP), FF (47MP), among others.
I don't currently have a Mitty 5X nor FF camera, but for what it's worth, I just did a quick comparison: S Fluor 4/0.20 vs CF N 4/0.13 @4X. Shoot with Pana G9 (20MP MFT sensor, 3.3 µm pixel pitch) in HR to make difference clearer.
Both images have been processed with the exactly same settings and have the same amount of sharpening applied.
300% crops:
CF N 4:
http://images2.imgbox.com/c8/07/aYD2ivg9_o.jpg

S Fluor 4:
http://images2.imgbox.com/ff/99/JFMc5neK_o.jpg
Those scales aren't easy to resolve at 4X, even the 0.20 is struggling, but the 0.13 barely resolves any detail.
So at least in this scenario the difference is clear.
Would be interesting to see real world samples/comparisons with other cameras/objectives.

- Macrero

I can't put my finger on the images, but this reminds me of my comparisons between Nikon 4x/ 0.13 and 4x /0.2 when I started, using a Nikon 700D. That's a full frame 12MP sensor, the image circle was nowhere near covering the whole sensor. I was covering maybe only 5MP of it. Outresolving the sensor perhaps, but the IQ was dramatically better with the NA 0.2 objective.

[mjkzz]

wow, this thread has grown a lot, learning more here . . .

the title of this thread is about "test results", like Scarodactyl has pointed out, what is "better"? Reading through the thread quickly, I see many things that are particularly important to their posters. To me, seeing more details is more important than anything else (just me) because if FOV is too small, I can always find a way to overcome that (stitching), if contrast is low, it might be OK, at least beats not seeing details, of course, to a point that contrast level can not be too low to distinguish features. [edit] image circle is important for a GIVEN camera, but this can be rectified by using a different camera, etc, etc.

Seeing details, in theory, depends on the resolving power of BOTH optics and sensor. So far the theories are always good to me, have not seen anything conflicting, if conflicting, it is always me not using equipment correctly. For optics, resolving power of a specific lens is something you can not improve, and that is theory, too. The resolving power of a specific sensor, too.

So to me, with resolving power of a system in mind, it would be great that "test results" include data for different sensors and when compare them, compare them in a setup that can distinguish the difference, in this case, pushing both up, maybe to 300mm, rather than pushing them down. I think if data/pictures done with 300mm tube, in theory, I predict the HR produces better image (ie, more details to me)

OK, back to crazy world

[mjkzz]

Another comparison for those who (like me) trust their eyes more than graphs and charts.

Mitty 10 vs Mitty 7.5 @5X on sensor. 100% crops.

M7.5:



M10:



Again, the higher NA objective shows clear advantage.

I doubt the 5X HR would do much better than the 7.5X @5X, but it would be nice to see a comparison.

- Macrero

Great comparison, I can see the 10x beats the 7.5x, do they have same NA? Somehow I think the 10x has higher NA.[edit] never mind, you already said so, I need to read

[ploum]

0.61*lambda/0.21
Another example... Let me assume that for smallest FOV, you have somehow chosen LP/mm to match the Rayleigh criterion. Then the MTF of the objective for that LP/mm will be only about 9% = 0.09, as it always is for Rayleigh. From that, I conclude that the overall system MTF = objective MTF * sensor MTF, can be no larger than 0.09. But Line 212 of your table, "MTF globale", shows a value of 0.42 . Now I am totally confused. How is the MTF globale 0.42 computed?

These are only examples. In general, I cannot figure out how most of the numbers in your spreadsheet might be computed.

I understand there are some language difficulties.

If you would like to post your spreadsheet, or send it to me some other way, then I will try to make sense of the formulas that I find in it.

But right now, all I see are some numbers that I cannot make sense of. (Anybody who can, please jump in and explain them to me.)

--Rik

0.61*lambda/0.21 i use 0.5*lambda/0.21 for me. You could see than for objective it's that. For example 0.5x0.55/0.21 = 1.3 um for resolution of this objective.

I do not calculate the MTF for the cutoff frequency of the lens otherwise you will not see anything .... The contrast is zero in this case. I therefore choose certain criteria. I choose as a criterion that it takes about 3 pixels for this to be visible on the sensor for example. To follow the evolution of the MTF of the objective I chose 3 diffraction points. The MTF is therefore relative and shows the evolution of the MTF.

For the sensor when I chose 3 pixels as the tilting zone and its associated magnification to say that the sensor can no longer sample correctly, I took the size of this task and I observed its evolution on the sensor by doing vary the higher magnifications. This varies his MTF. For lower magnifications, the MTF remains fixed but the sampling remains blocked at 3 pixels. In this case, the sampling frequency will drop. In fact the objective will sample at frequencies which will also decrease and the MTF of the objective will increase with the downsampling.

[ploum]

wow, this thread has grown a lot, learning more here . .

OK, back to crazy world

I'm happy for that.

In fact, those who want to understand my curves will understand everything they need to upgrade their equipment and better adapt their sensor and their objectives! In fact if we can optimize in the same way a 10x, a 7.5x or a 5x mitutoyo, it would be necessary to take a shorter tub lens for the stronger objectives. My curves give ideas .... For my part, I have already taken these measurements. it is not theory. This is theory confirmed by practice. Whoever does not want to understand the theory risks losing many years. My curves are confirmed by test charts. I will one day make a complete article. But I don't have that time yet. Language is also a barrier. I currently have no one in my country to understand me in my native language in this small world where scientific formalism is not understood. This being the case, for a given sensor it is necessary to adapt the tube lenses to obtain a good sampling. I can tell you from experience that taking 3 pixels per diffraction task is a great thing. However if you have the means for a large sensor and you prefer high contrast images you can take 5-6 pixels for a diffraction task. You will hardly have a better resolution but a better contrast !

[rjlittlefield]

0.61*lambda/0.21 i use 0.5*lambda/0.21 for me. You could see than for objective it's that. For example 0.5x0.55/0.21 = 1.3 um for resolution of this objective.

OK, then you are not using Rayleigh, despite what you said.

Worse, 0.5*lambda/NA gives the cutoff point of the lens: the spatial wavelength where MTF goes to zero. See equation (4.6) in the references that I quote at viewtopic.php?p=124831#p124831 . So this does not match your words.

Further, if I calculate using 0.5x0.55/0.21, I get a frequency of 763.6 line pairs per mm, not 752 LP/mm. So I still do not know where the 752 comes from. Nothing quite matches.

In your other post, you wrote "I will one day make a complete article."

I do not wish to be harsh. But I will tell you bluntly that your complete article will be rejected by any competent reviewer, unless it is a lot more clear than your postings here at PMN.

This bothers me, because I am confident that your approach is useful. The problem is that I cannot tell exactly what you are doing, so I cannot tell exactly how your approach can be useful.

I am happy to help you improve this. But to do that I need to see formulas, not just numbers and word descriptions of what was done.

--Rik

[ploum]

I don't write an article here.... I must write that in French and for private books.

An article need 150 pages.... and i have not time to do in english... I give some ideas for people.

In fact i can use 0.55 um or 0.56 um for lentgh wave of green colore. it's not very important for me

And if i do ARRONDI(;;;;2) it's possible.

In fact i do some help for others i dont' search an understanding more than what i give with my curves.

I can't calculate MTF FOR CUT FREQUENCE . I have choice some criterium for that. 3 pixels or 3 diffraction spot.

Regards

[rjlittlefield]

I give some ideas for people

I would like to understand your ideas. But I do not understand your ideas yet, because I cannot see how your numbers are produced.

Let me try this a different way.

I have an FF 36 Mpix sensor, and a 10X NA 0.25 objective.

Can you tell me how to construct graphs like yours, for myself, for this combination?

--Rik

[ploum]

Yes i do when i finish my work

All this to conclude that if I get a more contrasted image with a lens with a higher numerical aperture it is simply because the sensor can be in subsampling which means that the system will analyze lower frequencies through the objective and therefore give a more contrasted image. This is not because of the lens but the sensor. To say that a goal will be better in this case is a false answer. To know if the objective is better, it is necessary to obtain a correct sampling of the image with an MTF on the identical sensor and therefore an identical sampling. Taking a lens with a larger numerical aperture than another will therefore give an image with the same resolution if the sensor is downsampled but with a higher contrast. This will not show that the objective is "better" since we do not ultimately test the objective directly and its resolving power but its superior MTF. Because the sensor will be in a more accentuated sub-sampling and therefore the useful frequencies through the lens will be lower. This fact improves the MTF of the lens and therefore of the system.

Also to compare an objective with another with different numerical apertures without affecting the MTF of the objective due to the more or less pronounced sub-sampling of the sensor, it is necessary to modify the tube lens so that the sampling on the sensor be the same. in this case it is possible to compare the contrast or the colorimetric correction.

[Macrero]

When my renovations are done I can take a better head to head with my 10x vs 10x hr. I am not sure where I stuck the source files and the edited one I did dig up is kind of ugly.

That would be very interesting comparison. Looking forward to seeing the result.

I can't put my finger on the images, but this reminds me of my comparisons between Nikon 4x/ 0.13 and 4x /0.2 when I started, using a Nikon 700D. That's a full frame 12MP sensor, the image circle was nowhere near covering the whole sensor. I was covering maybe only 5MP of it. Outresolving the sensor perhaps, but the IQ was dramatically better with the NA 0.2 objective.

That's my experience as well. The higher NA objective always outresolves the lower NA one, regardless of sensor size, resolution and pixel pitch.

Image circle/coverage is another story.

Great comparison, I can see the 10x beats the 7.5x, do they have same NA? Somehow I think the 10x has higher NA.[edit] never mind, you already said so, I need to read

Thanks Peter. Yes, the 7.5X is NA 0.21 and the 10X is 0.28. The difference is smaller than in the 0.13 vs 0.20 comparison, but it is still clear.

[ploum]

I have an FF 36 Mpix sensor, and a 10X NA 0.25 objective.
Can you tell me how to construct graphs like yours, for myself, for this combination?
--Rik

[ploum]

Here two objective APO PLAN on the same systeme (sensor, lens tube...)

What is the best ?