Mitutoyo M Plan APO HR 5X 0.21 Test Results

[dmillard]

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.

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

Hello Macrero,

Your images speak more loudly than words! I agree that it would be nice to see a comparison of the Mitutoyo M Plan Apo HR 5X against the M Plan Apo 7.5X both at the same magnification. The Mitutoyo 7.5X has been my favorite objective, but I think the HR 5X may edge it out. A 200mm and a 135mm pair of good tube lenses would work to give 5x on both.

However, I think if you want to see the relative performance of the objectives, rather than that of the tube lenses used, you should use tube lenses of similar construction that scale up or down in focal length. Examples would be the Apo-Gerogons, which have a wide range of focal lengths, and which were originally shown to be superb performers by Charles Krebs and others, and again more recently by Robert OToole.

Earlier this year, I bought a Mitutoyo HR 5X in an early morning Buy It Now. I had been looking for one for the past decade, and this is the first one that I had seen, sold as new from a Chinese seller who had good ratings, accepted returns, and was offering it at about 10% below the price that chris_ma mentioned earlier in this thread. After considerable shipping delays, it arrived, and I was very relieved when it worked even better than I had anticipated . I have subsequently seen several of these objectives offered from different sellers in China, but at about double the price that I had paid.

I have two tube lenses of similar construction that I could use, a 180mm f/9 Fujinon A with the Mitutoyo 7.5X, and a 270mm f/9 Apo-Gerogon with the Mitutoyo HR 5X. Each would then have the same magnification of 6.75x, and the same effective aperture of about f/16. A couple of caveats - the multicoated Fujinon A has already shown itself to perform well as a tube lens. Definitive literature is hard to find on the Fujinon A lenses, but they were sold as being ideal for close-up use, yet still good at infinity. The Apo-Gerogon is single-coated, and is optimized for use from 1x (optimum) to 1/3x. I haven't yet tried it as a tube lens, because it wouldn't mount in a Copal 1 shutter, and it was too large to fit in a Nikon bellows. I now have an SM2 tube set from Thorlabs, but I need to tap the tube mounts to 1/4-20 to fit to the clamps. That will be my project for today.

Please let me know if anyone has any ideas or suggestions before I do the comparison.

Best regards,
David

[Macrero]

Hello Macrero,

Your images speak more loudly than words! I agree that it would be nice to see a comparison of the Mitutoyo M Plan Apo HR 5X against the M Plan Apo 7.5X both at the same magnification. The Mitutoyo 7.5X has been my favorite objective, but I think the HR 5X may edge it out. A 200mm and a 135mm pair of good tube lenses would work to give 5x on both.

However, I think if you want to see the relative performance of the objectives, rather than that of the tube lenses used, you should use tube lenses of similar construction that scale up or down in focal length. Examples would be the Apo-Gerogons, which have a wide range of focal lengths, and which were originally shown to be superb performers by Charles Krebs and others, and again more recently by Robert OToole.

Earlier this year, I bought a Mitutoyo HR 5X in an early morning Buy It Now. I had been looking for one for the past decade, and this is the first one that I had seen, sold as new from a Chinese seller who had good ratings, accepted returns, and was offering it at about 10% below the price that chris_ma mentioned earlier in this thread. After considerable shipping delays, it arrived, and I was very relieved when it worked even better than I had anticipated . I have subsequently seen several of these objectives offered from different sellers in China, but at about double the price that I had paid.

I have two tube lenses of similar construction that I could use, a 180mm f/9 Fujinon A with the Mitutoyo 7.5X, and a 270mm f/9 Apo-Gerogon with the Mitutoyo HR 5X. Each would then have the same magnification of 6.75x, and the same effective aperture of about f/16. A couple of caveats - the multicoated Fujinon A has already shown itself to perform well as a tube lens. Definitive literature is hard to find on the Fujinon A lenses, but they were sold as being ideal for close-up use, yet still good at infinity. The Apo-Gerogon is single-coated, and is optimized for use from 1x (optimum) to 1/3x. I haven't yet tried it as a tube lens, because it wouldn't mount in a Copal 1 shutter, and it was too large to fit in a Nikon bellows. I now have an SM2 tube set from Thorlabs, but I need to tap the tube mounts to 1/4-20 to fit to the clamps. That will be my project for today.

Please let me know if anyone has any ideas or suggestions before I do the comparison.

Best regards,
David

Hi David,

congratulations on your acquisition, sounds like a great deal!

7.5 vs 5HR @5X would be very interesting. It would be also nice to see the performance of the HR pushed up and down (especially down).

I agree about the tube lenses. Almost all of my tube lenses are enlarging/repro lenses of similar/same design/series, except the Makro-Symmar 120. I find them to work well enough and I see no need for spending $$$ on expensive tube lenses.
The only "exotic" tube/rear lens I use is an Apo-El-Nikkor 105/5.6N, but it is a part of my collection, I haven't bought it for the sole purpose of using it as tube lens. A stellar lens btw.

I will look forward to see your tests and comparisons.

Best,

- Macrero

[lothman]

just a question:

The Sony A7Riv has the same pixel pitch than the APS-C models like A6400, see here.

So "pushing down" should be the same than recording a fullframe at higher magnification with a wider field of view, then shrinking down to APS-C-size (making pixels smaller) and the recalculate to the Pixelnumber of the APS-C chip. Is that correct? What will give better results?
a) Pushing down the lens on a smaller sensor or
b) using a bigger sensor with more pixels and downsampling the result.

If b) wouldn't it be easier/cheaper to record with a bigger high resolution sensor? I assume a Sony A7riv is cheaper than a Mitutoyo 5x HR.

[rjlittlefield]

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

<answer consisting of graph and table>

I had hoped for a method, but I got another graph and numbers.

But that's OK, because enough information has now accumulated that I am confident about the math of Pluom's method.

In particular, all of Ploum's graphs and numbers for LP/FOV and LP/mm can be closely reproduced by the following formulas:

• LP/FOV = minimum_of ( pixels_per_FOV / min_pixels_per_LP, mm_per_FOV * max_LP_per_mm)
• LP/mm = LP/FOV / mm_per_FOV

where

• LP/FOV is number of line pairs across the field of view
• LP/mm is number of line pairs per mm, measured on the subject
• pixels_per_FOV is the number of pixels across the field of view, that is, the sensor width in pixels
• min_pixels_per_LP is the minimum number of pixels per line pair, a somewhat arbitrary number ranging in value from 2.5 to 3
• mm_per_FOV is the width of the field of view, measured in mm at the subject
• max_LP_per_mm is the objective's maximum resolvable resolution, expressed in line pairs per mm, = 2*NA/lambda or a little less

Note that these formulas have a very simple interpretation:
(a) to retain all the details, you need around 3 pixels per cycle, and
(b) you will never see any details finer than the objective's best resolution.

Expressed in this form, it should be clear that this is old wisdom. I know of long discussion threads in this forum as early as 2007, and I'm confident that other people wrote similar rules long before that.

So, the novelty of Ploum's method must lie in his generating graphs of FOV versus LP/FOV, essentially repackaging the old wisdom as a relationship between FOV and number of line pairs.

I assume this particular packaging is useful, once it has been thoroughly understood and lived with for a while.

--Rik

[mjkzz]

Thanks @ploum for your insights, need to learn more about all of these and what you are saying. Long time ago, there was a long discussion about something similar and eventually, I "had" (ie, no choice in face of facts) to agree with Rik who was trying so hard to teach us that fact. During that discussion, some experiments I did behind the scenes was not set up correctly.

To me, I tend to look for intrinsic properties of things. For objectives, one of the intrinsic property is its resolving power, the rest, like magnification, image circles, etc, etc are probably nomenclatures under a specific settings, for example, magnification of 5X only means when the objective is put on a specific tube lens, that is magnifying power. So is image circle. In fact, from this forum, I learned that an infinite objective is just a optical device that bends lights in certain way and you can use it the way you like as long as it is within certain range or limit.

Then there is this engineering part where all these nomenclatures can be massaged, if you will, to fit into certain application. In this case, the 5x 0.21 could be better than a 7.5x 0.21, though both might provide the same resolving power, the 7.5x 0.21 might not have enough image circle when pushed down to 5x. Looking at the other way around, the 5x 0.21 probably provide larger image circle when pushed up to 7.5x, thus making it more flexible for more applications.

Why is this? The reason lies in yet another intrinsic property -- how they are constructed/designed to bend lights, and the reason it is intrinsic is because that is what theories dictate, beside the NA number (ie, a 5x 0.14 can be pushed up, benefiting from larger image circle regardless of NA), follow the theory, you get what you want and use it (engineering) by its limitations.

[RobertOToole]

Hi Ploum,

Just read your post and all the replies just now, about 5 days late! (thanks Chris!)

I'll just add some small notes in separate replies, wish I could have read the posts earlier.

One of the Mods should spilt this post though!

And with i long focal lenght with 400 mm that's could be great !

Personally I would not bother with any long lenses to pull the magnification that high on the Mitutoyo HR 5x.

Why?

I own a 5x 0.20 Qioptiq mag.x lens and I have not published the test yet, but I spend way too much time and money testing 240-270mm tube lenses (the Qioptiq standard TL is 250mm FL) only to be disappointed with excessive CAs in the corners. FYI the mag.x is very well corrected similar to the HR 5x Mitutoyo, and much better than a typical APO objective like the Nikon Plan APO 4x 0.20. (these two are tested head to head on my site).

I did plan on going longer to 300mm+ but from what I found in my recent 240mm tests, unless I find a Qioptiq tube lens, I'm not wasting my time pulling any HR objectives anytime soon. BTW Qioptiq does offer 400mm and 500mm tube lenses for the mag.x.

Best,

Robert

[RobertOToole]

Also rather than buying a 5x 0.21 I prefer with a 61 Mpix, it will be more economical to use a shorter tube focal length to get the same sampling (and the same resolution) with the 7.5x 0.21.

the original post was about the 5x 0.14 vs the 5x HR 0.21, where I do see a difference between the two on a 61MP FF sony camera, whether the theory predicts otherwise or not.

now it might well be that the 7.5x 0.21 pushed down to 5x is better then the 5x 0.14, and maybe even close to the HR 5x 0.21 (although I'm sceptical about corner performance). that's for another test though, and probably a tricky one considering that the tube lenses will need to be different.

I totally agree that the 7.5 0.21 is more economical at 1/3 the cost

FYI guys:

Qioptiq 5x 0.20 Mag x Vs Mitutoyo 7.5x pushed down to 4.5-4.6x.

+ the M Plan 5x 0.14 at 5x.

All on APS-C.

The Mag.x beats both!

https://www.closeuphotography.com/5x-lens-test

Also, this might also be interesting:

Mag.x 5x 0.20 APO vs Mitutoyo 5x 0.14 M Plan at 3x

APS-C 24 MP

The HR lens (5x 020) is a lot sharper and CAs are much better controlled.

https://www.closeuphotography.com/3x-lens-test

Best,

Robert

[ploum]

Thanks @ploum for your insights, need to learn more about all of these and what you are saying. Long time ago, there was a long discussion about something similar and eventually, I "had" (ie, no choice in face of facts) to agree with Rik who was trying so hard to teach us that fact. During that discussion, some experiments I did behind the scenes was not set up correctly.

Thanks to this forum. Because I wondered about the fact that despite the downsampling, the most open lenses gave better images despite a downsampling in all cases. However, the first idea is to give a constant MTF to the objective which is no longer the case here ... I hope that this discussion will allow everyone to progress and better adapt the material at their disposal.

As for Robert, I thank him for his tests, which I watch regularly, even if I interpret them a little differently when a more open objective is declared "winner". It would take other tests (and it's time consuming) as I have already detailed. This being the case when we take tests on tube lenses, where the magnifications and the objective are the same, then the tests are 100% in accordance with my way of thinking.

Rick summed up well what there was to be summarized. He has a synthetic thought . Hopefully this intelligence will return to the future that awaits humanity.

[JKT]

Here's what I've tried to use:

The only special part should be the combination of the two resolutions. What has so far been mentioned equals n=infinity in this formula. Otherwise it should be the same as what has been discussed ... unless there is yet again another error. Let's not go into details why I wrote "tried"...

[chris_ma]

Here's what I've tried to use:

great to see some formulas, but I agree with robert that this probably is best split into a different thread. maybe somethings like:
"Formulas to calculate the approximate resolution or lenses (and photographic systems)"

[Chris S.]

. . . I agree with robert that this probably is best split into a different thread. maybe somethings like:
"Formulas to calculate the approximate resolution or lenses (and photographic systems)"

We will split it. I had originally contacted Robert to ask if he wanted us to do so.

It's a surprisingly long job to split a thread like this, if one wishes to do so without creating a mess. I'm looking for a block of maybe three hours without interruptions. An unthoughtful effort of splitting could, of course, be done with much less time.

During the splitting process, I may pull the thread offline temporarily. So if anyone suddenly finds it missing, have no fear.

Chris_ma, I may well borrow from your suggested title.

--Chris S.

[Macro_Cosmos]

I would like to chime in with tests but I no longer own a 200mm tube lens. All my ITL200s are gone and the clone CMH-200 is with a good friend.
Well, maybe I can use that fluorite/semiapochromatic tube lens I have.

The theories, I can understand, calculations? Not without confusing myself. I will have to run my own calculations.
In theory, there should be no difference between a 5x NA0.14 and a 5x NA0.21 objective if they are both out-resolving the sensor. I don't have my spreadsheet to calculate exactly the kind of pixels required. In practice, the latter always seems to be better. You must note that this still isn't an apples-to-apples comparison, obviously an HR being many times more expensive than a normal version translates to tighter quality control and better aberration correction. The "clarity" one might see could just be a consequence to better aberration correction.

Contrast can be defined as "CNR", or contrast-to-noise ratio. We would like to discriminate between the signal and the background (signal). The final image usually consist of two types of signals and various factors of noise.

In microscopy, the signal we want to preserve is the subject's signal, ideally we want to completely eliminate destructive background signals. Methods such as SDCm, LSCm and TIRFm[1] completely eliminates the background signal, which yields very high contrast and beautiful imagery free from weird out-of-focus artefacts. In combination with deconvolution, the result is much better than standard widefield fluorescence microscopy. Another method is precision localisation with SIM (structured light illumination), which uses a grid to take 3 images, eliminating the background signal algorithmically. You can get twice the NA out of your objective, this is very similar to pixel shift methods translating to clearer images.

Low contrast can be defined as "background and signal being hard to distinguish" -- spatial pixel averaging occurs which causes apparent reduction in temporal resolution.
Example: cityscape on a very foggy day.
The SNR is high due to plentiful photons, but the skyline can be barely seen, why? Our SNR is strong here, this shouldn't be the case. Well, the subject and background is similar.

When we talk about contrast in a scientific manner, it's better to have it defined.

A practice to extract the most out of an objective would be to simply outresolve it, and then downsample. The process is tiring and laborious with little to no benefits in my experience. I believe this ties in with @ploum's statements. A secondary component is the type of light used, which is another factor that renders many comparisons unreliable at best.
When using LED lights, what's the type, how's the spectrum like? This applies to other light sources too.
White LED: blue dominant, everything else is subpar.
Warm white: green dominant, barely any blue component.
Daylight LED: ideal
Halogen: Strong in green, very strong in red and IR components
Xenon (with VIS bandpass): Strong in blue, slight drop off afterwards, very good light source

Most CMOS sensors offer the highest quantum efficiency in the green spectrum. So with a green dominant LED, we get high SNR. With a standard one, the SNR is lower. Ideally, we'd like rSNR to stay the same -- plentiful light! This should in theory eliminate variations due to the light source. Either way, we are still testing the performance of an entire optical system. Let's not mention sensor linearity and ways consumer camera companies like to play with their FPGAs passing the result off as "raw".

Want to see an extremely egregious example? Oh it's also Sony, is anyone surprised?
https://petapixel.com/2021/10/29/could- ... e-alpha-1/
I'm not going to entertain sensor speculations. What's happening here is a 48MP sensor being reduced to 12MP through readout binning. Basically, 4 photosites are considered as one pixel during readout (my understanding). This is different to digital binning methods which combine pixels algorithmically to garner lower readout noise and higher sensitivity at the cost of temporal resolution, thus leading to better low light performance. With 4 pixels read out as one, chances of errors and readout noise is quadrupled, I think, so it's actually worse than digital binning. Sony's a genius, hats off, they basically spun what could be one camera into two and marketed one of them as "superior for low light" when the other model can more or less be the same with a firmware update. Now, why not just offer us digital binning...

Anyhow, enough babbling. What I'll attempt to do is compare my Edmund Optics 5x objective to an HR one with higher NA (0.2), I'll use my semiapochromatic 200mm tube lens and Olympus projection eyepieces to extract detail. Hopefully I can offer something to this fun discussion. I will be using wafers and a diatom test slide. The NA here is 0.2 maximum, so coverslip wouldn't matter.

While it's frankly infeasible to optimise such tests, I believe we can at least strive for purity.


[1]: Spinning disc confocal microscopy, laser scanning confocal microscopy, total internal reflection fluorescence microscopy.

[rjlittlefield]

In theory, there should be no difference between a 5x NA0.14 and a 5x NA0.21 objective if they are both out-resolving the sensor.
...
When we talk about contrast in a scientific manner, it's better to have it defined.

It's also better to define what theory is being talked about, and I don't see where you've done that.

For comparing 5x NA0.14 and 5X NA0.21, the theory that I use is based on MTF and diffraction, both of which are well defined and discussed at length in standard texts.

Using that theory, optically perfect NA 0.21 gives higher MTF ("more contrast") than optically perfect NA 0.14, at all spatial frequencies above 0 and below cutoff.

So, the statement you make about "no difference" can only be true if your theory somehow compensates for the differences in MTF.

Can you be explicit about what theory you're using, that does that?

--Rik

[Macro_Cosmos]

licit about what theory you're using, that does that?

--Rik

Resolution matching, 5x at NA 0.14 matches a 36MP 35mm format sensor.
So with say a 24MP sensor, the objective resolves much more. Moving up to 0.21 should yield no difference, right?
Not the case of course.

See lines in letter B.

[rjlittlefield]

Resolution matching, 5x at NA 0.14 matches a 36MP 35mm format sensor.

For the sake of discussion, can you say exactly what it means for an objective to "match" a sensor?

Moving up to 0.21 should yield no difference, right?

I can't tell, are you actually puzzled by the observed difference, or are you trying to make some point by asking a rhetorical question?

--Rik