Teleconverters in macro

[Scarodactyl]

I don't think the resolution limit is news to anyone, but sometimes this is a practical solution in a real world shooting situation.

[Guppy]

Hi

I deliberately mention camera (pixel pitch 2.2µm and 4.9µm) and lens (10X/0.28 and 20X/0.42, they are my measured values!
At 5:1, 1:1 or in reducing photography it behaves differently.
Concerning teleconverters there is no simple general rule.

Kurt

[rjlittlefield]

I deliberately mention camera (pixel pitch 2.2µm and 4.9µm) and lens (10X/0.28 and 20X/0.42, they are my measured values!

I do not disagree with your numbers, but I am curious how they were measured. Is that described somewhere?

--Rik

[Guppy]

Hi Rik

I often disagree with numbers I read, too,
so I make my measurements several times on my setup with the Zeiss Resolution Test Object 3000.
These results are authoritative for me.

Kurt

[Lou Jost]

"Concerning teleconverters there is no simple general rule."

Agreed, or at least, the general rules are very complicated.

Here's one general rule: the resolution (on the subject) of the aerial image coming out of the teleconverter can't be higher than the resolution (on the subject) of the aerial image coming out of the prime lens.

Here's another: The photon flux coming out of the teleconverter (photons per sq mm) can't be higher than the photon flux coming out of the prime lens divided by the magnification factor squared.

Here's a tentative, surprising, and potentially controversial one I'd like to suggest (a generalization of my earlier comment): for a perfect teleconverter (one with no optical aberrations), the resolution (on the subject) of the final "teleconverted" image made by the sensor is always higher than the resolution of the image made by the sensor using only the prime lens.

My reasoning is that the final resolution (on the subject) of an image on the sensor is some increasing function of the resolution of the aerial image of the lens and the resolving power of the sensor. If you increase either one, the resolution (subject-side) of the final image will increase (though the increase may be so small as to not matter). When you add a teleconverter, you are now greatly increasing the number of pixels under each detail of the subject, so the subject-side resolving power of the sensor increases, and so must the total resolution (on the subject) of the whole system.

Does that make sense? Or have I forgotten something important?

Edit: I just read the entry by Kurt about his ability to actually measure exact resolutions. Maybe he could detect this effect if he had a sharp-enough prime lens.

[Scarodactyl]

It is interesting that these numbers exceed the classic NA to LP/mm formula (lp/mm = na * 3000). Of course that's probably by some particular definition of linepair.

[rjlittlefield]

I make my measurements several times on my setup with the Zeiss Resolution Test Object 3000.
These results are authoritative for me.

I cannot find documentation about that target. Can you point me to a description of it?

In the meantime I expect that it's a high resolution line target, and you are looking to see if the lines can be separated at all, even though low contrast.

If that's correct, then the number will be near the diffraction-limited cutoff frequency.

It is interesting that these numbers exceed the classic NA to LP/mm formula (lp/mm = na * 3000). Of course that's probably by some particular definition of linepair.

The general formula for cutoff frequency is nu_0 = (2*NA)/lambda. That's the spatial frequency where MTF drops clear to zero and stays there.

For lambda = 0.00055, that means cutoff at NA * 3636. Probably the number 3000 comes from Rayleigh criterion, NA/(0.61*lambda), which corresponds to a slightly higher MTF achieved at lower lp/mm.

So, the difference is not so much a definition of "linepair" as "how clearly do I need to see them?"

--Rik

[Scarodactyl]

I assume 3000 is also favored because it gives nice looking numbers with typical NAs. IIRC I've seen Leica Nikon and Olympus use it for stereo/macroscope systems.

[Guppy]

Hi Rik

I am not a laboratory for microscopy optics, I am a photographer.
For me the result with real objects in the processed image is important.
The resolution drop to the image corners should not be disturbing.
In general, a good Fotolens with a low resolution drop to the corners (large good quality image circle) has a lower resolution in the center of the image than a lens with a small good quality image circle.
Photo lenses, low resolution drop to the corners.
Microscope lenses, very high resolution in the image center, strong resolution drop to the image corners.
For the microscopist, resolution in the center of the image is important.
For the photographer the resolution in the image corners is important, if it is good there, it is usually also good in the image center.
In the demands on a lens there is a clear difference between microscopy and photography, I am a photographer!

Depending on the setup and the object you are photographing, a clear difference is often not visible in the processed image.

Example:
Nikon D810 (full frame) with NIKON M Plan, 60/0.7 ELWD, 210/0, image cropped by 10%.


Original see:
www.focus-stacking.ch/B/05629_00.JPG

I know some formulas and calculations.
To know it more precisely in reality, I also make resolution measurements.
With real objects in reflected light (e.g. computer microchip) the illumination can falsify the results!
With the Zeiss resolution test object 3000 and diffuse transmitted light there are no significant deviations due to illumination.
For measurements I use flash light with short burning time (close to daylight and without blurring by vibration).
I evaluate the images with the Zeiss resolution test object 3000 by eye.
An image detail of the MITUTOYO M Plan Apo 20X/0.42 ∞/0 f=200

I see a resolution of 1400LP/mm
A laboratory for optics can do better.

The difference of the resolution lines with the Zeiss resolution test object 3000 is about 10%, (800, 900, 1000, 1100LP/mm etc.). Average values in the image can be estimated. Accuracy about 2.5 to 5%.
When making the series of images, I change the distance (Focus) between the lens and the object by 0.001mm with the StackShot. From the series, I select the sharpest image for evaluation.

With this method, I detect differences between two MITUTOYO M Plan Apo 10X/0.28 ∞/0 f=200, for example.

My setup:








The tube around the lens is made of Hi-tack flocked light trap Self-adh flock material



Ich bin kein Labor für Mikroskopie-Optik, ich bin Fotograf.
Für mich ist das Ergebnis mit realen Objekten im bearbeiteten Bild wichtig.
Der Auflösungsabfall zu den Bildecken soll nicht störend sein.
Allgemein besitzt ein gutes Foto-Objektiv mit einem geringen Auflösungsabfall zu den Ecken (grosser qualitativ guter Bildkreis) in der Bildmitte eine geringere Auflösung wie ein Objektiv mit kleinem, qualitativ guten Bildkreis.
Fotoobjektive, geringer Auflösungsabfall zu den Ecken.
Mikroskop Objektive, sehr hohe Auflösung in der Bildmitte, starker Auflösungsabfall zu den Bildecken.
Für den Mikroskopiker ist die Auflösung in der Bildmitte wichtig.
Für den Fotografen ist die Auflösung in den Bildecken wichtig, ist sie da gut, ist sie üblicherweise in der Bildmitte ebenfalls gut.
In den Ansprüchen an ein Objektiv besteht ein klarer Unterschied zwischen Mikroskopie und Fotografie, ich bin Fotograf!

Je nach Setup und Objekt das man fotografiert, ist ein eindeutiger Unterschied im bearbeiteten Bild oft nicht erkennbar.

Beispiel:
Nikon D810 (Vollformat) mit NIKON M Plan, 60/0.7 ELWD, 210/0, Bild um 10% beschnitten.


Original siehe:
www.focus-stacking.ch/B/05629_00.JPG

Einige Formeln und Berechnungen kenne ich.
Um es in Realität genauer zu wissen, mache ich auch Auflösungsmessungen.
Bei realen Objekten im Auflicht (z.B. Computer Mikrochip) kann die Beleuchtung die Ergebnisse verfälschen!
Mit dem Zeiss Auflösungstest-Objekt 3000 und diffusem Durchlicht sind keine bedeutenden Abweichungen bedingt durch die Beleuchtung vorhanden.
Für Messungen verwende ich Blitzlicht mit kurzer Abbrennzeit (nahe Tageslicht und ohne Verwischung durch Erschütterung).
Ich beurteile die Bilder mit dem Zeiss Auflösungstest-Objekt 3000 von Auge.
Ein Bildausschnitt vom MITUTOYO M Plan Apo 20X/0.42 ∞/0 f=200

Ich sehe eine Auflösung von 1400LP/mm
Ein Labor für Optik kann es besser.

Der Unterschied der Auflösungslinien beim Zeiss Auflösungstest-Objekt 3000 beträgt etwa 10%, (800, 900, 1000, 1100LP/mm u.s.w.). Mittelwerte im Bild können geschätzt werden. Genauigkeit etwa 2.5 bis 5 %.
Beim Anfertigen der Bildserie verändere ich mit dem StackShot die Distanz (Fokus) zwischen Objektiv und Objekt um 0.001mm. Aus der Serie wähle ich zur Beurteilung das schärfste Bild aus.

Mit dieser Methode erkenne ich z.B. Unterschiede zwischen zwei MITUTOYO M Plan Apo 10X/0.28 ∞/0 f=200.

Mein Setup:








Der Tubus um das Objektiv ist aus Hi-tack flocked light trap Self-adh flock material


Kurt

[JKT]

Having fun with the numbers it seems that the resolution in that test should then be k*Lambda/NA, where k must be between 0.477 (0.509 if there was 1000 line target, which was not resolved) and 0.536. This naturally assumes perfect target and objectives perfectly according to spec, but it was still an interesting test.

[Lou Jost]

"With teleconverter the resolution of the object is not higher (empty magnification)."

It is also not lower than the prime image. I think this is noteworthy. Many people are prejudiced against teleconverters. I've seen such complaints even on this forum. But these measurements show that the teleconverter DOES NOT DEGRADE THE IMAGE AT ALL. That means the best teleconverters today are acting almost as perfect lenses. So I think with a good prime lens, my prediction above should be testable: the subject-side resolution of the image (as captured by the sensor) using such a teleconverter could actually be detectably higher than the resolution of the image using the prime lens alone!

[Scarodactyl]

Thanks for doing these measurements, they are really interesting!

[ray_parkhurst]

"With teleconverter the resolution of the object is not higher (empty magnification)."

It is also not lower than the prime image. I think this is noteworthy. Many people are prejudiced against teleconverters. I've seen such complaints even on this forum. But these measurements show that the teleconverter DOES NOT DEGRADE THE IMAGE AT ALL. That means the best teleconverters today are acting almost as perfect lenses. So I think with a good prime lens, my prediction above should be testable: the subject-side resolution of the image (as captured by the sensor) using such a teleconverter could actually be detectably higher than the resolution of the image using the prime lens alone!

I believe I showed this to be true in a thread some years ago by shooting with a 105PN (or maybe a 95PN) with and without a 2x teleconverter. By downsizing the teleconverter image 2x, the resulting image had higher resolution than the image without teleconverter. But all I think this experiment says is the lens is out-resolving the Bayer sensor. Or maybe I am not understanding your hypothesis.

[Lou Jost]

I believe I showed this to be true in a thread some years ago by shooting with a 105PN (or maybe a 95PN) with and without a 2x teleconverter. By downsizing the teleconverter image 2x, the resulting image had higher resolution than the image without teleconverter. But all I think this experiment says is the lens is out-resolving the Bayer sensor. Or maybe I am not understanding your hypothesis.

Ray, I probably saw that post...yes, that result is perhaps not surprising when the lens greatly out-resolves the sensor. My hypothesis asserts that the resolution (on the subject) with a perfect teleconverter will be higher even when the prime lens does not out-resolve the sensor. A perfect teleconverter always improves resolution. That is a bit surprising but I think it is true, for the reasons I gave. However, the improvement will probably be undetectably small if the prime lens resolution is low relative to the sensor's ability to resolve.