16 replies to this topic

[UlfW]

I made a new measurement with my pair of lenses, Prinz Galaxy 35/3.5 and the Soligor 35/3.5 FA
They are the same types as discussed in Jonathan's topic:
https://www.ultravio...__fromsearch__1

My measurement method has limitations as I lack a suitable strong light-source and integrating sphere.
I use a collimator-based setup instead.
This leads to the following: different kinds of magnifications due to lens effects which change the total intensity which passing the lens and reaching the spectrometer.
With this method I cannot measure the absolute transmission, effects from the aperture settings or system vignetting.
The cutoff wavelength measurements still seem to be quite valid.

---

The Setup:
A DH-2000-BAL connected to a small collimator, from Avantes, COL-UV/VIS via a 400um UV-VIS optical fiber.
This collimator gives a beam of ca 3mm diameter.
After a gap of ca 10cm, a second bigger collimator COL-UV/VIS-25 is placed facing the first collimator.
It is connected to the Ocean Optics Flame spectrometer via a second 400um UV-VIS optical fiber.
The collimators are aligned for a maximum throughput, and the light source is stabilised for at least half an hour.
Before the measurements, both offset and 100% calibrations were done.
After the measurements they were verified and if needed re-calibrated.

---

Testing:
By placing the lens in the light path close to the second collimator and gradually aligning it I can find a position and angle where the signal has a maximised level after the short-wavelength cutoff knee.
The good graph is stored. This is repeated several times until the best possible throughputs are recorded several times.
As aligning is difficult, some sub-optimal spectra are recorded, especially at first, until the optimal are found. Afterwards he sub-optimal spectra are pruned.

To be able to compare lenses with different designs and magnifications, I normally normalise the different spectra, for an equal level, against wavelengths around 400nm.
This time I would like to show the "raw" result without any normalising, as I assume the optical design is identical.

---

The Results:
My pair of lenses seem to behave almost identically

green: Prinz Galaxy 35/3.5 at f/3.5. red: Soligor 35/3.5 FA at f/3.5
Four spectra from each lens.
Maximal transmission due to system magnifications: ca 25%

The test beam is rather narrow.
Because of this the aperture setting will not affect the result as long as the entrance pupil is wider than the light-beam's diameter.
This test verifies that:

dark green: Prinz Galaxy 35/3.5 at f/8. red: Soligor 35/3.5 FA at f/3.5.

The emitted beam from the first collimator is ca 3mm in diameter.
The entrance pupil at f/8 is ca 4.7mm

When the aperture is close to the beam width, it becomes very difficult to align the lens properly.
I began to see problems aligning at f/11.

red: Soligor 35/3.5 FA at f/3.5. purple: Soligor 35/3.5 FA at f/11.

Both my lenses show a 50% cutoff wavelength ( 50% of 25% ) at 335nm.
This is equivalent to the "three stops down value", sometimes used in the lens sticky, stating a 320nm cutoff.



Edit: a few punctuation/tense edits by the Editor.

[Andrea B.]

Thank you, Ulf, for this work. Your efforts, along with others, help us very much in understanding the use of UV-capable lenses which are not UV-dedicated. We have known the Prinz Galaxy 35/3.5 and Soligor 35/3.5 to be valid UV-capable lenses. Your work further confirms this.

I would particularly like to thank you for fully explaining your test protocol.
So often in the past nobody did that.

I think we should emphasize that the cutoff wavelength measurement is valid, but that you could not measure absolute transmission. And I have a question for you about that: should I put the "half-maximum transmission" value into the Lens Sticky if the absolute transmission could not be measured? How far off 25% do you think the absolute transmission is?

I am always so pleased and happy that our forum has members who are inspired to test and experiment with UV-gear and share their results. We have all learned so much! UVP forum members rock!!!

[enricosavazzi]

I have been thinking of a simple but reasonable way to eliminate (or at least greatly reduce) the biases and errors introduced by system noise and uneven spectral distribution of the illumination. I think that one can use a pragmatic way to get a measurement that is not too different from the relative transmission of a lens at a given wavelength, here sketched as a sequence of steps. It begins with the reasonable assumption that lens transmission in green light (540 or 550 nm according to different sources) can be regarded as a 100% baseline transmission of the lens at a given aperture and focus position. This assumes that the measuring apparatus does not introduce significant leaks/contamination from outside the optical path of the lens and measuring system. Another reasonable assumption is that we only need the transmission of the lens along its optical axis, i.e. at the center of the image circle. We just assume that off-axis effects are negligible, so we don't really need an integrating sphere and can use collimated illumination.

• Use the measuring apparatus to record a spectrum between 300 and at least 550 nm, without a lens in the optical path. Let's call this spectrum S_L (or light spectrum). We will use this spectrum as a baseline to compensate for a non-flat spectral distribution of the illumination and measuring system. In some systems, you may need to use a diaphragm to attenuate the light source and avoid saturation of the spectrometer's sensor. A diaphragm is far better than an ND filter, because the former has a guaranteed flat spectral response.
  
• Place the lens in the optical path, select a reasonable lens aperture and a reasonable focus distance (e.g. the lens focuses a point on the subject (illumination source) side to a point on the image side, which is where the entrance of the optical fiber of the spectroscope (usually behind a small diffuser) is located. Record a third spectrum, S. You may need to change the illumination intensity between S_L and S, but this is all right as long as it is done without changing the illumination spectral distribution.
  
• Place a black stop in the optical path and record a corresponding dark spectrum, let's call it S_D. We will use this spectrum to compensate for the dark noise of the system. I normally do this with my spectrometer, and it is a significant improvement over the non-compensated readings. The main effect is that dark is now at the true y=0 value of the spectral graph at all wavelengths. Without this compensation, with no illumination I get instead a slightly jittery line that slowly slopes upward toward higher wavelengths. This procedure does not compensates for random noise, but this type of noise is reasonably attenuated by using a long exposure time and/or averaging among multiple recordings and sensels (spectrometers and/or their software often can do this).
  
• Compensate for the dark noise by subtracting S_D from S: S = S - S_D, where the result is the de-noised transmittance spectrum.
  
• Normalize S by using the S value at 550 nm: S = S / S[550].
  
• Compensate and normalize S_L in the same way: S_L = (S_L - S_D) / S_L[550].

Now S and SL can be directly compared. The relative transmittance spectrum of the lens in % is given by S / S_L * 100, where the lens transmittance at 550 nm is always 100%.

In practice in the computation one must also handle the case where a few values in each normalized spectrum can be zero or slightly negative (mainly because noise is not exactly the same from reading to reading).

Am I forgetting something?

Edited by enricosavazzi, 28 July 2019 - 15:35.

[UlfW]

Andrea B., on 28 July 2019 - 13:06, said:

Thank you, Ulf, for this work. Your efforts, along with others, help us very much in understanding the use of UV-capable lenses which are not UV-dedicated. We have known the Prinz Galaxy 35/3.5 and Soligor 35/3.5 to be valid UV-capable lenses. Your work further confirms this.

I would particularly like to thank you for fully explaining your test protocol.
So often in the past nobody did that.

I think we should emphasize that the cutoff wavelength measurement is valid, but that you could not measure absolute transmission. And I have a question for you about that: should I put the "half-maximum transmission" value into the Lens Sticky if the absolute transmission could not be measured? How far off 25% do you think the absolute transmission is?

I am always so pleased and happy that our forum has members who are inspired to test and experiment with UV-gear and share their results. We have all learned so much! UVP forum members rock!!!

I am sorry for not describing the method earlier, as I have presented many relative transmission graphs for lenses on the forum before.
I have hesitated as the method has evolved and I have gradually gained understanding and confidence why the results are really valid.
If needed, I think I can now motivate the method and its limitations.

IMHO the half maximum value is an interesting parameter for the Sticky, regardless of the absolute attenuation, as it describes how much of the UV-spectra the lens is passing.
The absolute transmission is also interesting, especially at 365-370nm and above as that parameter gives a hint about the exposure time that can be expected for a given motif and light-situation.

Please do not be confused by the numerical value of 25%! It is just an intermediate raw value and cannot be compared with similar values from other lenses.
It is caused by scaling of the lens and the ratio of entrance- and exit-pupil dimensions and how well they relate to collimator and fiber-end dimensions.
Normally it should be normalised to the long wavelength amplitudes. I just kept it un-normalised to show how similar the two lenses are.

I am not yet fully aware about the exact relations causing the intensity scaling and think that the errors recalculating the measured value theoretically is too uncertain.

I think that these two lenses, Soligor and Prinz Galaxy, have a higher transmission at 400nm than the real reference lens, the UV-Nikkor 105mm.
These lenses have fewer lens groups and at 400nm the transmission of the used glass materials are good.
Possibly the AR-coating can be slightly better too, that the wide-band used coating in the UV-Nikkor.
I would guess that the real absolute transmission at 400nm would be between 80% and 90% like the Meritar Jonathan measured at the second attempt.

It might be possible to find the correct absolute transmission at the 400nm end, and use a simple low cost lens as a transferrable reference.
The idea that David suggested for t370 measurements might be a way to go, ( t400 or t420 instead). Then a strong LED could be used as a light source.

I will see if I can set up a usable metod.
It would be wonderful if Jonathan could include his EL-Nikkor 80/5.6 in his absolute measurement experiment.

Edited by UlfW, 28 July 2019 - 16:20.

[UlfW]

enricosavazzi, on 28 July 2019 - 14:44, said:

Am I forgetting something?

If I didn't miss something in your method description, I think you forgot the scaling effects due to entrance/exit-pupil relations of the lens and how well the light is coupled back into the fiber.
That is the magic performed by an integrating sphere to eliminate those effects.
I also think that 100% transmission at 550nm is optimistic for older lenses and coatings.

However using an easy wavelength for corrective scaling is a good idea.
For my measurements I would like that wavelength to be included in the range of my deuterium lamp where I have spectrogram measurements.
Naturally a split measurement with the tungsten lamp could be performed.

[JMC]

UlfW, on 28 July 2019 - 16:12, said:

It would be wonderful if Jonathan could include his EL-Nikkor 80/5.6 in his absolute measurement experiment.

Hint noted Ulf. When I've finished modifying and tweaking the method, I can do the El Nikkor 80mm f5.6 on it. Feel free to hassle me occasionally about it

[UlfW]

UlfW, on 28 July 2019 - 16:12, said:

I am not yet fully aware about the exact relations causing the intensity scaling and think that the errors recalculating the measured value theoretically is too uncertain.

My verbal description of the test setup in the first post might be difficult to follow.
A picture say more than a thousand words and and I used less than 100 words in the description above.

I sketched a diagram with the excellent free drawing program Inkscape to clear my mind, in hope of getting some ideas for the reason to the intensity scaling.
https://inkscape.org/

Please see the resulting simplified setup schematics:

In the ideal alignment I tried to find the parallell collimated beam travel along the optical axis for a maximum transmission.
The lens system affect the beam by changing the diameter and the system transmission by attenuating some wavelengths more than others.
https://en.wikipedia.../Entrance_pupil

I took some closeup pictures of the entrance- and exit-pupils of the Prinz Galaxy-lens at f/5.6 to find the pupil diameter relation.
The measured distances between the same features in the images was 1837px and 2413px ±2px.
The exit beam diameter is expanded by 1.389 x 3mm = 4.17mm.
This is not a problem as diameter of the lens in the big collimator is more than 4x the small collimator lens.
The cone of light into the second collimator is well within the acceptance angle as the NA is the same for the two fibres.
https://www.rp-photo...ber_optics.html

I have still no valid theory why only 25% of the light is seen passing the lens compared to the light in the thinner calibration beam, defining 100%.
Suggestions are welcome!!

Edited by UlfW, 29 July 2019 - 13:40.

[Andrea B.]

[Off Topic: Thank you for mentioning Inkscape. I can use something like that. The diagram is excellent.]

[enricosavazzi]

UlfW, on 28 July 2019 - 17:18, said:

If I didn't miss something in your method description, I think you forgot the scaling effects due to entrance/exit-pupil relations of the lens and how well the light is coupled back into the fiber.
That is the magic performed by an integrating sphere to eliminate those effects.
I also think that 100% transmission at 550nm is optimistic for older lenses and coatings.

However using an easy wavelength for corrective scaling is a good idea.
For my measurements I would like that wavelength to be included in the range of my deuterium lamp where I have spectrogram measurements.
Naturally a split measurement with the tungsten lamp could be performed.

Thanks for your comments. Entrance pupil, exit pupil and coupling between light exiting the lens and entering the fiber remain constant through the test of a given lens at a given aperture and focus setting. If we assume that these variables are not wavelength-dependent, they should not affect the results.

The initial test without a lens is only carried out to obtain a "transmission" spectrum of the measuring setup alone, to use it as a comparison with the transmission spectrum with a lens in the system. The two spectra are then normalized each on its own against their respective intensity values at 550 nm, so the absolute values in either spectrum are not relevant (and therefore the absolute intensity values read with lenses of different speed, aperture and focus settings are automatically compensated for). In practice, this method gives results that eliminate lens speed from the transmission diagram. If a comparison that takes lens speed (and/or effective aperture) into account is desired, then the computation must be modified by incorporating the effects of different effective apertures by scaling the relative transmission spectra (this is also wavelength-independent, so quite simple to implement).

[dabateman]

Ulf,
For clarity, is the whole DUT system the subject lens to be tested? Or is the subject lens just the rectangle marked system transmission?
I think the loss occurs due to air glass interface and back scattering within a lens. Some lenses will reflect back light of certain wavelengths.

[UlfW]

David,
The DUT system is a schematic representation of a lens to be tested.
The Soligor- and Prinz Galaxy-lenses are constructed with five separate lenses, having ten air to glass transitions.
The lenses have some kind of AR coating that I expect still have at least some effect at 400nm.

Assume that the coating wasn't there.
Then transmission left after the loss due to air glass interfaces would be something around 0.96¹⁰ = 66%, assuming realistic refractive indexes for glass.
I expect the transmission left to be even better, maybe between 70 and 80%
As the cutoff wavelengths seam to agree with many other measurements. it seams to be some kind of scaling effect.
When I measure I see 25%. That is what bothers me. I want to know the reason for the scaling.

When Jonathan measured last time with an integrating sphere the Meritar showed ca 70%. That is a much more realistic.

[JMC]

Ulf, there seem to be loads of different variants of the Prinz Galaxy 35mm f3.5. Would it be worth us both putting up images of our copies to see whether they are even the same? If they are different this may explain our different experiences.

[UlfW]

OK, the top one here:
https://www.ultravio...dpost__p__17233

[JMC]

Ah right, yes mine doesn't look like either of those 2.

[enricosavazzi]

JMC, on 30 July 2019 - 17:54, said:

Ulf, there seem to be loads of different variants of the Prinz Galaxy 35mm f3.5. Would it be worth us both putting up images of our copies to see whether they are even the same? If they are different this may explain our different experiences.

Some of my copies can be seen here (you should also follow the links from the following page to other pages):
http://savazzi.net/p...phy/35mmuv.html

[UlfW]

JMC, on 31 July 2019 - 10:57, said:

Ah right, yes mine doesn't look like either of those 2.

That is fine, your Prinzgalaxy and my Prinz Galaxy are really different versions.
I think your your Prinzgalaxy is possibly slightly older than my lenses, and having a different coating.

The UV-cutoff wavelengths seams to be the same, so the optical design except for the coating can be the same.

[Andy Perrin]

Heh, I'm glad the mystery is solved now. I guess what I learned from this is that I don't want to buy a Prinz Galaxy/Prinzgalaxy without knowing the exact spelling.