Lens transmission in the UV - latest update

[JMC]

Since getting into UV I've been interested in finding a way to do lens transmission measurements - something to give me % transmission as a function of wavelength, to allow absolute comparisons between different lenses.

 

I tried the straightforward approach of using a fiber and collimating lens as the input, and then a fiber a collimating lens as the collection setup, but got some weird results as the test lens focussed the light. This can result in values above 100% transmission, which doesn't make sense.

 

Next was an integrating sphere, but with my little Ocean Optics fiber and light source, I struggled to get enough light to get any measures from the sphere at all, even without a test lens.

 

I decided to stick with the integrating sphere (to collect all the light coming through the lens), but needed a more powerful light source. In the spirit of bodging something together, I made a box from wood laying around in my workshop, and painted the inside black. I made a hole at one end, and fitted a Bausch and Lomb quartz/calcium fluoride condenser lens over that. I made a moveable baffle inside, and fitted a Thorlabs adjustable iris to that, so I could vary the diameter of light beam if needed. Then at the other end of the box fitted my integrating sphere, with a 1" hole for incoming light, and a collection port at 90 degree for the spectrometer (Ocean Optics FX).

 

The light source is my LC8 Hamamatsu light (quite 'peaky' as it is a mercury xenon lamp, but lots of light between 300nm and 400nm). The aim is just to be able to get good data between 300nm and 400nm. Longer wavelength than this and there are other people that can generate the data for me.

 

This is what the setup looks like.

 

 

 

 

 

 

I got it all finally assembled yesterday and did some initial tests. Firstly, light on and light off but no test lens in place;

 

 

With the light on I get about 100% transmission, and with it off 0%. Not too noisy (this was 20ms per scan, 50 scans, and a boxcar width of 3).

 

Next some lenses;

 

 

I tested my Rayfact 105mm, as that is the only one I have some manufacturer data for in the UV. According to the manufacturer data, it goes from about 76% at 400nm to 78% at 300nm, so relatively flat, but goes up at shorter wavelength (data from Rayfact, although the PDF says 'data values are design values. It is not a guaranteed value'). With my measurements it looks to be about 83% transmission between 300nm and 400nm. Not the same as the data from Rayfact, but not a million miles away, and importantly, showing good transmission at 300nm as expected.

 

The other lenses were ones I had laying around and at hand to test.

 

As expected the Canon 50mm macro is pretty rubbish, and the 40mm pancake version is better (I have seen this time and again). The Meritar 50mm and Soligor 35mm were quite similar. The Prinz Galaxy seemed to let more light through than the previous 2.

 

I will emphasize this - this is very preliminary data. It certainly isn't a finalised method. After collecting it, I realised I hadn't optimised the collection fiber for signal, so I may be able to get better signal to noise. Also, there is a weird dip around 333nm, and a bump around 365nm which occurs across the lenses. I'm struggling to believe this is a real signal, and more likely due to the nature of the light source (being very peaky). Also I didn't move the internal baffle closer to the integrating sphere for the smaller lenses.

 

Nevertheless I think the approach has some potential, so as time allows I will try and optimise it further.

[dabateman]

Excellent!

This looks really good. For your Rayfact you are exactly at its specification. 5% tolerance error allows for that. Also all the brochures I have seen guarantee a minimum transmission of 70% for that lens. So you have a good copy.

You maybe for getting about the 335nm and 365nm Hg peaks. If I smooth your curve in my mind it looks like those spikes are causeing your increase decreases. Fit a line through the curve at all points and it may be more representative.

However, I have no idea whats going on at 308nm. The 330 area I can rationalize with peak suppression, but there is no 308nm line I am aware of, and its bump is more than that of the 365nm line.

 

[JMC]

Hi David. I think it has potential, but as of now I don't know all the issues with it (such as what is that 308nm bump as you mentioned). I do have a different bulb for my Hamamatsu light source, which doesn't have the mercury peaks, but gives out much less light in total. That may be worth a try.

 

It was a hard thing to build. Deceptively simple, but difficult to get right. If anything it's been a harder challenge than the sensor sensitivity measurement.

 

I'm also not sure that the condenser lens I have is ideal in this current setup (perhaps the focal length is not ideal). I may try some of the Thorlabs collimating lenses at some point.

[ulf]

Congratulations to a promising test setup.

Looking forward to following the progress.

 

The 308nm bump and the levels below 350nm for the two worst lenses (Canon) are most likely crosstalk in the system, either in the spectrometer or the wooden box.

The high intensity of the peaks scatter around inside the spectrometer and give false readings at other areas of the detector array.

I agree with David that the Hg-peaks cause the disturbances in the graphs around 335nm and 365nm.

 

I would have expected that the Prinz Galaxy 35/3.5 and the Soligor 35/3.5 FA were more similar.

If I remember correctly the two of those lenses I have show a quite similar fall-off slope.

[SteveE]

Very very nice!

 

I too have an integrating sphere and have also found that a DH-2000-BAL is nowhere near bright enough for lenses.

At the moment, I am investigating using a modified AD-200 for a light source. If that does not work out, I may fall back on my original idea of using cloudless midday sky for illumination. We have a lot of that here.

[Andy Perrin]

Ulf:

I would have expected that the Prinz Galaxy 35/3.5 and the Soligor 35/3.5 FA were more similar.

If I remember correctly the two of those lenses I have show a quite similar fall-off slope.

Could that be due to accumulated gunk (dust, residues, etc.) on one or both of the lenses? Or maybe an issue with the aperture not being set quite right (like dabateman's flawed tests the other day)? The curves Jonathan measured were similar in shape.

[dabateman]

Ulf:

Could that be due to accumulated gunk (dust, residues, etc.) on one or both of the lenses? Or maybe an issue with the aperture not being set quite right (like dabateman's flawed tests the other day)? The curves Jonathan measured were similar in shape.

 

They look to differ by about 1/3 of a stop or less. So yes look very similar and could be down to left over coatings or aperture closer. But I do think Jonathan is testing everything wide open. Past results you remember may have been flawed testing them at f4? Which then if one didn't close properly would acount for the exact difference.

[ulf]

Ulf:

Could that be due to accumulated gunk (dust, residues, etc.) on one or both of the lenses? Or maybe an issue with the aperture not being set quite right (like dabateman's flawed tests the other day)? The curves Jonathan measured were similar in shape.

 

They look to differ by about 1/3 of a stop or less. So yes look very similar and could be down to left over coatings or aperture closer. But I do think Jonathan is testing everything wide open. Past results you remember may have been flawed testing them at f4? Which then if one didn't close properly would acount for the exact difference.

 

There are many things that can affect the result and my memory about this or earlier tests might be wrong too.

 

The curves seam similar in the short end. Both are directed at ca 315nm.

The thing I reacted about is that the Soligor line is steeper in the wright side of the diagram even if it is lower.

That indicates that the lens design might be different. I believed they are similar.

 

However these are the very first results and future refinements might give different results if there are any methodical flaws.

[Timber]

Meritar is being one of my favourite lenses for UV, it has 2 version, same UV performance but the silver one has MUCH better build quality than the zebra. Unfortunately it's only easy to find with EXA mount...

[JMC]

I'll be working on this method, on and off, when I get the time, so will provide updates when I can.

 

I've got a Xe only lamp for the Hamamatsu light source as well HgXe one that was in there for the initial experiment. The Hamamatsu LC8 light source is sold in about 5 different options, each designed for different jobs. However when I spoke with them, I was told that I can use the Xe lamp in my unit, although the output would be expected to be lower than from one designed specifically for it.

 

I tried it, and yes the output was much lower, I needed about 400ms per scan instead of 20ms with the HgXe lamp. However it gives a relatively even output, without the mercury peaks in the spectra. It also gives out light up to about 700nm.

 

A couple of scans below from todays quick experiment. 2 runs of the Rayfact 105mm f4.5 UV and one of the Meritar 50mm f2.9. First between 300nm and 700nm.

 

And between 300nm and 400nm.

 

Above 400nm the line gets noisy, although each scan was only an average of 2 scans, and with a boxcar averaging of 3. The 2 runs for the Rayfact were pretty close to each other. The Rayfact values are lower than with the mercury xenon lamp, but more in keeping with the manufacturer data.

 

The scans no longer had the odd features around 333nm and 365nm, which confirms the idea that it was a feature of the peaks of the mercury xenon lamp.

 

Overall the Xe lamp is better than the HgXe lamp. However it does require much longer running times, which may to lead to an issue with baseline and lamp intensity drift.

 

The adjustable iris inside the device is useful, but easy to knock and adjust by accident. I need to make sure I can lock that, or replace it with a fixed aperture.

 

The biggest issue is holding the lens in place. I need a set of adjustable stands that fit inside the box. Ideally I'd set the whole thing up vertically, and just sit the lens on the integrating sphere, but I'd need a flexible light guide for that, and I don't have one at the moment.

[ulf]

Jonathan, this looks very promising indeed!

 

I think the reason for not reaching zero below 310nm might be due to scattering / crosstalk in the spectrometer from longer wavelengths.

I assume that you have done a proper dark level calibration, eliminating the system offset.

[Andy Perrin]

I think the reason for not reaching zero below 310nm might be due to scattering / crosstalk in the spectrometer from longer wavelengths.

Overall the Xe lamp is better than the HgXe lamp. However it does require much longer running times, which may to lead to an issue with baseline and lamp intensity drift.

Emphasis mine. Maybe this is the issue? Despite the dark level calibration initially, what if it drifted? I know my TriWave has to do the dark level calibration every few minutes or it starts misbehaving, and that camera is cooled, so it maintains constant temperature with feedback.

 

On the other hand, I know several people (JCDowdy especially) have mentioned scattering / crosstalk issues in UV-C as a major cause of error.

[JMC]

Andy, Ulf,

 

Yes dark level was calibration was done. I've seen issues before with stray light effects, increasing apparent signal to the short wavelength end of transmission curves, and I think this is a problem. It's a known issue with CCD spectrometers.

 

The Meritar scan in post #10 was done immediately after calibration. I shall make sure I do this in future as well, as I was seeing drift over time.

[ulf]

One way to decrease the scattering problem with an array spectrometer is to limit the wavelength range of the light source in a suitable way.

 

To study the transmission in the limited range of ca 270nm to 370nm I would have placed an UG11 in the light path before the spectrometer.

Take care to not overheating the filter if it is placed close to the light source.

From that point of view the best position would be after the integrating sphere.

[ulf]

I made a new measurement with my pair of lenses, the same types as Jonathans's above, Prinz Galaxy 35/3.5 and the Soligor 35/3.5 FA.

I will start a new topic about those measurements.

 

To show the effect of scattering, discussed above, I measured with and without the tungsten lamp activated in my DH-2000-BAL.

The intensity of the halogen lamp is adjustable and set a bit low to prolong it's life.

 

The intensity variation as seen by the spectrometer is not very even, but quite normal. Smoothing: 3 pixel boxcar.

It is much less peaky than the first lamp Jonathan mention above in post#1.

Even here some crosstalk can be seen at the left bottom corner when the magenta graph with the tungsten lamp light is added.

These curves are the ones used to normalise against and get 100% when measuring the light-source alone

 

The measurement of my Prinz Galaxy 35/3.5 at f/3.5 with the two alternative light-source configurations:

 

The offset is less than seen in Jonathans measurements because the method I use gives a max level out from this lens of just 25%.

This scales the offset to ca 1/3, compared to what would be seen if the max level seen from the lens were 75%.

The offset also depend on the intensity of the off band light causing the offset and the light-sources are different.

 

The reason for the lower level than seen in Jonathan's measurements is due to the different methods.

Without a strong light source and integrating sphere the measurements become affected by the different kinds of magnification and it is difficult to collect all light passing the lens.

This actually makes sense, but makes it impossible to measure the absolute transmission in the lens.

The cut off characteristics found is still valid.

[JMC]

I would have expected that the Prinz Galaxy 35/3.5 and the Soligor 35/3.5 FA were more similar.

If I remember correctly the two of those lenses I have show a quite similar fall-off slope.

 

Hi Ulf, I just did a quick test on the 2 lenses outside in the sunlight. Using my UV converted Nikon d810 (ACS filter). Nikon to M42 adapter. Both lenses run about 30s apart. F5.6 for both. ISO2000. Handheld. Saved as jpg and RAW. Image of 20% Spectralon diffuse reflectance standard on the ground.

 

Soligor (KA serial number) image (jpg from camera)

 

Prinzgalaxy image (jpg from camera)

 

If I open the RAW files in RAW digger as RAW composites, and extract the red, green, blue and green 2 channel data for the 20% diffuse reflectance standard, and then average them for each lens, I get the following values;

 

Soligor KA serial number.......4678

Prinzgalaxy lens....................8613

 

The Prinzgalaxy lens is letting through about 1.8x as much light as the Soligor (both at f5.6).

 

As a rough guess the lens transmission work suggest about 1.5x more light through the Prinzgalaxy compared to the Soligor. Mine aren't similar to each other - much more UV through the Prinzgalaxy.

[ulf]

OK, confirmed then that for the same 1/f-setting your lenses transmit different amounts of light.

It is a rather big difference, not likely to be caused by internal contamination.

I do not think that the UV-cutoff point is very different.

The might instead be caused by an incorrect reassembly, leaving the aperture settings in the wrong position.

I tinkered a bit with some lenses like these, two years ago.

I have a hazy memory of that this might happen sometimes if you are not careful.

[JMC]

Ulf, Thankfully I have 2 Soligor 35mm lenses, one KA serial number, which has been tested here up until now, and one H serial number, which I haven't done any transmission testing on yet.

 

I just got the 20% diffuse reflectance standard out again and did a similar test to post #16 above.

 

Looking at the average of the channels for the two lenses I get;

 

Soligor 35mm H serial number........1350.4

Soligor 35mm KA serial number......1534.4

 

Absolute values are different as it's a different time of day to the last test. But the lenses are very close together in their scores (both were tested at f5.6), with the KA serial number one letting slightly more UV through at the same settings. It was only about 1.14x more though, so I wouldn't read much into that without more testing.

 

I consider it rather unlikely that both my Soligors have been worked on and incorrectly re-assembled, so it does look as though my Prinz Galaxy 35mm lets significantly more UV through than the Soligor(s).

[dabateman]

Jonathan,

I now find the aperture to be sticky with old lenses. Before setting a specific value. Cycle through all values twice. My UAT measured differently after doing that.

Also,

Since you have them in hand and they are the same focal length. Hold them next to each other and look down from the front of the lens. At the same f5.6 setting the aperture blades should look to be closed to the same amount. If not than you know something is up.

Ulf, you should do the same. Yours may be sticky or different.

[JMC]

David, Done a visual check and the two Soligors look to be similar at the same f stop setting. I tend to run my lenses through from open to closed before using them.

 

A quick update on the method. I've made a stand to put the test lens on so I don't have the hold it in place. Slightly awkward, as it means I have to make a new stand for each lens. As mentioned before, having the whole thing upright would be easier for positioning the lens. I've tested it with the Rayfact 105mm, and it helps. it keeps the values very stable, and allows me to take it out and reposition it easily. I shall make a few more now for different lenses. When I have more results I shall share what I find.

 

I have to do a baseline correction and calibration immediately before each run, as I'm getting drift in signal over the course of a few minutes. Slightly awkward, but I'd rather trust the data.

 

I need to check my UAT and UV Sonnar as well as the El Nikkor 105mm f5.6 when I get chance. It'd be nice to compare the Rayfact and UV Nikkor too. My real life testing suggested the Rayfact was letting slightly more light through than the older UV Nikkor. Lot's of stands to make.

[JMC]

Today, a UV lens shoot-out on the transmission measurement.

 

I've again modified the method slightly. I now run a baseline with the lamp off immediately after the lens and then subtract that from the lens data. Helps deal with the drift I've been getting.

 

4 lenses;

Rayfact 105mm f4.5

Asahi UAT 85mm f4.5

UV Sonnar 105mm f4.3

Coastal Optics 105mm f4.5

 

 

As expected the Rayfact 105mm lets the most UV through, and is flat across the range.

 

The UV Sonnar is also a fairly flat across the range but at a lower transmission level than the Rayfact. Perhaps not a big surprise - the Rayfact is 6 elements in 6 groups, the UV Sonnar is 7 elements in 7 groups so more loss would be expected.

 

The Coastal Optics 105mm lens has more of a wave to the transmission profile, and is lower than the Rayfact in the UV. When I tested this lens before I noted it had a lower transmission than the Rayfact, and that is confirmed here. Here's the link to my earlier test - https://www.ultravio...dpost__p__23317

 

The wavy transmission profile for the Coastal Optics 105mm lens is also mentioned by the manufacturer.

 

The 85mm Ultra Achromatic Takumar is interesting. Transmission is almost as good as the Rayfact at 400nm, but then gradually drops as the wavelength gets shorter, and with a pronounced dip in the profile at around 308nm. I've seen this type of dip before when looking at transmission through calcium fluoride windows. Certain grades of CaF2 give a dip around there.

[dabateman]

Thank you for the data. Too bad you don't have your 60mm f3.5 macro. I bet it would have the worst transmission even though f3.5 and only 4 elements. But its a fun macro lens.

[Andy Perrin]

Yeah, I would love to know the transmission of that one, if you take requests, Jonathan! :D

[JMC]

I'll be doing that one, but not until I get it in November. I'm planning on running as many of my lenses as I can.

[JMC]

A few more lenses to add to the data, using the reworked method (including baseline subtraction).

 

Canon EOS 50mm f2.5 macro

Canon EOS 50mm f1.8

Canon EOS 40mm f2.8 pancake

 

Soligor KA serial number 35mm f3.5

Prinz Galaxy 35mm f3.5

 

El Nikkor 80mm f5.6 old chrome version

 

 

The Canon lens data backs up what I have seen before, with the 40mm pancake being the best I have for both transmission and how far into the UV it goes. The 50mm f3.5 macro is the worst.

 

The El Nikkor 80mm f5.6 enlarger lens is very very good, going almost as deep as my 35mm Prinz Galaxy, but with better transmission at 350nm and above.

 

Remember these are transmission graphs, all done at maximum aperture (this value should not be aperture dependent).