• Ultraviolet Photography
  •  

Characterizing the Transmittance of 7 Lenses using a Simple Method

UV Camera UV Lens UV Lighting White Balance
29 replies to this topic

#1 SteveCampbell

    Member

  • Members
  • 42 posts
  • Location: Australia

Posted 27 September 2017 - 04:42

In a dark room I sequentially mounted 7 lenses on a tripod-mounted camera and photographed a 365nm UV light source, focusing on the LED. I started with the longest (200mm) lens, and moved the camera closer to the subject so that the LED was of framed to approximately equal size between photos.

All photos taken in the center of the frame @ f/5.6, 1/8000, ISO1600 on a Canon 5D mark II + UG11 + BG40, and displayed at color temperature = 2000K and tint = 0.

In Photoshop I resized the resulting photos to compensate for any minor differences. I reduced each 45px photo (row 1) to 1px to get an average of the photo, then resized back to 45px (row 2) to sample the resulting color. I normalized the RGB values to those of the 80/5.6 (chosen as this is a well-characterized lens) to obtain % transmittance relative to the 80/5.6 (graph 1, table 1). I also added the estimated brightness of each lens at their maximum apertures. Perhaps the Blue/Green ratio can give us some insight into the depth of the UV spectrum transmitted by each lens.

The technique is imperfect for several reasons, but perhaps effective enough for a simple comparison.

Attached Image: UV-Transmission-Test-1---point-values---larger.jpg

Specific lens versions:
Takumar 28mm/3.5: Super Takumar 49mm filter version
Takumar 35mm/3.5: Auto Takumar version
EL-Nikkor 80mm/5.6: metal version
Takumar 105mm/2.8 original Takumar version (model 1)
Takumar 135mm/3.5: original Takumar version (early)
EL-Nikkor 135mm/5.6: metal version
Takumar 200mm/5.6: Tele-Takumar version

I had high hopes for the beat-up old 1958 105/2.8, but alas, the RGB results seem to suggest that it's rather strongly anti-UV coated. While it's not much use to me in its current shape, perhaps the simple 4-group/4-element design might make it amenable to uncoating if I buffed each element with a fine abrasive.

Edited by SteveCampbell, 27 September 2017 - 05:33.

STEVE CAMPBELL
DeviantArt | 500px | Tumblr

#2 JMC

    Member

  • Members
  • 128 posts
  • Location: UK

Posted 27 September 2017 - 08:47

Wouldn't you need to the keep the distance between the camera and light source the same for this to be able given comparisons between the lenses (given the way the light intensity drops as a function of distance)?

#3 SteveCampbell

    Member

  • Members
  • 42 posts
  • Location: Australia

Posted 27 September 2017 - 09:06

Forgive me if I'm wrong, but I thought that to achieve a common F-stop you need a larger absolute aperture diameter in proportion to the increase in focal length, compensating for the drop in light. I assume this is why when photographing a given subject with a 400mm f/2.8 lens you don't have to change shutter speed if changing to a 40mm f/2.8 lens and shooting the same subject (assuming no disparity in T-stops).

View PostJMC, on 27 September 2017 - 08:47, said:

Wouldn't you need to the keep the distance between the camera and light source the same for this to be able given comparisons between the lenses (given the way the light intensity drops as a function of distance)?

STEVE CAMPBELL
DeviantArt | 500px | Tumblr

#4 JMC

    Member

  • Members
  • 128 posts
  • Location: UK

Posted 27 September 2017 - 09:13

You may well be right Steve, I wasn't thinking about f stops, just the way the light dropped as a function of distance. Perhaps others can offer a perspective as well.

#5 SteveCampbell

    Member

  • Members
  • 42 posts
  • Location: Australia

Posted 27 September 2017 - 10:50

To put it another way, if you've moved away from the subject, you can increase the focal length to achieve the same composition you had before (ie image height). Since the smaller angle of view of an increased focal length lets in less light, you need a larger aperture diameter to compensate. This relation between focal length and aperture diameter is where f-stops come from; increasing focal length with a proportional increase in aperture diameter results in a constant f-stop. By using f/5.6 across all photos, I think I kept the total amount of light constant.
STEVE CAMPBELL
DeviantArt | 500px | Tumblr

#6 OlDoinyo

    Member

  • Members(+)
  • 454 posts
  • Location: North Carolina

Posted 27 September 2017 - 11:47

Using a narrow-band light source means you are checking transmittance only near one wavelength (365 nm). In the real world, this is not the whole story, and the three channels will be giving you just triplicate information, not independent data. It would be better to use a broad-band source.

#7 SteveCampbell

    Member

  • Members
  • 42 posts
  • Location: Australia

Posted 27 September 2017 - 11:56

I don't think it's truly a narrow-band source, otherwise it would be impossible to explain the RGB deviation in the Takumar 28/3.5 and 105/2.8. However, I could add a 395nm light source for additional tests.

View PostOlDoinyo, on 27 September 2017 - 11:47, said:

Using a narrow-band light source means you are checking transmittance only near one wavelength (365 nm). In the real world, this is not the whole story, and the three channels will be giving you just triplicate information, not independent data. It would be better to use a broad-band source.

STEVE CAMPBELL
DeviantArt | 500px | Tumblr

#8 JMC

    Member

  • Members
  • 128 posts
  • Location: UK

Posted 27 September 2017 - 12:04

Ok, I think we may have 2 different things going on here. Yes, you are keeping the f number constant, but you are changing the distance to the light source. If you halve the distance between the camera and light, the intensity of the light goes up 4 times. So even keeping the f number constant, the LED will appear brighter the closer you get to it. I'm happy to be proved wrong on this, as these are the types of experiments I love doing myself, but it could be easily checked if you have a zoom lens you can use (even if not ideal for UV you should still see something) - set it at a constant aperture and repeat the experiment going from one end of the zoom to the other, moving closer to the camera the wider it gets.

#9 SteveCampbell

    Member

  • Members
  • 42 posts
  • Location: Australia

Posted 27 September 2017 - 13:00

I took a few photos at f/2.8 with focal lengths ranging from 300mm (8 degrees) to a 15mm Fisheye (180 degrees). Each time I recomposed the photo (more approximately this time) such that the laptop took up more or less the same % of the photo

[each photo below was taken at ISO800, 1/350th]

Attached Image: 300mm-to-Fisheye.jpg

It makes sense when you consider that the relationship between the aperture diameter and the aperture area is non-linear - that is SA = πr². At a constant f-stop, as the focal length decreases linearly as your approach the subject, the aperture area decreases exponentially. At the same time, intensity of the light source increases exponentially. The effects cancel each other out, and the overall brightness of the scene remains the same.

View PostJMC, on 27 September 2017 - 12:04, said:

Ok, I think we may have 2 different things going on here. Yes, you are keeping the f number constant, but you are changing the distance to the light source. If you halve the distance between the camera and light, the intensity of the light goes up 4 times. So even keeping the f number constant, the LED will appear brighter the closer you get to it. I'm happy to be proved wrong on this, as these are the types of experiments I love doing myself, but it could be easily checked if you have a zoom lens you can use (even if not ideal for UV you should still see something) - set it at a constant aperture and repeat the experiment going from one end of the zoom to the other, moving closer to the camera the wider it gets.

Edited by SteveCampbell, 27 September 2017 - 13:09.

STEVE CAMPBELL
DeviantArt | 500px | Tumblr

#10 Andy Perrin

    Member

  • Members(+)
  • 818 posts
  • Location: United States

Posted 27 September 2017 - 13:35

OlDoinyo is correct that those torches are pretty narrowband. (Not lasers, but still narrow.) Usually we care quite a bit about the shorter wavelengths in testing lenses, and a wideband source would be best for this. Adding a 395nm isn't going to do much -- nearly everything is good at 395nm!

Edited by Andy Perrin, 27 September 2017 - 13:37.


#11 SteveCampbell

    Member

  • Members
  • 42 posts
  • Location: Australia

Posted 27 September 2017 - 13:40

Here's an emission spectra of a similar UV LED. A google search for various 365nm LED spectra yielded a similar distribution for different LEDs.

Posted Image

View PostOlDoinyo, on 27 September 2017 - 11:47, said:

Using a narrow-band light source means you are checking transmittance only near one wavelength (365 nm). In the real world, this is not the whole story, and the three channels will be giving you just triplicate information, not independent data. It would be better to use a broad-band source.

STEVE CAMPBELL
DeviantArt | 500px | Tumblr

#12 SteveCampbell

    Member

  • Members
  • 42 posts
  • Location: Australia

Posted 27 September 2017 - 13:49

Here's a representative spectrum from a 395nm LED - like the graph I posted above, it's probably not as narrow-band as people would expect. Not quite quantum-dot precision.

The idea of the including both 395nm and 365nm sources would be to normalize each to 80/5.6 and then look at the ratio of 365 to 395 to estimate the depth of transmission into UV. Eg if the 395 source was substantially brighter than the 365nm source it would represent a relatively poor performing lens. If the 365 and 395 were relatively similar, it would represent a good performer. I would compare the other lenses to the 80/5.6's 365-395 ratio to estimate performance.

Posted Image

View PostAndy Perrin, on 27 September 2017 - 13:35, said:

OlDoinyo is correct that those torches are pretty narrowband. (Not lasers, but still narrow.) Usually we care quite a bit about the shorter wavelengths in testing lenses, and a wideband source would be best for this. Adding a 395nm isn't going to do much -- nearly everything is good at 395nm!

STEVE CAMPBELL
DeviantArt | 500px | Tumblr

#13 JMC

    Member

  • Members
  • 128 posts
  • Location: UK

Posted 27 September 2017 - 14:15

View PostSteveCampbell, on 27 September 2017 - 13:00, said:

I took a few photos at f/2.8 with focal lengths ranging from 300mm (8 degrees) to a 15mm Fisheye (180 degrees). Each time I recomposed the photo (more approximately this time) such that the laptop took up more or less the same % of the photo

[each photo below was taken at ISO800, 1/350th]

Attachment 300mm-to-Fisheye.jpg

It makes sense when you consider that the relationship between the aperture diameter and the aperture area is non-linear - that is SA = πr². At a constant f-stop, as the focal length decreases linearly as your approach the subject, the aperture area decreases exponentially. At the same time, intensity of the light source increases exponentially. The effects cancel each other out, and the overall brightness of the scene remains the same.

Well, every day is school day. I didn't realise that it would cancel out like that. Thanks for taking the pictures to explain it.

#14 SteveCampbell

    Member

  • Members
  • 42 posts
  • Location: Australia

Posted 27 September 2017 - 14:20

No worries! It was fun working through it

View PostJMC, on 27 September 2017 - 14:15, said:

Well, every day is school day. I didn't realise that it would cancel out like that. Thanks for taking the pictures to explain it.

STEVE CAMPBELL
DeviantArt | 500px | Tumblr

#15 Alex H

    Oleksandr Holovachov

  • Members
  • 651 posts
  • Location: Sweden

Posted 27 September 2017 - 14:33

Spectrometer tests show El Nikkor 135mm to be far inferior to El Nikkor 80mm: http://photographyof...ses-for-uv.html

#16 SteveCampbell

    Member

  • Members
  • 42 posts
  • Location: Australia

Posted 27 September 2017 - 14:45

If they used the chrome 135/5.6, it's a curious discrepancy. I don't think I can explain it with the information I have on hand, but would be interested in finding the answer.

View PostAlex H, on 27 September 2017 - 14:33, said:

Spectrometer tests show El Nikkor 135mm to be far inferior to El Nikkor 80mm: http://photographyof...ses-for-uv.html

STEVE CAMPBELL
DeviantArt | 500px | Tumblr

#17 Andy Perrin

    Member

  • Members(+)
  • 818 posts
  • Location: United States

Posted 27 September 2017 - 15:28

You can't just project the 365 and 395 results further into the UV. The ratio will tell you zilch. Filters behave very differently in different wavelengths, so it's meaningless to project a trend more than a few nanometers. And those graphs you posted of the UV torches are reasonable but they do demonstrate how narrowband it is! They fall by around 50% within 10nm of the center wavelength on each side.

Edited by Andy Perrin, 27 September 2017 - 15:29.


#18 SteveCampbell

    Member

  • Members
  • 42 posts
  • Location: Australia

Posted 27 September 2017 - 23:16

Virtually all UV transmission curves I've seen so far have been hyperbolic with slight sigmoidal character. Perhaps beyond any given data point may lie dragons, but I'm suggesting that if a lens performs poorly at 395 and worse at 365, it's more likely to perform poorly at shorter wavelengths too. If a lens performs admirably at both 395 and 365, is that sufficient to say that it will perform admirably at 320nm? Nope. It's a test that likely has good negative predictive value for deep UV performance, and poor positive predictive value. Afterall, the whole idea here is a screening test to exclude lenses that are already performing poorly. I'm not looking to displace the role of the spectrophotometer here - just to make available a simple method to weed out poorly performing lenses and find possible candidates for good performance.

Looking at the previously posted spectrophotometer results, the steepest (and most linear) point in the curve is almost invariably around 365. The overall transmission plus the RGB distribution may therefore give us the x-translation and gradient respectively, while overall 395 transmission could give us the flat part of the curve. Should this be shown to be the case, I would be happily convinced that my previous comment on PPV is incorrect.

While 95% of the 365nm torch's spectrum lies within a 25nm interval, it lies in a critical part of the spectrum, an available UV spectrum about 80nm in total, and of which deeper UV has little additional value due to the diminished response of silicon at such wavelengths.

View PostAndy Perrin, on 27 September 2017 - 15:28, said:

You can't just project the 365 and 395 results further into the UV. The ratio will tell you zilch. Filters behave very differently in different wavelengths, so it's meaningless to project a trend more than a few nanometers. And those graphs you posted of the UV torches are reasonable but they do demonstrate how narrowband it is! They fall by around 50% within 10nm of the center wavelength on each side.

STEVE CAMPBELL
DeviantArt | 500px | Tumblr

#19 SteveCampbell

    Member

  • Members
  • 42 posts
  • Location: Australia

Posted 28 September 2017 - 03:20

Any idea if there are any affordable variable-wavelength UV light sources? All the variable xenon sources seem to start ≈$1,000

View PostOlDoinyo, on 27 September 2017 - 11:47, said:

Using a narrow-band light source means you are checking transmittance only near one wavelength (365 nm). In the real world, this is not the whole story, and the three channels will be giving you just triplicate information, not independent data. It would be better to use a broad-band source.

STEVE CAMPBELL
DeviantArt | 500px | Tumblr

#20 Cadmium

    Member

  • Members(+)
  • 893 posts

Posted 28 September 2017 - 08:25

My favorite full range UV light source is natural sunlight.
I suggest going outside, and shooting actual photos with these lenses, and comparing them.
I would suggest that LED's are not a good way to test much of anything, because they can often provide very strange results.
Other than the ultra achromatic Takumar 85mm, I am not aware that any of the Takumar lenses are close to the superior UV transmission of the the El-Nikkor 80mm, which is one of the best non specific UV lenses.
I don't have any Takumar lenses, but I have the El-Nikkor 80mm and 135mm, so I can do a Sparticle comparison test of those two lenses, once I find my 43mm to 52mm step up ring for the 135mm.
Of all the El-Nikkor lenses I have tested/compared so far, the El-Nikkor 80mm is the best transmitting UV lens. I have not tested the 135mm version however.
I will test/compare the 80mm and 135mm next week and post the results.

For now, here is a comparison of the El-Nikkor 80mm and the new/old El-Nikkor 50mm versions using the Sparticle.
Still, not as deep transmission as the Kuri 35mm. :-)

Here I used a full spectrum modified Cannon 199A flash. It is consistent and wide range UV-A.. With Baader U on lenses.
Attached Image: Sparticle_199A_Diffuser_El_Nikkor_50p_50m_80c.jpg

Edited by Cadmium, 28 September 2017 - 08:36.