3
\$\begingroup\$

I want to compare and quantify the light transmission of various lenses (with some custom modifiers). So I have a target setup with constant illumination, and a camera body for them ready to shoot RAW with a fixed ISO, aperture, shutter speed, and distance.

Suppose I shoot some comparison images controlled in this fashion. Now, I want to pick some points on the target and see, relatively, how bright they are through each lens. How should I do this?

I could open the images in Adobe Camera RAW and get a HSL lightness value using the eyedropper. Is the Adobe "luminance" number linearly proportional to transmission? E.g., if lens A has luminance of 90 and lens B on the same point has luminance of 81 then is B transmitting only 90% as much light as A?

What if I want to express this in terms of relative Exposure Values or Stops?

And are there any other variable that I would have to control to fairly compare the transmission of different lens configurations?

\$\endgroup\$
3
  • \$\begingroup\$ You probably want to read about T-Stops, which are much more common in the video scene. T-Stops represent the actual light transmission of the lens, as opposed to f-Stops, which represend the depth-of-field capabilities. \$\endgroup\$
    – Vertigo
    Commented Nov 2, 2015 at 8:51
  • \$\begingroup\$ It really depends on what you are trying to find out and how accurate you need the results to be. Can you be more specific about what you're trying to find out by having the numbers/comparisons? \$\endgroup\$ Commented Nov 2, 2015 at 11:47
  • \$\begingroup\$ @JamesSnell - My current project is comparing spotting scopes. Using resolution charts I can get their resolving power, but I also want to be able to characterize their relative "brightness." Whether that's in relative luminance, T-Stops, or whatever, I just want objective numbers that can be given some intuitive meaning. As long as it's more accurate than what a human eye could discern, looking at two scopes side by side, it would be fine. \$\endgroup\$
    – feetwet
    Commented Nov 2, 2015 at 13:45

2 Answers 2

5
+50
\$\begingroup\$

In a nutshell: no, you cannot use your camera to accurately measure light transmission. The problem is with the way your ISO control works on a digital camera. ISO is not a generic "amplify signal by X amount"; it is designed as a compensation adjustment to match the exposure index of the equivalent film stock. Therefore ISO 100 with one lens is not the same as ISO 100 with another. The ISO setting is far more esoteric than most people realize.

Let's say you were shooting at ISO 3200 at f/1.4 and 1/100, and with this camera, anything above ISO 3200 looks really terrible. If a lens with terrible light transmission was shot at those settings, the ISO portion of the exposure would boost the signal amplification to match whatever it determines "ISO 3200" should be at, meaning you're introducing a ton of signal noise. If on the other hand you shot with a lens with great light transmission at the same settings, you'd will still have the same exposure, but it would look cleaner: this is due to lower signal amplification.

Even worse, although ISO is standardized, you can measure it in a number of different ways; this means light transmission values will vary across different sensors, and not in a linear fashion. And this is just the tip of the iceberg; ISO values are computed in a myriad of complex ways that take into account the sensor, the sensor size, the resulting brightness of the image, and all sorts of computational processing.

Instead of using a camera's imaging sensor to measure light transmission you need to use something designed to measure actual light reception, not gather light into an image.

Edited for clarity.

\$\endgroup\$
3
  • \$\begingroup\$ Wow, that is a great answer! Now I see I shouldn't have cheated the question: I'm actually holding the lens constant also and taking pictures through spotting scopes, trying to measure relative transmission of various scopes. Does this ISO variation carry through to different images through the same lens? If so I guess this approach doesn't work. If not, feel free to add another answer to address that variation of the original question. \$\endgroup\$
    – feetwet
    Commented Nov 3, 2015 at 23:21
  • 1
    \$\begingroup\$ Ah, I see. If you're only going for relative comparisons of spotting scopes, if you keep everything else the same (camera, light source level, lens, exposure settings), you can get a decent comparison across each one. It won't be the most accurate, but you can at least get a relative comparison. Measuring light transmission can get really crazy really fast: here's a post that discusses it. \$\endgroup\$ Commented Nov 3, 2015 at 23:34
  • 1
    \$\begingroup\$ One more thing about the luminance portion of your question: on a relative luminance (0-100) scale, it is indeed linear, so your lens a vs lens b comparison would be correct. However in RGB space (0-256) you must convert to the RL scale to linearize it. There are various calculations on how to do this. \$\endgroup\$ Commented Nov 4, 2015 at 0:41
2
\$\begingroup\$

As has been said, for most lenses you can find the T-stop measured on dxomark, depending on the camera used. For example find the 70-200 VRI Nikkor here: http://www.dxomark.com/Lenses/Nikon/AF-S-VR-Zoom-Nikkor-70-200mm-f-2.8G-IF-ED-mounted-on-Nikon-D300__440

To answer your question more specifically, you can either compare the histograms for full image light transmission information, or place a linescan with a software of your choice through the image (ImageJ is what I would use) and compare numbers, which you can translate into stops.

You do need to take vignetting into account, so you can also transfer images in Photoshop from polar to rectilinear and then measure along one axis. However, this might introduce faults due to the re-calculation of pixels.

The vignetting will also be in the linescans, seen as falloff in the scan.

\$\endgroup\$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.