I decided that, if I”m going to be able to compare various lenses for hotspotting potential, I need to make the lighting repeatable. That means out with the old, natural light approach, and in with a completely artificially lit technique.
I started out by creating this target:
I printed it on a piece of C-sized Exhibition Fiber paper with an Epson 4900 and lit it with a Paul Buff Einstein strobe. Using the variable output control on the strobe, I made an aperture series with the 28mm f/1.4 Nikkor-D. When I looked at the images in Lightroom, I could see right away that I had a problem. The Epson 4900 inkset is not completely opaque to infrared light that the LifePixel standard IR filter responds to.
I found a crude solution; I covered the black dot with two pieces of gaffer tape. It wasn’t a perfect IR sink either, but it was a lot better:
In order to calibrate out some of the lens transmission differences — and, in the IR region, these very quite a bit across the samples that I’ve tested — I measured both the central black square, and the piece of gaffer tape on the right in Lightroom’s percentage with the sRGB tone curve.
Here’s the raw central data:
The Nikkor seems relatively immune to hot spotting. The CO lens might have a little at some of the wider apertures, but not wide open. The Zeiss lens starts out not so good, and rapidly gets worse as you stop down.
I linearized all the Lr gamma corrected values, took the ratio of the center gaffer tape readings to the right side gaffer tape readings, and plotted the result using a log scale :
Well, the results are different, but not in a way that heightens enlightenment, at least for me. There is one possible exception: you could say that the Zeiss results say that the lens is usable at f/8 and wider.
All in all, this is not a totally satisfying outcome. There are lots of things that could be done to improve the quality and the repeatability of the results.
- Go to a backlit target, with a piece of metal or something else truly IR-opaque occluding the central area.
- Develop algorithms to measure and compensate for lens falloff.
- Use light area rather than dark ones to normalize the illumination.
I know of no simple way to compensate for different light transmission vs wavelength characteristics in the lens,
I think this could easily turn into a project that’s a lot more trouble than it’s worth. For now, I think I’ll use this test as a first cut screen for hot spotting, to be followed up by field testing if the results are the least bit inconclusive.
By the way, although I did the above measurements after conversion to greyscale, I did look at the images in LR-converted false color. Here are thre
Note the spectral responses are wildly different in the light areas, They are also different, but in different ways, in the gaffer-tape-covers areas, whre the reflected spectrum in apparently different from the incident spectrum