I recently picked up a Sigma 10-20mm f/4-5.6 ultra-wide-angle lens for my Pentax K-5 II DSLR. On the first day that I took it out, I tried using a 1000x neutral density filter to blur water in a stream. The results were disappointing:
Even though the light was pretty dull, I attributed the bright horizontal streaks to lens flare from the cheap, uncoated filter, and I didn’t think much more about it.
Today I realized that I never tried my cheap Polaroid 720 nm IR filter on the K-5 II. The results were pretty bad on my K-7, because that camera had terrible noise in long exposures. But since the K-5 II has much better noise characteristics, I thought it might be worth trying.
I shot the Sigma 10-20 at about 30 seconds, f/11, ISO 100. I created an IR-optimized profile in Adobe DNG Profile Editor, and the results were encouraging in terms of noise and appearance:
There’s that streak again! I decided to investigate it further. My findings:
- The streak occurs in the same place on the sensor, with two different (admittedly cheap) filters.
- The position of the streak does not depend on the zoom setting of the lens, suggesting it’s not lens flare.
- An equally long exposure in a dark room with the lens cap on does not produce the effect. So it’s not a sensor problem like amplifier glow.
Then a hypothesis hit me: The problem is caused by light entering the viewfinder during the exposure, and bouncing around into the sensor. It made sense: A horizontal edge (like the edge of the mirror, maybe) could cause the linear streak. And since light coming in the viewfinder bypasses the strong filter on the lens, and the small lens aperture, even if not much light makes it from the viewfinder to the sensor, it could still be substantial compared to the image-forming light.
Pentax Hoya Ricoh includes a viewfinder cover with the camera, and they probably have a reason for doing that.
I tested the hypothesis, and it was a great success:
These images were processed identically, so why the big difference in color? A quick check with RawDigger provides the answer:
The left (open-viewfinder) image has higher RGB counts throughout the image; the stray light isn’t just in the bright streaks. Since the stray light is unfiltered, that’s a lot of green and blue light that’s not present in the IR image, and it dramatically affects the color rendition.
The moral of the story: When using a strong filter that blocks a lot of light (whether an IR filter or a strong neutral density filter), cover the viewfinder to prevent stray light from ruining the image.
This post is dedicated to @thingnamer.