UV/IR response of modified D70s

On this page, I discuss the UV and IR response of a Nikon D70s camera modified as discussed here. I am unable to provide quantitative data on the spectral response of this camera, because I don't have a light meter calibrated for such a wide range of wavelengths. In addition, I was also forced to use four different light sources for this test. Nonetheless, I am providing the exposure times needed to achieve a comparably uniform exposure with this camera, at a constant lens aperture (f/22) and with the same subject (an orchid flower). In the case of visible-light and IR pictures, these exposure times do allow a partial comparison of relative sensitivity because the pictures were taken with the same light source (albeit the intensity spectrum of the light source would also need to be known for a more precise comparison).

The lens used in this test is a Rodenstock UV Rodagon 60 mm f/5.6, which is designed for near-UV photography. It also happens to display little focus shift across the entire range of wavelengths used in this test, and therefore the setup used in the test required no adjustments, other than changes in exposure time. I used a few filters to restrict the range of wavelengths reaching the camera sensor:

• Schuler UV filter, transmitting near-UV in a rather narrow range centered at 360 nm. It also has a very low, but sometimes troublesome, peak in the near-IR.
• B+W 486 filter, which cuts UV, IR and a little of the deepest red. I use this filter to restrict the spectral sensitivity of the modified camera, roughly bringing it back to its original sensitivity.
• Hoya R72, an IR-pass filter with cutoff at 720 nm.
• No brand/no model filter, with cutoff at 820 nm.
• No brand/IR 950 filter, with cutoff at 950 nm.
• No brand/no model filter, with cutoff at 1,000 nm.

In addition, the camera was tested also without a filter. In this case, as well as with all IR-pass filters and the B+W 486 filter, the light source was a 40 W incandescent lamp.

The three UV pictures were taken, respectively, with a 395 nm LED lamp, and a HID lamp and a UV-C fluorescent tube producing a mixture of UV and visible light. The physical placement of these lamps resulted in different shading of the subject and background. The HID and fluorescent picture also required an increase in brightness in post-processing, because of the low UV intensity of these light sources in the band passed by the Schuler filter.

The following pictures (a detailed view of the same flower, at f/22) were converted to BW by turning the color saturation to zero. This is the best way to convert pictures to BW while avoiding biases due to different dominant colors. All other picture properties, including contrast, were left unchanged. This allows for a better comparison of the actual contrast and detail level of different pictures.

Oddly, the UV pictures (including the color ones in the preceding series) seem to convey an inverted relief of the petals, with respect to the other pictures. In the UV pictures, the petals look convex toward the camera, not concave like in the other pictures. This is particularly obvious in the BW pictures. Apparently, this is due to the fact that both the visible/IR and UV light sources were placed above the flower, and illuminated both the front and the back of the petals. While visible and IR light easily shines through the petals (thus changing the light modelling of the relief), the petals are instead opaque to UV, and only UV light hitting their front is recorded in the picture. Some of the observed differences may also be due to UV light reflecting differently than visible/IR from the front of the petals.

An obvious application of this camera is in sun-lighted landscapes. The enhanced sensitivity allows the simultaneous capture of visible and IR information with short exposure times. In the above example, which combines visible and IR, note the blue sky, pink/purple trees, ripples on the water and birds standing on a rock - all these things are impossible to capture at the same time with an unmodified camera.

Results

This camera is not very sensitive in the near-UV. The Schuler filter further absorbs 40% or more of the available UV, and transmits a little visible indigo. This is the most likely cause of the indigo tones obtained with HID and fluorescent light sources. The LEDs used for this test, instead, are essentially monochromatic.

With an incandescent light bulb as light source, the modified camera is about 15-20 times more sensitive to near-IR than to visible light. The preceding statement needs to be qualified, because I don't know the relative contents of visible versus near-IR light produced by this source. Nonetheless, these relative contents should be fairly constant among 40 W incandescent light bulbs for domestic use. Not surprisingly, in this test the IR washes out visible light information almost completely. The color pattern of the flower, very obvious in visible light, is now barely visible. The subsequent picture, taken with a 720 nm filter, shows no such pattern, which therefore must be absent in the IR.

The 720 nm filter gives a pink/mauve cast, and some color information is still present (e.g., the small bluish diseased spot top-left of centre). This may be due to some visible light "leakage" of the filter and/or to a differential spectral response of the camera to different IR wavelengths (the spot is visible also at longer IR wavelengths, see following pictures). Exposure time is doubled with respect to the picture taken without filter.

The 820 nm filter gives a bluish cast. No color information is visible. Exposure time is only marginally higher than with the 720 nm filter.

The 950 nm filter requires a five times exposure increase with respect to the 820 nm filter, and the 1,000 nm filter a ten times increase. The longer wavelengths penetrate more deeply into the subject and give it a slightly more translucent effect, but the difference is not substantial. With the 1,000 nm filter, we are probably close to the lower sensitivity limit of the camera. With the last filter, the lens is also beginning to show a detectable focus shift. I don't have filters with longer cutoffs to test, but probably the results with such filters would not be substantially different.