Working against the system?

Digital image sensors are quite sensitive to near-infrared, i.e., wavelengths just above the 700-nanometer human eye response limit. Even the microscopic RGB filters placed over every photosite to make it record only the red, green, or blue component of the image do not drastically affect that sensitivity (at least not in the red component). This means that, as long as we use an IR filter to cut off the visible light, taking infrared images should be a snap. Well, it should, but it is not.

It turns out that this infrared response may hurt the color rendition. Because 99.9% of digital camera users will never try infrared photography, the manufacturers solved the problem in the most economical way: by placing an IR-absorbing filter just in front of the sensor (usually it is combined with the anti-aliasing filter; see my article in Quest 16 on the Leica M8).

The bottom line is that this anti-IR filter blocks about 99% of the infrared light from reaching the sensor. The remaining 1% (which is still here just because it is not easy to make a filter changing the transmittivity from 100% to 0% within just a few tens of nanometers) is all what's remaining for us infrared enthusiasts to work with. No wonder that a typical exposure with the most common R72 IR filter is about 1000x longer than in visible light (E-510). This usually means shutter speeds of one second or longer, which excludes moving subjects and makes a tripod necessary.

Still, even working around the system and against the designers' intent, the results can be quite pleasing and I have a hundred or so frames to prove it.

Modifying your camera for infrared use

Fuji and Canon attempted to give us some choice here by offering versions of their models with the anti-IR filter being replaced with a flat piece of glass (to keep the optical path the same). These cameras, obviously, were never addressed at a wider audience, and, being priced significantly higher than their visible-light siblings, did not sell well; I believe they are not made any longer.

Another option for a photographer who wants to shoot in infrared without suffering excessively long exposure times, is having a camera modified. I have seen a number of small companies (or individuals) offering such services on the Web, priced anywhere between $300 and $600. These modifications are nothing else than removing the anti-IR filter and replacing it with a plate of glass of the same thickness. Really, it is the product of thickness and refraction factor which should remain the same; if not, the autofocus will have a systematic shift and you may lose the capability to focus all the way to infinity (both effects may be, if not excessively large, masked by the depth of field).

The obvious disadvantage is that the color rendition of your camera in visible light may become way off, but this is to be expected, unless you try one of the expensive and hard to find anti-IR filters which can be put on a lens. I've never tried this, but it should work.

Another minus may be that the replacement glass plate is no longer providing the functionality of an anti-aliasing filter (making the image a bit fuzzy to avoid Moiré patterns). I do not consider this a real disadvantage, though.

Most importantly, many of us may be reluctant to have a camera modified by someone without a clear track record, a kind of gamble.

Update of March, 2008: Another good news. I have collected as much information as I could about a New Jersey company, LDP LLC (better known as MaxMax.com who deal in all things infrared (and ultraviolet, too), including IR modifications of digital cameras, done in their own "clean room" facilities. Their Web side answered most of the questions I had, and the company president provided answers to (most of) the remaining ones. They also seem to gave a good track record, therefore I'm converting my old Olympus E-500 SLR to infrared-only (this has the advantage of being able to use SLR viewing). So far I've placed just a stub of my E-500IR story, where you will be able to trace the progress and my impressions.

It would be real nice if camera makers offered a special-order option, a camera variant without the IR-blocking filter. Certainly, this is not a mass-market item, but it might be good for a company image, and doing the swap at the production stage is safer and easier than afterwards. Unfortunately, I don't think any of the industry players is ready to do that. This is simply not where the money is.

(The Sony 707 and some of the models following it allowed the anti-IR filter to be moved aside; unfortunately, there was a provision to make that impossible in outdoors light. Some of clothing may be partially transparent to infrared and Sony decided to play it safe, not to offend the feelings of the family-oriented US market. Here a glimpse of cleavage slipping out on a prime-time TV becomes a state affair, but showing a guy massacring dozens of people with a chainsaw is a perfect family entertainment.)

A B&W, IR camera — a pipe dream?

For me, however, even removing the anti-IR filter is just a half-measure. If you remember, what we refer to as a 10-megapixel camera really has just about three megapixels. Because of the RGB mosaic filter, each photosite records just one component; or about 30% of (white) light reaching it. In particular, most of the IR response is concentrated in "red" photosites, i.e., ones under a red microfilter. Your infrared image is based on just these, and in the standard Bayer grid pattern "red" photosites are exactly 25% of the total (another 25% respond to blue, and 50% to green).

From a purely technical standpoint, the simplest solution would be to get rid of the RGB filter mosaic altogether. This would not only make each photosite generate a full pixel information (so that 10 "megapixels" would become 10 megapixels without quotes), but they would also become more sensitive: a base sensitivity of ISO 100 would become ISO 200 or 300.

Such a camera would become an ideal tool for a demanding B&W photographer: more resolution, higher sensitivity, and no need for an anti-IR filter, as there would be no color information to be negatively affected by infrared light. Thus, the added benefit would be a vastly improved usefulness in infrared.

Can we expect something like this? Not in a predictable future, I'm afraid. While getting rid of the IR-blocking filter is a relatively simple matter, switching to another sensor is not. Sensors are as inexpensive as they are only thanks to the huge production volume, and B&W is destined to stay a niche area. The mass market does not need or want B&W-only cameras, and this means nobody is going to have them.

Last year Kodak has revealed a new sensor type, not using the Bayer pattern: out of each 16 photosites eight are unfiltered, recording all light components, and another eight deliver the color information, using the RGB microfilters. This may be at least a partial improvement, but I have yet to see a camera using this sensor.

Olympus E-510, 14-42 mm F/3.5-5.6 ZD at 14 mm, R72 filter, 0.6 s at f/3.5, ISO 100 (+5 EV compensation)