I did my own conversion on a Nikon D80, including cutting my own rectangular filter into a 77 nm round hoya filter.
I did something a little unusual and I went so far as to remove the infrared filter from the AF sensor module and meters, as well as the CCD itself.
Really, the exchange of pure CCD filters is quite easy.
Pulling the filter on the sensor is a bit complicated.
Removal of the filter on the AF sensor is a nightmare – You have to take a lot All other components out of the camera, as well as several parts glued and spaced a few millimeters of the delicate circuits.
Reattaching autofocus is also quite simple. Basically, three screws adjust the distance and inclination of the sensor support. It's just a bit slow, takes a lot of shows and is rather tricky.
On the other hand, I have an infrared camera that AF perfectly in IR each time.
The meter also seems to work well. I have to dial a bit of exposure compensation, but I do not think I've changed it since. I think that adding visible light filters to the sensor (rather than simply removing the IR filter, as I did) would likely fix the measurement problem.
Really, infrared photography is really one of the nicest things I've done.
This seems to bring some of the art and imagination back to photography. You really have to think and imagine what your shows will look like, because you can not really see them again before you go to a computer, to make whiteness corrections.
I took pictures of the conversion process, including camera internal elements:
Note that the sensor filter is accessible very early in the photoset (_MG_1030.jpg, about 15 images in the format), and that the process usually consists of taking the back of the camera, unscrewing a circuit board, and then unscrewing the sensor. The rest of the set is what needs to be done to reach the AF sensor (ie remove ALL of the camera body.) It seems that the AF sensor is one of the first parts installed during assembly).
I've also destroyed a Sony alpha A-200, although I'm not able to convert it to infrared because the filter is deposited directly on the CCD's protective glass. The images are also on the link above, or here.
I thought to strip the sensor (remove the glass) or try to cover the filter (it was clearly vacuum-deposited on the surface), but I decided that it was not worth it, but I still wanted to try a Nikon camera. . I still have a bag of cerium oxide after experimenting with polishing the glass for the break-in process.
If I understand correctly, the really striking color photos, as in Matt Grum's answer, usually take a filter of 900 nm or more ( a lot post), rather than 720 nm. In addition, the photo above has not undergone any treatment, except for a simple balancing of whites and clicks. The weak blue colors come from the sensor itself.
Anyway, I recently bought a Canon 5D, going from a Canon 30D (the Nikon was an experiment, I decided that I liked Canon's UI much more), so I'm going probably convert the 30d to IR soon because I have a lot Canon lenses, while I have only one lens for the Nikon.