I've had a problem with my Canon 50mm f/1.8 II lens for some time now. The lens didn't autofocus accurately. I thought that I was the problem, moving the camera between autofocus and shooting, but after eliminating the human factor by using a tripod, the issue persisted. Then I found this video:
Please note that the video was not made by me.

There is not much information about the video and I didn't find any similar tip on the internet. The video tells about some unusual way of calibrating autofocus via connecting some soldering points. Despite no guarantee it works, I was very tempted to try it.

Here is the focus test before the calibration:
http://dl.dropbox.com/u/288566/img/IMG_5586.jpg

I retried the shooting many times, refocusing to infinity and back, again elsewhere and back, and every time the result was the same. The lens focuses a bit before the object.

To compare (and eliminate any problem with the camera body), I focused the chart with my 55-250 lens:
http://dl.dropbox.com/u/288566/img/IMG_5590.jpg

You can use a guide on how to disassemble the lens, for example:
web.media.mit.edu/~bandy/rgb/disassembly.pdf
There are also video guides on Youtube.

After having disassembled the lens enough to get to the soldering points mentioned in the video, I actually had a problem understanding which two contacts to join. There were some contacts on the left side of the circuit board and some on the right side. In the end of the video there is a picture showing to join points on the right side; the guy in the video had had the some autofocus issue as me.

Now, you will probably stone me (using stones, not weed). Recently I sold both my soldering irons so the Murphy's law had me think "creatively". I searched in the circuit board where the lines from the soldering points were going. I was happy to find that they ended in two neighboring pins of a nearby chip. So I found a thin wire and I wrapped the two pins with it. Wrapped.

Wrapped.

I'm sorry but I was very eager to try this and at the same time it was a non-destructive method where I could undo it easily if it wouldn't work. The lens was so deeply taken apart and I was so close to resolution that I had no other choice as to wrap those pins with a wire.
dl.dropbox.com/u/288566/img/IMG_5613.jpg

I then cut away the excess wire, cleaned the glass and blew off the dust and put the lens back together. I might have been lucky here because according to guides on the internet it takes much time and patience to put the guts back into the cylinder, however I made it under 5 minutes.

I mounted the lens back onto the camera and immediately turned the camera on, hoping not to see any error message. Everything started as usual, autofocus motor playing all its well-known music hits when focusing. I used the same method - focus chart - as before the modification. I know you won't believe me but this time, the autofocus works accurately! I made MANY MANY MANY testing shots, still not believing that this worked, but aside from how weird and unbelievable way of calibration it was, it just worked.
dl.dropbox.com/u/288566/img/IMG_5617.jpg
dl.dropbox.com/u/288566/img/IMG_5628.jpg

I could not believe that such a lame attempt to fix a lens just worked. Later I bought a soldering gun and soldered those points on the circuit board, removing the wire from the chip.
I am attaching the image, sorry this time with Slovak description.
dl.dropbox.com/u/288566/img/2012-02-14-22.29.28b.jpg
White text: Join these two if the lens focuses behind the object.
Yellow text: Join these two if the lens focuses in front of the object.

This awesome lens (especially after calibration and stopping down to f/4.5 on a fullframe DSLR) produces images of just crazy laser-sharp pixel level details. For its price it's perhaps the best value lens and one of the most popular lenses ever made.