In order to correctly perform AF microadjustment, it is important to understand a little about the performance of the phase detect AF system of your camera. It’s almost certainly not as good as you think it is!
The biggest issue is the repeatability of the results. The camera manufacturers must make compromises between accuracy, speed, repeatability and performance in different situations (e.g. different light levels). They are also limited by the focus performance of different lenses.
Take the following example: if a manufacturer decided that ultimate accuracy was the critical factor, they could gear their AF algorithm towards very small step changes when comparing the phases of the two different images. They would choose to use an autofocus sensor with a very high resolution, and get great sensitivity by averaging the results over a relatively long period of time. Finally, the system would only work with lenses that were very accurately controllable by the camera. All this would lead to fantastic accuracy… but it would take a comparatively long time to reach the result on a limited subset of very expensive lenses, and in reality the system would be useless in most situations as it would be too slow or never reach a satisfactory focus due to movement of the camera while trying to focus!
Conversely, you could target speed. Take a few very quick measurements and make a best guess at the focus position. The performance could be further improved with predictive algorithms which track the change between focus operations and guess where the subject is going to be by the time the shot is actually taken – and this is what happens in a lot of high-end cameras. But you will compromise overall accuracy.
The accuracy of the complete system is dependent on all of the components:
Each time an AF operation is performed (i.e. from half-pressing the shutter button to the lens being focused), each component above has a little inaccuracy which can lead to slightly different results each time you press the button.
It would be perfectly reasonable to assume that with a good setup – bright lighting, high contrast target, static camera firmly mounted on a tripod – that the AF result across a number of shots would be identical.
Well… it isn’t!
As explained above, the point that is actually in focus is dependent on the performance of the lens getting the light to the AF sensor, the AF sensor doing its job, the AF processing system calculating the result and sending drive signals to the lens, the lens motors driving the lens elements appropriately and finally the optical performance of the lens when everything is in the final position.
If all of these elements behaved perfectly 95% of the time, then adding all those small errors together you would only achieve perfect focusing only around 70% of the time.
You can check this for yourself: Mount your camera on a good tripod. Stick a high contrast target (black and white squares generally work well as long as they are not too small) to a solid wall some distance away from the camera, under good lighting conditions (e.g. a nice sunny day). Make your setup really solid – use mirror lockup, use a remote release. And take 10 shots, but in between each shot make sure you manually move the focus to infinity so the AF system has to work fully each time (otherwise the AF system may see that the focus is “good enough” and not change it in any way).
Now inspect the images at 100% magnification. Are they all focused identically? I’d suggest not. Maybe 6 of the 10 are similar quality... maybe 7 or even 8 if you have a really good camera and setup (and some luck!). But you will almost certainly not see 10 identically sharp images.
Continue on to How does FoCal help?