autofocus accuracy

Feeling bored, so I decided to test the autofocus accuracy of my lenses and camera once and for all today. Probably would've been smarter to do this a month ago when my camera was still under warranty, but I digress.

Due to its not being able to produce the shallow depth of field necessary for the chart I used, I did not test my EF-S 18-55mm f/3.5-5.6 kit lens. But if you can't see any misfocus in the chart, then who cares if it misfocuses?

The results for my other three lenses, though, are revealing. I tested each lens at its widest aperture, taking five shots with the focus manually preset at infinity, and five shots preset at the minimum focus distance. All three lenses were tested close to their minimum focus distance, due to the small size (less than letter-size paper) of the chart.

EF 50mm f/1.4 USM

This is an ancient (1993) model with Micro USM, and it shows in build quality, optical performance, and now autofocus accuracy. Focus distance varied between 5mm and 10mm of front focus, with the majority of shots near 10mm. There was no apparent correlation between focus distance and whether the starting point was infinity or the near limit.

EF 85mm f/1.8 USM

Even though this is even older (1992) than the 50mm, it has Ring USM. Autofocus accuracy was spot on every time, with all ten frames between perfect and 1mm of front focus. There was no correlation between focus distance and starting point.

EF 70-300mm f/4-5.6 IS USM

This is one of Canon's most recent redesigns, with supposedly improved focus accuracy over its predecessor - although both use Micro USM. Due to my tripod being only so high, the maximum focal length I could test on this lens was 190mm. At this focal length, the maximum aperture was f/5, which was shallow enough to read the chart, once a large amount of sharpening was dialed in. The frames starting from the near limit all exhibited between 2mm and 6mm of front focus. The frames starting from infinity varied from 1mm back focus and 1mm front focus, which is essentially spot on. All in all, a better performance than the 50mm f/1.4.

Conclusions

I find it interesting that three Canon lenses on the same Canon camera have three different autofocus behaviors.

The 50mm f/1.4 consistently front focuses. 5-10mm is not enough to make me send the lens in for calibration, but it's just one more reason to get rid of the thing when my 17-55mm f/2.8 arrives.

The 85mm f/1.8 is perfect, which is no surprise, since it's essentially perfect at everything. If only I had a full-frame camera to make proper use of its focal length.

Finally, the 70-300mm f/4-5.6 put in a fairly decent performance, but I now know to prefocus at infinity for best results. I have to do this anyway because the autofocus is too slow to lock onto moving objects unless it starts close to the right distance (which is near infinity for something like a bird in flight). So it's a happy coincidence that its accuracy is near perfect with this technique.

Overall, I have enough evidence to conclude that there's nothing wrong with the autofocus accuracy of my Digital Rebel, but I need to test new lenses more carefully when I buy them from now on. You better believe I'm going to test the $1179 EF-S 17-55mm f/2.8 IS to death.

Afterword

BTW, I kept all these lenses because after casual observation, they all had reasonably accurate autofocus, and perhaps as importantly, they all easily achieved focus lock in low light situations. I've tried (and returned) several lenses that either didn't focus accurately, had trouble achieving focus lock in low light, or both on my Digital Rebel. The models were: Tamron's 28-300mm f/3.5-6.3, Canon's EF 28-135mm f/3.5-5.6 IS USM, and Tamron's 28-75mm f/2.8.