I noticed it while hand holding and using AI servo to lock focus. I have pretty steady hands, but I really hope that I'm just an idiot and it was because of the hand holding. I haven't tested it using a tripod yet but I will tonight. I was hand holding because I was doing some AF testing that involved movement. When I noticed the inconsistency I did the tests hand holding still but not moving the camera or the subject. I'm still hoping it was just shaky hands! But I felt like it would be more honest to write it, I was pretty vocal about my 35 being so great...
AI servo isn't designed to lock focus, it tries to achieve focus quickly and constantly refocus to track action at the cost of accuracy. You shouldn't expect AI servo to get accurate focus in low light, switch to one shot in low light situations, it's a night and day difference in those situations.
If that's how you have set up your Servo you're doing it wrong. I can keep servo on a target for half an hour if I want to and move a little back and forth, or stay still and it's always locked right. I never ever use One Shot.
I do the same with my AI S, it even stayed on target when I jumped at it and pressed the shutter mid-air!
I got some worrisome results
I used normal objects as targets in all cases, not printed targets, but that shouldn't matter. See:
3/11 FF 5-10mm
6/11 BF 7-10mm
10/11 BF 20mm (??)
PF: 8/11 (72.7%)
5/20 0-2mm BF
9/20 misF (prev.FP)
15/20 slight FF
17/20 slight FF (same target as 15)
18/20 PF (white on white? really? that works?)
1/30 PF (white on white)
5/30 misF (??)
11/30 misF (previous FP) *
12/30 massive FF (what's going on?) *
13/30 misF *
14/30 PF *
15/30 PF *
16/30 PF *
17/30 misF *
18/30 misF *
19/30 misF * ** these were all overexposed, I forgot to
adjust when moving to the brighter areas
26/30 misF (prev FP)
PF: 23/30 (76.6%)
Wow that's bad. I admit I didn't always choose the perfect (most contrasty) targets, but this from a lens that manages to AF in low light mid-air no problem.
The previous were all on a tripod. I just tried handheld (AI S), 8/11 (72.7%) in the dim room. That's the exact same result as with the tripod (with AI F), and I used different targets and the camera was in a different place.
It's never been that bad before, it's been reliable even in lower light levels. Now I'm a little worried
How can the AF just suddenly get so much worse? Or did I just try to focus on impossible things? What really gets me is that it nailed all the white-on-white targets
. That doesn't even make sense (unless it nailed them because they were the only things emitting enough light?)
All tests were done wide-open. I was hoping it would just be a problem wide-open but I tried AFing on my black lens cap on bubble wrap on white in the dim room and it front focused
even though i had the middle point at the edge of the cap so there must have been plenty contrast.
I should note that both rooms have like the dimmest lighting ever. Many normal homes are probably a lot brighter. This lens still focuses perfectly in bright light. It only seems to have problems in low light.
I love you Sigma 35... Please don't give up, stay with me...
edit: I talked it over with Viggo and realized that's not bad results at all because the light I tested in is actually terrible! (I just never realize how terrible because I'm used to it (really need to replace the light bulbs!))
In bright light it works great so I was just worrying for nothing