I did a lot of shooting with this lens last night and this morning. In my normal conditions shooting environmental portraits (focusing distance less than 10 feet) the focus seems to very accurate. If there's any problem it's at a distance. I did a focus test with a chart and it was perfect (again, from a short distance).
My findings thus far reflect what badlydrawnboy has found. I did tests from 1.4-4.0 in half stops at about 4m. Perhaps refocusing a few times gets it a more accurate focus?
How far away were you from the subject? Have you tried taking any shots at 1.4 from a distance of more than 10 feet? That's where people seem to be having the most trouble.
I did a lot of shooting with this lens last night and this morning. In my normal conditions shooting environmental portraits (focusing distance less than 10 feet) the focus seems to very accurate. If there's any problem it's at a distance. I did a focus test with a chart and it was perfect (again, from a short distance).
Chris
Mine is maybe the opposite. It front focuses pretty bad on a 450D (the only Canon I have available). This is confirmed with a controlled test. I have used the center AF sensor only. The results vary somewhat depending on the lens starting position; when set to minimum distance before focusing the results is a bit worse than if the lens is set to infinity whe half pressing the shutter button. Testing distances have been 0.5-4 meters.
Interesting comparison. This photo from that link demonstrates the interesting bokeh characteristics of the Sigma. The foreground bokeh is harsh with sharp edges and doubling of elements, whereas the background elements are very soft.
It isn't as nice as the background but it's still pretty good.
f/1.4
Backfocus is bad, and doesn't improve between f/1.4 and f/4 I doubt it would change with smaller apertures either. It's only visible at large distances, so once the focus mechanism passes a certain point the issue occurs, probably meaning it's a software issue in the autofocus algorithm in the lens cpu.
Excuse the white balance issues, these are raws converted in lightroom with no processing (like the shots above) to jpeg at 80% quality.
I cropped out the centre rectangle from the following shot, lots of mess to show focus. I had the focus point set to the spraycan again and re-focused several times before each exposure
EDIT: I just tested and every frame at f/1.4 is sharp until the focus distance gets to about 10' so my initial can tests must have been below 10'. the 5' setting is perfectly sharp.
It's almost like the focus markings and values are wrong, maybe it changes gearing or moves a different element for the more distant focusing and these values were put into the cpu wrong.
Ok here are uploads at each focus marking on the lens up to 10' (the last shot).
Thanks for performing these tests, which confirm what I've also noticed with my copy. Focus is perfect with short distances, but misses badly from a distance.
The question now is, can Sigma fix this when a lens is sent back in for "recalibration"? It sounds like this may be a more serious flaw in the design that is not likely to be corrected with recalibration. If that is the case, I will return mine (reluctantly - it's such a great lens otherwise!) to B&H so I don't get stuck with it.
Ye gods, that is astonishing, the sudden shift at 10ft. I wonder if this behaviour also occurs on other people's lenses?
It scarcely seems possible that Sigma could have failed to spot an anomaly like that during prototyping and pre-release testing. And we do know that the guy from Japan who posted his early photos -- Chez Wimpy -- had no focusing problems at all, at any distance. So at least one fully functional example of the model exists.
So I think it's more likely to be a production glitch involving something mechanical, or a wrong electronic component value, rather than a faulty algorithm.
Mine also seems to have the same problem. I'm away from home this weekend so can't do a controlled test until Monday, but in casual shooting I noticed exactly the same thing as Thrice.
I wonder if Chez Wimpy had a chance to test focus from 10+ feet between f/1.4 and f/4?
Took it out for a quick shoot, and even with me compensating for the back focus by MF after one-shot it's so hard in the dark coupled with such a terrible distance scale.
I'm likely to do the same as badlydrawnboy as I'm in Australia and the distributor here won't touch imported Sigma lenses so I may have to send it back to B&H... also reluctantly.
f/1.4 ISO100 1/10s <10'
f/1.4 ISO100 1/8s >10'
f/5.6 ISO100 1.5s >10'
f/1.4 ISO100 2.0s <10'
f/1.4 ISO100 1/10s >10'
You see I missed focus on every one of the f/1.4 >10' shots. I was saved by the smaller aperture on the third shot.
I had a look at Chez's shots but hard to tell when he's over 10' and whether or not he hit his mark using AF. The "landscape" sorta shot it's even harder to tell cos the plane of focus is so far away it gets effectively larger, so it may look like he hit the house but the tree behind and the hills in the distance look equally or more sharp to me.
Thrice, sorry to say, that looks like AF hell. I would have guessed that the Sigma 50, having apparent residual spherical aberration and lacking a floating element, would be optimized for a certain f-stop and focus distance (which might vary from copy to copy if QC were not good) and not perform very well outside of that sweet spot. However, the samples from Chez Wimpy's copy show great performance near and far, wide open and stopped down, so there goes my theory . Based on his results, Sigma has accomplished something extraordinary with this design. Still, I'm not willing to take my chances on this lens. I went through that with the Sigma 30, and it was a real pain in the neck.
C.W., I know how you feel about the Sigma 30 and the Sigma 50, that people find problems because they go in looking for problems, etc. In that case, please consider selling your copy of the Sigma to me at a small premium and then buying another copy for yourself. Only joking - I sold my Canon body. Otherwise, I'm not sure it would have been a joke request .
On a serious note, it's still very early days for the Sigma 50. Hopefully it will prove to be a less controversial lens than it's crop sibling!
asabet, it is possible I, and several just got very unlucky with a certain batch. Maybe we should try and find some correlation between serial number and back focus? I got 1001407.
Planning to speak to my friend @ Canon, and the Sigma distributor boss whom I'm on friendly terms with and see if they can help me out.
Otherwise I guess I'll send it back to B&H for refund and purchase locally through the official distributor so I can get it calibrated free. Given that I've spent more than USD50 on shipping to me already, and will likely have to pay around USD30 to ship it back, the cost of calibration would have to be well over USD100 for me to consider sending it back and waiting.
I just did a test and can confirm that I have the same backfocus issue at distances over 10 feet that Thrice has. FWIW my serial # is 1001680 - pretty close to Thrice's.
The biggest question here, though, is whether Sigma will be able to fix the problem if we send it in for recalibration. Any theories on this? I don't want to send it in to Sigma only to find out they didn't fix it, and then be stuck with it because I've missed the return window at B&H.
When I bought this lens I was also considering the 35L. As much as I like the Sigma, I'm leaning back towards the 35L because 1) I generally prefer that FL on FF and 2) it seems there's a lot less sample variation with that lens.
I e-mailed Sigma as well, asking the hard questions like:
a) is sigma aware of the issue
b) is the issue repairable via calibration
c) does sigma know the cause of the issue
In as polite and professional a tone as possible You get a few business days with B&H anyway. If it's fixable via calibration I have done the maths and it's worth just paying for the calibration here, through official sigma agent or another capable service center. Might want to request an RMA number from B&H in the meantime? So you don't go beyond the DOA period.
) with interesting results:
.