Guy,
I can see where an update in firmware can help in recalibrating a lens whereby it no longer extends beyond the -20 (or + 20 for that matter) with regards to the AF fine tune adjustment in a particular body. Yet when there is asymmetry and optical misalignment causing causing softness on one side of a image, I can't imagine anything helping except an actual physical adjusment of the lens.
Dave (D&A)
That would seem intuitively correct but I wonder of sometimes it might work. These asymmetries can come and go according to really very slight changes of focus, or rather, be masked as the inclusion or exclusion of subject areas into areas of marginal DOF changes. We don't expect sharp edges on a lens like this until f4 for sharp-ish, f5.6 for sharp and f8 for really sharp, and so slight asymmetries can sometimes be largely veiled at wider apertures by general peripheral de-focus and largely hidden in DOF at f8 and possibly f5.6 so in my experience it is f4 and 5.6 that is the danger zone and even there, if the asymmetry is slight, there's a chance that an AFFT can deal with it.
Do a Reikan test on some lenses and the top of the curve can be quite flat. F&R zooms, for example, you might find that the short end needs a plus 12 and the short end a plus 6, but that both of those values are at the apex of a flat-ish part of the curve, in which case a plus 9 might work for both ends. So it is with mild asymmetries. The joker here, of course, is that you can't easily use Reikan with this bug. However, there is some stuff on the Internet to the effect that specifically this bug affects only the second and subsequent sigma lenses you enter an AFFT for: so if you Delete Saved Value and then start again, you can make the camera forget that there was ever a previous sigma lens and think only about your current one.
I am sure that most techs, when faced with a mild asymmetry, will try and tune it out with firmware and see if that makes the customer happy...
EDIT: adding sample images after using Guy's technique for getting round the FW bug.
Here are two shots that show pretty clearly what I mean. They are handheld at F4 and f5.6 and then a focus mask is added in Capture One. This camera is now accurately adjusted to +20 for my new copy of the Sigma, which was previously showing a very slight weakness on the left.
Observations: firstly, ignore the tower, it isn't in great focus on either shot (better in F5.6). Focus was on the nameplate of the little monument, centre. The grass 'mask' is pretty accurate but at the very granular level, the lens focusses
very slightly closer on the left. I have confirmed this somewhat with other tests but need to do some with extreme rigour. Nonetheless I am fairly confident.
Note that the field of focus is croissant-shaped, as is very often the case. What this means is that there IS a line, parallel to the sensor, in a perfectly adjusted lens, at F4 and tighter, where everything is in pretty good focus in the same plane as the subject. But if your AF FIne Tune is off and the croissant is placed too far forward in the field, then the edges of that line will fall into the defocus area. getting AFFT just right has shifted the croissant backwards so that all of that parallel line is 'in the zone' but this is very very finely judged. For example, this lens now passes a brick wall test, just, because I have placed the croissant as far back as I safely can (the lens still focusses pretty well on centre wide open). However if I switch to Live View AF, the camera moves the croissant forwards and the edges of the wall, even at F5.6, are a little out of focus and slightly more so on the left because the slight decentering means that the croissant is twisted very slightly closer to the camera on the left.
My conclusion, subject to further testing at other distances, is that this lens does need attention but more because the AFFT value required is +20. However, an adjustment to the slight decentering would mean that I could notch the fine tune forward a touch and get slightly more reliable results in overall use. For me this shows that careful AFFT can under certain circumstances 'hide' slight lens asymmetry and possibly even negate the need for a repair. Given that almost no lenses have perfect alignment of all elements and groups, it's a judgement call as to when the effect can be worked around and when it needs fixing. SO I will be interested to see what happens with Guy's.
All this goes to show that to get the very best results from these systems requires quite an anal retentive approach, a lot of thought and quite some work! Luckily, few subject 'shapes' will show an effect like this noticeably. Regarding the fact that Live View tends to place the croissant too far forward, I will experiment with using magnified focus on a point half way to the edge: that should get a good compromise.