My tests now show convincingly that my main zoom, Sigma 17-50 OS, have a problem - a pretty wide range of MA going from 17 to 50mm, and also moving from closer (~2.5m) to farther (8m) distances. This is the same behavior (only less extreme) I observed in my previous zooms, Tamron 17 -50 non-VC, and my first copy of the Sigma.
I also can see (from other zooms, and couple of primes I tied, including with freshly SC-calibrated 135L), that my camera itself (50D) is a bit off - I consistently get MA around -3 point. My theory is that because the camera is off, my zooms have this complex MA behavior. The idea (and hope) is that if my camera was mechanically calibrated to MA=0, then MA for zooms will change much less with FL and distance.
But this is a 100$ + two two-hour round trips to Toronto idea. Before I do it, can anyone who used Focal for a mix of primes and "standard" zooms (17-5x for crop, 2x-7x for FF) on a single body comment on their findings here?
Specifically, if your primes are randomly scattered around MA=0 (suggesting the body is well calibrated), how do your standard zooms behave at different FL and distance - are they well behaving? Same for if you observe a situation similar to mine (primes are clustered around some non-zero MA, suggesting a mis-calibration in the camera) - how do your standard zooms behave?
A side note - it'd be terrific if the alternative firmware Magic Lantern added support for focus MA. (For now it is listed as "highly unlikely" or "not possible"). Then I can imagine one could easily add advanced MA features - e.g. it could be calibrated at two FL extremes in a zoom, and then interpolated in-between (like 1DX bodies do), or even do a two-D interpolation - at different FL and distances. I'd be willing to donate some money if Magic Lantern would even consider working in this direction. I think there is a place on their website where you can vote for new features. Probably this one: