After having visited Sigma and got the lens adjusted I was ready for some real life use, and also some further testing: It has been discussed if there is a problem with the AF depending on the shooting distance.

Earlier I tried to see if there is any focus shift and concluded there should be no problem with a moving focus plane as the lens is stopped down (as opposite to the EF50/1.2L).

With the lens adjusted and hitting a target 2000mm away with an error between 0 and 2mm everything should be fine.

Real life use indicates there still is a problem, or at least an intermittent problem.

Here is what I trying to find out:

I picked a flat target where I taped an USAF test chart. It looks like this:
http://photos.imageevent.com/jonas_b/hometests/sigma5014/428_Siggy50_AF_test_setup.jpg

Then I took several series of images from several different distances, all in all 250+ images. To get a starting point I took images for this standard USAF test chart. I placed the camera at a distance of 55 times the FL (47mm x 55 = 2585mm) and got this:

http://photos.imageevent.com/jonas_b/hometests/sigma5014/Siggy50_450D_standard_USAF_chart_47x55.jpg

I interpret the slight misfocused chart this way; f/1.4 gives us at least 70 lp/mm, f/4 gives us at least 78 lp/mm.

For each tested distance I took five images using AF at f/1.4, then 2x a series of AF going from f/2 to f/5.6. After that I focused manually to get something to compare to and then, still in MF, I stopped the lens down step by step to f/16 to see if any focus shift threw everything out of the window.

1: 0.5m

The 5 AF f/1.4 images were all very similar to each other:
http://photos.imageevent.com/jonas_b/hometests/sigma5014/710_0.5m_f1.4_AF.jpg

Thereis no point in "measuring" the resolution here as the laser printed USAF chart isn't up to the task. We are somewhere around 40-45 lp/mm. The point is that all the images, AF and MF, were pretty much the same. Here is the result from the manually focused image (focusing wide open using Live View) exposed at f/4:
http://photos.imageevent.com/jonas_b/hometests/sigma5014/194_0.5m_f4_MF_at_1.4.jpg

We are looking at the resolution of the laser printer.

The series of images going from f/1.4 to f/16 were all the same; no image got blurred due to focus shift.



2: 1.0m

The 5 AF f/1.4 images were all very similar to each other, and the result was again nearly the same as when focusing manually; 68 and 76 lp/mm respectively with a small amount of back focusing.

The series of images going from f/1.4 to f/16 were all the same; no image got blurred due to focus shift.




3: 2.0m

Resolution is now 76 lp/mm for both the AF and MF images, maybe the manually focused image is a tad better being on the verge to 85 lp/mm:
http://photos.imageevent.com/jonas_b/hometests/sigma5014/240_2m_f5.6_MF_at_1.4.jpg




4: 3.0m

Something has happened. The AF results, at f/1.4, are varying between 34 and 64 lp/mm. When using AF I needed to stop down to f/5.6 to achieve 80 lp/mm (the expected resolution).

http://photos.imageevent.com/jonas_b/hometests/sigma5014/255_3m_AF_f_1.4.jpg
Above AF, below MF:
http://photos.imageevent.com/jonas_b/hometests/sigma5014/268_3m_MF_f_1.4.jpg

The series of images going from f/1.4 to f/16 were all the same; no image got blurred due to focus shift.

Hmm. So, here we have the AF error people have been talking about. What I see in my images i a back focusing problem.




5: 4.0m

With AF and f/1.4 we are now down to 42 lp/mm. Using AF stopping down we get the best result at f/5.6 (as long as I measured using AF) with 75 lp/mm.
Using MF (and Live View) f/1.4 gives us 75 lp/mm while stopping down to f/4 reaches the expected level of 85 lp/mm.

So, AF at a distance of 4m gives us a resolution worthy of a 3MP camera, but worse (as the per pixel sharpness is worse and some CA is added as well).

Back focusing.





6: 6.0m

With AF and f/1.4 we can read between 35 and 40 lp/mm, When setting the camera to f/2 and f/2.8 we get the same and at f/4 we reach values between 45 and 50, at f/5.6 we get 60-64 lp/mm.
The manually focused images shows again between 70 and 80 lp/mm at f/1.4 and at f/5.6 we have the expected max value around 85 lp/mm. The AF images are showing back focusing.

MF f/1.4, here we get an imgae looking the way we are used to:
http://photos.imageevent.com/jonas_b/hometests/sigma5014/319_6m_MF_f1.4.jpg

Below we have the AF version:
http://photos.imageevent.com/jonas_b/hometests/sigma5014/301_6m_AF_f1.4.jpg

The images are downsized to a stanbdard web format, 800 pixels on the long size and there is no PP. Already here we can clearly see a difference. The crops show the back focusing problem we are used to now:

http://photos.imageevent.com/jonas_b/hometests/sigma5014/301_6m_AF_crop_f1.4.jpg

and

http://photos.imageevent.com/jonas_b/hometests/sigma5014/314_6m_MF_crop_f1.4.jpg




7: 8m, 12m, 24m, 32m

AF, f/1.4 typical 30-40 lp/mm
MF, f/1.4 70-75
AF, f/5.6 typical 60-65 lp/mm with the exception 32m where I got 75
MF, f/4 80-85 lp/mm

The error was always a back focusing problem and when using AF I had to stop down to f/5.6 or f/8 to get the target clearly within the DOF.




Conclusion:
With a 450D and the AF adjusted by the Sigma rep AF works fine up to somewhere 2.5 and 3m. From that distance and onwards we have a back focusing problem which is pretty consistent.

There are no problems at shorter distances, at longer distances we can't trust to get a reasonably sharp image unless stopping down to at least f/5.6 with a few lucky exceptions wher f/4 was OK.

The question now is if it helps that the contrast wide open is excellent, that there is very little flare problems and that the bokeh is wonderful?

The longitudinal CA and the AF issues may be too big drawbacks.

--
Jonas