Here are some results from the August 10th issue of ColorFoto (Germany) for longer tele-zooms tested on a Canon 20D. The evalutation gives points for 3 focal lengths based mainly on resolution/contrast (the higher, the better; max=150pts). The last figure is the grand total.
Sigma 80-400mm OS
: 80mm/69, 180mm/72, 400mm/66.5, total=69
this test-lens was 5% decentred
: 100mm/73.5, 200mm/72, 400mm/69, total=71.5
This test-lens was 14% decentred, meaning that
1. Even a Canon lens can be decentred
2. The Canon could have yielded better resolution-values had it be well-centred
3. Still it's better than the Sigma.
Canon 70-300mm IS
: 70mm/73.5, 120mm/73, 300mm/67.5, total=71.5
This test-lens was 11% decentred, and came out clearly better than its "DO" cousin (62.5 total pts.) that costs twice as much.
: 100mm/71.5, 170mm/72.5, 300mm/67.5, total=70.5
this test-lens was 7% decentred. It costs more than double that of the Canon 70-300 and has no stabilization, which more than compensates the larger aperture of the Sigma.
As an asside for this Canon section, but nonetheless interesting: In combination with a Nikon D200 the results of the Sigma were:
Sigma 80-400mm OS
: 80mm/70, 190mm/69, 400mm/59.5, total=66
this test-lens was 15% decentred (!)
Although the Sigma was mounted to a body with higher resolution capabilities the results came out worse! This should be attributed to the quite heavy decentering, as everything else should be more or less equal.
So again the ugly quality control issue raises it's head: not only with Sigma (again) but also with Canon (max decentering 16% in a fixed focus lens) and Nikon (max decentering 19%!)
What I didn't get though with this test: why didn't ColorFoto take a 10MP-Canon body, to make it a little more comparable to the lenses tested on the Nikon D200 or the Oly E400
And no, there were no lenses tested with 500mm focal length