astrolucida Offline Upload & Sell: Off
|
Fred Lindsey wrote:
I tested an 18-200 OS in Jessops a while back to see what it would be like in terms of performance for travelling.
I WAS SHOCKED AT HOW BAD IT WAS. The centre didn't touch my 70-200. The corners well. I don't know if I tested a bad one but considering I was comparing 200mm F2.8 to the sigma OS at F11 I would say it was conclusive.
I tested both lenses at 200mm, at the same apertures, using a 30D. Even wide open at f6.3 the Sigma was sharper than the 70-200f2.8L IS. Imatest results showed, for the whole frame, 1860 for the Sigma and 1818 for the 70-200, so not a big difference, but considering the price differences of the lens, a significant one. In the corners the difference was larger, with Sigma again being sharper.
However, the situation reversed at 135mm, where the center sharpness was about the same but Sigma lost clearly at the corners (75% of the sharpness only). At 100mm, again the center sharpness was about the same while in the corners Sigma had only 50% of the sharpness of the 70-200!
When compared with the 17-55, again the center sharpness was almost the same (90-100%) but in the corners only 20-30% of the sharpness, except at 18mm, where it got 85-100%. Even worse, at the midpoint between center and corner, the Sigma had only 50-70% of the sharpness - thus the very sharp area was extremely small in the center.
The lack of sharpness in Sigma seemed to be astigmatism-related. At least it was blurry in an ugly way, not just soft, but somehow distorted. Thus, even if it showed equal sharpness in the center to the 17-55 and 70-200, at the same apertures, it was very soft at some focal lengths, for most of the image area.
This is why I sold the lens. It was a very nice low light 18f3.5 OS and near-macro 200f6.3 OS lens, but I could not stand its low image quality as an 35f11 OS lens! Thus, a very uneven performer, restricting the available apertures far too much. This does not seem to be a calibration issue, because others have experienced similar performance.
|