My lens comparison graphs are based on data from lens tests meticulously performed by Klaus Schroiff in his excellent Reviews at his stellar site PhotoZone.de. Klaus tested all these lenses under identical conditions on his own Rebel XT, and all results were obtained through the same exact processing, i.e. RAW-to-JPEG conversion and Imatest analysis - and Klaus is a very exacting German engineer who I believe has kept the margins of error pretty tight. (See his Lens Test FAQ for more details.)

The only real caveat when making these comparisons is the reality of sample variation. Hopefully the tests are representative of at least the middle (or upper) 50th percentile of the respective lenses. The world may never know. But I tend to think Klaus's results are right on track -- the general consensus of photo community reviews do seem to me to support these data as indicative of a typical "good" copy of each of these lenses.

Some of these graphs include data from John Nevill's meticulous Lens Tests at his very cool site EOS-Images.com - these data are shown by DASHED lines. Like Klaus, John performed all his tests under identically standardized conditions, so comparing within John's data set should be completely legit as well. Interestingly the PZ & EOSpix curves seem to match up pretty well most of the time, but other times apparently reveal the great extent of lens sample variation.

HOWEVER, absolute comparisons BETWEEN these two data sets of course involves some fudge room for the differences in testing/analysis conditions and human error. A different RAW convertor was used, but John says the settings he uses bias the results to be in line with the sharpness output of Klaus's PZ tests. Other factors skewing these two data sets differently may include different lighting, slight sensor differences between the 20D and 350D, and slight differences in testing chart design and the regions of interest (ROIs) chosen from each test image for Imatest analysis for Center, Border, and Extreme Border MTF50 and Border CA. So when you compare between these data sets, try to think in relative (e.g. trends) rather than absolute terms. And don't forget the salt! :)

PLEASE NOTE that none of these graphs is to be used or distributed for any commercial purpose. Klaus and John naturally want to protect their ownership and the accuracy of their hard-earned test data and prevent them from being distributed without credit given to the sources. So they have given me permission to freely share these graphs but not the tabulated raw data files. But if you have specific advice on improving/correcting the data presentation, please let me know via email at lincolnwestfall AT gmail DOT com.

BTW, if you want to exclude certain lenses from any chart to yield a cleaner, personalized version for your own use, just open the chart image in Photoshop, select those specific curve colors, and replace them with the background gray! (Thanks to DPR user Gordon S Walker for the idea for this work-around!)

Thanks for visiting, and enjoy your stay! --Lincoln :)