The Great 400mm Canon Shootout
Posted: Tue Mar 18, 2008 2:09 pm
Several months ago I promised a sharpness test of the various 400mm options in the Canon lineup. My efforts were delayed some as I wanted to be sure my 100-400 lens was operating optimally. After allowing Canon Irvine to fully optimize the sharpness of my 100-400 I was finally able to do the test. All tests were done on a Canon EOS 1Ds Mark II at f/5.6, f/8, and f/11.
The candidates are the EF400 f/4DO, EF400 f/5.6L, EF100-400 f/4.5-5.6L, EF300f/2.8L + 1.4x, EF 70-200 f/2.8L + 2x.
Unfortunately I did not have a 400/f.2.8L lens at my disposal which I suspect would win the test.
I used the popular money test. A clean $5.00 bill was flattened and taped to a white background which was in turn taped to cardboard and then taped to a wall. This insures flatness. A tripod was set up 15 feet from the subject. Everything was perfectly squared up and level. All images were taken at 1/60 sec with mirror lockup and bounced flash to eliminate any exposure variables - all shots received at least 93.75% of the light from the flash.
The results are not very surprising and my visual observations are 100% correlated with file size. The images that I found to be more detailed visually also had the largest CR2 and JPEG file sizes. This is due to the fact that a more detailed file is less compressible than a less detailed file.
The conclusion:
At f/5.6, the EF400/5.6 is the clear winner, only the 400DO challenges it for supremacy presumably due to it being stopped down by one stop where all other candidates are wide open at f/5.6. A significant step down and nearly identical are the 100-400 and the 300/2.8+1.4x. The 70-200 with 2x is awful relative to the others losing almost all fine detail.
At f/8 the order is the same but the 300/2.8+1.4x is almost on parity with the 400/5.6 and the 400DO. The 100-400 is still significantly behind and the 70-200+2x has closed the gap to the 100-400.
At f/11 all lenses and lens TC combos should be performing as close to optimum as possible. The 400/5.6 is still the winner with the 400DO and 300+1.4x coming in tied for a close second. The 100-400 is significantly behind the top 3 but also significantly better than the 70-200+2x. It is also interesting that the 70-200 with 2x falls off at f/11 in image quality. The test was repeated three times with identical results so this is not an artifact of the testing but an actual condition.
Overall, this tests rules out, for me anyway, the 70-200 +2x as a viable option and confirms my previous experience with many different versions of the 100-400. The 100-400 is not up to the level of the 400 primes or even the 300/2.8+1.4x even on a finely tuned sample. Certainly I am not ruling out the possibility that there are 100-400 lenses out there that are better than the sample used here but it is most certainly not the norm and is likely rare. Analyzing the histograms also shows that the 400DO has slightly lower contrast than the 400/5.6.
Another observation also confirms my field experience. The 400/5.6 can be shot wide open with almost no fall-off in image quality.
The following are the 100% swatches all unsharpened and have only gone through an identical RAW conversion in ACR 4.4 and have been cropped. They are in order of their finish in the test:
The candidates are the EF400 f/4DO, EF400 f/5.6L, EF100-400 f/4.5-5.6L, EF300f/2.8L + 1.4x, EF 70-200 f/2.8L + 2x.
Unfortunately I did not have a 400/f.2.8L lens at my disposal which I suspect would win the test.
I used the popular money test. A clean $5.00 bill was flattened and taped to a white background which was in turn taped to cardboard and then taped to a wall. This insures flatness. A tripod was set up 15 feet from the subject. Everything was perfectly squared up and level. All images were taken at 1/60 sec with mirror lockup and bounced flash to eliminate any exposure variables - all shots received at least 93.75% of the light from the flash.
The results are not very surprising and my visual observations are 100% correlated with file size. The images that I found to be more detailed visually also had the largest CR2 and JPEG file sizes. This is due to the fact that a more detailed file is less compressible than a less detailed file.
The conclusion:
At f/5.6, the EF400/5.6 is the clear winner, only the 400DO challenges it for supremacy presumably due to it being stopped down by one stop where all other candidates are wide open at f/5.6. A significant step down and nearly identical are the 100-400 and the 300/2.8+1.4x. The 70-200 with 2x is awful relative to the others losing almost all fine detail.
At f/8 the order is the same but the 300/2.8+1.4x is almost on parity with the 400/5.6 and the 400DO. The 100-400 is still significantly behind and the 70-200+2x has closed the gap to the 100-400.
At f/11 all lenses and lens TC combos should be performing as close to optimum as possible. The 400/5.6 is still the winner with the 400DO and 300+1.4x coming in tied for a close second. The 100-400 is significantly behind the top 3 but also significantly better than the 70-200+2x. It is also interesting that the 70-200 with 2x falls off at f/11 in image quality. The test was repeated three times with identical results so this is not an artifact of the testing but an actual condition.
Overall, this tests rules out, for me anyway, the 70-200 +2x as a viable option and confirms my previous experience with many different versions of the 100-400. The 100-400 is not up to the level of the 400 primes or even the 300/2.8+1.4x even on a finely tuned sample. Certainly I am not ruling out the possibility that there are 100-400 lenses out there that are better than the sample used here but it is most certainly not the norm and is likely rare. Analyzing the histograms also shows that the 400DO has slightly lower contrast than the 400/5.6.
Another observation also confirms my field experience. The 400/5.6 can be shot wide open with almost no fall-off in image quality.
The following are the 100% swatches all unsharpened and have only gone through an identical RAW conversion in ACR 4.4 and have been cropped. They are in order of their finish in the test: