As I have the three Canon lenses (100-400mm, 70-300 DO IS and 70-200IS/2.
I thought it would be useful to do a quick comparison of their performance. This was in part something I have meaning to do for a while, but was also stimulated by some interest on this forum about the these particular lenses and their relative performance.
Now, before I begin, I will put in the usual disclaimer that everyone will have their own opinions - some more strongly held than others - and I don't intend to get into a great debate about methodology or the fact that what I am seeing may be different compared with other posts on this or any other board. This is for information only, what you do with that information is entirely up to you.
The question I wanted to answer for myself was when to use each of these lenses. They all cover a similar range (particularly when the 1.4x TC is added to the 70-200), therefore, should I dispose of one or two of these lenses or do they each serve a particular niche. As usual, one of the major criteria in selecting a lens is image quality - primarily measured by the amount of detail captured (this is ignoring distortions at this point in time, though distortion has never really been quoted as a problem with any of these lenses). Going into this test I had the pre-conception that the 70-200 was top dog and that the 100-400 was crap and the 70-300DO was double crap (based upon that font of all wisdom the internet).
The test scenario was as follows:
Test camera - Canon 1DII, mounted on tripod, mirror lock up, cable release, yadda yadda.
Test subject - Opposite side of the valley from where I live combination of trees and brick walls, roofs (plenty of fine detail in both highlights and shadows).
Shooting at all apertures - though evaluated at f5.6 and f11
Processing in ACR - Same processing parameters, same White balance, exposure, etc...no sharpening in ACR or noise reduction.
Sharpening done in Photoshop using a two step sharpening algorithm:
USM 1.6/75%/0 followed by USM0.3/500%/0 (This is my typical DO lens fixer uppper, but applied it consistently to all images to make the comparison a bit more even - some of the images may be a bit oversharpened, but I wanted to see if there was detail that could be extracted not determine the most appropriate sharpening parameters).
Tests were conducted at two focal lengths 200m and 270mm (this corresponds to furthest reach of the 70-200 with and without the 1.4x teleconverter).
The resulting files can be found here (NB these are JPEGs, they are close to what I can see in photoshop but may have artifacts related to the compression process).
Comparison imagesObservations:
1/ Colour cast/Exposure
Relative to each other the lenses show no difference in colour cast. Other than the fact that the 70-300DO IS has a peculiar habit of compressing the image into the right hand end of the histogram (useful for ETTR afficionados) there is little to differentiate between these lenses.
2/ Captured detail
First, and most important comment, on my teeny 8 Mpixel camera the differences bewteen these lenses are so slight as really makes no difference. Yes, there are some differences, however, relative to other problems in making images (being in the right place at the right time, sharpening technique, composition, exposure, etc...) these differences are so small as to be not worth worrying about.
Now for those that like to pick nits.
Centre Performance:
At 200mm and 270mm at f11 there is virtually no perceptible difference between any of the images. You could argue that the 100-400 is slightly up and the 70-300DO slightly down but on an 8Mpix camera the difference is slight. If I was using a 1DsII - or the yet to be announced zigapixel 1DsIII then this difference may be greater. You may wonder why the 100-400 is slightly up on the 70-200? at the tested focal lengths the 70-200 is at its extreme and at 270mm I had the 1.4x converter on; if the test was repeated at 100mm then the 70-200 may have beaten the 100-400 - perhaps a test for another day.
However, once the aperture is opened up to f/5.6 things change. At 270mm the 100-400 starts to soften up quite quickly, the 70-300DO stays unchanged (surprised? I was) and the 70-200 plus teleconverter is the worst of all three. At 200mm both the 100-400 and the 70-300DO soften up (the 70-300DO is worse in the shadows cf 100-400 but better in the mid-tones - see note below), the 70-200 maintains a consistent performance compared with f/11 and, therefore, exceeds the performance of the other two lenses.
Corner Performance
Again at f/11 there is very little to choose bewteen these lenses. The 100-400mm is perhaps slightly better otherwise not really a great deal in it. The 70-200 plus teleconverter shows aberrations/purple fringing against highlights in the corner, its noticeable but not distressing.
As the lenses are opened up to f/5.6 both the 100-400 and the 70-200 start to soften up and whereas the 100-400 had a slight lead at f/11 I would say that the 70-300DO has perhaps a nose at f/5.6 in the corners (another surprise result - this lens does not appear to soften between f/11 and f/5.6).
Perhaps I am not picky enough or I need a now camera to show all the faults, but I can't get too excited about the small differences that exist between these lenses and all of them can be recommended in various situations. So, what did I decide?
1/ The 70-200IS/2.8 is my working lens when I need a telephoto with wide aperture for close work/low light (typically portrait in a controlled environment).
2/ The 100-400 when I need to get extreme reach and need flexibility from 200-400mm. I would agree with Michael that this lens is overdue a replacement, though I would suggest a compact 200-400 with a ring zoom with a newer IS would complement the rest of the range better than a 100-400mm with a push me pull you.
3/ The 70-300DO fills in when I need to be out and about and inconspicuous. The Image stabilisation on this lens is significantly better than the previous generations on the other two lenses.
Otherwise, if I need to bump up the image quality then it is time to think about primes again.
Note : An observation for the technically minded.
One of my observations is that the 70-300DO performs worse in shadows than it does in mid tones in terms of retaining detail. If anyone has a better rationalisation than the one I am about to write below then I would be interested in hearing it.
All lenses cause light to disperse by the time it hits the sensor - in DxO terms they measure this spreading in terms of Blur Units. Typically this spreading can be removed/minimised with a bit of unsharp mask. A technical term that I can think of is that the spreading of the light is a function of the Point Spread Function (PSF) of the optical system, and provided that this spreading is small and gaussian in nature then light USM is sufficient to recover detail in the image. If the PSF is small (relative to the signal) and the system is noise free, then it should be possible to recover the original image perfectly if the inverse PSF is known - exactly what I believe the DxO software does.
However, in the case of the 70-300DO/IS the PSF spreads the image over a larger area (smaller signal per pixel over larger number of pixels), therefore, if the signal is weak (dark areas of the image) then the signal gets lost in the noise of the sensor relatively quickly and shadow detail becomes unrecoverable. For mid tones the signal is much greater than the noise and, therefore, it is statistically much easier to recover the original signal accurately.
If this theory holds true, then the net result is that as pixel density increases and sensor noise decreases (with improvements in technology) then it should (with quality software like DxO) be possible to recover more detail produced by the lens as camera technology improves - particularly in the shadows. The other point to note is that this lens should only be used at low ISO, as at higher ISO more of the signal will get lost in the noise and become unrecoverable.
If nothing else, I have given you something else to think about and debate
