Reikan Photography Blog: Length vs Quality - Canon 70-200 f4L IS vs Sigma 120-400OS

A few months back I bought myself a Sigma 120-400OS lens - previously, my longest lens was the Canon 70-200 f4L IS, and having gone to a full-frame camera I realised that 200mm was not actually that long. I was offered an excellent price on the Sigma, which meant the Canon equivalents were coming on for 4 times the price, so in the interest of testing out, I opted for the Sigma.

The Canon 70-200 f4L IS lens is renowkned for being one of Canon's sharpest lenses. The Sigma 120-400OS is a decent lens, but with a book price of about 70% of the 70-200 f4L IS, it is not sold as being a superior quality lens. Since owning the Sigma lens I've often wondered if an upscaled image from the Canon would match the Sigma in quality - I'm talking about a super-sharp image at 200mm from the Canon lens doubled in size (or quadrupled in area), against what could be a softer image from the 400mm of the Sigma. So I decided to do a quick test.

I will mention that this point that it is a quick test - I'm an engineer by trade, and there are a lot of flaws with this test. But it answered the question that I was investigating first of all - that being "is there a clear winner among these two"?

The moon was out, and it struck me that you can't get a much better target for testing a lens (well, other than a lens target, but I didn't have one of those lying around, unlike the moon which happened to be out and about the night I did the test!). I stuck the 7D on a tripod (7D as if I was genuinely going to take a shot of the moon, that's what I'd use as the image would be spread across more pixles than with the 5D2), and took some shots with the 70-200 f4L IS, and then similar shots with the 120-400OS.

I'll be honest, I'm not sure I was really awake when I did the testing. Part way through the test I realised I'd taken the 70-200 shots in JPEG, so I switched to RAW for the 120-400OS pics, and on reviewing the metadata I found that I'd in-fact had the 70-200 at 169mm rather than 200mm (yeah, I know, don't ask!). Whilst this doesn't equate to any kind of valid test now, it does put the disadvantage firmly in the 70-200's side of the court, so with the added fact that it was going to have to be scaled 2.38 times to compete against the 400mm image, I definitely thought I was going to have to redo the test (and I will have to - I just didn't want the data I have so far to go to waste).

Anyway, enough waffling from me - here are some results:

The image above is 6 slices of the same portion of the moon, all scaled to the same size. They are taken with the 2 different lenses, at different apertures and different focal lengths (hence the scaling). You'll need to click on the image above to view the 100% image, but if you fancy a test, then have a look at them and decide (before reading below) which one you think is sharpest...

OK... here are the details. From left to right:

70-200, 169mm, f8
70-200, 169mm, f5.6
120-400, 214mm, f5.6
120-400, 214mm, f8
120-400, 400mm, f8
120-400, 400mm, f10

I should point out that this wasn't the clearest of nights - it wasn't thick with cloud, but there was a definite cloud layer, so these images should not be thought of as the best these lenses can produce. You can notice the effect of the cloud layer as a halo around the moon, more pronounced on some images than others.

From a processing point of view, I did what I could to each of these images to bring the best out in them. They have had levels adjustment, some contrast changes and sharpening as I saw fit - different for each image. Now I know that a lot of people will complain about this, but I'm quite frankly not interested in what the lens is capable of, or the camera is capable of, but what the combination of both of those and the processing software is capable of - after all that's what makes an image that you can print and sell.

Finally, I need to come clean about the focussing. I initially used live view contrast detection AF to focus as this takes out any risk of mirror-to-AF-sensor error (the error that you correct with microfocus adjustment), but I found that I could do a better job by manually focussing using magnified live view. However, I can't remember exactly which images I used which technique on - I know I focussed using one method or the other for sets of images and from that point didn't adjust the focus, but I can't absolutely guarantee that they were all focussed using the same method. This, in itself, is probably the biggest reason to rerun this test.

And, my conclusions of this somewhat dodgy test:

I think my personal best is the 120-400 at 400mm, f8. This pleases me because it means the 400mm is worth something, and also I tend to only use this lens at f8 as the reviews I read before I bought it suggested at f5.6 (max aperture) it was very soft.
The 70-200 f4L IS is a sharp lens - even scaled up nearly 2.5 times, it is moderately difficult to see the differences between the images. I will redo this at 200mm properly and make a better assessment.

I will re-run this test and spend more time doing it to get a more representative result, and hopefully it will be a clearer result, but who knows? The 70-200 f4L IS is a superb lens, and is considerable lighter than the 120-400OS and easier to handle, as well and possibly faster focussing, but it's also nice to see a cheaper, 3rd party lens produce quite usable results under the typical conditions of use.

And here are the images used to create the slice image above - they are all full-moon images, scaled to approximately the same pixel size. Click on each image to view at 100% pixel size: