The Great 400mm Canon Shootout

Rocky Sharwell wrote:

Ron Niebrugge wrote:Thanks for taking the time to do this test.

I know my 100-400 is barely adequate at 400. Too bad, because it is such a convenient lens otherwise.

Just drop it--then send it to Canon to be fixed--Mine came back so much better !

LOL. You know Rocky I did. Well not on purpose, but I was hopeful that it would be sharper on return. Unfortunately they said it was "within specs" - there was no improvement.

Super test EJ, much appreciated! Reference quality!

Gene

Cliff Beittel wrote:

Ed Erkes wrote:Bob, Before you decide to sell your 100-400, check out past winners of the Nature's Best and other magazine contest winners. You'll find that there are a lot of winnining images made with the 100-400. It is a very good lens and you have the advantage of zoom capability.

Actually, I don't think that proves it's a good lens. It proves you can make great images without great glass. Many factors other than sharpness are involved.




Cliff, Isn't it the final result that really counts!! If you can make Great Images with it, then the lens has to be good enough. I'd be willing to bet that I could display 20 12" X 18" prints of outstanding images, 10 using the 100-400 and 10 using the 400 f5.6 and nobody would really be able to tell which were made the prime lens vs the zoom. Plus a zoom will give the possibility to zoom out if the bird comes closer or another bird flies in to interact---giving you increased chances for more Great images!

upnorthfar wrote:
I have a feeling there might be a focus issue in the 300+1.4x pic here, otherwise this lens is HUGELY overrated.

Again, these are 100% crops unsharpened - most people never even look at an image like that. rest assured that there is no focus error and that the 300 by itself is the sharpest and fastest to focus lens I own.

Just surprised of the result. Also the fact that the 300 is still benefiting from going to f/11 (more DOF) also supported somewhat theory of a focus problem but I am now assured that no such error happened. My original post to this thread was a unnecessary blunt, sorry. My own lens testing shifted to angled flat objects and comparing sharpest focus areas instead of same image areas. But my test camera was a 400D that was not yet calibrated to anything.

I know how laborious it is to perform this kind of test, so thanks EJ for generating more data points particularly on the 400 5.6L vs 100-400 IS comparisons.

My own tests with my 100-400 vs 400 5.6L differ from your results, but I of course understand the variability of copies, and even methods, used in the tests. I believe no test, even how rigorous and careful these are done, can be perfectly representative of the majority of copies of each lens model. Due to the nature of its more complex design, I also believe that performance variation from copy-to-copy would occur more often with the zoom than with the prime.

So..... the 100-400 IS vs 400 5.6L debate rages on....

Romy

liquidstone wrote:So..... the 100-400 IS vs 400 5.6L debate rages on....

I actually don't think its much of a debate
As you state, the complexity of the 100-400 is much higher than the 400/5.6. Its really more of a debate of how much if any worse is the 100-400. I think unless you have a really bad copy of the 400/5.6, it is going to be better in the majority of comparison.

I'm getting more "heat" about the result of the 300/2.8+1.4x at f/5.6 - many owners of that combo feel or at least want it to be sharper than the 400/5.6. At least my version of this combo is not at the wide aperture end. I am fairly certain that its the TC, not the lens though.

E.J. Peiker wrote: As you state, the complexity of the 100-400 is much higher than the 400/5.6. Its really more of a debate of how much if any worse is the 100-400. I think unless you have a really bad copy of the 400/5.6, it is going to be better in the majority of comparison.

My 100-400 is as sharp as my 400 5.6L at 400 mm wide open. Here's my test/user report between these two:

http://birdphotoph.proboards107.com/ind ... thread=352


I too had doubts if my 400 5.6L might just be a so-so copy, so I compared the zoom to my 500 f4 IS which I think is an excellent one. I equalized the angle of view in this test, and I'm pleasantly surprised that the 100-400 performed not so badly. BTW, the relative performance of my 100-400, 400 5.6 and 500 f4 in the field mirrors those I get in controlled tests.

http://birdphotoph.proboards107.com/ind ... thread=353


I certainly have no doubts on your test being representative of your copies of these lenses. My tests reflect a different result, but these are likewise representative only of my copies.

MR's comparison of the two lenses favors the prime, so does user reports from birders I highly respect. But the test by http://www.photozone.de shows the zoom is slightly sharper at the center of the frame than the prime, both at 400 mm, f/5.6 (the prime however slightly edges out the zoom in the border of the frame at the same focal length and aperture).


Romy

Romy, you appear to have a truly exceptional 100-400.

I believe I have a truly exceptional 400/5.6 BTW - I believe it is on the high end of the distribution for sharpness because there is little difference between f/5.6 and smaller apertures - not at all what I expected.

All I can say is - whatever you all decide, feel free to toss the "junk" lens my way... I've only once shot with an L lens, a 300 f4 non-IS... that was 8 years ago & I still miss it (it was a loaner)!

Interesting, thanks for posting, E.J.!

akclimber wrote:Thanks EJ. I've been consisdering a used 400 f/5.6 and this may push me over the edge but dangit, I bet your report results in used prices for 400 f/5.6s going up!
Cheers!

Yeah, and the used price of the 100-400 will be going down now...

And since my new 400 5.6 should arrive next week, this effects me.

'course mine is one of the good ones

toomas wrote:I have shot with both, the 400 5.6 and 300 2.8 (owned the 300 for a long time before selling it for 500 f4) and I have to say the sharpness of your 300 + 1.4 @ 5.6 resembles more the results I've gotten with my 300 + 2x wide open.

I agree completely. I'm very suprised by the results of your 300 2.8 + 1.4x tests, as I've had much better results with both the 1.4x and 2x converters with my 300 2.8 IS. Actually, I've had better results stacking both converters at f/11 than what your test shot at f/11 shows (of course getting really sharp shots with converters stacked is rather hit or miss, but still).

Kari Post wrote:I agree completely. I'm very suprised by the results of your 300 2.8 + 1.4x tests, as I've had much better results with both the 1.4x and 2x converters with my 300 2.8 IS. Actually, I've had better results stacking both converters at f/11 than what your test shot at f/11 shows (of course getting really sharp shots with converters stacked is rather hit or miss, but still).

Again, you can not compare your normal shots to completely unsharpened 100% crops unless the shots were taken under identical circumstance with identical RAW processing and an identical workflow. Most people never look at images like this at 100% - you would be surprised! Virtually all RAW converters sharpen unless you specifically turn it off. The f/11 shots here for the 300+1.4x are exceptional for a 100% crop with zero sharpening of any kind anywhere in the workflow, especially for a teleconverter.

Here is what the 300/2.8 looks like with typical sharpening applied and the original unsharpened version below:

Thanks for the repost and explaination E.J. I'm also using a 1D Mark II N, which has about half the pixels of the 1Ds Mark II and therefore would resolve a lot less detail.

EJ, based upon your most recent post (follwing up on my earlier post), is it accurate to conclude that the test you did shows that the differences, while visible at 100% crops are effectively negligible when sharpened? In other words, there is no effective difference between the lenses (regarding center frame sharpness) when simple PS is taken into account.

I would expect more significant sharpness differences at the edges, vignetting, CA, etc. that may not be easily corrected with simple PS. Your thoughts?

ColorChange wrote:EJ, based upon your most recent post (follwing up on my earlier post), is it accurate to conclude that the test you did shows that the differences, while visible at 100% crops are effectively negligible when sharpened? In other words, there is no effective difference between the lenses (regarding center frame sharpness) when simple PS is taken into account.

I would expect more significant sharpness differences at the edges, vignetting, CA, etc. that may not be easily corrected with simple PS. Your thoughts?

If the detail is there but it is simply a bit fuzzy then you can effectively deal with it through sharpening. That is basically what you are doing anyway to overcome the effects of the anti-aliasing filter. However if the detail is just obliterated such as is the case with the 70-200+2x, no amount of sharpening will return all of the detail although it will help.

ColorChange wrote:[...] is it accurate to conclude that the test you did shows that the differences, while visible at 100% crops are effectively negligible when sharpened? In other words, there is no effective difference between the lenses (regarding center frame sharpness) when simple PS is taken into account.

I agree with E.J.'s response to this. Sharpening will help any image, but what sharpening is doing is boosting edge contrast. If the edge detail has been smeared together too much, or the real edge detail is barely distinguishable from noise, then the software can't effectively recover edge contrast.

As an example of this, look at the original 100% crops again and compare the tightly packed horizontal lines in the top of the "5" with the more loosely spaced horizontal lines in the bottom curve of the "5". This was actually a really good example for E.J. to show because it illustrates the impact of the lenses' "native" sharpness on higher and lower frequency subject detail.

In the 100-400mm and 70-200mm + 2X examples, especially the latter, the separation of the horizontal lines in the top of the "5" is heavily hammered. You're not going to easily -- or at all -- sharpen that and get it back. Compare this to the same section in the 400mm lenses and the 300mm + 1.4X -- in all of those cases, there is clearly visible separation of the horizontal lines and so sharpening will have no trouble sprucing up edge contrast in the region.

In the bottom curve of the "5", all lenses show separation in the horizontal lines and so sharpening can easily work on all of them in that region. But even if sharpening can improve results, it can't necessarily equalize them all. Consider the 300mm + 1.4X where the f/5.6 result was not at the same level as f/8 and f/11. Look at E.J.'s examples that show sharpened results. While all 3 sharpened crops are very acceptable, the f/5.6 still looks a bit softer at 100% -- slightly less edge contrast with some minor loss of extra-fine detail in the top section of the "5", compared to the f/8 and f/11 examples. To equalize a bit more, you'd have to apply a bit stronger sharpening, which then risks amplifying any noise and creating stronger artifacts like halos. So there's a trade-off.

Now add one more set of factors, and that's conditions. E.J.'s tests were shot in a controlled situation -- excellent light, perfectly flat subject in a static setup, parallel to the sensor plane so DOF is a non-issue, tripod with mirror lockup, moderate distance, low ISO, flash, and everything else. If any of these conditions is not held in "real world" shooting -- subject motion, camera/lens motion, less light = poorer contrast, higher ISO = more noise, subject not perfectly flat to the sensor plane, or greater distance -- then the subpar lenses are going to fall apart faster in areas of fine detail because they have zero slack and no margin for error.

Yes, great images can be made with any of these lenses. But setting aside copy-to-copy variations, the results tell you something about the performance envelope within which you can operate if you want absolute top shelf results -- especially in normal usage scenarios for the resulting images, where people are not 100% pixel peeping to try to see if some recognizable detail is in there or not. PS can't correct everything; a lot of compensation for a much less sharp lens would have to be based on keeping distance, light, ISO, stability, etc. on a much shorter leash...

One thing to add to what Royce said, on the f/5.6 example of the 300+1.4x I could add a bit more sharpening. I purposely applied a global sharpening to all three frames not a photo specific sharpening for each. If I had done that, then even the 300+1.4x at f/5.6 could be made very similar to the other two.

Nice job Royce.

Of course the next step is to say what level of lost high frequency sharpness is visible on printed output. Detectable differences while pixel peeping is one thing, printed output is another.



And no, I'm too busy/lazy to do the analysis myself.

Absolutely, Tim. "Good enough" or "not good enough" always begs the question -- good enough for what, and for whom? One should always evaluate this stuff in terms of the final output form. If it's small web JPEG's for friends, medium size images for digital slideshows, moderately sized salable prints e.g. at 11x14, or big, high spec fine art prints at 20x30 or 30x40. Or, if licensing digital high res files to customers, then it is their final output form that matters...

Just to amplify a bit on Royce's (characteristically thorough) comments, the "5" here is a nice substitute for the sorts of MTF test charts that are often used for resolution tests, eg at DPreview, Imaging-resource, Photozone, etc. The varying spacing of the horizontal lines does the trick.

Any lens has an MTF curve, which is the amount of contrast it transmits as a function of spatial frequency; simply put, for a set of closely spaced black lines (0) on a white background (255), what is the white/black contrast difference after passing through the lens, as a function of the line spacing? The higher the contrast difference, the greater the resolving power of the lens and hence the sharpness of the image. The line spacing where the contrast drops by 50% (so-called MTF50) is a standard measure of resolving power. The lines on the "5" serve as a nice sample of lines of ever narrower spacing from which to judge the MTF curve on a semi-quantitative basis.

As long as there is enough residual contrast, sharpening (which enhances edge contrast) can partially restore detail, but eventually as the line spacing decreases to a fine enough level there is no contrast to work with (or trying to restore the contrast starts introducing unwanted artifacts) and the detail at that spatial frequency and beyond is lost. The higher the MTF, not only is the resolution higher, but the amount of sharpening required (and therefore sharpening artifacts introduced) is lessened, and the image quality higher.