Canon 24mm TS-E not sharp!

[bob_prichard]

As I am planning a week long trip to Yosemite, I thought I would check

the focus and sharpness on my wide angle lenses. The Contax 21,28 and

35 were all razor sharp, as was my 17-40L zoom (surprisingly so). But

the 24mm TS-E was soft at all distances (near to infinity)and all

apertures. I don't recall this lens being soft in the past. Is this

something that can happen because of the tilt-shift? Is it something

that Canon can correct?

[peter_white2]

The tilt/shift mechanism wouldn't cause a problem. But if the lens has been dropped or otherwise damaged, who knows?

[ray robertson]

Well, it's difficult to say whether or not your TS-E 24mm has always been soft. It's much easier to appreciate the performance of a lens through a comparison with another lens than it is to get an accurate impression without comparison, depending on your experience.

 

My TS-E 24mm is also a bit soft and disappointing in this regard. The one that Photodo tested a few years ago was also not too good, getting a rating of just 3.3 out of 5.

 

One advantage, however; at least it should be equally unsharp from corner to corner.

[robin_sibson1]

No problem with my TS24 lens. I did a quick comparison test at f/4 against the 24~105 and the 17~40. The TS24 and 24~105 were comparable, the 17~40 a bit better.

[lsaavedra]

Mine is also soft, but only at the corners when shifted 3mm or more than that. When used as regular wide angle with no perspective correction I haven't noticed relevant softness, but no test has been made in this matter

[ray robertson]

..."No problem with my TS24 lens. I did a quick comparison test at f/4 against the 24~105 and the 17~40. The TS24 and 24~105 were comparable, the 17~40 a bit better."...

 

Robin,

Do you mean your TS-E 24mm is comparable to the 24-105 in the centre, at f4? I would think the TS-E24 would be noticeably better than the zoom towards the edges. Even my Sigma 15-30, at 24mm, is sharper than the 24-105 towards the edges, but about equally sharp in the centre.

[robin_sibson1]

Raymond, this was only a very crude test, so I am not going to defend it against more scientifically conducted evaluations. But for what it is worth, ON THE 20D, I did not see much difference across the frame. Of course, it might be a different story out there in the for-me uncharted wastes of FF digital territory. It certainly would be with vignetting; the reported heavy FF vignetting of the zoom at 24mm f/4 hardly shows up at all on the APS-C sensor. Both the 24~105 and the TS were slightly less good than the 17~40. The 10~22 at 22mm was very competititive with the 17~40 at 24mm.

[ray robertson]

Ah! Of course. For me, life has become more complicated now I have a 5D. I have to make decisions such as; is it worth changing lenses from 24-105 to 15-30, or TS-E 24mm for this shot for the sake of sharper edges? Never has to worry about such matters with the D60 and 20D.

[robin_sibson1]

Raymond, I think we have a consistent view of the situation! For me, there was one thing that killed the 5D as a one-body solution, which is my need to be able to use the 100~400+Extender1.4x combination with AF on FF for wildlife shots. I did that for years on my -1V. I now get even better results without the Extender on the 20D, where 400 actual is the equivalent of 640 on FF, so I even win a bit compared to an actual 560 on FF. At the wide end, the 10~22 is a great little lens, but it would be nice to have the full capability of the TS24 back again. Maybe I will succumb at some point to a two-body solution when the 5D price drops! As for the various 1D cameras, I quickly discovered that the 1V in HS configuration was far too heavy for comfort, and those are even a bit heavier. Just as well, since spending that money on what is still rapidly evolving technology is not top of my priorities.

[yoyo_szeto]

Hi,

Michael Reichmann has done a comparison of several Canon lenses at 24 mm ( http://luminous-landscape.com/reviews/lenses/24-vs-24.shtml ). The review may give you part of the answer.

[bob_prichard]

Yoyo--thanks for the link and getting back on topic.

 

It seems that softness on this lens has been a problem for others. At this point, I think I will send it in to Canon for collimation. Does anyone have any experience with Canon on problematic lenses?

[mark u]

I note you have a recently acquired 5D. I think it's possible that the lens performs less well with this body than with other bodies you may have used it with - whether film or digital crop. This is because sensor optics can lead to reduced sharpness, additional vignetting and even additional CA close to the edges of the frame, especially when you use movements. It may be worth testing it on a film body if you still have one.

[ray robertson]

Mark,

You might be right but not for the reasons you are implying. Film grain can obscure lens defects such as CA, and lens 'softness' might also not appear so obvious with film. Digital sensors can be quite ruthless on lenses, which is why first owners of the 1Ds began to realise some of the lenses they had used for years with a film body were not good enough.

 

I'm back again to my assertion that, in the critical frequency range of 30 to 40 lp/mm where lens contrast is critical to the perception of sharpness, DSLR sensors do better than most color film.

 

On the other hand, film can record, albeit rather faintly, much higher frequencies than any currently available DSLR, which is an advantage for micro detail in large prints.

 

I'm not trying to resurrect a war between film and digital, just offering an explanation for what I've observed.

[ray robertson]

Well, owning three 24mm lenses (15-30, 24-105 and the TS-E 24) I decided I'd better find out what their relative performance is on the 5D.

 

I photographed a scene from my verandah, consisting of a variety of textures such as fine leaves, mottled tree trunks, papaya fruit, rough and smooth concrete surfaces, fine grass and small stones on the ground, car number plate and model name at a distance so that they were barely legible.

 

Surprise! Surprise! All lenses performed essentially the same except at the edges at full (and close to full) aperture. At the edges at f4.5, the ranking was TS-E 24mm, Sigma 15-30, Canon 24-105. By f16, those differences had completely disappeared. At f8 the differences were too marginal to bother with.

 

In the broad central area, equivalent to the APS-C format, there were no significant differences at all, anywhere. However, if one wanted to put a very fine point on it, the ranking would be TS-E 24mm, 24-105, 15-30.

 

One relevant point; this comparison was done on a 19" CRT monitor at 1280x1024 resolution and 200% enlargement of crops. At this degree of enlargement, fine leaves were pixelated and the equivalent print size of the whole image would have been 8ftx12ft.

 

Since it's very unlikely I'll ever make an 8x12ft print from a single 5D image, small differences that I see at this degree of enlargement can reasonably be said to be irrelevant.

 

I was also surprised at how well resolution held up at f16 for all 3 lenses. There was very little contrast fall off at f16 compared to f8. I think I might use f16 more often. I suspect this is one advantage of the larger pixels of the 5D.

[robin_sibson1]

To bring myself firmly back on topic (slaps wrist), the view that is emerging here is that there is nothing systematically wrong with the TS24 design, although it may not be quite as sharp as some 24mm alternatives. Thanks for your interesting comments, Raymond. That's consistent with the conclusions of the Luminous Landscape review. So, Bob, perhaps you should have your lens checked.

[olivierp]

I am in the process of buying a TSE lens, either 24mm or 45mm.

I will be more comfortable with the 24mm but I have read many comments on the Internet about the 24mm softness.

 

I have also read a test in a journal where the TSE 24mm got 4 stars out of 5 in sharpness.

The TSE 45mm is everywhere mentioned as a reference of sharpness.

 

I came to a shop who had the 2 lenses.

I made some shots on a chart.

 

All shots at f8 or f11 since long exposure is not an issue to the job I have to do. At these apertures, sharpness was verygood. The same excellent sharpness for the two lenses whatever the shift. I will be glad to share the files whit anybody interested in. (too big to post here).

 

Cheers,

Olivier

[ray robertson]

Oliver,

No need to post the files. I believe you. In any case, they wouldn't be meaningful without reference to a known standard. The problem is, there's always some quality variation between lenses even from the same batch, and certainly from different batches. One can only hope that the more expensive the lens, the more consistent the quality control.

 

The TS-E 24mm is an expensive lens so this obvious variation in quality is a bit disturbing. Photodo had a high reputation for MTF testing of lenses, when they were doing them. Most of the lenses they tested are still being made, but I suppose there's no reason that, over time, improvements to the manufacturing process could not take place and perhaps even design modifications, so I really have no idea if the Photodo rating of 3.3 is typical for a recently manufactured TS-E 24.

 

A rating of 3.3 is low for a prime. The medium quality Canon 28-135 IS zoom gets a rating of 3.5 for example. The remarkably cheap Sigma 24/2.8 gets a rating of 4.

 

Since my copy of the TSE24 is at least as good as my new 24-105 IS, which in turn is on a par with the highly regarded Canon 24-70/2.8, I would think that my TSE24 would merit a higher rating than 3.3.

 

Robin,

If lenses in general were more consistent in quality, then Michael Reichmann's lens reviews would be more useful. Unfortunately they are not consistent in optical quality, which is the cause of much confusion and difference of opinion. For example, Reichmann's comparison between the 100-400 IS zoom and 400/5.6 prime showed the prime was clearly sharper, especially at f5.6. However, the copy I bought (brand new and pristine in its box) was WORSE than my 100-400 at 400mm. I can only assume that I was lucky to get an above average 100-400, but unlucky to get a below average 400/5.6 prime. Of course I returned the lens for a refund, but before giving me the refund the shop sent the lens to the Canon service department for adjustment. It came back marginally better, but still worse than the 100-400.

 

My first EF-S 10-22mm wouldn't focus accurately close up. I returned it but no replacement was available. They were in short supply at the time. I tested another EF-S 10-22 later in Singapore, in the shop, using my laptop for immediate feed back, comparing it with my Sigma 15-30 at 15mm. The 10-22 clearly wasn't up to scratch. The shop owner could see it for himself. A few days later I tested a third lens in Kuala Lumpur. That was better. Still not quite as good as my Sigma at 15mm but close enough so I took it.

[qtluong]

There is no way a normal 100-400 would outperform a 400. I have both, and as expected, there is a significant difference ! As for 24mm, both my 24-70 and 17-40 are sharper than the TSE, but that doesn't make the TSE an "unsharp" lens, especially is T or S are required for the shot. <a href = "http://www.terragalleria.com/">Terra Galleria Photography</a>

[ray robertson]

..."There is no way a normal 100-400 would outperform a 400. I have both, and as expected, there is a significant difference !"...

 

Luong,

I think you meant so say. 'There's no way a normal 100-400 would outperform a NORMAL 400/5.6 prime.' And that is probably true. However it's very clear to me that there's a quality control variation amongst lenses and it's probably true to say that the cheaper the lens the greater the QC variation is likely to be, on balance.

 

I also owned both of these lenses. I still have the 100-400 of course. I tested the 400/5.6 thoroughly, using tripod, cable release, mirror lock-up, autofocussing and manual focussing, with and without 1.4x extender. There was no instance where the 400 prime shots were better (ie. sharper, more contrasty, more detailed) than the 100-400 shots, but there were plenty of examples where the 100-400 shots were clearly better.

 

I copied the test results onto CDs and handed them to the shop together with the lens. I even made A4 size prints of crops to demonstrate clearly the differences. I did the whole testing procedure again after the lens had been returned from the Canon service department. I wasted a heck of a lot of time just to be sure my results were accurate.

 

The interesting thing about this saga is that Canon never admitted the prime lens was sub-standard. After all, the 400/5.6 is slightly cheaper than the 100-400 zoom and probably not a big seller. Without IS its usefulness is limited. It could well have been the case that the entire last shipment of this lens to Australia at the time was sub-standard. Suspecting this might have been the case, I didn't pursue the matter.

[ray robertson]

I should also add that there was nothing clearly wrong with my copy of the 400/5.6. If I hadn't already owned a 100-400 and had something to compare it with, there would have been no reason for me to return the lens. I would just have assumed that that's what you get at that price point. If I'd bought the 400 prime first and the 100-400 zoom later, I would of course have been extremely pleased with the good value and excellent sharpness of the zoom lens.

 

As regards general concerns about sharpness, all we can say is that one 'particular' lens is sharper than another and it's not always clear if the sharper lens is typical or just an aberration. I'm a firm believer that all lenses should be individually tested in the factory and should ship with a set of MTF charts as testament to their quality. Each model of lens could be categorised as 'standard', 'premium' or 'deluxe' and priced accordingly.

[pbizarro]

There may indeed be a consistency problem betwwen different samples of the same lens. But I believe that inconsistency between "users reviews" or "tests" is even bigger, and leads to very dubious "conclusions".

[ray robertson]

Paulo,

That's just a belief. Where's the evidence? It's true that some folks just don't have a clue as to correct testing procedures, but there are also vastly different results from people who do have a clue.

 

A few examples, the Photodo testing of the TS-E 24mm. 3.3 is really lousy for a prime. Michael Reichmann's testing of the Sigma 12-24. He had a lemon. My testing of the 400/5.6. I had a lemon.

 

Those are perhaps extreme examples. Look at all the shades in between. Total confusion, period.

 

In the absence of individual MTF charts for each lens that's produced, the confusion will continue.

[robin_sibson1]

Raymond, while we are off-topic again in the area of long-ish lenses, a couple more comments. Yes, of course I am aware of sample variation; I should be - I am a statistician! I don't EVER recall seeing a published test which assessed it for camera lenses, and that's really overdue. However, we are where we are. What I wanted to draw your attention to was the review at www.photozone.de of the 300/4L IS. They are quite critical of its performance on the Extender 1.4x, and even comment that this flies in the face of received wisdom on the forums. Sample variation in the Extender, perhaps, since that same lens on its own was excellent? Any thoughts?

[ray robertson]

Robin,

The Canon 300/4 IS is another lens which, over the years, seems to have shown great variability in optical quality. I notice that Photodo have included again the MTF charts for all their tests (for some reason they were missing for a while).

 

The Photodo rating for the 300/4 IS is a mere 3.4, hardly better than the TS-E 24, yet their rating for the non-IS version of the 300/4 is an excellent 4.3. The MTF curves at 40 lp/mm and f4 tell the story. The IS version shows rather low contrast at full aperture and 40 lp/mm, well below 50% and even below 40% across much of the image circle, whereas the non-IS version is well above 60% MTF.

 

When using a 1.4x converter with the 300/4, you'd probably want to use it at f4 to maintain autofocussing capability. With the IS lens that Photodo tested, it's easy to see that it wouldn't perform well with a converter at full aperture.

 

I can only assume that, when a professional organisation such as Photodo test a lens and the results are either far above or far below expectation, they grab another copy from another batch and test that as well. I would therefore tend to believe there are quite a few 'sub-standard' copies of the 300/4 IS around.

 

Thanks for the link to Photozone. I hadn't looked at the site for so long I'd forgotten about it. Below is a revealing extract from their 100-400 review.

 

...."The 100-400mm L IS is one of those lenses which are discussed frequently in various user groups specifically in comparison to the 70-200mm f/2.8 L IS with 2x converter and the 300mm f/4 USM L IS with 1.4x converter. Well, here?s our share adding to these discussions ... :-)

The lens exhibits a very even performance throughout the zoom range. Translated to the 8 mega pixel context of the EOS 350D the lens can be rated very good(+) at 100mm to very good(-) at 400mm which is very respectable. As to be expected the 70-200mm f/2.8 as well as the 300mm LS IS are slightly better within their native range. The 70-200mm is also marginally better with the Canon EF 1.4x II converter. However, the 100-400mm L IS outclasses the 300mm fix-focal with 1.4x by quite a margin and, while we haven?t tested it yet, it is quite safe to assume that the 70-200mm with 2x converter will also fall short at 400mm (google around a bit and you will a comparison between these two setups). Consumer lenses such as the Canon EF 75-300mm USM IS are vastly inferior at their long end setting anyway."...

[mark u]

Since this thread is still alive I think it's worth adding these comments to it.

 

It's good to know that a good copy of the 24 TS-E performs well on a 5D - there's nothing like actual testing for resolving the potential issues. However, I don't think that it's true that the apparent underperformance of some lenses on full frame digital bodies is due to inadequate performance of the lens per se and that film is masking that.

 

One interesting counter-example is that a number of users have noted that the 70-300 IS DO lens actually performs better on a 5D than it does on a 20D, where the new non-DO 70-300 seems to be the better performer. This strongly suggests that interactions of the sensor and lens optics have an important bearing on image quality.

 

The issues with a number of short focal length and very fast lenses used wide open were carefully explored by Chasseur d'Images, who discovered that bench test performance is clearly affected by sensor optics. Indeed, performance of the same lens with different sensors showed variances. Moreover, it's easy to understand that the limited acceptance angle of microlenses and the extra dispersion of oblique rays through refraction and reflection in the sensor optics can cause problems. I think that some confusion has arisen because the lenses that are least affected by these problems tend also to be top performers when bench tested - such as the Zeiss 21mm.

 

What seems to be clear is that the business of lens testing has now become even more complex - not only are there sample variation issues, but also the need to compare performance with particular sensors as well. Mean time, it seems that Bob may benefit from having his lens checked over by Canon.