Canon EF lenses: some reflections about current state of the art.

I bought the EF 24-105 and the 70-300 DO lenses as main lenses for my EOS 5D. I have tested both of these lenses and I have to say that I have mixed feelings about the direction of current developments.
The EF24-105 has the famous L designation, which implies that the lens belongs to the Canon top of lens designs. On the positive side we have an effective image stabilisation, a dustproof and drip-proof construction, impressive flare reduction wide open at all focal lengths, and excellent image quality stopped down. On the negative side we find three stops vignetting wide open at the wide-angle side, very strong distortion at the wider angle settings and quite observable decentring at the tele-position. Many comments assume that the vignetting is caused by the large sensor size, but I am not sure and I am busy doing additional tests (with film based EOS bodies) to check whether the distortion is optics-related or sensor-related. To be continued.
The 70-300 DO follows the same pattern. It has acceptable performance at the wider angle settings, delivers outstandingly good imagery in the range 100-200mm, and falls off a bit at the tele position. Canon claims that this lens should perform in the L class, but that is only the case when operating the lens stopped down one or more stops. The IS does a very good job, but the colour fringing at the tele-setting is quite objectionably. The DO element reduces the physical dimensions, but introduces flare at the shorter end of the focal range. The DO however, reduces a 300mm focal length to a very small physical size.
It is fair to note that both of these lenses deliver high quality imagery when used within the boundaries of normal (even professional) photographic requirements. And they provide those options that are valuable in field service, like image stabilisation, flare reduction and high contrast delineation of major subject outlines. Many years ago, Geoffrey Crawley reviewed the EOS 1 with a 1.8/50mm and a zoomlens 28-80mm. He remarked that both lenses fitted the requirements of professionals working within the limits of 35mm film format. But he noted that persons obsessed with the ultimate of resolution requirements might be disappointed.
Testing lenses
If we were to take a very canonical view, we should state that the overall wide open performance of the two EF lenses reviewed here, is not very good. If we had reviewed these lenses in this state of the art a few years ago (referring to the requirements of silver halide recording quality) we could have noted that the image quality was not first rank. Especially as the apertures are not very wide to start with (f/4 in the case of the 24-105) and f/4.5 to f/5.6 in the case of the DO 70-300) we should have demanded a better overall quality. With IS in place one could use the argument that is easy to stop down and get much better quality. In fact the 24-105 would make an outstandingly good lens when you restrict yourself to the narrow range of 35 to 90mm at f/5.6 to f/8. And the DO 70-300 performs impeccably in the range 90 to 200 at f/8 and smaller. The IS factor would make these apertures comparable with lenses operating at f/2.8 or f/4. But this is not a good argument. IS makes it possible to take pictures in low light situations where non-IS lenses, unless of high speed, would not operate. IS does not change an f/5.6 lens into a f/2.8 lens.
Bird's eye view?
Taking a broader historical perspective, the design characteristics we see here are strongly reminiscent of what we had in the thirties of the previous century. Then designers could deliver good performance only at moderate apertures and focal lengths. Exotic designs were of course offered to photographers, but lenses with apertures of f/1.5 or even f/2 were not good and needed to be stopped down substantially to get passable results. Companies like Leitz/Leica and Zeiss devoted lifetimes of designers to improve the image quality and now we may say that high-speed lenses deliver better quality than their counterparts with less speed.
Recently I bought two Canon New F1 bodies with two FD lenses, the 1.4/50 and the 3.5/50 macro, for the price of the battery holder for the EOS 5D. Such is life!
The FD lenses deliver really good quality: the 1.4/50 is a bit disappointing wide open and the 3.5/50 is outstandingly good. In those days the companies prided themselves on the uncompromising search for the best possible performance.
The design focus nowadays is no longer on the sheer image quality as delivered by the optical cell, but on the total package. No one, not even Canon or Leica, can create zoomlenses with an extended range without compromises. And the optical balancing (vignetting here, distortion there, low contrast with fine detail here, low edge performance there) is complicated by the economics of manufacture (18 lens elements in one lens are not easy to align precisely and hold down the price of the lens) and the ergonomic requirements of the user (small size, high speed AF, extended vario range, macro possibilities).
It is evident from current designs that one has to accept a number of compromises and one also has to accept that some optical parameters (vignetting, distortion, colour fringing, even edge contrast) will (have to) be compensated during the post processing stage on the computer.
Margins all over the place
Popphoto remarks that they are not involved in sensor or optical testing, but in an assessment of the digital imaging chain. This approach matches the current trends: and I may ad that the overall quality of the image will be delivered by the imaging chain as an integrated system. The prime role of the optical cell as the main component responsible for the ultimate quality of the image is over. This is easy to explain: if we were to match the classical demands for optical performance to the requirements needed for the current sensor sizes for professional use (APS-C and APS-H and full 35mm area), the lenses would have to become huge in physical size. This is not a feasible option, so designers have to work along the two lines of lens compromises and distributing the performance parameters over the imaging chain. They can do so, as they now control all parameters in the chain from sensor sizes, microlens design, optical design, and software algorithms in the camera.
From a pragmatic viewpoint there is nothing wrong with this approach. After all we are only interested in the final result on our A3 and larger fine art prints, are we not?
Final results count!
And since the prints can only be produced with the assistance of computer programs (our digital lightroom), there is no basic problem with programs that optimize the result just before printing the image file. The main argument for the 4/3 Format is the fact that the small sensor area allows for superior optical designs according to the classical rules within a reasonable physical size. Current 4/3 designs do offer high speed lenses with excellent optical performance. But the drawback of the system is the small sensor size that limits the size of the prints that can be made with this system.
For traditionally oriented individuals (like myself) who have been trained to see the lens as the decisive element that defines the imagery possible with a camera system, this slow eroding of the primacy of the optical cell, may draw a tear or two. But we should look to the future not the past if we do not want to become Luddites.
Leica should deal its cards
How Leica will position their future lens designs to accommodate the promised range of digital cameras will be interesting to study. The statement by Leica officials that future lenses will continue to be designed for use on new digital as well as old film-based cameras, may be interpreted as an indication that the company is involved in a delicate balancing act between conflicting philosophies.
The new role of the imaging chain
We should indeed concentrate on the imaging chain that is responsible for the final print and we should choose systems that deliver what we want to show. This is not a new insight: in the silver-halide era of photography, we analysed the separate components of the imaging chain to get the best system. In the fifties and sixties of the previous century, the notion that resolution (number of lines per spatial dimension) was the best predictor of image quality ruled the waves. We quickly found out that this was not so. But it seems that history lessons are lost on the current generation of equipment testers. In almost every magazine and website, we are being drowned in resolution figures, now as lines per image height (and not lines per mm as in the previous period).
These figures are taken very seriously in an almost over-hyped way (as happened in the past!). Resolution figures have never had a fair correlation with good imagery and this rule holds true in the digital era too. In another article I noted that these figures are not only meaningless (without a proper context for interpretation), but also prone to a large margin of variation: the famous 15%. Popphoto posted a comment on their website stating that the original resolution figures for the EOS 5D were off by 8%, based on the test chart employed. This is an error on the minus side, but it can also be on the plus side. Thus resolution figures can indeed by different by a margin of 15% (as I predicted). Another 5% should be added due to differences in monitor resolution, so in fact the figures quoted might vary by almost 25%, making these test results quite useless in a quantitative way, in addition to their limited value in a qualitative way.
But the insights of the past will not be remembered by current test methodology: the magazines have invested most of their reputation in the test equipment used. Remember the German magazines that claim that their computers analyse more than two million measurements per lens to arrive at a figure of merit that is quite useless. And the websites cannot allow themselves to use more sophisticated methods to evaluate the products: the time required to do this is so long that the next generation is already knocking at the door.
So we continue with testing methods that are the equivalent of the fast-food products in the real world: it is quick and cheap and superficially it even tastes quite well. But after eating slow food in a real restaurant, you will surely understand the difference.

More Canon lenses reviewed

After a few months of heavy shooting with the Olympus E-1 and processing the files with the new Photoshop CS2, I am now eagerly awaiting the new Epson R2400 as a complement to my Focomat V35 for chemical printing. A full report of my experience with the E-1 and a detailed comparison between the processing of digital files and the processing of film emulsions and the image quality of both media will be published shortly. In the meantime I picked up my Canon F1 with a (new) FD 1.4/50mm to enjoy the simplicity and directness of filmbased shooting. As it so happened, I could review a series of Canon EF lenses, the 2.8/24, the 1.4/24, the 1.4/35, the 1.4/50, the 2.8/100 and the 2/100.
There was not enough time to do a full test of every possible lens, so this report should be seen as a review, not a test.
But the general conclusions might be of some interest.
The slide film used was the Fuji Provia 100F, one of the best in its class.
I had used the FD 1.4/50mm already some time ago, but with the fresh experience of the new Leica 1.4/50mm ASPH lens, it is of some interest to see whether the verdict changes. There is after all a generation of difference between the two (more than 35 years in optical design). Overall the FD lens draws a bit softer all apertures. This can be clearly seen in the highlights where the Leica lens has excellent reproduction of the specular bright spots, where the Canon FD is duller. The Leica defines very fine detail with more crispness; with the Canon this level of detail is simply not reproduced at all. It is hidden in the grain pattern. Colour reproduction of the Leica is neutral, but with a higher level of saturation. Some photographers might prefer the more painterly reproduction of the Canon above the scientific definition of the Leica. Put in the perspective of the seventies the Canon lens was at that time an excellent, if not superior design and one that would have challenged if not surpassed the Leitz product. Even today one could be very happy with the overall performance if you are not spoiled by the image quality offered by the new Leica lens.
The FD lens and the Leica Summilux are fully mechanical lenses and there is no compromise whatever in solidity of mounting, even if the Canon already has some non-metal parts to reduce the weight.
With the AF lenses we are entering a totally different world. Not reviewed here is the Canon 1.8/50mm, because it fell apart after trying to put the lens on the body. It was a heavily used version and therefore I will not make a point here. The incident however gave the occasion to look inside the lens and I was amazed about the totally non-metal design of the lens. Even the several lens elements were fitted in plastic mounts and the overall design was geared to automatic assembly. The image quality of the lens was excellent, but that does not surprise, as the good old double gauss design is a bread and butter design, that has been studied to the extreme by all manufacturers.
The lenses reviewed
1.4.50mm AF design delivers excellent performance, has a well-built quality and is only wide open inferior the current Leica R design. The Canon version has a remarkably high level of distortion (more so than the Canon 1.4/35mm), a trace of coma and a strong spherical error in the sagittal direction. The designers balanced the distortion with the spherical aberration to get good performance at full aperture.
Canon 1.4/35mm is an outstanding design with very little distortion and some traces of coma. The central portion of the image is very good, with the outer parts of the image filed dropping visibly in performance. Stopped however it is an impressive lens. There is some decentring, but not enough to worry about.
Canon 1.4/24mm has excellent performance wide open, but with the corners visibly less good. There is curvature of field and relatively large colour fringing at the edges. Given the wide aperture the lens is a good example of the Canon design capabilities.
2.8/24mm is a more modest design and performance is very good: overall contrast is high, and the corners of the image are much better than in the case of the 1.4/24mm Again we note decentring on axis.
The Canon
2.8/100mm offers very high contrast and is almost as good as the Leica version, but we note mechanical errors in the focusing unit.
The Canon 2/100 is very good, but clearly no APO design as we see small colour fringes at the outlines of details in the outer zones of the image. Again we note that the focusing unit has mechanical errors in alignment.
general characteristics
This is a characteristic aspect of several of the Canon lenses reviewed here. The alignment and decentring errors however are stronger in the skew and vertical direction, than in the horizontal direction. In other words: when the photographer is using the camera in the normal (horizontal) position, the errors are hardly important. This could be noticed with all Canon lenses. It seems plausible that the Canon designers have adapted the design to the autofocus requirements and limitations. The AF focusing needs components with as little weight as possible, and with this demand one has to accept that the focus group has a slight axial and lateral movement. The accuracy of the AF has its limits and a wider depth of focus region can counteract these, and this can be accomplished by allowing some lower contrast wide open. The location of the circle of least confusion is not always the best candidate for the positioning of the plane of sharp focus and by clever manipulation of the options available the designer can create quite satisfactory imagery, even wide open.
The focal lengths, reviewed above, are classical designs in the Canon range. Optically the difference between the FD and EF lenses is not as big as with the new Leica designs compared to the previous generations.
Canon has clearly set all its cards on the design and optimization of AF lenses with an extended zoom range and with image stabilisation. Within these constraints the quest for maximum image quality is not possible. And we may remark that IS is not always the best option. You may need up to eight additional lens elements to create a potent IS ability. But these lens elements cause loss of transmission that can be compensated by the automatic exposure system, but then the advantage of the IS, compared to normal non-IS lenses, is less evident.
Leica on the other hand operates with a different paradigm: the definition of best image quality is located on a higher level, but to guarantee this performance, the tolerances have to be very tight and this can be accomplished only with fully mechanical designs, that are more expensive to produce.
For the overwhelming number of working photographers, the Canon parameters are fine as can be inferred from the ubiquity of Canon equipment. The photographer who wants the optical clarity and visual sensuality and richness of the classical large format pictures, combined with the dynamics of 35mm photography has that option too: current Leica lenses can easily produce this quality when one uses the high definition ISO100 films from Kodak and Ilford.