Canon 400mm DO lens and Canon EOS D-60 camera.
                                    Lens resolution and image quality issues...

Douglas Goodell, July, 2002

Question - Does it make sense to put a lens capable of 90 lines/mm resolution on a digital camera limited to 50 lines/mm? Answer - yes it can. But, like all simplifications, there's much more involved. Here is what I found.

Background

I recently had the opportunity to evaluate the new Canon 400mm, f-4.0, DO lens. This lens introduces a fascinating new technology that appears to have many advantages (one of which was claimed to be lower price, but obviously not yet). At the same time, I was also trying to learn more about digital camera behavior (I was ignorant) and had a Canon EOS D-60 for that purpose. One attractive aspect of this camera/lens coupling was that it might provide a compact, light weight, readily travelable system for nature photography, bird work in particular. After all, there are times when it is impractical to have the big guns at your disposal.

I appreciated that the D-60 would not have all of the capabilities I was use to with the Canon EOS 1-V body. However, I was not yet ready to pay $5000+ for a Canon 1-D body at 4 Mega-pixels since I was fundamentally concerned that a even a 6-Mega-pixel camera (the D-60) could not have the capability for detail capture of high resolution slide file, or even of scanned film at 4000dpi (approximately 24 Mega-pixels). This was a significant concern as my main product is large size fine-art prints.

I decided to couple the evaluation of the DO lens with that of the D-60 camera. I developed several image comparisons from shots of birds in the field, drawing in part from my archives. I also developed some quantitative resolution comparisons.

Quantitative testing can provide insight into the capabilities and limitations of systems. My feeling, however, is that such tests are useful only in comparison to other systems (lenses and/or cameras) with which one has familiarity. To provide a comparative framework, the test matrix got a bit large. I included four lenses, two multipliers, two camera types, two support systems, and two exposure conditions. I structured these tests as real-life field tests, under typical good working conditions. These are not controlled optical lab bench tests; they are influenced by my field technique and by my interpretation of the resolution charts. In these respects they are self consistent and should be useful to others. As they say, 'your milage may vary', but the comparisons should be relevant and instructive.

Test Structure

The resolution test matrix included the following components.

Lenses -- Four lenses were included along with two multipliers (EF 1.4-II, EF 2.0-II):

       Canon EF 400mm f-4.0 DO IS UMS 4.3 lbs, 9.4 in. (Weight, length)
       Canon EF 100-400mm f-5.6L IS UMS 3.0 lbs, 10.3 in. (7.4 in. closed)
       Canon EF 400mm f-5.6L UMS 2.8 lbs, 10.1 in.
       Canon EF 600mm f-4.0L IS UMS 11.8 lbs, 18.0 in.

Cameras -- Canon EOS D-60 and EOS 1-V HS. The D-60 was used at ISO-100, factory defaults, and high/fine jpg image settings (which a few tests confirmed were equal to RAW images files). The 1-V was used with Fuji Provia-F film at ISO-100. Tests with the D-60 body were made at the same distances as those for film with the 1-V body because this gives the correct image size from which to determine the lines/mm resolution data. However, recognizing that there is a 1.6X magnification factor for the D-60, some measurements were also made at distances giving frame filling comparable to the film-base system. (This distance is typically listed as about that for the film system with a 1.4X multiplier.)

Support -- Systems were either tripod mounted or hand-held. Tripods were the small Gitzo 1228 for the smaller system (one of the objectives was to keep the system small) and the Gitzo 1548 for the large system, i.e., the 600mm lens. A Wimberly side-kick was used to mount the smaller systems; a standard Wimberly head was used for the large system. Variability of hand held tests, as expected, was much greater than that of tripod based tests.  Statistical analysis was not done to characterize the data error sensitivity, but judgment suggests plus or minus about 3% for tripod based tests and 6% for hand held tests.

Exposure -- Two conditions were used: (A) to provide a 'modest' shutter speed of about 1/180sec, and (B) to provide maximum shutter speed (i.e., full aperture). Tests were made outside (NY latitude) in conditions varying from full sun to slightly overcast (it took eight days to complete all the tests). The 'A' exposure condition at 1/180sec was intended to represent a typical value and one where hand holding and stabilization effects would show up. The corresponding aperture was f-11 and was intended to provide sufficient depth-of-field to minimize focusing errors. (Auto-focusing was used where ever possible.) The 'B' exposure condition was at full aperture of the particular lens and was intended to provide the highest shutter speed, minimizing any system vibration effects. Typical shutter speeds were in the range of 1/750 to 1/2000 second for lenses without multipliers.

Test Measurements

To characterize system performance, I used a Koren 2002 lens resolution test chart [see acknowlwdgements]. This chart is a strip with lines increasing in spatial frequency from 2 to 200 line pairs per millimeter. A "5mm" chart at a gamma of 1.5 was used in these tests, printed as a 250mm strip. The chart-to-film distance is set to give the necessary 50X magnification reduction providing a chart image size of 5mm on the film plane. In the case of a 400mm lens this distance is 68 feet; for other lens combinations the distance was adjusted accordingly. Any small variations in the actual size of the test chart image were dealt with by measuring its length in Photoshop and adjusting the results to the requisite 5mm size.

For my tests I used only center image locations since with super telephoto images the edges are almost always off focus. I included both horizontal and vertical test strips. Three exposures were made for each condition giving a total of six readings for each condition. Also included was one of the 1951 USAF test charts, a Kodak gray card, color patch, gray scale patches, a Rodney test chart, and some B&W focusing bands. Some bird feathers were also included for reference.

A Nikon LS-4000 was used to scan the film; scanned and D-60 images were analyzed in Adobe Photoshop (with no sharpening). Direct film reading was done with a microscope.

Results and Discussion

Many of the quantitative test results are reduced and summarized in Table I. They indicate that the resolution limit of the D-60 is about 50 to 55 lp/mm, that of scanned film (4000dpi) is about 65 to 70 lp/mm, and that of the base film is above 95 lp/mm. These are reasonable values for this test, and in fact the value for scanned film is about that reported by Norman Koren. Clearly the resolution capability of the 6 Mega-pixel digital camera is not (yet) as good as that for film based systems. But, that turns out to be only a small part of the story. The other considerations involve system application and perceived quality of real images.

System Characteristics

Considering the system application first, it is apparent that the 'penalty' for the D-60 is greatly reduced if the lens in use dominates the resolution performance. For example, the 100-400 zoom lens has inherently lower resolution than the prime lenses. This is no surprise and no one will fault the 100-400 lens; it is a highly respected lens capable of giving excellent results. Similarly, adding a multiplier to the lens system lowers the resolution performance, but can still give excellent results (more later). Of interest for bird photography, two very practical system combinations giving essentially the same field coverage: D-60, 1.4X, 600mm compared to 1-V, 2.0X, 600mm, have essentially the same resolution performance. Obviously the system structure is important.

Some specific characteristics of individual lenses are also noteworthy. First, it is clear that the 400mm f-5.6 prime lens is an outstanding performer. Whether or not the other prime lenses would be equal at f-5.6 rather than at their f-4.0 maximum aperture was not evaluated here. On the other hand, it appears that the IS system reverses the advantage when hand holding the system so that the 400/5.6 while very good, may no longer be superior to the other prime lenses when hand held. Second, it appears that the 400mm DO lens is optically in the same class as the 600mm L lens, and this includes performance with multiplier use. Third, as noted above, the 100-400mm zoom lens is optically the weakest of the set. Finally, the one peculiarity of the DO lens behavior was in hand holding at moderate shutter speed with a 1.4X multiplier and the D-60 body. This behavior was confirmed in three re-checks. I must guess that the IS behavior is adversely affected by the mass distribution in this particular combination.

Not reflected in these resolution test results are the now well documented limitations of the D-60 camera with respect to focusing and frame speed. These seriously limit the camera for action shooting. On the other hand, the f-4.0 aperture of the 400mm DO lens compared to other 400mm lenses is a real asset for the D-60. The extra stop is a real advantage in focusing and makes the combination very attractive, though not a barnstormer.

Perceived Image Quality

This part of the discussion really relates only to the digital versus film comparison. Also, it relates only to enlargements of 20X and greater -- that's an 11x14 print from the D-60's 0.768x0.512 frame size and a 20x30 print from a 35mm film frame. But, if you have to pull out an image from less than a full frame, you may quickly exceed 20X magnification at much smaller print sizes. (Viewing slides is a different situation due to the generally great viewing distance.) For D-60 images pixelization becomes evident starting at about 20X and that alone may affect image quality. Therefore, for high magnifications it is recommended that pixel size be reduced using PhotoShop's bi-cubic interpolation, Genuine Fractals, or Fred Miranda's D60_SI_PRO. For the following discussion, pixelization has been adjusted out of the range of concern.

Figures 1 through 4 present sections from both digital and film based images. Figure 1 shows a mockingbird based on the 600mm lens and equal frame filling with the 1-V and D-60 cameras using different multipliers. The images shown represent about 40 percent of the full frame, and are at 9X and 14X magnification, respectively from each camera. At these magnifications there is really no difference in quality between the two images.

Figure 2 shows a highly magnified section of the head from each image, and here differences are apparent. Note that these are rather extreme magnifications. Also note that optical performance from resolution tests of each system is essentially the same, namely 50 lp/mm. In this case the digital image is clearly superior because of its absence of grain. There is no difference in detail between the two images, but there is a big difference in image quality! The detail is controlled by the resolution; the quality perception is controlled by the grain differences.

Figure 3 shows sections of the eye and bill of a great egret. The color differences reflect different specimens and are not relevant to this discussion. What is relevant are the detail and quality characteristics. Each image, in the context of any reasonably sized print, would be considered to be of good quality, but the details are clearly different. The top image is from the D-60, 600mm, 2.0X combination for which the measured resolution performance was 45 lp/mm. At essentially the same magnification, the middle image is from the D-60, 600mm, 1.4X combination for which the measured resolution performance was 52 lp/mm. In this case, there is clearly more detail in the eye, bill, and under-bill feathers. The difference is almost surprising for a resolution difference that is just above the that suggested to be significant.

While both the top and middle images in Figure 3 seem 'good', the bottom image shows what's missing from them. This image was shot at close range (requiring only low magnification) to show the full details of the eye, bill, and feather structure. Had the resolution capability of the D-60 been higher, the middle image especially would have exhibited more of this detail.

Finally, the middle section of Figure 4 shows a film-based image at a magnification comparable to that of the upper two images in Figure 3. Here, even though at higher resolution, the image is seriously degraded due to film grain. Admittedly, this is an extreme magnification for a film-base image, but at the same magnification, the digital image gives a more pleasing image quality. The bottom image of Figure 4 shows a more likely magnification of a film image from a system with relatively high measured resolution (near the limit for a scanned image). In this case, the detail level is high, though somewhat less extensive than that in the bottom image of Figure 3.

Overall, the images in the four figures suggest that image quality from the D-60 can be quite high. Also, the 1.6 multiplying factor of the D-60 can be an advantage by permitting use of optically better lens combinations. Nevertheless, there is a point at which the limiting resolution of the D-60 can restrict capture of detail, and this point is very dependent on the particular shooting conditions. On the other hand, the absence of film grain provides a clear advantage for image quality as image magnification is increased. It is at magnifications of 25X to 30X and higher where the different effects of resolution and grain become complicating. What is important quickly becomes very subjective and depends very much on the type of image.

Other Observations

Physically the D-60 with the 400DO lens is an attractive combination (and effectively 640mm). Its combined weight of 6 pounds is exactly the same as the combination of the 1-V with 100-400mm, which is a combination most people are use to hand holding. I did feel that the large diameter of the 400DO was more awkward to manage for flight shots than were the 400mm f-5.6 lenses, but I got use to it. On the other hand, as a hand held combination, the large aperture of the DO lens is almost essential for the D-60's focusing, which is limiting for action use. (The low frame rate is also limiting in action work.)

The D-60, 400DO combination can take up to a 2.0X multiplier with little loss in resolution (the D-60 is limiting). With its 1.6 multiplier effect, it's the equivalent of a 1280mm lens, and close to the resolution of the film based 1-V, 600mm, 2.0X combination (the multiplier limits resolution). Of course, the 400DO lens can provide much higher resolution if used with a film based body and without multipliers.

A few images were made to examine the D-60's noise level at different ISO settings. These included examination of bird's eyes in which noise could degrade quality in the black pupil. I was pleasantly surprised to find that settings of 100, 200 and 400 gave very good images; set at 800, however, there was significant degradation.

Conclusions

I initiated these evaluations to determine whether or not to purchase either the 400mm DO lens or the D-60 camera body. I became very impressed with each and especially with the combination. However, I would not opt for the 400 DO at its street price of around $5700 -- its just too expensive in relation to other lenses. Fortunately, I was evaluating and purchased a 'slightly' used lens as a much lower price. On the other hand, I felt that the D-60 camera was a good buy; it has very good potential even with its limitations. Moreover, it opened to me the capabilities of quality digital image capture. It especially demonstrated the advantage of a clean image base, with no film grain, in generating images of high perceived quality. The resolution limitations can be real, but can be offset by careful situation control and the grainless base. At this point I would welcome the functionality of a 1-V body with the sensor quality of the D-60 (or hopefully better).

Acknowledgments

The author is indebted to Norman Koren for the resolution test chart and his extensive tutorials on image sharpness. A series of seven articles (chapters) appears on his web site at normankoren.com. I highly recommend these articles. I must add however, that they are not light reading; I am just beginning to comprehend the material and I hope that my discussion has done no injustice to Mr. Koren. Also, an excellent article on digital camera image quality by Miles Hecker is available at luminous-landscape.com. I also wish to thank those providing reviews and tests of the D-60 camera and 400DO lens: Phil Askey at dpreview.com, Dave Etchells at imaging-resource.com, Michael Reichmann of luminous-landscape.com, Jef Graeber at photo.net. If I had found all of these resources before starting my tests, I might have saved myself some effort. It was, nevertheless, a very instructive exercise.