EOS big three as of December 2008

The original EOS 5D has been around for over three 3 years. While a full frame digital SLR at a $3000 price point was ground breaking in the fall of 2005, by the end of 2008 the 5D had become somewhat long of tooth. On the Nikon side, the D700 bettered the 5D's performance and was selling for $2700. The Sony A900 was the high resolution king at 24.5MP and had a class leading body for just $3000. The 1DSmk3 was still absurdly priced at $8000 and weighed as much as a light cruiser. What was mother Canon to do?

Megapixel lovers rejoice, the widely anticipated EOS 5D is finally here. Canon EOS shooters now have three SLR's with a pixel count equal to or greater than 15MP. That's two more than any other camera maker.

As some of you know, I am a landscape photographer. I migrated to full frame digital SLR's from medium format film. I used to shoot 6x4.5, 6x7 and 6x9 film with the end product being prints of 20"x30" and larger. As such, I chose to investigate only one area of the 5Dmk2 in this exercise, overall image quality or IQ for short. For this test, I decided to construct a still life with a wide range of color, lots of detail and very controlled lighting to obtain what for me are valid results.The test would produce 100% crops for pixel peeping. The final analysis of IQ however would be made using real life obsevers and detailed sections of 16"x24" and 20"x30" photographic prints. The ISO100 test prints would be made on my Epson 9880 at a printer resolution of 2880 x 1440 DPI. They would all be prepared in identical fashion for optimum printing with a final file resolution of 240 DPI. For those of you that may grimace, that I didn't use a resolution of 360 DPI, I can tell you that I have not been able to find a single person who can see the difference between a print at 240 DPI and 360 DPI without a magnifying glass. Since I make fine art prints for other artists professionly, I will stand by my decision. You are free to disagree. As a landscape photgrapher, I felt it unecessary to make prints at higher ISO's. The higher ISO 100% crops will be made available for you to judge, but actaul print files at higher ISO's were not made, printed and viewed.

The stage is set.

The test scene shown above was constructed to provide a wide range of detail, textures, colors, and tonal variations. The idea was to give the cameras a widely varied visual workout. The light sources were two 60W halogen photofloods in reflectors set left and right of the stage at about 45 degree angles.

The shots are fired.

I then proceeded to shoot the test scene with four camera bodies. They were a 5Dmk2, a 1DSmk3, a 50D and a 5D. I decided to throw in a 5D to see how the old boy holds up against the new kids on the block. I shot in raw mode, in Adobe RGB color space on a tripod, using my EOS 50mm f2.5 macro lens. I used mirror lock up, with the 2 second self timer to avoid any vibration problems. I also used aperature priority mode with the f-stop set to f8. The auto-focus point was president Grant's hairline.. Samples were shot at ISO's 100, 400, 800 and 1600. The 50D was moved further from the stage to compensate for the 1.0 vs 1.6 crop factors. Highlight priority where applicable was set to off. Noise reduction was set to off. I never shoot above ISO 1600 so higher speeds were not used in this test.

The resulting RAW files were then processed in Adobe Lightroom 2.2 . I used the Adobe Standard settings for contrast, color saturation and color tone. I used no sharpening in the RAW conversion, and applied a USM in Photoshop of 0.3, 300, 1. This is my standard sharpening to counter act the effects of the in camera anti-aliasing filter. It is also what's recommened by Canon for the 1DSmk2 and 1DSmk3. I adjusted the 1DSmk3 exposure slightly to match the metering of the other cameras.

The pixel peeping begins.

Thumbnail center crops of the resulting images are shown in the table below. To view a high quality 100% jpeg click on the appropriate row and column. Beware, these crops are large! Each one is about 2MB. Feel free to save the crops on your computer and view them side by side in Photoshop. To my eye the 1Dsmk3 and 5Dmk2 results were basically identical between ISO 100-800. The 1DSmk3 seemed to be cleaner at ISO1600. Between ISO 100-800 any difference you might see between the two,was due to slight anomalies in the actual focus point between shots, and slight differences in metering between the two cameras. I tried to correct for small exposure and color differences after the fact, but didn't spend a great deal of time at it. The 5D hangs in there pretty well, but with only 12.7MP, it has less micro-detail .

The conventional wisdom is that full frame SLRs will always produce the highest IQ. APS-C cameras while good, just can't match full frame sensors. At the pixel peeping level, that certainly was true. Even though the 15MP 50D had more pixels, the 5D out resolved it detail wise. Noise wise it faltered slightly above ISO400 and more so above ISO800. The small pixels are noisey, detail and tonality suffers, at least at the pixel peeping level. Don't misunderstand me, the 50D held up very well, especially between ISO 100-400. It just doesn't have the contrast, tonality, and color rendition at this level. Pixel peeping is not the end product, at least for me. I like to make and look at high quality prints. How do these cameras stack up at that level?

	1Dsmk3	5Dmk2	5D	50D
ISO 100
ISO 400
ISO 800
ISO 1600

The experts view the prints

I very rarely shoot landscapes at ISO's higher than 100. In that light, I used ISO100 test shots for making all the prints. This certainly helps the 50D with it's relatively tiny pixels. I defined two distances for viewing these prints. Normal viewing distance was between 20 to 24 inches. Hypercritical viewing distance was between 10 to 12 inches. Most people over fifty definitely needed to use reading glasses at this close range.

I chose two print sizes to work with. The smaller one was 16x24 inches, the larger one was 20x30 inches. Rather than work with the whole print, I cropped an 8x10 inch section out of the center for people to view. I just used the 5Dmk2 for this test and omitted the 1Ds3 files. They were so close at the pixel peeping level, I thought any difference would be invisible at the print level.

You can download the actual 8"x10" crop print samples for the 20"x30" prints by clicking on the table below. Print them at your leisure and see what you think. Beware, they are 5MB in size! Hit your browser back button when done to return to this article.

		20"x30" Print samples
	5Dmk2	5D	50D
ISO 100

I assembled a group of four expert photographers to judge the prints. Each member of the group had 20 to 40 years in the business.

At normal viewing distance, the 5D and 5Dmk2 8"x10" test crops from the 16"x24" print were basically indistinguishable. They both received a rating of A as to overall print quality. The 50D print was rated a half step below, we'll call it B+ in print quality.

At hypercritical distance, the 5Dmk2 was rated the highest quality, the 5D was a very close second. It had just a bit less microdetail and contrast. The 50D print was ranked last, it lacked the accutance and micro-detail of the other prints. Over all though, all the prints were rated as being perfectly acceptable for professional quality purposes.

At normal viewing distance, the 5Dmk2 8"x10" test crop from the 20"x30" prints was picked 1st. The 5D test crop was rated second, though very close and the 50D was picked last. It's color, tonality and micro-detail was a clear step below the others. We will give the 5Dmk2 an A grade, the 5D an A- grade and the 50D a B grade at the 20"x30" print plateau.

At hypercritical distance, the results were the same but the visible differences were greater. The 5Dmk2 was first, the 5D second and the 50D last. It was thought that both the 5Dmk2 and 5D produced 20"x30" prints that were acceptable for professional quality purposes. The 50D was thought to be not up to snuff for highly detailed work, but it would probably suffice for things like portraiture.

We did not run a test with 13"x19" prints, but the I strongly suspect that the prints would be indistinguisable at that size. Even when viewed at hypercritical distance.

You can download the crops from the 20"x30" test be clicking above and judge for yourself.

If we ran the test at higher ISO speeds, the results may have indeed been different. If you want those results, you'll have to run your own tests.

The Fat Lady Sings

So, to sum it up, it seems that the coventional wisdom holds. Big pixels are better pixels, at least when the final output is a relatively large print. Crop sensor cameras have advantages in the reach department, for telephoto work, but not for overall quality. We won't argue the point that full sensor cameras require higher quality lenses to resolve detail across the whole frame. It certainly is true, and lens selection is critical for best results from the 5Dmk2 or Sony A900. We won't debate the merits of a 1 series body versus the 5Dmk2, thats beyond the scope of this test.

Thanks to Rod Knowlton for the loan of his 1Dsmk3 and Barry Horn for the loan of his 50D.

Miles Hecker has been involved with photography for over forty years. He teaches digital photography at Casper College in Casper,Wyoming. His photos have won awards from Photo.net, The Luminous Landscape, Wyoming WIldlife and Natures Best magazine. Miles' photos have been published in American Vignette, Backpacker, Natures Best Images, Popular Photography, Wyoming Audubon, and Wyoming Wildlife.