What lenses did I test?

I tested the following three lenses:

• Canon EF 28-105 f/3.5-4.5 USM: I bought this lens with my camera in December 1997.
• Canon EF 28-135 f/3.5-5.6 IS USM: I bought this lens in December 1999, intending it to be an upgrade for the 28-105. The main purpose of this test was to make sure it was at least as good as the lens it was replacing.
• Canon EF 100-300 f/4.5-5.6 USM: I bought this lens used in November 1998 and decided to test it, too, while I had everything all set up.

Note that this test does not attempt to evaluate the image stabilizer in the 28-135 lens; that's the subject of a separate test, the results of which should appear on my Web page shortly.

How did I do my testing?

On the assumption that most people are more interested in the results than in the procedure, I'll break out of the correct order I was taught in science classes in high school. Details on the testing procedure, including some flaws in my procedure, can be found later in this document.

One thing I'd like to note up front is that my sharpness evaluations are based on viewing photographs of a newspaper. These prints are at approximately 1/4 the size of the actual newspaper, and I'm trying to read lines of text which, on the original page, were only about 2mm high; they're about 0.5mm high on the prints. How far away do you hold your newspaper when you read it? Move it to four times the same distance (probably about twice arm's length, so you'll need a helper for this) and try reading it. That's a fairly severe test of sharpness.

Results

Testing Notes

• The drawn rectangle, though it shows clearly in all prints, was much easier to see through the viewfinder with the 28-135 than with the other two lenses.
• The 28-105 and 28-135 are the same speed from 28-35, but then the 28-135 falls half a stop behind. Between 35-50, it loses a whole stop (3.5-4.5), whereas the 28-105 loses only half a stop (3.5-4).

Testing Procedure

For my test target, I used a traditional cheap lens test target - a sheet of newspaper taped to the wall. More specifically, a two-page broadsheet spread, measuring approximately 64 cm wide x 55 cm tall. On this newspaper I drew a rectangle in the same 3:2 proportions as the 35mm frame; this was approximately 60 cm wide x 40 cm tall. This makes it easy to spot curvilinear distortion (barrel or pincushion distortion). The target was taped on all four corners, and was attached fairly low on the wall so that the tripod's legs did not need to be extended very much (as this reduces the tripod's stability slightly).

The camera (EOS Elan II) was mounted on a tripod. All pictures were taken in Av (aperture priority), using 2-second mirror prefire to reduce mirror vibration; a corded remote shutter release was used so that I did not need to touch the camera itself in order to take the picture. For the 28-135 IS lens, IS was turned off, as Canon says that it should not be used in conjunction with a tripod. No filters were used on any of the lenses.

The film used was TMX (Kodak T-Max 100), an extremely sharp and fine-grained black and white film. I chose this film because I wanted to be certain that any lack of sharpness would be due to the lenses and not to my choice of film.

For each set of pictures with the same lens at a given focal length, the tripod was moved until the rectangle on the newspaper was approximately filling the viewfinder. Four pictures were taken - one wide open, with the next three at full-stop intervals starting at f/5.6 (if wide open is f/4.5 or faster) or f/8 (if wide open is f/5.6).

Notes were attached to the newspaper for each shot, listing the lens, focal length, and aperture used, in case the prints came back from the lab in a different order.

A Kodak Image Check lab was used to develop the film and produce 4x6" prints. Most of my evaluations were done from naked eye viewing of the prints. I also used an 8x loupe to check details on the prints and on the negatives.

For the scans to illustrate corner sharpness, I have enlarged the first paragraph in the top left corner of the page (right below the words "Too late"). This paragraph reads:

At the crack of noon on
Jan 1, every church bell
in Saskatchewan was
supposed to ring.  Normal-
ly every church bell in
Saskatchewan only rang when (1) an
air raid was about to occur, and (2) on
Sunday before church.  What a time
they would have had!

For centre sharpness, I have enlarged the paragraph below the bottom right side of the drawing near the centre of the photograph (yes, I know, this is not actually the centre of the image). This paragraph reads:

the trouble - that people don't want
to think about difficult things and
hope they'll somehow magically take
care of themselves.  Trust me, Virginia:
Santa knows it doesn't work that way.
The only way to fix things is to do
something about them.)

The sharpness images are not retouched. They are 600 dpi scans from 4x6" prints. The distortion images have been subjected to ruthless contrast enhancement (to the point just before the lines around the corner of the frame begin to disappear), and reduction to 32 gray levels from 256. They are 75 dpi scans from 4x6" prints. If the written words and the scans appear to disagree, trust the written words; I'm quite lacking in talent with a scanner.

Flaws in the procedure

• The use of black and white film means that colour saturation, colour balance, and chromatic aberrations cannot be evaluated. Chromatic aberrations, however, would show up in black and white as reduced sharpness.
• The newspaper is not a particularly high-contrast target, so an evaluation of each lens' contrast is not really possible.
• The newspaper, though taped to the wall at all four corners, was definitely not flat. It should all have been well within the depth of field for each picture, however.
• Lighting was far from even. Therefore, no evaluation of vignetting was done.
• No lens hoods were used, because I only have lens hoods for two of the three lenses and I wanted a level playing field. All three lenses have exposed front elements and many elements, making them all susceptible to flare.
• The camera was not at exactly the same height as the centre of the frame in all cases, leading to some trapezoidal distortion in the pictures. This is, however, easy enough to spot and ignore during evaluation of the pictures. The scans were done using prints and a flatbed scanner; I know I didn't always get the prints exactly square on the scanner, so some of the images are tilted.
• The only space in my home large enough to do this test is carpeted, meaning that the tripod was not on a hard surface. To counteract this, I manually pulled down on the hook on the bottom of the centre column before starting the two-second mirror pre-fire, giving a reasonably stable platform.
• The camera used does not show 100% of the frame in the viewfinder, and the prints also do not include 100% of the frame. The test does not therefore evaluate sharpness to the very edge of the frame.
• Only one picture was taken at each combination of lens, focal length, and aperture. A proper test would involve taking several pictures for each combination to reduce the chances of procedural error (e.g. the tripod not being held quite as steadily for some pictures as for others).
• These pictures were taken at relatively close distances. Most lenses, other than macro lenses, are optimized for shooting at greater distances. These tests really only check performance at the distances used. To test some of them at longer ranges would require huge newspapers - for example, to test a 28mm lens at 10m would require a test target several storeys tall.

Other lens tests on the Web

I'm not the only person who's put up a Web page with tests of Canon EF-mount lenses. As I run across others, I'll put pointers to them here.

• Comparison of the 17-35 f/2.8L USM, 24-105 f/4L IS USM, 28-105 f/3.5-4.5 USM, 28-135 f/3.5-5.6 IS USM, 50 f/1.4 USM, and 100 f/2.8 USM
• My comparison of the 28-135 vs. the 50/1.4
• Test of 28-105 f/3.5-4.5 USM, 28-135 f/3.5-5.6 IS USM, and 28-70 f/2.8L USM
• 50 1.4, 50 1.8, 28-135 f/3.5-5.6 @50mm
• Comparison of Canon's x-300 zooms
• photo.net reviews of many Canon bodies and lenses
• Derek Linney's list of numerous reviews and tests of EF lenses

	Web www.stevedunn.ca

people have visited this page since it was created on 2 January 2000.
Return to writings index
Return to photo page
Return to home page