(From Canon lens literature) Ultra-wide-angle lens for serious applications. Easy to hold and carry at 14.3 oz. (405g). Floating rear focusing system. Sharp images are obtained at all subject distances.
July 30, 2008
by Andrew Alexander
The Canon 20mm ƒ/2.8 is a small wide-angle lens, first introduced by Canon in 1992. It uses a design of 11 lens elements in 9 groups, and the use of a floating rear focusing system improves AF performance by not requiring the movement of many elements.
The 20mm ƒ/2.8 was released long before the advent of digital SLR camera bodies, designed to fit a full-size 36 x 24mm film frame. Thus, it will have no problems mounting on either Canon full- or sub-frame camera bodies. On a Canon digital body, it will have an effective field of view of either 32mm (1.6x) or 26mm (1.3x).
The 20mm ƒ/2.8 is still actively manufactured by Canon, takes 72mm filters, and it available from dealers for around $450.
Image sharpness provided by the Canon 20mm ƒ/2.8 when used at its widest aperture (ƒ/2.8) is somewhat soft, a fairly ''lumpy'' profile between 3 and 5 blur units. However, stopped down to just ƒ/4, overall sharpness improves dramatically. At ƒ/4, we note excellent central sharpness (just over 1 blur unit) and very slightly soft corners (around 2 blur units). Stopped down further, there's only marginal improvement at ƒ/5.6, the optimal aperture. At ƒ/8 and smaller, diffraction limiting is already setting in, but it isn't until ƒ/16 that we note any real reduction in image sharpness. Even at ƒ/22, we're back to a generalized softness across the image, between 3 and 4 blur units.
Mounting the 20mm ƒ/2.8 on the full-frame 5D shows off the wide-angle nature of this lens, warts and all. Wide-angle lenses shot wide open typically struggle with image sharpness in the corners, perhaps due to field curvature, and the 20mm certainly isn't an exception. At ƒ/2.8, central sharpness is good at between 2 and 3 blur units, but corner softness is extremely noticeable - 8 blur units, and even going off the chart in the bottom-left of the image. Corner softness is a factor at almost every aperture setting when shooting full-frame, with the possible exception of ƒ/11, where it is (only) 3 blur units in the corners. At ƒ/4, central sharpness is very good, even excellent, but the corners are still extremely soft - between 5 and 7 blur units.
Overall, sharpness performance is best on sub-frame cameras between ƒ/4 and ƒ/8; on full-frame cameras, corner softness is always an issue (more or less), with the most even results coming only at ƒ/11.
The 20mm ƒ/2.8 copes fairly well with chromatic aberration. On the sub-frame 20D, the lens exhibits more CA at smaller apertures than when it it shot wide-open. At ƒ/2.8, CA is rising in a fairly linear manner from 3/100ths of a percent of frame height generally and 6/100ths in the corners, to 5/100ths generally and 9/100ths in the corners at ƒ/22.
On the full-frame 5D, light comes in at a greater angle in the corners (sub-frame sensors focus on the ''sweet spot'' of the lens) and consequently, CA is more prominent in the corners of the image. At ƒ/2.8, the 20mm shows significant CA in the corners - 9/100ths of a percent of frame height - but this reduces significantly as the lens is stopped down. Outside the corners, CA is well-controlled, showing up at only 3/100ths of a percent regardless of the aperture selected.
Corner shading isn't a real problem with the 20mm mounted on the sub-frame 20D. At ƒ/2.8 the corners are 2/3EV darker than the center of the image, and stopping down to ƒ/4 removes the shading almost completely.
It's a different story with the 20mm mounted on the full-frame 5D, however. Corner darkening is some of the most significant we've seen: at ƒ/2.8, the corners are 2 1/4 stops darker than the center. This shading remains an issue at every aperture: 1.5 stops darker at ƒ/4, one stop darker at ƒ/5.6, and levelling out at about 2/3EV darker at all other apertures. If you're looking for this kind of built-in soft vignetting, this may be the lens for you, but otherwise, brush up on your anti-vignetting controls in your favorite post-processing software.
As the 20mm is a fixed wide-angle lens, we aren't surprised to see some barrel distortion, and the 20mm shows a fairly conservative 0.5% barrel distortion in the corners, and 0.3% generally.
The 20mm ƒ/2.8 is a USM lens, so it focuses very quickly and almost silently. Full-time manual focus override is available by simply turning the focus ring at any time.
The 20mm ƒ/2.8 isn't a very good macro lens, with a magnification ratio of 0.19x and a comparatively long minimum close-focusing distance of 25 cm.
Build Quality and Handling
The Canon 20mm ƒ/2.8 USM is a solid lens, coated with a black semi-gloss finish that resists wear and tear. It's built with a metal lens mount and plastic filter thread. It's small but stout, about as wide in diameter as it is long (2.8 in. x 3.1 in.).
The lens only has a few features to talk about, the most noteable being the windowed distance scale measuring distance in feet and meters. A depth-of-field scale is marked for ƒ/11, ƒ/16 and ƒ/22 apertures, and there's also an infrared index marker.
Apart from the focus ring, the only control structure is the manual focus / autofocus activation switch, a simple toggle on the left side of the lens. The focus ring is a ridged rubber, 1/2 inch wide. Manual focus operation on this lens is not as smooth as we have noted with other USM lenses, travelling 90 degrees to cover the entire focus range. The lens accepts 72mm filters, which do not rotate during autofocusing operations.
The 20mm accepts the optional EW-75 or EW-75II lens hood, a plastic petal-shaped hood which adds 1 1/2'' to the overall length and helps the lens to resist flare. It's worth it to use the lens hood, as flare is a noted problem with this lens, showing up as a string of hexagons due to the lens' 6-bladed diaphragm. Note that Canon's literature indicates the lens has 5 blades, but we've confirmed it has six.
Canon EF 24mm ƒ/2.8 ~$300
If you don't need to go as wide as 20mm, the 24mm is slightly less expensive, but lacks the USM focusing. Sharpness is slightly better at ƒ/2.8, but CA, corner shading and distortion are all about the same. Equally questionable on full-frame bodies.
Sigma 20mm ƒ/1.8 EX DG Aspherical RF ~$400
We haven't tested the Sigma 20mm, but it is a full stop-and-a-third faster than the Canon 20mm. The lens doesn't incorporate Sigma's hypersonic motor equivalent (HSM) and takes massive 82mm filters. Better macro, 35mm compatible, and sells for slightly cheaper.
Canon EF 16-35mm ƒ/2.8L II USM ~$1,500
This lens is in quite the different category than the 24mm prime, but it is indeed an option if price is no object. Sharpness, CA performance and corner shading are all much better, and distortion is about the same. Also a USM lens, as a zoom lens it adds the benefit of a wide range of focal lengths.
Canon EF 14mm ƒ/2.8L II USM ~$2,200
If you're a subframe Canon digital SLR camera user and you're desperate for a field coverage similar to the 20mm, the 14mm is the way to go, giving an equivalent view of 22mm. It doesn't come without a significant price tag - almost five times the cost of the 20mm - but the 14mm outclasses the 20mm in almost every way (sharpness, corner shading, distortion), with the exception of some CA issues when shot wide open.
Canon EF-S 18-55mm ƒ/3.5-5.6 IS ~$150
If you don't need the ƒ/2.8 aperture, and are working on a subframe digital SLR, this lens makes a viable alternative. The 18-55mm will provide ƒ/4 at 20mm, and the optical performance at this focal length setting is superb. CA is a little worse, but sharpness is very good, and distortion and corner shading are about the same. Add the benefits of image stabilization, and that you're also spending less for this lens.
The conclusion for this review comes down to recommendations, and we have to segregate the audience into a couple of camps.
For full-frame digital SLR or film users, the 20mm ƒ/2.8 shows off its poorer performance issues, especially for sharpness and light falloff. Whether this performance is poor however, depends on your point of view; if you're looking for a lens which isolates a centrally-located subject by means of sharpness and light falloff, the lens does it spectacularly. If you need even light distribution and even sharpness, the 20mm prime will disappoint, especially at ƒ/2.8. At other apertures, you have a better chance of getting what you need.
For subframe digital SLR users, the lens performs a bit better, but at ƒ/2.8 image sharpness is still poorer than we'd like. At ƒ/4 and smaller, the lens performs wonderfully, so if you don't mind using it at this setting, you really can't go wrong.
The VFA target should give you a good idea of sharpness in the center and corners, as well as some idea of the extent of barrel or pincushion distortion and chromatic aberration, while the Still Life subject may help in judging contrast and color. We shoot both images using the default JPEG settings and manual white balance of our test bodies, so the images should be quite consistent from lens to lens.
As appropriate, we shoot these with both full-frame and sub-frame bodies, at a range of focal lengths, and at both maximum aperture and ƒ/8. For the ''VFA'' target (the viewfinder accuracy target from Imaging Resource), we also provide sample crops from the center and upper-left corner of each shot, so you can quickly get a sense of relative sharpness, without having to download and inspect the full-res images. To avoid space limitations with the layout of our review pages, indexes to the test shots launch in separate windows.