Canon currently has a number of choices for a general purpose zoom lens. Three
of the choices are the EF 28-90mm f/4-5.6 (USM and non-USM), the EF 28-105mm
f/3.5-4.5 USM and the EF 28-135mm f/3.5-5.6 IS USM and these span the range from
around $100 to $400 - so one of these Canon zoom lenses will probably fit your
There's also a EF 28-80mm f/3.5-5.6 lens which I believe is out of production
and a EF 28-105mm f/4-5.6 USM (not to be confused with the
28-105/3.5-4.5 USM). Both are "low end" lenses, comparable to
the 28-90/4-5.6 USM, but since I have access to neither, I didn't test them!
The Zoom Lenses
Left to right: Canon 28-135/3.5-5.6 IS USM; Canon
28-105/3.5-4.5 USM; Canon 28-90/4-5.6 USM
All lenses shown zoomed to maximum focal length
Canon EF 28-90mm f/4-5.6 (USM) - Price class $100
This is Canon's low end basic zoom lens, often sold in "kits" with an EOS
body. It's small, it's light and it's cheap. The construction uses lightweight
plastic, even for the lens mount - though the optics are all glass. The lens
feels fairly flimsy and I don't think I'd want to try drop testing it!. Though it
has manual focus, it has no manual focus ring. To focus manually you rotate the
front end of the lens barrel. Focusing is loose and undamped and quite sensitive
to small rotations of the barrel. In addition the barrel wobbles up and down
quite a bit, which isn't a good sign for optical alignment! There is no distance
scale (hence no IR focus markings). A basic lens at a low price. Though the lens
is available in USM and MM (Micro Motor) versions, this isn't the ring USM motor
with FTM (full time manual focus). There really isn't a significant difference
between the MM version and the USM version, though Canon claim the USM version is
quieter and focuses faster. It probably is and does, though I'm sure the
difference isn't very significant.
EF 28-105mm f/3.5-4.5 USM - Price class $200
This Canon zoom lens is a huge step up. It's solidly constructed with a metal
lens mount. It has a ring USM motor with full time manual focus, a distance scale
and IR focusing marks. It's a little bigger, a little heavier and about twice as
expensive as the 28-90, but it's a much better constructed lens. Manual focus is
smooth, well damped and geared so that fine adjustments can be made, as you would
expect from a ring motor FTM USM based lens. The current version is the "II"
model, though it is optically identical to the original model which was tested
here. The differences are mainly cosmetic.
EF 28-135mm f/3.5-5.6 IS USM - Price class $400
This Canon zoom lens is significantly larger and heavier than the 28-105 but
is of a similar solid construction with a metal lens mount, ring USM motor, FTM,
distance scale and IR focusing marks. It also has image stabilization which
allows it to be hand held at shutter speeds about two stops slower than would
otherwise be possible and still get sharp images. It costs about twice as much as
the 28-105. It takes a significantly larger filter than the other two lenses
(67mm vs. 58mm). Manual focus is smooth, well damped and geared so that fine
adjustments can be made, as you would expect from a ring motor FTM USM based
lens. This lens uses one aspheric element (neither of the other lenses tested
have aspheric elements) which may contribute to better performance - at least
that's the idea behind using aspheric elements
LENS (all USM)
Size (length x diam)
Ring USM with FTM
68mm x 71mm
72mm x 75mm
97mm x 78mm
CAVEAT:Testing was done using a Canon EOS 10D
DSLR body. The Canon 10D has the smallest pixel pitch (7.7 microns) and so the
highest native resolution of any current DSLR. It can resolve around 65 lp/mm at
the sensor. Film can resolve more, typically up to around 80-90 lp/mm with slow
speed, high resolution films and the best lenses - though unless you get drum
scans you probably won't see that level of detail in scanned images. Digital
testing with a 10D tends to reduce the difference seen between lenses as a
consequence of this resolution limit, so if you see a difference in performance
between lenses when tested on a 10D, you'll see at least that much difference,
probably more when using film.
Also, the 10D has a sensor significantly smaller than 35mm film (about
22mm x 15mm vs. 36mm x 24mm) and so does not show edge and corner defects which
you might see on film. For example most lenses vignette slightly when used wide
open, so the corners of the image on film are often slightly darker then the
center of the image. Since the 10D image sensor is smaller then film, such
vignetting is very much reduced. Similarly many aberrations increase as the
distance from the optical axis increases. For 35mm film, the corners are 21.5mm
from the optical axis, for the 10D the corners are 13.3mm from the optical axis,
so corner image quality is expected to be higher for 10D digital images than for
full frame 35mm film images.
Obviously if you are shopping for a lens for your 10D (or other Canon
DSLR with a similar sensor size), the the results of the tests presented here
will be directly applicable. If you are shopping for a lens for your film camera
then bear in mind the limitations outlined above. However the zoom lens that
tests better on a 10D will almost certainly be the better performer when used
with a 35mm film body too.
One problem with testing zooms, and comparing three different zooms, is that
you can generate huge amounts of data. You can test at 5 different focal length
settings at 6 different apertures at both center and edge for each of the zoom
lenses. That's 180 different images to compare. Rest assured I'm not going to
present 180 images here or it would take me a month to write the review and all
day to download the page.
Instead I'm going present a subset of the images, illustrating some basic
points, but I'm going to draw conclusions based on my own assessment of a larger
set of images and you'll just have to take my word for the results I present
Note that all of the image here represent very close examination of the image,
much closer than you'd normally see in a print. Some of the crops are 100%, in
that case a 200x200 pixel sample would be a 1" square section of a 10x15 print.
Some of the crops are at 400%, in which case a 200x200 sample would be a 1/4"
square in a 10x15 print.
Focusing is the most critical issue in any lens test.
If you miss focus even slightly you can draw incorrect conclusions about lens
performance. You can't just blindly assume that AF works perfectly every time
with every lens. Usually it does, but this is a lens test, not an AF test. For
these tests the correspondence between autofocus and manual focus was checked.
Manual focus was assisted by the use of the Canon Viewfinder Magnifier. Focus
bracketing was used to make sure that the viewfinder screen gave an accurate
indication of the final image (it did!).
Here's the first set of images, all shot wide open at 28mm (f4 for the 28-90,
f3.5 for the 28-105 and 28-135). The image is cropped from the center of a test
target shot from a distance of about 6m. The target was in direct bright sunlight
and exposed so as to put the white background of the target close to a pure
white. Since the black areas are a small fraction of the total white areas, if
any flare is present, it will show up in reduced contrast
Looking at the three top images showing the performance wide
open, the gray scale strip at the top of each image shows that the 28-135 has the
best color/lowest flare. The 28-105 shot shows some evidence of flare since the
left most block is a little light and a little blue. The 28-90 performance is
commendable, with maybe a little less flare evident than the 28-105 (though it is
at f4 vs. f3.5). Remember that these crops are 138 pixels square out of the
original 3072 x 2048
At f9.0 and 28mm all three lenses are pretty similar in the
center as can be seen from the lower three images above. Any differences between
these images are clearly insignificant. All three Canon zoom lenses show some
improvement over their performance wide open. Contrast is higher and flare is
lower in the case of the 28-105 and 28-90 lenses as can be seen in the dark
patches on the left of the gray scales. The 28-135 is pretty good wide open, so
shows less contrast improvement, though a resolution increase can be seen
The following crops are from the top left corner of the image,
shot wide open at 28mm
It's evident here that the 28-90 is the worst performer of the
group, showing more flare and less sharpness than the others. The 28-135 seems to
have a small advantage over the 28-105, but again remember that these are 36x63
pixel crops from the original 3072 x 2048 image so the effects you are seeing
here are fairly subtle. They are reproduced here at 400% enlargement. I doubt
you'd see them in a 4x6 print, and I doubt they would be very noticeable in an
8x10. At 300 dpi in an 8x12 print, these crops would be about 3mm x 5mm
Here's a shot of a difficult subject at the edge of the frame, 28mm at f9.0.
White/black transitions show up color fringing you'd never see on a "normal"
image. The 28-90 is clearly not quite so sharp and shows more color fringing.
However note that these are 400% enlargements from the original (you can clearly
see the individual pixels), so the fringing isn't quite so bad as it may appear
in these images.
Below are the results of a test at 50mm focal length. These crops are taken
from the center of the image
Again it's clear that there is no big difference in resolution,
however it's evident that the 28-90 zoom lens shows lower contrast, especially
when used wide open. You can tell this by looking at the density of the
horizontal shadow at the top of each crop. The 28-105 and 28-135 are very close
in performance. If I had to pick one, I'd say the 28-135 has slightly better
contrast, especially wide open. When stopped down to f9.0, all three lenses are
Samples from the edge of the image are shown below. Again, wide
open the 28-90 has the lowest level of performance, with the 28-105 being
slightly better and the 28-135 slightly better still. At f9.0, all three Canon
zoom lenses are quite similar.
Below is another example of relative performance at 50mm, this time in the
rendering of specular highlights. The shot is of a polished chrome car door lock
with the sun reflected off it:
As you can see, the cheaper lens does not do very well wide open and renders
the bright specular highlight with a blue flare around it. Stopped down the three
zoom lenses are pretty equal.
50mm - how about a prime?
The question always seem to come up whether zooms are as good as primes. Well
lets look at the EF50/1.8 prime and see how it stacks up against the zooms at
50mm. To make this fair, all the lenses were set to f4.5, which is more or less
wide open for the zooms, but more than 3 stops closed for the 50mm prime.
I think the prime has a slight edge here. Contrast is certainly
slightly better than the 28-135 and 28-105 and significantly better than the
28-90. Resolution isn't very different though. So lets look at the edge of the
Well, not too much difference between the prime and the 28-135 and 28-105
zooms, but the 28-90 is clearly suffering pretty badly at the edge. Is the prime
significantly better then the two more expensive zooms - In my opinion, no , it's
not. But it is three stops faster and 3 stops for $80
is a pretty good deal!
Finally here are the images at 90mm. There's a slight difference in
magnification since 90mm isn't marked on either the 28-105 or 28-135 lenses, so I
had to make a guess as to where to set the zoom ring. These images show both a
normal resolution test pattern and a very low contrast test pattern. Again it's
evident that the 29-90 has the lowest contrast and resolution wide open, and this
time stopping down doesn't really help much. The other two lenses are fairly
equal, with the 28-135 perhaps showing just slightly better contrast, though it's
a tough call.
The images below show edge performance at 90mm. It's fairly clear that the
28-90 lags behind the other two lenses wide open. Both contrast and resolution
are low. It improves when stopped down but still isn't quite as sharp as the
28-105 or 28-135. It's tough to see much difference between the two better
lenses. If there is one it's too small to be of any significance.
What about the Canon zoom lenses you didn't test
Well, I didn't test them because I don't have access to them, but in the
spirit of photo.net I'll still make a few comments!
There is a 28-80 low end zoom which was, I think, replaced by the 28-90.
Plastic construction, no distance scales etc.etc. From what I've read and from my
general experience with Canon lenses, I'd assume it's pretty similar to the
There is a current 28-105/4-5.6 USM which is NOT,
repeat, NOT to be confused with the 28-105/3.5-4.5 USM
tested here. The 4-5.6 version is another low end plastic lens in the same family
as the 28-90. Since it's only about $50 cheaper than the 3.5-4.5, I can see no
earthly reason to consider buying one.
There is a 24-85/3.5-4.5 USM which is priced between the 28-105/3.5-4.5 and
the 28-135/3.5-5.6. It's a mid range lens like the other two, with full time
manual focus, a ring USM, distance scales etc. It has a good reputation and is
certainly an alternative if you are prepared to trade off the telephoto end for
the wide-angle end of the range.
There is a 28-200/3.5-5.6. I'd be very interested in testing this lens, but I
think it would be most fair to test it against 3rd party lenses of similar zoom
Then there are the high end "L" series lenses of which the 24-70/2.8L USM is
the current production version. However at $1400, it's not really a lens most
consumers will consider and it's not one that's in my camera bag, so a test on
that lens will have to wait until someone gives one to me...
Conclusions (about the zoom lenses I DID test)
The Canon EF 28-90mm low end lens does fairly well. True, it's not generally
as good as the more expensive lenses when used wide open, particularly at 90mm,
but it does quite well when stopped down. Edge performance isn't quite so good as
the better lenses, but the difference is pretty small, especially stopped down at
shorter focal lengths. All in all an OK performance and certainly capable of
yielding decent quality small prints and maybe even the occasional 8x10. If
you're looking for the smallest, lightest, cheapest lens and you don't intend to
use it wide open all the time then it could be a reasonable choice. It's not a
very sturdy lens and the ergonomics aren't great either (no separate focus ring,
no distance scale), but it's "only" around $100 and these days $100 doesn't buy
you much more than this in an autofocus lens. I don't think I'd go so far as to
actually recommend this lens, but if you have one I wouldn't recommend throwing
it away either!
The Canon EF 28-105mm is a very nice lens to use, it's well built and has good
optics, certainly better than the 28-90. At around $200 I think it's a bargain
and a much better buy than the 28-90 as long as you don't mind the
slightly larger size and increased weight. You also get a longer reach (105 vs.
90), which can be useful. I've owned this lens for many years and I've never been
disappointed with it. I can certainly recommend this lens to anyone wanting good
performance at a reasonable price.
The Canon EF 28-135mm is clearly the winner from an optical and functional
viewpoint. Optically it's as good as or slightly better than the 28-105, plus it
has an extra 30mm of reach. The IS function really works and allows you to
handhold at 1/8s at 28mm, whereas without IS you might need 1/30s for sharp
images. If you're stuck without a tripod this could be the difference between
getting a shot and losing it. The downside of the lens is its extra cost (about
$200 more than the 28-105) and its significantly increased size and weight, as
well as the need for a larger (and more expensive) filter (72mm vs. 58mm). If
size weight and cost aren't an issue, it's the better lens. I got this lens in a
deal where I bought a camera and this lens came with it. I had intended to sell
the lens, but after using it for a while I'm quite reluctant to part with it! The
IS feature is really nice...
What would I chose if I had no lens in this range? If my funds were limited I
wouldn't totally dismiss the 28-90, though I think if at all possible I'd skip
the beer and pizza for a month, save up another $100 and go for the 28-105. It's
a significantly better lens and well worth the extra cost. The major difference
between the 28-105 and 28-135 are in terms of functionality (IS and 135mm vs.
105mm), not so much in optical quality - though I would say the 28-135 does have
a slight edge, especially when used wide open. Which Canon zoom lens you buy
between these two is very much a personal decision depending mainly on just how
much you think you'll need the IS function and whether you think you'll use the
extra range. Be warned though, the IS function is seductive and once you've used
it it's hard to give it up!
Where to buy the Canon zoom lenses
Depending on where you buy, which version of the lens you buy and whether you
buy gray market or USA versions, the 28-90 will cost you $100-$120, the 28-105
will cost you $200-220 and the 28-135 will cost you around $400.
? Copyright 2003 Robert M. Atkins All Rights