This doesn't jibe with that:

(http://www.cps.canon-europe.com/kb/detail.jsp?faqId=1110)

Anyway, if a 50mm f/1.4 does misfocus under repeatable
circumstances, it and the camera should find their fastest way
to a canon service shot, as they do not act as they should.

-Wolfgang

Re above - the lens used to create phase detection signals (in focus
when at maximum amplitude and coincident in phase) is in one physical
position. They can't put two sensors at different places in the focal
plane but aim them at the same point in the image. The best that can be
done is two sensors working a bit like a split-image rangefinder (top
and bottom halves) and a lens system which works like the split-image
angled prisms, in effect.

There is always an optimum aperture. If the system is going to work at
f6.3 (necessary with modern zooms) it may not be very good at f1.4. If
it was optimised for f2.8, it probably wouldn't function at f6.3. That
applies to many early AF SLRs - they work down to say f5.6, and may JUST
scrape by with 28-300mm or something similar which has an f6.3 max at
the long end. Fortunately it's also connected to focal length and actual
physical aperture, hence Minolta being able to make a 500mm f8 mirror
lens which could autofocus when a 200mm f4 fitted with a 2X converter
(also f8) won't. The 500mm f8 doughnut aperture is actually like a dim
f6.3 circular aperture and fools the system.

David

ACK! Not hardly.

I compared the 70-300 IS when it came out with my 70-200 f/4, and the
image quality, while better than the older 75-300 models, is still not
up to the L glass even when stopped down to f/8.

It's not just sharpness either, colour and contrast, and CA are better
with the f/4 as well. Not to mention the other obvious features.

But needs vary, so the 70-300 has its uses.

I have one of the 70-300 IS f/4.0-5.6USM with the affected serial number.  I
am extremely pleased with this lens anyway.  It is so much better optically
than the Sigma 70-300mm f/4-5.6 DG macro and the Canon 75-300 F/4-5.6 lenses
both of which I have used.  I'll probably send it in for repair eventually
though.  The Sigma focuses closer for more magnification (up to 0.5) and has
distance, magnification and depth of field scales on printed on it and is a
bit shorter and much less expensive.  That is the limit of the Sigma's
benefits.  The Canon 70-300 is so much sharper, especially out at 300mm and
has only the tiniest bit of CA where the sigma has plenty.  Canon's IS on
that lens is excellent too, with none of the drift that I see on the
28-135mm IS at slow shutters speeds.  The 70-300 is really convienient and
does well enough in low light.  With a monopod, it is even better.

The max aperture changes as follows:

in 1/3 stop mode:
70- 89mm  F/4.0
90-134mm  F/4.5
135-234mm  F/5.0
235-300mm  F/5.6

in 1/2 stop mode:
70-109mm  F/4.0
110-234mm  F/4.5 (why isn't it F/4.8?)
235-300mm  F/5.6

So the 70-200mm F/4L is only 2/3 stop faster at the long end or 1/3 stop
faster if you set the camera for 1/2 stops.  I'm not suggesting the 70-300
is better than the 70-200 at all though but it is very good for many people
and better than most of the other consumer grade lenses in the xx-300 or
xx-200 range

The 70-300 IS has a few things that some people won't like.  It's not dust
or weather sealed.  It doesn't have distance, depth of field, or
magnification scales printed on the lens.  It's not completely parfocal.
The front element extends (but doesn't rotate) during zooming and extends
and rotates during focussing so you have to reset your polarizer after
focussing.  Sometimes the focus hunts a bit and isn't super fast.  The focus
ring is fairly fast with manual focussing so it makes manual focus not easy.
The focus ring and lens hood rotates during autofocussing.  Canon doesn't
include the lens hood in the package and charges a lot for it although it is
a very nice good.  The 70-300 is just a bit too long to fit in most camera
bags when mounted on the camera.  The Sigma fits nicely in most.  Canon has
included a zoom lock for 70mm but if it is disengaged, the lens suffers from
serious zoom creep. You have to keep your hand on the zoom ring when
pointing up or down.

So optically it is good, but the lens is not without its faults.  I think it
is reasonable value for the money though.

I've found that Photozone tests are not always balanced, and sometimes
they're just way off.

They don't always reflect real world experiences, for whatever reasons,
which is why I don't refer people to their tests. If you examine the
tests, they show the f/4 dropping off towards the long end, which is
backwards - it gets sharper.

And the sample images are shot wide open with the f/4, while the 300 is
stopped down to improve image quality. That makes sense if you're trying
to show the best it can do. But for comparisons, it makes the results
and data uneven.

Either they had a poor sample of the 70-200 and a very good sample of
the 70-300 IS, or their tests are flawed.

I've used both, and the 70-200 is substantially sharper than the 70-300
IS when used wide open. Not a little bit, a lot sharper. All of the
70-200 units I've used are very sharp especially toward the long end. I
don't think I've ever used a sharper zoom lense in a similar range,
including Canons own f/2.8 versions (wide open).

Once you stop down to f/8 they get close, but at that point sharpness is
not where the differences reside, it's contrast and colour, and CA,
which I also noticed in the tests is not what you get in real world
results.

I've also used several Nikon lenses, and on one hand, they indicate the
18-70 performs great, while the 70-200 VR less so. And once again,
personal experience shows their test results are not accurate.

So don't always put stock into what resolution tests indicate from
Photozone. Perhaps use it as a general guide, but I wouldn't make buying
decisions based on their data and info.

YMMV...