Reading mode: Light Dark

EOS450D XSI Focus test

Go to last post
V

Vikamy

Member
Messages
31
Reaction score
0
Location
IL
Like many others on this forum, I actively followed the exciting exchange regarding focusing issues in 450D. I was often getting bad focus with the kit lens particularly at large distances. The 45 degrees focus test proved to be OK.

Then I borrowed EFS 17-55 IS and made some shots of the resolution chart from the distance of about 6m. The shots were made with 450D, kit lens and 17-55, 35mm, F4.5, ISO 200, RAW, standard style, tripod.

Auto focus (central point) and live view (live mode) were compared for both lenses. See the results below

http://www.flickr.com/photos/arieltribe/2553903552/in/set-72157605451622117

Please note that the distance to the chart is not right for the absolute resolution measurement but gives a good idea regarding the relative resolution of the images.

It seems to me that there is a significant difference between the auto focus and live focus in the kit lens. While there seems to be little difference for 17-55.

I wonder if this is due to the fact that the second AF sensor (in the center) does not kick in for F less than 2.8.

If what I tested is correct, it would explain the focusing problems that I observed.

I also think that it is too bad that this problem exists in what otherwise are excellent camera and lens.

V

--
'When you have to shoot, shoot, don't talk' Tuco.
 
Sorry,

I guess the link did not work. Please see below



--
'When you have to shoot, shoot, don't talk' Tuco
 
that you refered to. So does the center point focus accurately ONLY if the lens opens to 2.8? If it does then both your test lenses do not open to 2.8. right?

However, your test clearly shows that the 18-55mm IS has a focussing poroblem and the 17-55mm is sharp. So all this focussing talk about the xsi is really the lens? What is your conclusion? I am in the market for a 18-55mm IS and I have an xti and do not want any focussing issues with that combo.
 
So the 17-55IS lens uses the extra precision AF sensor of the 450D, regardless the aperture used to take the shot.
NickSP said:
that you refered to. So does the center point focus accurately ONLY
if the lens opens to 2.8? If it does then both your test lenses do
not open to 2.8. right?
However, your test clearly shows that the 18-55mm IS has a focussing
poroblem and the 17-55mm is sharp. So all this focussing talk about
the xsi is really the lens? What is your conclusion? I am in the
market for a 18-55mm IS and I have an xti and do not want any
focussing issues with that combo.
Click to expand...
--
Slowly learning to use the 450D and and the Canon G6.
Public pictures at http://debra.zenfolio.com/ .
 
NickSP said:
that you refered to. So does the center point focus accurately ONLY
if the lens opens to 2.8? If it does then both your test lenses do
not open to 2.8. right?
Click to expand...
18-55 does not open to 2.8 so I tested them in the same conditions at 4.5. Aparently, Canon uses a second AF sensor for lenses rated less than 2.8.
NickSP said:
However, your test clearly shows that the 18-55mm IS has a focussing
poroblem and the 17-55mm is sharp. So all this focussing talk about
the xsi is really the lens? What is your conclusion? I am in the
market for a 18-55mm IS and I have an xti and do not want any
focussing issues with that combo.
Click to expand...
I am not sure that it is the lens or the combination of the lens and the AF sensor. Otherwise, I think 18-55 is a good and convenient lens. I did not hear of focusing problems with 18-55 and XTI

--
'When you have to shoot, shoot, don't talk' Tuco
 
Vikamy said:
Like many others on this forum, I actively followed the exciting
exchange regarding focusing issues in 450D. I was often getting bad
focus with the kit lens particularly at large distances. The 45
degrees focus test proved to be OK.

Then I borrowed EFS 17-55 IS and made some shots of the resolution
chart from the distance of about 6m. The shots were made with 450D,
kit lens and 17-55, 35mm, F4.5, ISO 200, RAW, standard style, tripod.

Auto focus (central point) and live view (live mode) were compared
for both lenses. See the results below

http://www.flickr.com/photos/arieltribe/2553903552/in/set-72157605451622117

Please note that the distance to the chart is not right for the
absolute resolution measurement but gives a good idea regarding the
relative resolution of the images.

It seems to me that there is a significant difference between the
auto focus and live focus in the kit lens. While there seems to be
little difference for 17-55.

I wonder if this is due to the fact that the second AF sensor (in the
center) does not kick in for F less than 2.8.

If what I tested is correct, it would explain the focusing problems
that I observed.

I also think that it is too bad that this problem exists in what
otherwise are excellent camera and lens.
Click to expand...
This is not a result of the lens not going to f/2.8. Both methods should give fairly comparable results. My lens behaved as yours did, and Canon Service agreed it had a problem. It is now at Canon on its second trip, since their first attempt to fix it made no change at all. They agree it still has a problem. I wish you better luck than I have had. For what it is worth, I also have Canon 50 mm f/1.8 II and 100-300 mm lenses plus a new Tamron 28-300 mm lens, and they all focus perfectly. For me the problem is the kit lens itself or possibly its interaction with the XSi. Single off-center focus spots work much better than using the just central one alone, something you might try out of curiosity if you are so inclined.

Joe

Joe
 
Vikamy said:
I wonder if this is due to the fact that the second AF sensor (in the
center) does not kick in for F less than 2.8.
Click to expand...
Yes, that might be part of it. Also the kit lens is not as well built mechanically, so there may be some tolerance issues there. This would not affect the incremental focus of contrast AF, but for the calculate and go of phase AF, it can be an issue. Think of trying to stop a car on a dime if one of your brake calipers is not working right.

--
Erik
 
Erik Magnuson said:
Vikamy said:
I wonder if this is due to the fact that the second AF sensor (in the
center) does not kick in for F less than 2.8.
Click to expand...
Yes, that might be part of it. Also the kit lens is not as well built
mechanically, so there may be some tolerance issues there. This would
not affect the incremental focus of contrast AF, but for the
calculate and go of phase AF, it can be an issue. Think of trying
to stop a car on a dime if one of your brake calipers is not working
right.

--
Erik
Click to expand...
... so you are saying that with Contrast AF you stop your car way before the dime, and then you push the car slowly until you have it exactly on the dime!

With phase detection AF you just push once, and pray that the car stops where you want?
 
Gerrit Smit said:
... so you are saying that with Contrast AF you stop your car way
before the dime, and then you push the car slowly until you have it
exactly on the dime!
Click to expand...
Close, it's more like you will notice when you run over the dime. If you are going forward, you stop and backup more slowly until you run over it again. Repeat the cycle going more slowly each way until you are are sure you are on the dime.
Gerrit Smit said:
With phase detection AF you just push once, and pray that the car
stops where you want?
Click to expand...
Pretty much. Your job to to drive a go-cart forward and back to center the cart on a line. You can't see out of the car. All you have is a throttle, a brake, odometer calibrated in meters, a watch and a calculator (or lookup table) that says "according to the factory, if you accelerate for X seconds and then slam on the brakes, you will go Y meters." Someone in the passenger seat who can see and has a rangefinder will tell you how far to go -- one time. You are trying to get there as fast as possible.

Phase AF: the passenger measures the distance using a rangefinder and says "go 25 meters." You look at the table and see that means accelerate flat out for 4 seconds and then slam on the brakes. So you do this. If the car and the conditions are exactly to factory specs, you'll stop right at 25 meters. If the brakes are cold or it's a cheap car, you might go 26 meters. You see this on the odometer and know to back up one meter. If your car is out of calibration (e.g. tires are too small), you may go only 24 meters but think you've gone all 25.

Contrast AF: the passenger tells you "I don't know how far. Drive until I say stop." He doesn't use the rangefinder, he just looks through a hole in the bottom of the car. You drive at medium speed until he sees the line go by and says "stop! backup!". Then stop and forward a little more. Repeat until done or you cannot make any smaller movements.

If you think about the Phase AF scenario, you can see where things can go wrong. If the rangefinder is only good for + - 1 meter, you may go 25 meters when you were supposed to go 24 or 26. If the wheelbase of the car is more than 1 meter, you may consider it close enough as the line will still be between the wheels. Obviously, if the passenger used a high precision rangefinder good for 1/3 of a meter you may not have this problem, but then your car is probably shorter as well. If the person sitting in the passenger seat has their seat adjusted forward or backward by a 1/2 meter, then all of the distances will be off by that much. I've already mentioned what happens when your car is out of factory specs for distance travelled. You can send it to a service center where they may test it and give you a new calculator and/or odometer. Or maybe it's just cheaply built and the odometer doesn't have any sub-meter precision so it says 25 meters but you've really gone 25.9 so you don't know you need to back up some to get it perfect.

One obvious question, is why not a hybrid AF that uses phase to get close quickly and contrast to fine tune? Well, it could be done and has been done -- although not for SLRs. The issue is how much time do you save and when does it come? You are looking through the viewfinder. You press the button. The phase AF system calculates the distance and moves the lens. The mirror flips up and the sensor switches into contrast view. It then has to rock the lens back and forth to see if the focus can be improved or not. All the while your mirror is up and shutter lag is increasing. Finally, it's ready to take the photo, so it can stop down the lens and start the exposure cycle. So you may get perfect focus (assuming the subject) has not moved too far), but at the cost of a lot more shutter lag and a lot more mirror up time. What are the failure modes? How long/how far does the contrast system search for better focus? What if the target is leaves blowing or water rippling? In the end, is this faster or easier to use then existing contrast AF?

--
Erik
 
Member said:
What if the target
is leaves blowing or water rippling? In the end, is this faster or
easier to use then existing contrast AF?

--
Erik
Click to expand...
... or my son who never stops ...

So basically, if you don't trust your phase detection AF, half-press the button 2 or 3 times and you may ( may ) get better results.
 
Gerrit Smit said:
So basically, if you don't trust your phase detection AF, half-press
the button 2 or 3 times and you may ( may ) get better results.
Click to expand...
It depends. If the limitation is that the conditions only allow 1 meter resolution for the range finder, it may not help because you are already inside that zone so it won't move. Defocusing and then refocusing may help.

--
Erik
 
jsmiller said:
This is not a result of the lens not going to f/2.8. Both methods
should give fairly comparable results. My lens behaved as yours did,
and Canon Service agreed it had a problem. It is now at Canon on its
second trip, since their first attempt to fix it made no change at
all. They agree it still has a problem. I wish you better luck than I
have had. For what it is worth, I also have Canon 50 mm f/1.8 II and
100-300 mm lenses plus a new Tamron 28-300 mm lens, and they all
focus perfectly. For me the problem is the kit lens itself or
possibly its interaction with the XSi. Single off-center focus spots
work much better than using the just central one alone, something you
might try out of curiosity if you are so inclined.
Click to expand...
Joe, I think this is very interesting input regarding your other lenses. It might be that there is a firmware problem that affects only a certain number of lenses. Hopefully in that case, we may get a firmware upgrade. I really like the kit lens, I think it fits 450 very well and generally is very sharp. I think that 17-55 is better suited for 40D.

V
jsmiller said:
Joe

Joe
Click to expand...
--
'When you have to shoot, shoot, don't talk' Tuco
 
Does anyone have problems like that with other lenses? It looks like I'll need to look for an alternative to my kit lens.
 
NickSP said:
So does the center point focus accurately ONLY if the lens opens to 2.8?
If it does then both your test lenses do not open to 2.8. right?
Click to expand...
AFAIK, the centre AF point operates as a cross-type (both horizontal and vertical arrays of sensors) with lenses of f/5.6 or better. It operates with "enhanced sensitivity" for lenses of f/2.8 or better. My 55-250mm IS focusses perfectly in any mode at f/5.6.
NickSP said:
However, your test clearly shows that the 18-55mm IS has a focussing
poroblem and the 17-55mm is sharp. So all this focussing talk about
the xsi is really the lens? What is your conclusion? I am in the
market for a 18-55mm IS and I have an xti and do not want any
focussing issues with that combo.
Click to expand...
In my experience problems appear with phase detection autofocus with specific instances of lenses, rather than with all instances of certain models. See -
http://forums.dpreview.com/forums/readflat.asp?forum=1031&thread=27977738

Maybe the 450D shows up lens errors more than other models?

Test your lenses following an accurate, objective, and repeatable method like the one I posted, and return any that don't perform correctly.
 
WilbaW said:
In my experience problems appear with phase detection autofocus with
specific instances of lenses, rather than with all instances of
certain models.
http://forums.dpreview.com/forums/readflat.asp?forum=1031&thread=27977738
Click to expand...
WilbaW said:
Maybe the 450D shows up lens errors more than other models?

Test your lenses following an accurate, objective, and repeatable
method like the one I posted, and return any that don't perform
correctly.
Click to expand...
Thanks for the link. Good tests. Please note that I do not observe any issues with either lens at close distance (45 degree test). It seems that the main problem is shooting at distances over 3 meters. Under these circumstances, it is very easy to reproduce the results. I performed this simple test (see the set up below) in my office.



I took 3 shots in each position under 2 different types of illumination. In every case ESF 18-55 shots were NOT in focus for AF with central point while Live View focus was always IN focus. I think this is very consistent with observations by many other on this forum.

IMHO the problem is related to a bug in Canon's AF firmware, which generates a false positive AF result or may be sends the wrong timing signal to the lens motor. In either case, the problem should be easily fixable in a new firmware release.

--
'When you have to shoot, shoot, don't talk' Tuco
 
Vikamy said:
Joe, I think this is very interesting input regarding your other
lenses. It might be that there is a firmware problem that affects
only a certain number of lenses. Hopefully in that case, we may get a
firmware upgrade. I really like the kit lens, I think it fits 450
very well and generally is very sharp. I think that 17-55 is better
suited for 40D.
Click to expand...
Yes, this is a very important point. You will read many posts laying blame on the phase method of focussing. While fundamentally it wouldn't be expected to be as accurate as the contrast method in Live View, it should be quite accurate, or Canon would have a hard time selling cameras. You example shows very poor focus, as mine did. As I said, my other three lenses, gave virtually identically sharp focus in both methods in many pictures with many aperture and FL settings and light conditions. They work properly. The kit lens does not. There is no other explanation than a faulty lens or interaction of lens and camera.

For a further test, take a picture of something quite distant with the lens wide open at 18 mm FL. Use the center focus spot only, then use an off center spot only, making sure that that off-center spot is only on very distant objects. On my camera the centrally-focused picture is out of focus, while the one taken with the off-center spot is almost as well-focused as the Live View contrast focus. Something strange is going on. Of course your camera could be different.

Joe
 
 
You must log in or register to reply here.

Keyboard shortcuts

f
Forum
m
My threads
Mobile site
About
Editorial content
Cameras & Lenses
Community
All content, design, and layout are Copyright © 1998–2025 Digital Photography Review All Rights Reserved.
Reproduction in whole or part in any form or medium without specific written permission is prohibited.
When you use DPReview links to buy products, the site may earn a commission.
©GPS Media - Guides, Products, Services.
Back
Top