auto focus issue of 70-300 IS

[lishen]

I just bought an used copy of EF 70-300/4-5.6 IS, and hoped it could provide more reliable AF than 55-250 IS does. Yes, it's AF capability seems to be superior, since I was able to get higher percentage of sharp images especially when shooting moving subjects. But however, I found it's still far from reliable, even when shooting landscape on tripod. Below are example shots with 70-300@300mm on tripod. I gave the same subject 4 shots, and you could obviously see the significant IQ difference in the 100% crops.

The original image in daylight:



The 100% crops:



The original image:



The 100% crops:



Such difference are most obvious at 300mm, but it can also be reproduced at 200mm and below. I'm sure the difference is not caused by camera shake, since I can get consistent sharp images if I use MF with the help of live view; in addition, I also turned off IS, enabled mirror lock-up, used self-timer, used the same center focus point of 40D... I think the inconsistent AF leads to these results.

I went to the Canon service center in China. They told me this phenomenon is probably normal, and we can't expect the micro-USM motor has the same accuracy as that of 70-200L does (even in bright light as the above shots show).

What do you think? Do I get a bad copy? Or it's just all right.

Thanks for your help!

 

[comeon]

--

well i dont know if you have a "bad copy" but i can say this ...it likes like you were a mile away when you took the photo ...you click these of 1 after the other ..?? and zooming in IE pixel peeping that tight ...when would ever do that in real life ... i would rather see two of the originals and a 100% crop ..are you sure about that ..looks like 400% or better to me ...

 

[lishen]

Yes, the subject could be 1 mile away, and I took these shots one after another (every shot was triggered by the 10s self timer). All of these crops are of 100% (not 400%). I think for landscape, the 100% details are important. I have 70-200L, 24-105L, etc, and they could give me consistent sharp images in such situation; maybe larger aperture and superior USM are also important to landscape shooting if I want to good AF accuracy?

 

[lishen]

In very bright sunlight, the performance of this lens looks really good to me.

Original image (260mm, F/8, directly from DPP)



100% crop:



original image (300mm, F/10)



100% crop:



I'm really satisfied with such IQ. If this lens could perform more consistently when shooting landscape, I will probably sell my 70-200/2.8 IS.

 

[dbm305]

Might not be focus. At 300mm even on a tripod you will get a lot of variation in how sharp the image is, even at moderstly fast shutterspeeds. Even with the mirror locked up! Wind, distant cars vibrating. If you care abotu 100% sharpness you need to take multiple pics.

 

[lishen]

Thanks! Yes, I agree. But as mentioned prevoiusly, I can get consistently sharp images when using MF at 300mm, so I may safely eliminate the possibility of camera shake here. Actually, I did all the test shots at home. In addition, even at 200mm or below, I'm able to reproduce this issue.

 

[lishen]

Below are 2 group of images taken by 70-300 and 70-200L respetively. Both are set to 200mm, F/8, 6s.

100% central crops from 70-200L



crops from 70-300



The IQ from 70-200L are constantly high, but only the first image in the group of 70-300 is good at 100% (might not that obvoius lilke the above shots at 300mm though)

Am I too critical here?

 

[Al Bond]

Another difference between the lenses is the maximum aperture. The f2.8 lens will use the high precision central AF sensor so should always be more accurate.

 

[lishen]

Yeah. I think so. But I also hope 70-300 could give good AF accuracy in landscape shots (not in very dim light), in which the DOF should be wide. Moreover, even if I use F/16, the IQ still obviously varies from shot to shot. This is really a bad thing to me, and the unreliable AF perhaps means that I may miss some important opportunities even if I don't care about the little loss in IQ as the above examples show, since I can get landscape shots totally out of focus when the contrast of the subject becomes even lower.

 

[Montana500]

For softness and focus issues at the long end. Make sure this uit has been fixed under that recall. I had to send mine in. It was referred to as "vertical orientation problem".

 

[lishen]

Thanks. I think my copy should be no problem since it's produced in 2008. You copy of 70-300 now performs very well?

 

[Montana500]

Yes, it was much better after being repaired.

 

[jpr2]

Thanks. I think my copy should be no problem since it's produced in
2008. You copy of 70-300 now performs very well?

if it has "3" or "4" as a third digit, then it should not be affected by the issue discussed, btw. mine performs rather well for landscapes too at 300, even handheld, but the light and contrast need both be really top notch - I think
you're right pinpointing it as AF (and not vibration) problem,

jpr2
--
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
wildlife, macro, B&W, and 'interactive' street:
http://www.flickriver.com/photos/qmusaget/sets/72157600341377106/
street candids (non-interactive):
http://www.flickriver.com/photos/qmusaget/sets/72157609618638319/
Comments and critique are always welcome!
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

 

[lishen]

That's great!

 

[lishen]

Thanks! Mine is "6", so I think it shouldn't suffer from the problem mentioned before.

Below please find some real-world landscape shots that might be able to prove the reliability issue of this lens.

sharp one



100% crop near the focus point



soft one



100% crop near the focus point



I got 3 sharp images and 2 soft images in total. The only difference between the 2 groups of images is the composition. Different compositions lead to slightly different focus positions (always using the central focus point of 40d), and this might affected the AF accuracy.

Actually, I found that when I switch the AF mode to "AI servo", this lens is often difficult to totally lock the subject at 300mm in bright light, even though both the camera and subject don't move at all. It just keeps trying, and I can hear continuous noise. Could this mean that in the normal "one shot" AF mode, at the long end there's chance that this lens actually don't perfectly focus on the subject even though the camera indicates the success of AF? So the subject may be totally out of focus in the final image. At 70mm, I seemed to rarely run into such issue, and it's quiet in the "AI servo" mode if no movement.

Moreover, I found when shooting landscape, I could often get better results if I have the camera automatically select the AF points if I can't make sure where is the best place to focus on.

 

[lishen]

Today I played another copy of this lens, and did some comparisons. That one seems to be notably sharper at 100%. I will probably go to canon service center again tomorrow.

 

[Blue II]

I have to wonder about your test subjects. With the building, you could sometimes be focusing on the face of the building and other times focusing inside one of the offices, causing the AF to focus differently.

With the moving cars, the focus point could also be changed?

Maybe try a static, solid subject such as trees, or even the standard brick wall?

--
Steve

 

[lishen]

Sorry for the confusion. Actually, all the cars didn't move (or very very slowly move) at that time. That was an usual traffic jam in Beijing China

But I can post more test shots that unlikely raise confusion later