Bother with back focussing

[stroberaver]

Hi everyone

I've had my A700 for nearly a year now, and love the performance. I upgraded from an A100, with which I enjoyed using a Sigma 70-300 DG APO. As I used the Sigma 70-300 on the A700, I began to notice poor performance. Blatant back-focussing, and washed out/overexposed images. I began to fear the lens had taken one knock too many in the bottom of my camerabag. After testing some other telezooms in a shop, I was 80-90% confident that they weren't backfocussing, so I treated myself to a Sony 70-300G in the Jessops sale.

Is it just me, or is there still some significant back focussing going on? I took some test shots in my flat's car park a couple of weeks ago.

The image below was taken at 70mm, wide open (f/4.5) using the right hand focus point in landscape format. The focus point was squarely over the numberplate. Yet the back of the car is blurry and the wall and brambles are nice and sharp. It's approximately a 67% crop from full size.


Focussing further away (but not zooming), the problem seems to get worse. The image below is supposed to be focussed on the Ford Focus, about 6-7 car lengths away.


But where's the actual plane of focus? Up on the building on the other side of the wall...

[stroberaver]

3 images per post...

The problem occurs in the real world too, not just in test shots. I've been following the guys that go out on the roads at work, shooting some documentary-type snaps to use on the company website. In both the images below (largely unprocessed other than converted from raw), I had achieved focus lock on the guy's back, which has some high contrast writing on the yellow jacket.

Check out the trees and brambles in the background - much sharper!






What can I do? I'm so disappointed my lovely new 70-300 G isn't focussing correctly either. I read that focussing problems are usually due to lens problems, but my A700 body backfocusses with both a Sigma 70-300 and Sony 70-300G. Yet other lenses are fine. I've got a Tamron 90/2.8 that I've used at longer lengths and not observed any focussing errors (although I've not specifically tested it as much as the 70-300s). Focus accuracy is fine at closer distances.

What can I do? The A700 still has a couple of month's warranty on it. Do I have to send it off to be calibrated, with all my precious lenses? I think I'd rather drive to wherever it needs to go and leave them the kit for an afternoon! I also need my camera for some more stuff for work coming up soon.

Any advice/suggestions/help greatly appreciated.

[David Kilpatrick]

If you have this degree of backfocus with long lenses, you would expect full aperture wide angle results (18mm or less) to be unusable.

Is that the case?

If backfocus shows with long lenses, it should be worse with shorter lengths. I think your camera does need calibrating.

David

[stroberaver]

Thanks for the advice - I'll try some shots tomorrow with my Tamron 17-50 and Sony 18-70. Can't say I've noticed any problems at wide angle, although I suspect most of them are taken well stopped down. Hopefully the weather tomorrow will be a little more favourable towards having the camera out.

[stroberaver]

Here's some similar shots from the same location, but this time at the widest angles I can create. The main image is a 100% crop, with the thumbnail in the corner of the entire frame. Both shots taken with the right hand focus marker squarely over the front numberplate of the white Clio. The widest possible aperture was used to minimise depth of field so as to avoid it masking any focussing errors.

Tamron 17-50/2.8 @ 17mm & f/2.8:


Sony 18-70 kit lens @ 18mm & f/4.5


I don't see much evidence of backfocussing here, but other opinions most welcome. I don't understand why I'm getting backfocussing at telephoto focal lengths but not wide angle.

Also worth mentioning is that with my 70-300G, the focus accuracy seems better towards 300mm. The telezoom samples posted previously were all at 70mm. Any advice is most welcome on how I go about getting my telezoom focus @ 70mm to work properly.

[David Kilpatrick]

First of all, you can not focus a wide angle f/2.8 lens at 17mm using an edge sensor - or an A100 etc. It is vital to use the centre f2.8 sensor if you have an A700.

I just did a test with the Tamron 10-24mm f3.5-4.5 on a Nikon body - using manual focus, at 10mm the focus light was 'yes' all the way from just next to infinity down to 1.5m on a subject at about 3m! But with the lens zoomed to 24mm, the focus light only came on at exactly the right focus.

This confirms what I have always suspected - focus using conventional AF system with wide angles, especially those slower than f2.8 or with non-f2.8 sensors, is vague.

David

[bakubo]

I just did a test with the Tamron 10-24mm f2.8 on a Nikon body - using manual focus, at 10mm the focus light was 'yes' all the way from just next to infinity down to 1.5m on a subject at about 3m! But with the lens zoomed to 24mm, the focus light only came on at exactly the right focus.

That is what I always do. My experience with the Sony 11-18mm on an A700 and the Sigma 10-20mm on the Canon 30D is that it is much better to focus at full zoom (18mm and 20mm respectively) and then go back to 11mm or 10mm.

[KevinBarrett]

First of all, you can not focus a wide angle f/2.8 lens at 17mm using an edge sensor - or an A100 etc. It is vital to use the centre f2.8 sensor if you have an A700.

I just did a test with the Tamron 10-24mm f2.8 on a Nikon body - using manual focus, at 10mm the focus light was 'yes' all the way from just next to infinity down to 1.5m on a subject at about 3m! But with the lens zoomed to 24mm, the focus light only came on at exactly the right focus.

This confirms what I have always suspected - focus using conventional AF system with wide angles, especially those slower than f2.8 or with non-f2.8 sensors, is vague.

David

What does it do to your argument if I point out that the Tamron 10-24 is actually an f/3.5 lens at best?

[David Kilpatrick]

First of all, you can not focus a wide angle f/2.8 lens at 17mm using an edge sensor - or an A100 etc. It is vital to use the centre f2.8 sensor if you have an A700.

I just did a test with the Tamron 10-24mm f2.8 on a Nikon body - using manual focus, at 10mm the focus light was 'yes' all the way from just next to infinity down to 1.5m on a subject at about 3m! But with the lens zoomed to 24mm, the focus light only came on at exactly the right focus.

This confirms what I have always suspected - focus using conventional AF system with wide angles, especially those slower than f2.8 or with non-f2.8 sensors, is vague.

David

What does it do to your argument if I point out that the Tamron 10-24 is actually an f/3.5 lens at best?

Sorry, I meant f3.5, I was thinking about f2.8 sensors! Of course, if it was an f2.8 lens, the focus would not show the same extreme latitude. At the long end, where the focus is more precise, it's only f4.5. No change to the argument, just an error in writing. Will change the original post!

David

[David Kilpatrick]

I just did a test with the Tamron 10-24mm f2.8 on a Nikon body - using manual focus, at 10mm the focus light was 'yes' all the way from just next to infinity down to 1.5m on a subject at about 3m! But with the lens zoomed to 24mm, the focus light only came on at exactly the right focus.

That is what I always do. My experience with the Sony 11-18mm on an A700 and the Sigma 10-20mm on the Canon 30D is that it is much better to focus at full zoom (18mm and 20mm respectively) and then go back to 11mm or 10mm.

The Tamron 10-24mm is strongly varifocal. You can't do that, the focus shifts miles between 24 and 10mm. But there is only one distance scale, which means the distances can only be a vague indication. With the 11-18mm and the 16-80mm there is some varifocal shift but it seems to be less than the added focusing accuracy gained by zooming in.

David

[stroberaver]

First of all, you can not focus a wide angle f/2.8 lens at 17mm using an edge sensor - or an A100 etc. It is vital to use the centre f2.8 sensor if you have an A700.

I just did a test with the Tamron 10-24mm f3.5-4.5 on a Nikon body - using manual focus, at 10mm the focus light was 'yes' all the way from just next to infinity down to 1.5m on a subject at about 3m! But with the lens zoomed to 24mm, the focus light only came on at exactly the right focus.

This confirms what I have always suspected - focus using conventional AF system with wide angles, especially those slower than f2.8 or with non-f2.8 sensors, is vague.

David

Thanks David. I'm a bit confused though as to why you say you can't use an edge sensor for accurate focus at wide angle & large aperture. I understand that focus detection will be less accurate, but in the examples I posted above, the focus accuracy looks fine to me, and this is using an edge sensor.

I've seen a few references about focus sensors only being able to "see" at f/5.6 (or sometimes less) but I don't really understand how. Surely if a lens is f/2.8, then it's letting through f/2.8-worth of light onto all the AF sensors? Are there any links or resources as to why the AF sensors don't benefit from this extra light? The wikipedia article on autofocus isn't particularly in-depth.

I'll reshoot the image when I get a chance but this time using the centre sensor. I presume it's best to keep the distance to subject about the same as the first telezoom shots?

Getting back to the original topic, if it turns out there's no focus problems at wide angle with the centre sensor but the telezoom regularly gives me backfocus, how can this be? If the camera body is corrected for the telephoto shots, won't this knock out all the shorter focal lengths? Or can it be programmed for specific lenses? What's involved in getting it calibrated?

Thanks for all the help.

[David Kilpatrick]

Sorry, I thought the focus did not look fine. There was every sign of strong back focus. You may underestimate the depth of field of w/a lenses, and how much error is actually needed to soften an image particularly when stopped down. When testing the Tamron 10-24mm f3.5-4.5, I was also using the Kodak DCS760C body - two advantages, one it has a 1.3X sensor so I could see beyond the nominal coverage of the lens, secondly it's a sharp enough 6 megapixels to check usefully on lenses, due to no AA filter:

Full image:



Now - manual focus, set the lens to beyond infinity at f11 - its furthest distance:


Next, infinity/10ft - there is a very small gap between these:


Continued in next post with closer distances.

David

[David Kilpatrick]

Closer distances in this series:

Next, 3ft/1m:


Then 2ft/0.6m:


Then 1.5ft/45cm:


Final example follows - DK

[David Kilpatrick]

The final step - 1ft focus, 30cm:



Now I appreciate these are at f11 (I wanted to see exactly what tolerance I had when using manual focus, within the huge range of the 'focus confirm' light. Only this last example failed to give focus confirm. The sensors on the Alpha models are approximately f/7.1 aperture sensitivity except for the central f/2.8 types on the A900 and A700. They have to be, in order to work with f/6.3 lenses - I don't know the exact effective aperture they 'see' all lenses as having, but it's smaller than f/6.3. The Kodak camera does not as far as I know have an f/2.8 central sensor, it comes from a period before these appeared, it's likely to be around an f/5.6 effective sensor as with most Nikons and in any case with the f/3.5 viewing aperture, only an f/5.6 or similar sensor would be enabled.

The point I'm making here is just that when you do see bad focus with a wide angle this short, the actual error in distance could be considerable, and masked by depth of field. With long lenses a much greater error is needed to produce the poor focus, but the depth of field is so much less that you become aware of miscalibration.

David

[stroberaver]

Many thanks for the reply and examples. As the A700 is only my second SLR (digital or otherwise) I'm not always totally confident about what I should be expecting when it comes to these things. I didn't realise DoF could be so all-encompassing at wide angles.

Unfortunately I didn't get a chance to shoot the same kind of scene using the centre f/2.8 sensor on the Tamron 17-50. I've also got a Tamron 90/2.8 which I must also test - with a shallower DoF than the Sony 70-300G, it should amplify any focussing errors even further.

But pending these further test shots, I presume I'm looking at getting the camera calibrated. At the risk of sounding like a complete newbie, how do I go about this? What's involved? Does the camera have to go back to Sony and disappear for weeks into their corporate service process or is it something I can do? (If only Sony had put AF-adjust in the A700 like they have with the A900!)

Thanks for the continued help. I've now got dreams of perfectly focussing lenses, G and otherwise!