Improving focus accuracy - AF micro-adjust or calibration dilemma?

[M.Chapman]

I've got the following equipment 2 x Canon 550D SLR bodies and Canon 15-85 EF-S, 10-22 EF-S, 55-250 EF-S and 60mm EF-S lenses.

 

My dilemma is that I find that phase detect focussing is just not reliable enough, and  so I tend to use Live View contrast AF which is always spot on, but slow and tedious. I've done some limited tests and it's clear that there's a little bit of front focussing going on with the phase detect AF suggesting that my lenses and/or camera may benefit from calibration or AF micro-adjust. Note that the errors are pretty small. If I use phase detect it's fine for 6x4". But at 100%, I can't be confident of reliably meeting Alamy's requirements.

 

So I appear to have a choice.

1. Send (both?) bodies and (all?) lenses off for calibration (any idea how much this might cost and who provides the best service?)


2. Upgrade to a body with AF micro-adjust feature (e.g. Canon 50D) and make the fine tweaks myself. Unfortunately the 50D is bigger and heavier than I would like.


3. Swap DSLR systems to perhaps a Nikon (D7000) which is not too much larger and has AF micro-adjust. Could be expensive with having to change lenses


4. Wait for Canon to add AF micro-adjust onto a smaller SLR .... 750D?

Any thoughts or advice appreciated.

 

Mark

[David Kilpatrick]

Even AF micro adjust will not solve your problem. You have to go right to the top end of cameras to get AF micro adjust which works across a range of focal lengths for a zoom lens (most like the 50D only have one calibration for each lens). Whatever calibration you make, it will only be good for - that final working aperture, the distance you chose to do the calibration, the focal length you chose (or at the best, wide and tele extremes of the zoom), the light level and also the colour of the subject. All these things change phase detect calibration.

 

No Canon body made can focus the 10-22mm accurately. The best they can do is get within depth of field. Test it yourself, set manual focus, watch the AF confirm, focus around a subject. The 10-22mm used at 10mm will confirm AF anywhere between 1.5m and infinity on most subjects! This is why Tokina designed the 11-16mm f/2.8 - it actives the f/2.8 centre sensors of many (now older) DSLR, and this increases focus accuracy enough to make it slightly better than a lottery.

 

Your best bet is to try the EOS-M. It's a superbly affordable mirrorless camera, and sells for almost nothing, including a lens, a decent little flashgun and an adaptor for your EF-S lenses. These then use contrast detect focusing and do it pretty well. You still have your 550D for times when you prefer a viewfinder, but the M gives you a later and better sensor, a second body for video if you ever do that, a valuable back-up and a nearly pocketable silent camera for those DSLR-unfriendly events. You will find the CD focusing miles faster and better than using live view on the 550D.

 

David

[M.Chapman]

David,

 

Many thanks for your reply. I hadn't thought of the Canon M series. I've already got a Canon G15 (sorry I didn't mention that) which works well as a discrete, carry anywhere P&S camera that I find can meet Alamy QC requirements, but only for stationary subjects in good light. 

 

In addition to Landscape photos, I also take wildlife pics (not many on Alamy, they're not good enough yet) so often take the 550D and lenses so I can cover both landscape and wildlife, for which speed of focussing and SLR viewfinder are preferable. But I find the phase detect focussing accuracy is just not quite accurate/reliable enough and so I'm often disappointed with the results

 

Based on your comment I took a quick look at info and reviews on the Canon M. Even though it has hybrid AF (mixture of phase and contrast) I note that some user reviews still complain the focussing is far too slow compared to a DSLR. I had hoped that a hybrid system would provide the best of both worlds. Do you have any hands on experience with the M series? What did you think to the focussing speed?

 

It's interesting you're fairly dismissive of basic micro-AF adjust, and yet some folks seem to swear by it (even the basic capability on a 50D). Given the sensitivity of phase AF to all the variables you mention, it's amazing it works at all, and yet most DSLR users rely on it. I agree that contrast AF will always provide greater accuracy, but feel that phase AF could work better than it currently does with my camera & lenses.

 

One option I'm considering is selling one of the 550D bodies on eBay and buying a S/H 50D to replace it. Then I can try it out, and still have 2 bodies.

 

I suspect this approach will be a lot cheaper than sending everything off for calibration.

 

I also took a look at Colchester Camera Repair Centre's website. They seem to have a fairly "rational" approach to resolving focussing problems, but I've no idea what they charge or what their quality of service is like.

 

Are there any DSLRs with known better phase detect focussing accuracy than others? For example I note the Sony SLT has a fixed mirror, that must remove another source of variability

 

Mark

[David Kilpatrick]

The reason most people are happy with calibration is because it only affects some of their work, and tends to be needed in specific conditons. For example, you might use your 55-200mm mostly at 200mm on subjects 10m away, wide open. You can calibrate for that. The EOS-M is of course slow to focus, every single contrast detect system made to date is much slower than the average phase detect system, such comments are normally made by users trying to catch action or very fast grab shots (I just don't even try to use these cameras that way, no matter how many frames a second they can capture).

 

Then again, you could wait for the right lenses to come out from Sigma, and buy those with the calibration port and kit. Then you can have zooms with multiple focal lengths and target focus distances custom tuned to match your camera exactly. The reason Sigma has invented such a complex and advanced system is because it is actually needed.

 

Even within each make, DSLRs vary in focus accuracy. So far the Sony SLT design has been far better than their previous DSLRs which had constant back or front focus issues. There's less to adjust and the sensor can be used to set the system up - the camera is its own calibrator during production.

[M.Chapman]

David,

 

Many thanks. The Sigma lens USB dock and Optimisation Pro software looks very interesting. I wasn't aware of this. I wonder how long it will take them to update their lens range so there's a reasonable range of lenses that support this feature? It could be really useful. In addition to the micro AF adjust It's a particularly neat idea to be able to restrict the focus range. 

 

Mark

[spacecadet]

So when I occasionally think my A350 has missed focus on the 18-55, that would be because it has, then?

[David Kilpatrick]

Mark - the A350 actually had a good reputation for AF. The A100, 200 series, notoriously the otherwise excellent 700 are all prone. The 300 and 500 series models rarely have calibration issues, the 900 of course offers adjustment (but it's really difficult to get right - I set mine up using fine graphic artwork at 3m with the 50mm f/1.4, then found that I'd made it be soft for landscapes). The later SLTs all seem very accurate generally.

 

The 18-55mm at 18mm doesn't focus well on any body - it has too much depth of field, and the camera is measuring focus at an effective f/7.1 - use any depth of field calculator, and you'll see what a large range of error that allows. Those who own bodies with f/2.8 capable sensors can buy alternatives like the 17-50mm Tamron or 18-50mm Sigma f/2.8 lenses, and not surprisingly, they tend to be really impressed. That's because instead of focusing at f/5.6 (Canon and Nikon, and new Sony from the A580 onwards) or f/7.1 (Sony prior to the new 15-zone AF module with three cross sensors) they focus at f/2.8, which actually works at 17mm.

[M.Chapman]

The best video I've come across on Sigma's new lens optimisation software is here http://vimeo.com/64665246

 

It looks very interesting. I could be tempted to buy new the version of 17-70mm lens just to try it out although a longer telephoto would be a more discriminating test, but the 120-300mm is too expensive for me.

 

Makes me wonder how this all works though. Suppose the focussing error is in the body because the phase detect sensor is misaligned relative to the image sensor. You fit the Sigma lens, take some tests shots and make the necessary adjustment to compensate. Presumably the lens has to "tell" the camera that optimum focus with this lens, focal length and focal distance is actually so many micro-steps forward or back relative to the "in phase" position on the focus detector. Mmmm..... But until it's focussed the lens doesn't actually know the focal distance. Maybe the lens passes all the numbers to the camera and the camera sorts it all out?

 

However, I've seen comments on forums stating that Canon's AF is "single shot". ie. determine how out of focus the image is by measuring the phase mismatch. Calculate the focussing movement required, and send it to the lens and then that's it. The system never checks if the resultant image is actually in focus, and if not repeats the process. There are comments about pressing the shutter 1/2 way twice to get the best focus. However, I've never noticed this. Maybe it's just "folklore".

 

Mark

[spacecadet]

Is there an answer? I bought it used to replace the crappy Sigma 18-50 that went west and it's a little better, but just occasionally it's off enough to keep it out of a submission. I won't be changing kit any time soon.

[spacecadet]

 

 

The 18-55mm at 18mm doesn't focus well on any body - it has too much depth of field, and the camera is measuring focus at an effective f/7.1 - use any depth of field calculator, and you'll see what a large range of error that allows.

Is there an answer? I bought it used to replace the crappy Sigma 18-50 that went west and it's a little better, but just occasionally it's off enough to keep it out of a submission. I won't be changing kit any time soon.

 

Mark - the A350 actually had a good reputation for AF.

 

The 18-55mm at 18mm doesn't focus well on any body - it has too much depth of field, and the camera is measuring focus at an effective f/7.1 - use any depth of field calculator, and you'll see what a large range of error that allows.

One reason I took your recommendation of it!

Is there an answer? I bought it used to replace the crappy Sigma 18-50 that went west and it's a little better, but just occasionally it's off enough to keep it out of a submission. I won't be changing kit any time soon.

[Vlad]

There are also issues you won't be able to fix with AF micro-adjustments or calibration, like "focus shift" on some lenses (for example the 50/1.2). The focus of the lens will depend on the aperture, so a lens focused correctly at open aperture will change focus (it's an optical design problem) while being stopped down by the camera when actually taking the image ...

[M.Chapman]

I was in the area, so I called in at Park Cameras new showroom (nice facility with lots of cameras on display all with fully charged batteries) to look at options. Several things came to light.

 

Whilst I was in the shop, 2 other photographers came in and, like me, were also trying to resolve focus accuracy problems with Canon gear. Both were using pro spec Canon bodies and lenses. OK it's a small random sample, but it didn't sound good. One said that Canon have recently reduced their body with one lens calibration fee to just £30. He wondered if this might be because they were getting so many complaints. He was considering switching systems.

 

Anyway, whilst I was there I had a hands on look at a variety of of options

 

1) Canon 50D (they had one S/H) too heavy and bulky

2) Canon M - the focussing maybe accurate but it's way too slow

3) New Canon 100D. Live view focussing is now hybrid phase/contrast and was much snappier than on my 550D. Really nice compact SLR.

4) New Sigma 17-70mm lens with USB dock and AF micro-adjust. Looks like a good option. Feels well made. USB dock not yet in stock.

5) Panasonic Lumix G series. Very impressed. Really fast (and supposedly accurate) contrast AF and pretty decent EVF and on Alamy's OK list. Good choice of lenses.

 

Now seriously considering option 4 (and sticking with my Canon bodies) or option 5 (switching over to Lumix with micro 4/3 lenses).

The speed of the Lumix contrast AF was pretty much the same (or better) than my 550D and lenses achieve. If better focus accuracy is "integral" to the design (same sensor for focusing and imaging) then where's the catch? I might wait for the Lumix G6 to settle at a reasonable price and switch systems.

 

In the meantime, if the £30 body + 1 lens calibration price is correct I'll try sending one 550D body and the telephoto lens to Canon UK for calibration to see what happens.

 

Mark