Okay then here's how I would test for AF focus accuracy. I've modified this in the light of recent debate on here, and have attempted to address the two potential shortcomings of that infamous A4-print-out-shot-at-45degs test that folks download and use, and then complain about their cameras. I can see why people use that test, because in theory it should work fine, but for various reasons it often creates more problems than it solves. I know for sure that manufacturer’s camera technicians hate it because it generates a steady stream of cameras and lenses in to them that the customer insists don’t AF properly yet they can’t find anything wrong with them. So they return them, and then the customer re-tests them, gets irate and sends it back again, and so it goes on.
But first a bit of waffle.
One problem with that test is it is generally shot at very close distance, deliberately so, in order to reduce depth of field to very narrow levels. This is so you can see any focus shifts more easily, but in doing so it pushes the lens right to the limit of its range. Lenses (and cameras) are analogue devises and they are optimised for normal focusing distances, plus they work within a tolerance range set within acceptable limits for depth of focus. When the camera detects focus within tolerance, it stops, even though focus might not be absolutely at the technical optimum. It's within acceptable range, and you'll never notice the difference in practise. There are photographic operational reasons for this, and it's not just sloppy design or manufacture - it makes the camera focus faster and more decisively under normal situations.
The second potential problem of shooting a target at an angle is that you sometimes cannot be certain exactly where on the slope of the target the camera is actually focusing. This is because the focus sensor often reads a slightly larger area than that depicted by the focusing spot in the viewfinder. So to be certain of avoiding this, shoot a flat target square on.
Some people have done AF tests, decided that their camera was fractionally off optimum, and had it adjusted by the manufacturer. They have then found that at more normal focusing distances it is out, when before the test it was fine. This is very rare, but can happen with equipment testing which is designed to provoke errors where often none really exists, by pushing it to unrealistic or irrelevant extremes – it makes you think something is wrong, when in fact it’s fine. Have you noticed in regular picture taking that the focus is out, and consistently out, with sharp focus either just in front of where it should be, or just behind? If there is a fault, then it will always show itself in the same way every time. And this test will confirm it.
Of course, it is possible that either the lens or camera could be slightly out. It happens, but before you go to the considerable trouble and expense of sending it back for calibration, make sure there really is a problem. If a lens is out of tolerance, then it will show exactly the same problem on different camera bodies. If a camera body is out, then it will also demonstrate the same fault with different lenses. It’s not too hard to cross-check like this, either with friends with similar equipment, or at the local camera shop if you ask nicely. You can surely devise a similar test to this one even in store. It’s got to be easier than sending your kit back unnecessarily.
Now the test. Nothing too tricky or technical, it just needs to be a flat target shot at a sensible distance – nothing more. If that distance is too far, then you’ll never notice any error as depth of field will cover it, so it has to be reasonably close up, just not to extremes.
I took three cereal boxes, and lined them up side by side on the kitchen worktop. I think this is a big enough target to be realistic, yet small enough to show up any significant errors. Position one box square to the camera in the middle, and one to the side and set at a slight angle so that part of the box is just behind the middle one, and part of it is just in front. Put a third box on the opposite side, square to the camera and lined up so that the front is in line with the back of the middle one. This third box serves no real purpose other than being something to lean a 12in ruler up against, standing vertically on end, right up close to the middle box. Lean it so that the top is behind the front of the middle box, and the bottom is a couple on inches in front. I hope that’s clear!
Set the lens to the widest aperture (brightest, low f/number) for minimum depth of field. Make sure you have an absolutely shake-free shutter speed and wind up the ISO as far as you need to get it, and use IS/VR if you have it. Flash is fine, in fact flash is good, so long as you don’t get nasty reflections off the boxes or the ruler, which is actually quite likely in this situation unless you use bounce flash. Stand back so that the three boxes just fill the frame (* see footnote) and make sure the camera is square to the target – in other words, the camera must be at exactly the same height as the centre of the middle box. Using just the centre focus point, focus on part of the middle box that has plenty of good detail for the AF to lock onto positively, and take a snap.
Zoom in on the LCD at max magnification and check the markings on the ruler. Scroll up and down it and decide where best focus is, and where you think it runs out, both in front of the middle box, and also behind it. Look carefully, because unless you have a fast lens like f/2.8 or wider, the focus fades very gradually and it’s hard to tell where ‘acceptable’ sharpness fades into unacceptable blurring. Ideally, the part of the ruler that is in line with the plane of the box you focused on should be perfectly sharp, and the parts in front and behind should be gradually fading equally out of focus.
I’ve just done it with a couple of lenses and both were spot on as far as I could tell. With one lens (Canon 100-400mm L, 180mm at f/4.5 and 2m, on a 40D) I estimated acceptable depth of field to be about 2cm, and when I checked this against what it should be (
www.dofmaster.com ) it was exactly right! Not bad just by blowing it up on the back of the camera.
To see what this looks like on something a bit more pictorial than a ruler, then you really need to make a big print and look at the box set at an angle. Look at the type, or even the tiny printing dots that make up the image detail. You will see the focus gradually fading towards the front and the back of the box face, with the sharpest point being in line with the plane of the middle box. This will give you a better idea of what you can expect to see in real picture taking.
* NB Don't rely on this test if you have to focus closer than 2ft from the target - see post #17 below.
I phone ahead and make sure the shop has got two samples of the lens I want, then check them side by side at the store. I've now got this down to a fine art, and it's always nice to know that you've spent your hard-earned on a good un. Having said that, and having done this with the last five lenses I bought (so that's ten lenses in total, eight Canons two Sigmas) I would be quite happy to take any of the samples I checked. Differences were there in a couple of them, but miniscule and irrelevant. There are always manufacturing tolerances, and perfection is very rare, but equally, duff lenses are also quite rare (despite what you hear on internet forums!).
You're welcome. The test described above is very simple - all you need to know is what the problem is with 'bad copies' (nothing usually 