Thursday December 17
 
19:27
 
Autofocus accuracy & shallow depth of field

Moving to an F2.8 or F1.4 lens, it’s easy to think the camera isn’t getting consistent focus – but most of the time, what you’re actually seeing is a limitation of the accuracy of the autofocus sensor which is made obvious by very shallow depth of field. It’s not a fault of the lens.

Let me elaborate.

Take the center (most accurate) focus point as an example. It’s not actually a point – it covers quite a large area. The camera can focus on anything within that area. Between two consecutive handheld shots, the camera may pick slightly different focus points.

For example, if I focus on *, in one shot the camera might focus on the left of the *, another shot the camera may focus on the right. This is especially true with handheld because the camera is moving slightly – even if you think it isn’t, there will be tiny variances and that could effect which point the camera picks to lock focus.

Here are two handheld examples with my Sigma f1.4 – these are 1:1 crops:

In these, the star was in the middle of the autofocus sensor. It didn’t completely fill the sensor – some of the background text was also in the box.

The first sample, the sensor has focussed a little to the right. The second it has focussed near the left edge of the sensor. Remember, these are 1:1 crops and are handheld – a slight tiny variation in the angle I hold the camera at, a little shift to the left or right which we might not notice, that can cause the camera to vary where within the autofocus sensor area it locks focus.

With a kit lens, this isn’t such a huge issue because the aperture is F3.5 to f5.6 which gives wider depth of field. However with shallow DOF of a constant f2.8 or f1.4, variances in where the camera focusses can really show. Compare f5.6 and f2.8 and there’s a huge DOF difference. If I look closely at my 18-200VR shots, I can see there’s a variance also, but it matters less because of the DOF – it’s not as noticeable.

One interesting test is to mount your camera on a tripod, turn of VR if you have it, and use the Nikon infrared remote to take consecutive photos of a static item. In that situation, the focus point chosen by the camera is very likely to be the same between shots, showing that the lens is actually focussing correctly. Now repeat the test handheld (with/without VR) – you’ll find that the focus is far less consistent.

This isn’t a lens problem – it’s the shallow depth of field showing the limitation of the camera’s autofocus sensor.

If you have a Tamron f2.8 or Sigma f1.4 or Nikon f1.4 or Nikon f1.8 and you think you’re getting inconsistent focus, there’s a good chance it’s not a fault of the lens but a limitation of the autofocus sensor. Try the tripod test above and see whether the focus is consistent. Handheld, I see variance between shots with all my lenses, it’s just more obvious when shallow depth of field comes into the equation. (I have a Sigma f1.4 30mm, Tamron f2.8 17-50mm VC, and Nikon 18-200mm.)

Knowing this variation is going to occur, how do I combat it? Because of the variance at shallow DOF, I typically take two consecutive shots of everything when I can – not continuous shots, but reframing each time, and I make sure to wait for the VC to lock. Then I check to see whether the camera focussed on the exact point within the sensor box that I wanted it to. With two shots, I usually get it.

As an aside, one thing I’ve always wondered is whether the camera starts autofocus after VC/VR locks, or whether it is focussing from as soon as you half press the shutter button. I’ve not quite worked that out yet.

Note: Obviously you should do what makes you feel comfortable. If you have a return window and you’re not comfortable with the performance of your lens (maybe you don’t have enough time to test it as extensively as you want), by all means return it. However, if you’re seeing focus variation, this page might help explain why it is occurring and spur you on to do some tests and compare with other lenses so that you become comfortable.

If you want more technical detail on autofocus accuracy, take a look at this webpage on the D70 autofocus system. Note: Inconsistent focus is different from front/back focus. If you do the tripod test above, you’ll be able to check for front/back focus quite easily – (all the pics will come out like sample 2 if you have back focus, for example). Here’s an explanation about back focus, and here’s a prinable chart. Here’s another test chart. You can also use the test charts for the tripod tests above.