Autofocus is not 100%

samsung · 11-07-2009, 11:21 AM

When im using my AF kit lens and autofocus on a subject i just started to notice that the focus is slightly off. I've done some tests by taking some pics with only AF and another set by using AF and slightly manually focusing and discovered that the latter was slightly more sharper.

Anyone else notice this?

*Wheatfield* · 11-07-2009, 12:24 PM

This is why they are now putting AF biasing into the menu system.

samsung · 11-07-2009, 12:38 PM

Originally Posted by Wheatfield

This is why they are now putting AF biasing into the menu system.

So you're saying this is normal? All cameras' AF are slightly off?

*LeDave* · 11-07-2009, 01:34 PM

Originally Posted by samsung

So you're saying this is normal? All cameras' AF are slightly off?

It depends on how off it is that makes it abnormal. I suggest you find a AF test chart and start testing your lens, if it's really off then it's got problems. Google autofocus chart.

*Wheatfield* · 11-07-2009, 02:17 PM

Originally Posted by samsung

So you're saying this is normal? All cameras' AF are slightly off?

Enough of them are that AF biasing is becoming standard equipment on cameras. If your camera doesn't have AF biasing, you can send the camera with lens in to Pentax and have them matched.
If you have a selection of AF lenses and all require biasing in one direction, then it's a good idea to send the camera in with one of the median lenses and have the camera tuned to it.

**Marc Sabatella** · 11-07-2009, 04:04 PM

My assumption is that it's always been normal for a little discrepancy, but with sigital and higher resolution sensors in particular, we can discern things by pixel peeping msot people never noticed before.

Anyhow, there are also lots of things that can cause tests like you describe to produce misleading answers - the camera may have simply chosen to focus on something other than what you intended. Also, a little front focus is normal under tungsten or other strongly yellow/orange light, just because of how different frequencies of light as bent by the prism int the camera's focusing system. I'd sugget downloading a good focus test chart, performing the test in good light (eg, outdoors) after carefully reading the instructions for the chart, and posting images of the results.

Rondec · 11-09-2009, 05:07 AM

The best thing to do, as Marc says, is to do some focus testing with a tripod, chart, and natural light. You may find that the camera and lenses are spot on, but maybe there if some front focus or back focus. The biggest problem with auto focus is that the auto focus points are quite a bit bigger than they look in the viewfinder. Particularly when using a wide angle lens, they may pick a point of contrast that you didn't want to focus on to focus on. If someone is wearing a black and white shirt, you may find the camera focused on their shoulder and not their face.

libbyh · 11-09-2009, 03:18 PM

I would tend to agree with Marc on this one in that I think there has always been an 'acceptable' level of calibration error, but as sensor resolution has increased to a point where it's testing the optical limits of the lens, slight focusing errors are now easier to spot when pixel peeping. Ie it was an acceptable level or error because the amount it was off by was imperceptible to the eye.

I found this to be the case when I was post processing images from the K20D vice my DS and finding images seemed to appear softer at 100% when this is clearly not the case, you're simply looking at a smaller area of actual print size magnified. I think that's why you'll find the most critical pundits of IQ verify in print and not on screen.

Mind you for cameras that are completely out, you could send in the body and lens(es) for the manufacturer to calibrate, but this can be a real pain if you don't have a second body as it leaves you without a camera for a while. The beauty of microadjustment programmed in is that you can standardize your entire arsenal of equipment yourself as new lenses are purchased or upgrading bodies. Now if only the manufacturers could figure out that they need to create some sort of testing procedure to include with the body for consistent results.

On a side note, I have to credit with Olympus as having the most though out system for this so far in that they accept input for corrections at both the wide and long end of a zoom and each of the individual focus points. They just need to realize that they need to provide a software option for entering this slew of information.

rkt · 11-12-2009, 11:41 AM

You could try out some tests... one is described quite well below (link)...

Nikon D70 Focus Chart

but it is easy to become obsessive about this... and the human & ambient factors play a critical role in every aspect of this topic...