autofocus accuracy

[wind30]

hi people,

There are so many reviews nowadays on the various DSLRs but is there a comparative study done between the different autofocusing system used by Canon/nikon/minolta/olympus? Which one is the most accurate and stable?

I read from dpreview, the guy who runs the photozone.de, he mentioned that focusing is very impt in measuring lens sharpness. He did his test on a 350D and E300. He said that the 350D results on the average vary 20% while the E300 vary 5-10%. This variance will be much more on the field.

Is there any other such tests around?

[lsisaxon]

Originally Posted by wind30 hi people, There are so many reviews nowadays on the various DSLRs but is there a comparative study done between the different autofocusing system used by Canon/nikon/minolta/olympus? Which one is the most accurate and stable? I read from dpreview, the guy who runs the photozone.de, he mentioned that focusing is very impt in measuring lens sharpness. He did his test on a 350D and E300. He said that the 350D results on the average vary 20% while the E300 vary 5-10%. This variance will be much more on the field. Is there any other such tests around?

It also greeatly depends on whether the body is a consumer, semi-pro or pro body. For pro bodies, the manufacturer will calibrate it until it is very accurate. for the consumer body, the accuracy is not so stringent, so sometimes, the focus may be slightly near or slightly far.

How sharp the lens can be also affects the accuracy. If the lens is sharp, the sensor is able to better find the correct focus, if the lens is not that sharp or contrast is low, there is a greater possibility that the sensor cannot discriminate properly.

[waileong]

Do a Google, you should find some stuff at medfmt.8k.com. While a bit dated, focusing technology has not really changed that much since the days of film SLRs. It's basically ranging technology (either active or passive) which is applied to move the lens focus to the correct position.

It's not about calibration but about the size of the focus points. It's obviously much harder to focus on the eye compared to the head as a whole, and manufacturers have different algorithms, some of which work better than others.

The magnification, aperture and the focusing steps of lens itself obviously plays a critical role. So it's hard to compare across manufacturers.

But basically, manual focus using split screen is most accurate if you have time for it; else autofocus with more points is better, of course. Low light focusing is usually terrible, must use focus assist lamp if available.

Originally Posted by wind30 hi people, There are so many reviews nowadays on the various DSLRs but is there a comparative study done between the different autofocusing system used by Canon/nikon/minolta/olympus? Which one is the most accurate and stable? I read from dpreview, the guy who runs the photozone.de, he mentioned that focusing is very impt in measuring lens sharpness. He did his test on a 350D and E300. He said that the 350D results on the average vary 20% while the E300 vary 5-10%. This variance will be much more on the field. Is there any other such tests around?

[LittleWolf]

Originally Posted by waileong It's not about calibration but about the size of the focus points.

For (d)SLRs, it is a calibration issue, too. The focusing is done using a separate sensor, and the only way to relate it to the focus in the plane of the film/image sensor is by knowing (or assuming) the optical path lengths. If there are manufacturing or calibration tolerances, they will affect focusing.

Another problem is that most dSLRs use a relatively thick colour correction and low-pass filter in front of the sensor. It acts as an additional optical element that affects focus. Since the autofocus sensor does not "see" this filter, there's another discrepancy that has to be taken care of.

In contrast, most non-SLR digital cameras use the same sensor that records the image for determining focus. This system requires in principle no calibration and is inherently self-correcting with respect to manufacturing tolerances.

[waileong]

I understand what you meant. However, what I meant is that calibration is not the issue here. To make a meaningful comparison, we must assume that the lenses are properly calibrated, the the autofocus sensors are functioning properly, etc. Obviously, if one is calibrated wrongly or off slightly, the comparison is pointless.

Originally Posted by LittleWolf For (d)SLRs, it is a calibration issue, too. The focusing is done using a separate sensor, and the only way to relate it to the focus in the plane of the film/image sensor is by knowing (or assuming) the optical path lengths. If there are manufacturing or calibration tolerances, they will affect focusing. Another problem is that most dSLRs use a relatively thick colour correction and low-pass filter in front of the sensor. It acts as an additional optical element that affects focus. Since the autofocus sensor does not "see" this filter, there's another discrepancy that has to be taken care of. In contrast, most non-SLR digital cameras use the same sensor that records the image for determining focus. This system requires in principle no calibration and is inherently self-correcting with respect to manufacturing tolerances.

[lsisaxon]

Originally Posted by waileong I understand what you meant. However, what I meant is that calibration is not the issue here. To make a meaningful comparison, we must assume that the lenses are properly calibrated, the the autofocus sensors are functioning properly, etc. Obviously, if one is calibrated wrongly or off slightly, the comparison is pointless.

Unfortunately, calibration is not something that anyone has access to. For example, I know that the first few batches of the Nikon D70 has AF sensors which are tuned slightly near or slightly far. Even for film cameras, cheaper models like Nikon FM10, the focal plane of the focusing screen and the focal plane of the film are also slightly off.

So when comparing AF systems, usually within a manufacturer, there are also different sensor modules, plus calibration inconsistencies. So , it is not as simple to grab a camera each to evaluate the accuracy and compare each of the AF systems in question.

Most of the AF sensors in DSLR works very similarly to split image focusing screens. only when an image is lined up then is it considered in focus, so I would say that the algorithm as well as the calibration are important but not so much on the type of sensor.

The worst case that can happen is when the focusing screen, the AF sensor and the film/sensor planes are all different. What I used to do was to put the camera under a microscope and compare the optical distances to the focusing screen, AF sensor and the film/sensor planes. If 3 are in focus without the need to refocus the microscope, they can be considered aligned since the focus of microscopes is usually within a few microns.

So between brands/models/individual cameras, it really depends a lot on the person doing the alignment/calibration and how tight the manufacturers define the tolerance to be. Usually for consumer models, the calibration accuracy specification is not as tight as for professional models.