Auto Focus Speed & Acquisition | 5D Mark III

[Mark Elberson]

A while back there was a thread about f/2.8 cross type points and how there was no benefit to auto focus speed when using faster than f/2.8 glass due to the contraints of the auto focus sensor itself. My 5DIII has many f/4 cross type points am I am wondering if it works under those same constraints? Simply put, should I see a difference in auto focusing speed when using faster than f/4 glass? My 24-105 f/4 has been hunting in low light and was wondering if it was worth my while to throw on some faster glass to speed things up.

For this dicussion I am not interested in the differences in accuracy.

[ahab1372]

Sounds like a Neuro question.
IIRC, the faster lens will let in more light, which will help the AF to acquire focus at all.
The different AF point types (f2.8 sensitive vs f4 and f5.6) impact focus accuracy, not primarily speed.

[HDNitehawk]

Since you are talking about hunting not how fast a lens will focus. More light from the lens should make the camera hunt less.

But you asked about a faster lens, I think it will probably depend on which fast lens you go to wheter it is "faster" or not. Since all lens are not created equal.

[Mark Elberson]

Sounds like a Neuro question

That's what I thought :-)

[Mark Elberson]

Since you are talking about hunting not how fast a lens will focus. More light from the lens should make the camera hunt less.

But you asked about a faster lens, I think it will probably depend on which fast lens you go to wheter it is "faster" or not. Since all lens are not created equal.

I just meant a faster aperture not necessarily faster focusing. I am going to see if my 24-70 f/2.8 makes any difference.

[HDNitehawk]

I thought that was your question.

I suppose you will find a dark room where the 24-105 has a hard time or is barely making focus, and then try the 24-70 performs?
You would think the f/2.8 should be able to go a stop darker.
We will wait to hear back.

[ahab1372]

Please share your results. I expect less hunting with the 2.8, unless it is even too dark for that lens-camera combo

[Mark Elberson]

Speaking of how dark it is in my cavernous family room....how do I determine that? If I take a meter reading is there some formula that tells me what the EV is? I will play today with the f/2.8 glass to see if there is a discernible difference between it and the f/4. I guess common sense says it should be who knows! If I'm feeling truly adventurous maybe I'll throw in a fast prime for good measure :-)

[HDNitehawk]

I think if a guy had a light with a dimmer switch he could adjust it up and down and use the meter on the camera to see what the exposure is. Set ISO at 100, aperture wide open and monitor with shutter speed.

Another thought is, if you had ND filters you might be able to cheat the light down using different combinations.

I am sure there are formulas but I am not sure you would need it for this real world test.

[neuroanatomist]

In theory, it shouldn't make a difference. Each sensor has a certain absolute sensitivity threshold, and as long as there is sufficient light to exceed that threshold, the sensor works. It's worth noting that the center point is usually more sensitive than the other points, and the specification for the overall sensitivity of the AF system is based on that center point. It's also worth noting that not all specifications are created equal - both the 5DII and the 7D have the same -0.5 EV spec, but in my experience, the 5DII did much better in very low light.

But that's theory. The other practical factor involved is contrast. The AF system needs not only a certain absolute light level, but also a certain amount of contrast in the subject to detect the phase difference. To the extent that the additional light from a lens with a wider aperture allows better detection of the contrast, it should mean less hunting.

In very low light, enabling the 4 or 8 point expansion can also help the system lock focus, if not at the exact spot you intended. Note that expansion will use the immediately adjacent points, even if you have it set to manually select only cross-type points.

Realistically, empirical testing is probably the best way to figure out what's going to work best for you, and it sounds like you're already doing that.