One technical difference we want to understand. Methods of autofocus.

[GerrySiegel]

Auto focus is here to stay, and the goals creep higher and higher. I find the Lumix 20mm 1.7 a nice lens but I

agree that compared to the latest and greatest ( it has got to be at least three years since introduced), we want

quick snap focus. Which has something to do with motors in lenses and processing power of camera CPUs. I get

that part, but I see a couple trends. The GX series and the GH series use conrast detection focusing. The

Olympus EM 5 and EM 5/ EM5 II use contrast detection. Fine. The flagship EM-1 OTOH added on sensor sites that are devoted to phase

detection. To permit good use of the ED Zuiko lenses which was promised, so fine.. So are we headed to a combination modalituy for best results? Or

a two tier system, where one is reserved for the more costly models.... As in the latest big ticket full frame Sony A7R has lots of on

sensor photo sites for phase detection.

 

Q:1)Has any one found a good written informed guide to these focus mode factors? 2) how much do we practically

need to care about same to keep up to snuff? Is one for bird shooters?. One for the gentry and one for the plebians? I joke,yeah but I

am interested in this particular technical dichotomy (not technobabble to me) as a harbinger of what to look for and understand the

technical reviews intelligently. Thanks to any smart heads that have it figured out.. Aloha. -g-

[Bill C]

Hi, my favorite sources tend to be paper books; Sidney Ray's (expensive) Applied Photographic Optics, for example, but I'm sure there must

be good internet material.

 

Basically, the contrast method is like you or I do, looking at the ground glass. Is it in best focus? Don't know until I try moving the lens a little

bit. If it gets better, I keep going til it gets worse, then back up a little. But I never know which way to start turning, or how far to go. Works

great with any subject material as long as we have the time to wait.

 

The phase contrast method is hard to understand. I didn't get it until I sketched out some ray traces. Essentially some built-in hardware

looks at a certain part of the frame through two halves of the lens, and spreads each out over a sensor array. The system can immediately

tl which way to turn the focus ring, and roughly how far. So it can immediately fire the lens motors to an approximate position. You can get

an idea how the two halves differ by putting a narrow opaque strip across the center of your lens and turning it in and out of focus (I won't

say what happens.) Best of luck on your further search for info.

[ellis_vener_photography]

Phase detection takes the signals from two adjacent AF sensors and sends them through separates pipelines to the AF

computer and compares the data to see how close they are to being in phase with each other and in which direction they

are out of phase To focus on that spot the computer drives the lens in the required direction. Phase detection has been

the tan dared manner of AF for SLR cameras.

 

When you use Live view on most dales the focus detection method is to compare contrast.

[Ed_Ingold]

<p>Because phase detection provides information on how much and which direction to turn, focusing can be much faster and complete in fewer iterations. There is often little or no hunting at the end point if focusing is done with linear (or piezoelectric) motors in the lens itself. That's because the action of the lens too is highly predictable. In engineering terms, this is called predictive or feed-forward control.</p>

<p>Contrast focusing attempts to maximize the contrast of edges in the field of view. To do this requires trial and error - if contrast improves do more of it, and vice versa.</p>

[steven_clark]

<p>It sounds like phase detection might actually work on the same physical principle as split prisms: reading a much narrower-aperture image from two locations going through opposite sides of the lens and comparing them to create a rangefinder.</p>

[peter_langfelder]

<p>Steven, correct - phase detection AF is in principle the automated version of the split prism focusing screen while the contrast detect AF is the automated version of ground glass focussing. The details of the implementation differ though. I don't know the details of modern contrast detect AF implementations, but I would be very surprised if it were just trial and error. One can certainly quantify contrast and after about 3 measurement one can predict where the contrast maximum will be, move the lens directly there, make another measurement and refine the prediction.</p>

[GerrySiegel]

Thanks for the replies. Since the contrast detection you say is like ground glass focusing then it would have trouble in low light conditions and yet we are at same time moving to high ISO. Apparently it is more sensitive than I realized. True, some subjects may be harder to grab hold of, and at I see the built in light come on. I guess I wonder if we will be looking more and more at a combination of them. Thinking out loud on this. All I know is that in the tests I saw by David Thorpe in his videos, the camera focuses faster than I ever could. But t there were some speed of AF variances with same camera and different lenses. Not much, but noticeable, I guess the motor in the lens is getting "smarter."