AF types (passive) Contrast vs Phase

[pellicle]

Folks

I'm wondering if anyone has ever seen a site which proports to measure the speed of AF + lens systems on cameras?

I've found that the G1 focuses its 14-45 as fast as my EOS cameras focus the USM 28-105 lens (to pick a comparable lens) and often wonder if anyone has done anything about quantifying this.

I for one can't see any major reasons why Phase detect would have a superior advantage over contrast (with modern processing systems being faster than they were ...)

I feel that all too often people get hung up about Contrast Detect systems as they compare them to stuff in compact point n shoots which don't have effective lens control motors to make AF quick to begin with ....

 

[Ranger 9]

Phase-detect AF (PDAF) has an inherent advantage over contrast-detect AF (CDAF) because at the moment it starts up, it can determine the direction of defocus and the approximate degree of defocus. This means it starts out knowing which way the lens needs to be moved and roughly how far to move it.

CDAF has to focus by "trial and error." It has to turn the lens a bit... see if contrast goes up or down... reverse direction if it went down... turn the lens a bit more... see if contrast is still going up... repeat until contrast stops going up and starts going down again... then back up to the point that gave best contrast.

So PDAF has the advantage of starting out with more information and should achieve faster results -- all other things assumed equal. In practice they usually are not equal, so it would be difficult to quantify the advantage.

For example, a very important factor in AF speed is the mass of the lens components that need to be moved during focusing. Panasonic evidently has optimized this very carefully in its µ4/3 lenses and this helps them to achieve fast CDAF results. A typical DSLR lens will have more focusing mass and this will cost it some of the speed advantage of its PDAF system.

So to make a comparison, you'd need to find a system that includes two camera bodies taking the same lens, both equally optimized, but one using PDAF and one using CDAF. I don't think there is anything like that. The closest you can get is to look at some of the DSLRs that offer CDAF in "live view" mode. Most of these are very slow... but it would be hard to tell how much of that is because CDAF is inherently slower vs. how much is because the systems are poorly optimized for CDAF performance.

 

[kevinparis]

would i be right in saying that phase detect is more like rangefinder focusing in that it tries to match two offset images and that contrast detect tries to be smarter and actually focus on the content of the picture?

not trying to start a war ... just curious

K

 

[Ranger 9]

Value judgments aside, that's right in principle. Basically phase detect works exactly like the split-image focusing aids you used to find in pre-autofocus SLRs. Each AF detector looks through a pair of opposed prisms that pick up images from opposite edges of the lens.

By comparing the patterns of detail in the two images, the AF system determines if they're "in phase" or not, exactly the same way your eye would match up the two halves of the SLR's prism aid. If the images are out of phase, the direction and distance of the offset tell the AF system which way the lens needs to move and roughly how far it has to go.

The contrast-detect system also analyzes image details. But since it has no way of knowing whether it's "seeing" a high-contrast but out-of-focus subject, or a low-contrast but in-focus subject, it can't make an absolute determination of correct focus the way the PDAF system can. Instead it keeps moving the lens until contrast peaks, then stops it.

One advantage of CDAF is that it can deal easily with repeating "picket-fence" types of subjects, which can confuse PDAF because the pattern seems to be in phase at multiple points.

 

[pellicle]

Hi

Phase-detect AF (PDAF) has an inherent advantage over contrast-detect AF (CDAF) because at the moment it starts up, it can determine the direction of defocus and the approximate degree of defocus.

I think the key in this is "from the moment it starts" so it then has to stop and reverse. Why then does this not confer to CDAF?

I have certainly had times where a USM200 f2.8 focused the wrong way first and didn't stop (I tried a white Tshirt in the focus point specifically because I wanted to see how fast the lens hunted one way or another.

do you have more details on the operations of each system that you can explain the engineering reasons why or perhaps a link to a site?

on my rangefinder I've turned it the wrong way plenty of times first too

 

[cap'n bill]

There are considerations apart from speed. Contrast autofocus as implemented by Panasonic seems to be able to use the any part of the field of view for focusing, not just certain predefined points.

The G1 for example can lock on to an object and follow the focus as it moves across the picture.

 

[pellicle]

There are considerations apart from speed. Contrast autofocus as implemented by Panasonic seems to be able to use the any part of the field of view for focusing, not just certain predefined points.

100% agreed on that point too ... but I'm wondering on which points is Phase advantaged?

It seems to me that while being gear heads Photographers seem to know little about the technical issues of their gear. So while cameras such as the EOS 1 and 1V and HS models are hailed as "fast" AF cameras does anyone actually make measurements (or do we just believe the makers specs?)

I know that when I was doing occasional sporting event coverage that with the same lens on the camera I could not tell my EOS 630 apart from a 1v (which I borrowed for the day) except that the 1v had more focusing points and had both horizontal and vertical focus points ... (although not all are available with all lenses and at all focus points, so you can get caught out when you depend on it).

so assuming I got AF it was no problems and both were fast.

As you mention Contrast Detect has no such issue of horizontal or vertical measurement (try focusing on a powerline against the sky in the wrong orientation using a Phase detect system).

its looking like noone has done any measurements ... interesting

 

[Vivek]

It seems to me that while being gear heads Photographers seem to know little about the technical issues of their gear. So while cameras such as the EOS 1 and 1V and HS models are hailed as "fast" AF cameras does anyone actually make measurements (or do we just believe the makers specs?)

That is a broad brush steeped in prejudice and ignorance of facts or consideration of selected facts.

If a seasoned photographer chooses a particular gear based on how it performs in his/her hands there is no need for a rationalization, understanding of how things work in detail and a blog about it.

All it comes down to is the aperture, shutter speed and the focus (of the subject of interest in a frame) for a given ISO speed.

 

[pellicle]

If a seasoned photographer chooses a particular gear based on how it performs in his/her hands there is no need for a rationalization, understanding of how things work in detail and a blog about it.

unless one learns by writing, and hopes that what one has learned can be of benefit to others on the same path... though my doubts are decreasing on the latter ;-)

 

[Vivek]

unless one learns by writing, and hopes that what one has learned can be of benefit to others on the same path... though my doubts are decreasing on the latter ;-)

That saves the trouble for everyone doesn't it?

Cameras are for photography and not for blogging.:sleep006:

 

[pellicle]

That saves the trouble for everyone doesn't it?

Cameras are for photography and not for blogging.:sleep006:

I suppose you would also argue that photography is for taking pictures and that attending workshops, reading books and exploring the science is equally :deadhorse:

 

[Vivek]

I suppose you would also argue that photography is for taking pictures and that attending workshops, reading books and exploring the science is equally :deadhorse:

I can't help you with your limited imagination. Assume what you can.:ROTFL:

 

[Ranger 9]

Short, slightly-oversimplified summary:

-- PDAF is an automated version of the same process you use to focus an old-style SLR with a split-image "rangefinder" focusing aid. The disadvantages are that you can only focus in one area (without recomposing) and only on subjects that have linear details that can be "split." The advantage is that these types of details can be focused very quickly and positively, since linear misalignment can be judged more acutely than overall detail contrast.

-- CDAF is an automated version of the same process you use to focus an old-style SLR with a plain matte groundglass. The disadvantage is that it takes more time and more sawing back and forth to be sure that you've achieved peak detail. The advantages are that you can focus anywhere on the screen and with any type of subject that has enough detail to be visible.



Designing some kind of valid "comparison test" would be an almost impossible proposition, since there's an almost infinite range of conditions (light level, amount of detail contrast, size of detail contrast, repeating vs. non-repeating details, stationary or moving subject, etc., etc.) that would advantage one system over another.

The system that was the "winner" under one scenario would likely be the "loser" under a different scenario, so photographers would have to choose which from a large number of scenarios were most important to their own photography, and weight the results accordingly.

Also, not everyone might agree on what the "winning" criteria would be: is it more important to your photography to get pretty good focusing accuracy really quickly? Or to get extremely high focusing accuracy even if it takes a relatively long time? Or something in between?


Because of all these variables, I don't think there's a meaningful way to give a simple "report card" of AF performance!

 

[pellicle]

Sensei

I can't help you with your limited imagination. Assume what you can.:ROTFL:

time and space are limited ... as are what I can post here ... some of what I might imagine would surely get moderated

well ... it *should* get moderated ...

:salute:

 

[Vivek]

I've found that the G1 focuses its 14-45 as fast as my EOS cameras focus the USM 28-105 lens (to pick a comparable lens) and often wonder if anyone has done anything about quantifying this.

I for one can't see any major reasons why Phase detect would have a superior advantage over contrast (with modern processing systems being faster than they were ...)

I feel that all too often people get hung up about Contrast Detect systems as they compare them to stuff in compact point n shoots which don't have effective lens control motors to make AF quick to begin with ....

You start with a conclusion and despite the answers/explanations, you keep harping on what you started with as your "finding" with a sprinkling of your "scientific" wisdom on everything in your limited imagination.

Yes, you should get moderated!

 

[pellicle]

Vivek

You start with a conclusion and despite the answers/explanations, you keep harping on what you started with as your "finding" with a sprinkling of your "scientific" wisdom on everything in your limited imagination.

I started with an observation, and have made no findings or conclusions. I've asked for information, I notice you haven't supplied much (except mis-construing what I write). You having a bad day there?

 

[Vivek]

Pellicle,

Sigma 30/1.4 on D300 (using one of the cross sensors in the center) (contrast detect would have gone hunting elsewhere).

Computar-TV 25/1.3, G1 ("upgrade" would not do any good as it was manually focused)

Oly-D 17/2.8 on G1 (contrast detect in action).

All taken under similar lighting conditions.

If you want to ask questions, kindly refrain yourself from extra characterizations of the participants and such.

 

[pellicle]

Vivek

this appears to have come from my statement:

It seems to me that while being gear heads Photographers seem to know little about the technical issues of their gear

I think it requires a distinct effort to assume that I was referring to this group, rather than a generalisation. If you have for some reason taken it to mean that then please be assured that was not my intention.

However generalizations are often reflections of reality, and always there are exceptions. For instance as I'm an Australian it is a generalization that we are all descendants from English stock. That would be correct in 70% of the cases.

But you'll note I did not use "all" in my sentence ...

I participate in a number of other forums and meet a number of people. It has been my experience that many people with a professed interest in photography (and a penchant for buying expensive equipment) have no idea of how to use it.

Again if you take offense at that then I'm sorry, but it was not intended as a reference to you (and as such with your obvious penchant for technical exploration would be mistaken anyway).

Perhaps it might be wise to ask what did you mean by that before making assumptions and directing anger at me.

If there was something else which you took offense to then please let me know what it was.

But if you would kindly refrain from telling me what you find boring I would appreciate it too, as quite I'm often not interested in things that are discussed here. The difference is that I only respond to the things I'm interested in.

PS if it was "gear heads" that disturbed you, then please rest assured it was not something which I consider has a negative connotation. I consider myself a gear head (in areas such as photography, camping, cars and motorbikes), nd yes I do engage in all of them not just blog about them. That is where I get to take pictures like this from my motorbike in Tokyo:

and this (while camping here in a place called Repovesi) with my 4/3rds camera: