Panasonic's contrast focusing puts them miles ahead of any competition

RichA writes:
But the focusing system in the Panasonic's is better than anything in
DSLRs at the moment.

Is it? How does that work anyway?

My understanding is that using the information read from the focusing
sensor in a phase-based system, you can deduce the direction and amount
of adjustment needed (to some degree anyway), and also whether or not
you're at the point of best focus; this information helps a lot to
reduce hunting.

The information available from the sensor in a contrast-based system, on
the other hand, is much harder to interpret -- you can't even really
tell if you're at the point of best focus or not without adjusting the
lens and seeing whether the contrast is reduced or not.

That makes it sound _very_ difficult for a contrast-based system to be
as good as a phase-based system, since there's going to inherently be
some hunting with the contrast-based system. The only way I can think
of is to make the focusing motor so fast and powerful that the hunting
becomes unnoticeably fast...

-miles

--
P.S. All information contained in the above letter is false,
for reasons of military security.

In article , Miles Bader
wrote:

RichA writes:
But the focusing system in the Panasonic's is better than anything in
DSLRs at the moment.

Is it? How does that work anyway?

it's better than other contrast detect systems and comparable to entry
level dslrs, but it is certainly *not* 'better than anything in dslrs
at the moment' nor is it anywhere close to top end dslrs such as the
nikon d3.

That makes it sound _very_ difficult for a contrast-based system to be
as good as a phase-based system, since there's going to inherently be
some hunting with the contrast-based system. The only way I can think
of is to make the focusing motor so fast and powerful that the hunting
becomes unnoticeably fast...

part of it is the lens motor. the panasonic g1 does not focus as fast
with older 4/3rds lenses.

John Navas writes:
My understanding is that using the information read from the focusing
sensor in a phase-based system, you can deduce the direction and amount
of adjustment needed (to some degree anyway) ...

That's the theory, but it's not that simple in practice. The direction
and amount of focusing depends on the lens and focusing mechanism, and
to prevent focus error, a good phase detection system will touch up
(hunt) focus.

The information available from the sensor in a contrast-based system, on
the other hand, is much harder to interpret -- you can't even really
tell if you're at the point of best focus or not without adjusting the
lens and seeing whether the contrast is reduced or not.

Again, it's not that simple in practice. A good contrast detection
system can use such things as predictive focus, focus tracking, and
continuous focusing, making focusing very fast in many cases.
....

Yeah, of course things are always more complicated in practice, but the
point is that there's inherently more information available with
phase-based auto-focus systems. Even very rough information that lets
you guess how far out of focus you are, and in which direction, is very
useful (especially the "which direction" bit!), and contrast-based
sensors don't provide anything like it.

Of course you can use the techniques you mention to try and speed up the
"hunting" process, but they're at best educated guesses, and a poor
substitute for getting more useful information from the sensor.

Moreover, you can of course use the same techniques to make a
phase-based focusing system even better, so the original claim that a
contrast-based system on a relatively low-priced camera is better than
phase-based systems on expensive cameras seems ... rather dubious.

-Miles

--
Genealogy, n. An account of one's descent from an ancestor who did not
particularly care to trace his own.

In article , John Navas
wrote:

The reality is that both systems have pros and cons, that good
implementations of both systems involve some hunting, that both systems
can be made to work very well, and that claiming one is "better" than
the other is silly and pointless trolling.

nonsense. it's not trolling. phase detection is faster.

currently, the fastest contrast detect autofocus is about as good as
the slowest phase detect autofocus.

until contrast detection is as fast or faster than the fastest phase
detection, phase detection is very clearly better.

whether someone wants fast autofocus is another story. some people
prefer manual focus, which can be 'better' in certain scenarios, such
as macro.

Yeah, of course things are always more complicated in practice, but the
point is that there's inherently more information available with
phase-based auto-focus systems. Even very rough information that lets
you guess how far out of focus you are, and in which direction, is very
useful (especially the "which direction" bit!), and contrast-based
sensors don't provide anything like it.

Of course you can use the techniques you mention to try and speed up the
"hunting" process, but they're at best educated guesses, and a poor
substitute for getting more useful information from the sensor.

A servo system like a contrast-based autofocus need not "hunt"
when focusing on a static subject.

It would need dither to keep focused on a moving target.

But so would a phase-based SLR ... and mine,. a Canon 30D,
actually hunts when doing follow-focus. Actual hunt, as
opposed to dither. I'm surprised it does not do dither.

Doug McDonald

In article , John Navas
wrote:

The reality is that both systems have pros and cons, that good
implementations of both systems involve some hunting, that both systems
can be made to work very well, and that claiming one is "better" than
the other is silly and pointless trolling.

nonsense. it's not trolling. phase detection is faster.

currently, the fastest contrast detect autofocus is about as good as
the slowest phase detect autofocus.

Simply not true.

it's exactly true.

John Navas writes:
Yeah, of course things are always more complicated in practice, but the
point is that there's inherently more information available with
phase-based auto-focus systems. Even very rough information that lets
you guess how far out of focus you are, and in which direction, is very
useful (especially the "which direction" bit!), and contrast-based
sensors don't provide anything like it.

Of course you can use the techniques you mention to try and speed up the
"hunting" process, but they're at best educated guesses, and a poor
substitute for getting more useful information from the sensor.

They work actually very well in practice.

I'm predictive methods help, but let's just say, that I'm skeptical that
they help enough to make contrast-based focusing as good as phase-based
focusing.

Remember that the original poster's claim was: "the focusing system in
the Panasonic's is better than anything in DSLRs at the moment."

_That_ seems very surprising, and hard to understand. Since the
original poster didn't offer much in the way of evidence or explanation,
I'm trying to figure out if what he said is actually true, and if so,
how they pulled it off given the inherent speed deficiencies of the
contrast-based method.

Do you see _why_ a phase-based system offers more information? The
sensors in a phase-based system give you information about "out of
focusness". That makes it easier to know (1) when you should stop
focusing, (2) the likely direction the focus should change (3) how fast
you should try to change the focus to correct the focus.

The sensor in a contrast-based system doesn't tell you any of that. It
merely gives you a contrast measurement, which is only meaningful for
focusing purposes if you compare it with a contrast measurement at
another focus setting -- so the contrast-based system cannot actually
tell you anything without at least some movement of the focus, and
perhaps more crucially doesn't give you information about when you
should stop or how fast you should go. By taking measurements at
multiple points of focus you can build up the required info, but it
inherently requires the focus motor to keep moving beyond the point of
correct focus (because it can only tell if it's gone too far by noticing
that the contrast starts to drop again). It also doesn't know how fast
it should move, and if it drives the focusing motor very fast, it can
easily overshoot and may not realize it!

Anyway, it certainly seems _possible_, with a very fast focusing motor
and high sampling rate, to make this work reasonably well, but it's very
surprising if it works better than the phase-based system used in DSLRs.

-Miles

--
"Suppose He doesn't give a ****? Suppose there is a God but He
just doesn't give a ****?" [George Carlin]

In article , Miles Bader
wrote:

Remember that the original poster's claim was: "the focusing system in
the Panasonic's is better than anything in DSLRs at the moment."

_That_ seems very surprising, and hard to understand. Since the
original poster didn't offer much in the way of evidence or explanation,
I'm trying to figure out if what he said is actually true, and if so,
how they pulled it off given the inherent speed deficiencies of the
contrast-based method.

that claim is not true. the panasonic g1 is among the fastest contrast
detection systems and comparable to low end dslrs, but it is not faster
than every dslr available.

http://www.dpreview.com/reviews/panasonicdmcg1/

And our initial tests would suggest that they have solved at least
one of the technological problems mentioned earlier (the
contrast-detect autofocus is easily as fast as any other entry-level
DSLR).

In rec.photo.digital John Navas wrote:
On Thu, 24 Sep 2009 17:22:14 +0900, Miles Bader wrote in
:

RichA writes:
But the focusing system in the Panasonic's is better than anything in
DSLRs at the moment.

Is it? How does that work anyway?

My understanding is that using the information read from the focusing
sensor in a phase-based system, you can deduce the direction and amount
of adjustment needed (to some degree anyway), and also whether or not
you're at the point of best focus; this information helps a lot to
reduce hunting.

That's the theory, but it's not that simple in practice. The direction
and amount of focusing depends on the lens and focusing mechanism, and
to prevent focus error, a good phase detection system will touch up
(hunt) focus.

The problem with doing that is that it might involve a change of
direction, which would involve taking up the backlash in the focus
motor drive. A switch of direction which involves drive backlash is
too much for an efficient tightly wound servo control loop to handle,
so it's done as separate a invocation of the "start focussing again
from scratch" general focus routine. Which all takes a little time.

That's why even better phase focus systems will drop the focus motor
speed down in the final approach to the focus destination so that the
destination can be approached with maximum precision by monitoring the
phase readings during the final approach. That way the best focus is
achieved without having to restart (and possibly reverse) the
focusing. Also copes with minor relative movements of camera or
focussed object during the focus procedure.

--
Chris Malcolm

Miles Bader wrote:

Anyway, it certainly seems _possible_, with a very fast focusing motor
and high sampling rate, to make this work reasonably well, but it's very
surprising if it works better than the phase-based system used in DSLRs.

True for certain definitions of "better", but not all. You seem to be
speaking only of speed and efficiency.

In my experience with getting the phase detect AF of a 450D sorted, the PD
system is never more accurate than the contrast detect system (with
reasonably well lit subjects). The accuracy of the PD system is directly
dependent on careful calibration of all components, whereas the accuracy of
the CD system is not.