EF 50/1.8 AF Experiment?

David J Taylor wrote:
"Chris Malcolm" wrote in message
...
David J Taylor

wrote:
"Chris Malcolm" wrote in message
...
[]
Note by the way the the in-focus band defined by the acceptable focus
error band isn't a hysteresis band. But there's no reason why it
should have fuzzy edges either.

--
Chris Malcolm

Would you prefer the term dead-band?

A hysteresis band is a kind of dead band, but a dead band isn't
necessarily a hysteresis band, because a hysteresis band is
directional.

--
Chris Malcolm

Thanks, Chris. I thought that a directional effect was what Wilba was
seeing? Different focus when coming from nearest than when coming from
farthest?

Different kind of directional. When a system is to seek a target and
stop when it gets there there can be a problem of dithering or hunting
close to the target due to tiny movements over the threshold, possibly
due to noise. So some kind of delay between on-target and off-target
conditions is imposed, often by giving some kind of snap action over
the target, so that it has to run past the target to be triggered as
target reached, and then under-run it by some amount to be triggered
as off-target.

It's a kind of snap action. In the past often implemented with
snap-over springs on a mechanical switch. Now often implemented by
control software.

Wilba was originally describing something very like that, but since
other cameras behave the opposite way in that situation I wasn't
convinced he hadn't got it back to front, and didn't want to get
involved in that argument until it got a lot clearer what had in fact
happened.

--
Chris Malcolm

"Chris Malcolm" wrote in message
...
David J Taylor

wrote:
"Chris Malcolm" wrote in message
...
David J Taylor

wrote:
"Chris Malcolm" wrote in message
...
[]
Note by the way the the in-focus band defined by the acceptable
focus
error band isn't a hysteresis band. But there's no reason why it
should have fuzzy edges either.

--
Chris Malcolm

Would you prefer the term dead-band?

A hysteresis band is a kind of dead band, but a dead band isn't
necessarily a hysteresis band, because a hysteresis band is
directional.

--
Chris Malcolm

Thanks, Chris. I thought that a directional effect was what Wilba was
seeing? Different focus when coming from nearest than when coming from
farthest?

Different kind of directional. When a system is to seek a target and
stop when it gets there there can be a problem of dithering or hunting
close to the target due to tiny movements over the threshold, possibly
due to noise. So some kind of delay between on-target and off-target
conditions is imposed, often by giving some kind of snap action over
the target, so that it has to run past the target to be triggered as
target reached, and then under-run it by some amount to be triggered
as off-target.

It's a kind of snap action. In the past often implemented with
snap-over springs on a mechanical switch. Now often implemented by
control software.

Wilba was originally describing something very like that, but since
other cameras behave the opposite way in that situation I wasn't
convinced he hadn't got it back to front, and didn't want to get
involved in that argument until it got a lot clearer what had in fact
happened.

--
Chris Malcolm

My background to this is electronic, Chris, in data communication where
the hysteresis is intended to prevent noise from false triggering - it's
in normal logic circuits as well. For example, in serial communication
such as RS-232, a line may have to rise above +3 volts to register has
"high", but it can sink down to zero and still show as "high". It won't
show as "low" until it drops below -3 volts, and then it will need to rise
above +3 volts before it registers as "high" again. You may know all
this, some readers may not. It's not the exact figures which matter, but
that the circuit has a memory and doesn't change state at a single
precisely defined level.

So in the auto-focus case which end of the "in-focus" dead-band it
finished didn't matter, just that it could be different when approached
from infinity or from closest focus. the impulse response of the AF
system system and the dead-band width would determine where it stopped. A
system with overshoot could behave differently than a more heavily-damped
one.

Give me a 'scope and an electronic system over a mechanical one any day!
I feel I can see much better what happens.

Cheers,
David

Chris Malcolm wrote:
Wilba wrote:
Paul Furman wrote:

But if you manually focus from infinity it won't confirm focus till
you've
over shot it, even when moving very slowly on a focus rail?

This is a question I needed to be asked! :- )

I swear I checked this a dozen times, in the original thorough tests and
since, but testing on Sunday and this morning proves that I did get the
beep
test results arse-about, and I continued to trust my records of the
original
tests. I'm very sorry for confusing you. I blame the way the distance
scale
is arranged on the macro rail... no, I can't, it's all my fault. :- )

That's what I always thought. But I didn't relish the task of trying
to convince you to take another look! Congratulations to Paul Furman
on his extraordinary patience :-)

Yeah, sure. :- )

I took many 'nother looks, but once you get an idea stuck in your head...
:- )

That wasn't the only bit, and far from the most important bit, that Paul
wasn't getting.

Chris Malcolm wrote:
Wilba wrote:
David J Taylor wrote:
Wilba wrote:
David J Taylor wrote:

the fact that there is a dead-band where focus will be deemed to
be good enough if you are anywhere within that band. Approach
from different starting points, infinity or close up, and you will
likely
stop at a different position, always within that dead-band, though
(at least in theory).

Like Paul said, my AF system focuses precisely and repeatedly at the
ends
of a band, never in the middle. But if you put the subject in that
band,
the system will confirm focus.

Exactly what you would expect when the focus system has a dead-zone
or hysteresis, and when you are within that dead-zone the focus system
reports "in-focus".

Yeah but, "likely stop at a different position, always within that
dead-band", is not true in this case. My focus does not stop _within_
that
dead-band, it always stops at one or other of the ends of it. It's quite
precise, not fuzzy as you'd expect from a hysteresis band.

Why would you expect fuzziness in a hysteresis band? I've seen plenty
of systems with hysteresis with very precise boundaries to the
hysteresis band.

I don't, but David seems to, unless I'm reading him wrong.

Note by the way the the in-focus band defined by the acceptable focus
error band isn't a hysteresis band. But there's no reason why it
should have fuzzy edges either.

Whatever. Not my term.

Chris Malcolm wrote:

Wilba was originally describing something very like that, but since
other cameras behave the opposite way in that situation I wasn't
convinced he hadn't got it back to front, and didn't want to get
involved in that argument until it got a lot clearer what had in fact
happened.

The AF behaviour and the beep test finding the same focus points are still
as originally described, only the directionality of the beep test that has
been corrected.

"Wilba" wrote in message
...
[]
Why would you expect fuzziness in a hysteresis band? I've seen plenty
of systems with hysteresis with very precise boundaries to the
hysteresis band.

I don't, but David seems to, unless I'm reading him wrong.

I don't. I don't think I have ever used the word "fuzziness".

However, given that there is noise in any electronic or optical system,
and given that there will be tolerances in any mechanical system, I would
not be surprised if given a fixed starting point, the AF system did not
stop at /precisely/ the same finishing point every time.

David

Chris Malcolm wrote
Different kind of directional. When a system is to seek a target and
stop when it gets there there can be a problem of dithering or hunting
close to the target due to tiny movements over the threshold, possibly
due to noise. So some kind of delay between on-target and off-target
conditions is imposed, often by giving some kind of snap action over
the target, so that it has to run past the target to be triggered as
target reached, and then under-run it by some amount to be triggered
as off-target.

It's a kind of snap action. In the past often implemented with
snap-over springs on a mechanical switch. Now often implemented by
control software.

Thanks very much for your input to this thread: I read it all yesterday,
looked up the references (plus many others) and tried to understand how AF
works.

I thought I was doing well until I started thinking about the distance from
the exit pupil to the sensor/film plane. E.g. a long distance gives the
"narrow angle of attack" required by sensors to reduce vignetting - that
implies a wide baseline would be inappropriate in this case. I guess that
means I've missed a fundamental point (as happens quite often).

The reason for my great interest is attempting to understand an AF problem,
under very obscure lighting, with a non-Canon system: f/1.4 and f/1.8 lenses
refused to focus anywhere near the subject; f/2.8 and f/4 lenses were always
spot on. Tests shots made on each lens from wide-open to f/5.6 showed the
problem did not vary with shooting aperture. I certainly don't want to bore
everyone to death trying to understand it all. Indeed, I'm happy to have
learnt which lenses to avoid under certain lighting.

Pete

David J Taylor wrote:
Wilba wrote:
Chris Malcolm wrote:

Why would you expect fuzziness in a hysteresis band? I've seen
plenty of systems with hysteresis with very precise boundaries to
the hysteresis band.

I don't, but David seems to, unless I'm reading him wrong.

I don't. I don't think I have ever used the word "fuzziness".

But you did say, "you will likely stop at a different position, always
within that dead-band", which doesn't happen with this system (it always
stops at the ends of the beep band).

However, given that there is noise in any electronic or optical system,
and given that there will be tolerances in any mechanical system, I would
not be surprised if given a fixed starting point, the AF system did not
stop at /precisely/ the same finishing point every time.

Correct. Even CDAF has shot-to-shot variation. PDAF is a touch more
variable. But we're not talking about micrometric precision, just not
focussing between the end points of the beep band. The beep test shots
match, so it seems that PDAF triggers focus confirmation on the transition,
on entering or leaving the beep band.

Pete wrote:
Chris Malcolm wrote:

Different kind of directional. When a system is to seek a target and
stop when it gets there there can be a problem of dithering or hunting
close to the target due to tiny movements over the threshold, possibly
due to noise. So some kind of delay between on-target and off-target
conditions is imposed, often by giving some kind of snap action over
the target, so that it has to run past the target to be triggered as
target reached, and then under-run it by some amount to be triggered
as off-target.

It's a kind of snap action. In the past often implemented with
snap-over springs on a mechanical switch. Now often implemented by
control software.

Thanks very much for your input to this thread: I read it all yesterday,
looked up the references (plus many others) and tried to understand how AF
works.

I thought I was doing well until I started thinking about the distance
from the exit pupil to the sensor/film plane. E.g. a long distance gives
the "narrow angle of attack" required by sensors to reduce vignetting -
that implies a wide baseline would be inappropriate in this case. I guess
that means I've missed a fundamental point (as happens quite often).

The reason for my great interest is attempting to understand an AF
problem, under very obscure lighting, with a non-Canon system: f/1.4 and
f/1.8 lenses refused to focus anywhere near the subject; f/2.8 and f/4
lenses were always spot on. Tests shots made on each lens from wide-open
to f/5.6 showed the problem did not vary with shooting aperture. I
certainly don't want to bore everyone to death trying to understand it
all. Indeed, I'm happy to have learnt which lenses to avoid under certain
lighting.

That sounds like exactly what Chris has been talking about - a static focus
error due to the mismatch of the lens exit pupil and the AF sensor's
required exit pupil ("effective aperture"). I wonder what would happen if
you put on an external aperture which reduced the exit pupil to that
f/2.8-f/4 range.

"Wilba" wrote in message
...
[]
But you did say, "you will likely stop at a different position, always
within that dead-band", which doesn't happen with this system (it always
stops at the ends of the beep band).

So your system doesn't behave in the way which I would consider "likely".
Maybe the software is very clever and works out the absolute minimum
distance to move the focus to be /just/ within that dead band? Without
the design details it's difficult to be sure.

However, given that there is noise in any electronic or optical system,
and given that there will be tolerances in any mechanical system, I
would not be surprised if given a fixed starting point, the AF system
did not stop at /precisely/ the same finishing point every time.

Correct. Even CDAF has shot-to-shot variation. PDAF is a touch more
variable. But we're not talking about micrometric precision, just not
focussing between the end points of the beep band. The beep test shots
match, so it seems that PDAF triggers focus confirmation on the
transition, on entering or leaving the beep band.

Consistent with moving the minimum possible amount - i.e. getting the
fastest AF.

Cheers,
David