From: Wilba on
Paul Furman wrote:
> Wilba wrote:
>> David J Taylor wrote:
>>>
>>> 1 - "I would expect the "f/2.8" precision focus sensor to have a smaller
>>> range of "in-lock" indication [than the "f/5.6" standard focus sensor]."
>>
>> Correct. Canon say their "f/5.6" AF sensors should focus "within the
>> DOF", and the "f/2.8" sensors should focus "within the inner 1/3 of the
>> DOF".
>
> I think that's a simplification.

It may be but it's their specification (AFAWCT). If you can prove your
system doesn't achieve that, Canon will calibrate it for free under
warranty.

> Assuming you are shooting at f/5.6,
>
> -If you have a zoom that only gets as wide as f/5.6, they promise the
> focus will at least be just barely inside the DOF.

More precisely, the _subject_ is at least barely inside the DOF.

> -If you have an f/2.8 lens, they promise the focus will at least be just
> barely inside the DOF of f/2.8 (smaller than f/5.6). That would guarantee
> that the focus will be within the inner 1/3 of an f/5.6 shot.

No, the spec says the subject should be within the inner 1/3 of the DOF if
the "f/2.8" AF sensor is invoked. That should be the case if your lens's
maximum aperture is f/2.8 or better, and at the DOF of the maximum aperture.

> I don't think it gets within the inner 1/3 of an f/1.8 shot or even an
> f/2.8 shot.

Why not?

> I suspect these cameras are not promising to always be inside the DOF of
> an f/1.8 shot and you are lucky to get that at all.

That's Canon's spec., and under the right conditions it seems to be
consistently achieved.

> Manual focus in live view is guaranteed to work though.

With sufficient illumination, yes (at least in the case of the 450D).

> Well, my D700 doesn't zoom to 100% in live view so even that is
> not perfect, although better than most people can see with their eyes
> through today's DSLR viewfinders with focus screens optimized for
> AF use.

Yeah, 10x on a 3" screen is pretty handy. :- )


From: Wilba on
Paul Furman wrote:
> Wilba wrote:
>>
>> But the fact remains that I _do_ get images in which the the centre is
>> well focussed, so why should I care that they might, theoretically, not
>> be?
>
> To seek out new knowledge.
> To boldly go where no man has gone before!
> <star trek theme.wav>
>
>>> I would hope that an f/1.8 lens from a reputable manufacturer would have
>>> very little spherical aberration, rendering the effects we have been
>>> discussing to be very small, so they will be present but all being well
>>> below the limits you can observe in normal circumstances.
>>
>> ...do I need to modify my test procedure to
>> control for this effect? (That's the bottom line.)
>
> No, it looks like you are testing beyond the specifications of the camera.
> It's only guaranteed to f/2.8. Any further investigation would likely
> prove frustrating <g>.

Yes, my test procedure would not be considered normal use of the camera, but
OTOH, trap focus is a standard technique for some. :- )


From: Wilba on
Paul Furman wrote:
> David J Taylor wrote:
>> Wilba wrote:
>> []
>>> Does this effect explain my test results? (I haven't seen anything so
>>> far that tells me so.) Therefore, do I need to modify my test procedure
>>> to control for this effect? (That's the bottom line.)
>>
>> It's been so long back that I can't recall what your test results were.
>
> Backfocus.
>
> Barely in the DOF coming from foreground focus (but consistent), usually
> outside the DOF coming from initial background focus and more random.

Optimal focus when the initial focus is on the near side of the subject,
_front_focus_ when coming from an initial far side focus.

If you've been thinking back focus all the way through, you need to rethink.
:- )


From: Paul Furman on
Wilba wrote:
> Paul Furman wrote:
>> Wilba wrote:
>>> David J Taylor wrote:
>>>> 1 - "I would expect the "f/2.8" precision focus sensor to have a smaller
>>>> range of "in-lock" indication [than the "f/5.6" standard focus sensor]."
>>> Correct. Canon say their "f/5.6" AF sensors should focus "within the
>>> DOF", and the "f/2.8" sensors should focus "within the inner 1/3 of the
>>> DOF".
>> I think that's a simplification.
>
> It may be but it's their specification (AFAWCT). If you can prove your
> system doesn't achieve that, Canon will calibrate it for free under
> warranty.
>
>> Assuming you are shooting at f/5.6,
>>
>> -If you have a zoom that only gets as wide as f/5.6, they promise the
>> focus will at least be just barely inside the DOF.
>
> More precisely, the _subject_ is at least barely inside the DOF.
>
>> -If you have an f/2.8 lens, they promise the focus will at least be just
>> barely inside the DOF of f/2.8 (smaller than f/5.6). That would guarantee
>> that the focus will be within the inner 1/3 of an f/5.6 shot.
>
> No, the spec says the subject should be within the inner 1/3 of the DOF if
> the "f/2.8" AF sensor is invoked. That should be the case if your lens's
> maximum aperture is f/2.8 or better, and at the DOF of the maximum aperture.

I call them for fibbing then.


>> I don't think it gets within the inner 1/3 of an f/1.8 shot or even an
>> f/2.8 shot.
>
> Why not?

Because it's optimized for f/2.8, not f.1,8.


>> I suspect these cameras are not promising to always be inside the DOF of
>> an f/1.8 shot and you are lucky to get that at all.
>
> That's Canon's spec., and under the right conditions it seems to be
> consistently achieved.

Yes, it seems reasonable and any better focus would require manual live
view approaches.


>> Manual focus in live view is guaranteed to work though.
>
> With sufficient illumination, yes (at least in the case of the 450D).
>
>> Well, my D700 doesn't zoom to 100% in live view so even that is
>> not perfect, although better than most people can see with their eyes
>> through today's DSLR viewfinders with focus screens optimized for
>> AF use.
>
> Yeah, 10x on a 3" screen is pretty handy. :- )
>
>


--
Paul Furman
www.edgehill.net
www.baynatives.com

all google groups messages filtered due to spam
From: Wilba on
David J Taylor wrote:
> Wilba wrote:
> []
>> Yes, but I'm thinking the "f/2.8" AF sensor is only going to see the rays
>> that it would see from an f/2.8 exit pupil, so it doesn't see the rays
>> from the extremes of the f/1.8 exit pupil. No?
>
> I don't know - it would depend on the precise optical design.

Have you read http://doug.kerr.home.att.net/pumpkin/Split_Prism.pdf? It
describes exactly what I said (AFAICT).