From: David J Taylor on
"Wilba" <usenet(a)CUTTHISimago.com.au> wrote in message
news:00ceb424$0$15607$c3e8da3(a)news.astraweb.com...
[]
> This document - http://doug.kerr.home.att.net/pumpkin/Split_Prism.pdf -
> describes your idea as an "artifice" ("Getting real", p. 7), since it
> doesn't actually exist.
>
> In "Aperture dependency" (p. 17) the author describes the relationship
> between the position of the two "virtual AF apertures" (note: virtual =
> not in actual fact, form, or name; existing in the mind, especially as a
> product of the imagination), and the lens aperture and exit pupil. That
> is the basis of the "f/2.8" and "f/5.6" designations for the AF
> different AF sensors - denoting only which lenses are compatible with
> which sensor, because of the distance between the elements comprising
> each AF sensor.

What he is saying is that the "f/2.8" AF sensor will not respond to rays
form a greater exit pupil than "a typical f/2.8 lens". Hence it is quite
possible that the extreme rays from an f/1.8 lens will not be used by the
AF sensor, and hence if those extreme rays focus at a different point than
the central rays, the focus point indicated by the AF sensor from the more
central rays will be incorrect.

Cheers,
David

From: David J Taylor on
"Wilba" <usenet(a)CUTTHISimago.com.au> wrote in message
news:00cea8a8$0$15576$c3e8da3(a)news.astraweb.com...
[]
> But do you know why a particular AF sensor will have a range of lens
> apertures with which it will work? It seems to me that you're assuming
> that it's fundamentally about matching an aperture through which the
> sensor receives light. It absolutely isn't, it's about the base of the
> rangefinder - essentially the distance between the two elements of the
> AF sensor.

According to your reference:

http://doug.kerr.home.att.net/pumpkin/Split_Prism.pdf

it's about the exit pupil of the imaging lens.

> Here's a photo of the AF chip fitted to my camera -
> http://a.img-dpreview.com/reviews/canoneos450d/images/whatsnew/canon_450d_afchip.jpg
>
> See the four bars in a row through the centre? In discussions with
> others, we believe that the inner two bars form the "f/5.6"
> standard-precision sensor, and the outer two form the "f/2.8"
> high-precision sensor.
>
> When talking about AF sensors, "f/5.6" and "f/2.8" refer ONLY to the
> apertures of the lenses appropriate for each sensor. I assume that an
> f/5.6 lens can illuminate both elements of the "f/5.6" sensor, but can't
> illuminate both elements of the "f/2.8" sensor in the way it needs to
> work. AFAIK, that's all there is to it.

If your assumptions are correct, the point is that there is light from the
f/1.8 lens which is not used for the focus measurement, and the fact that
this unused light come from the extremes of the lens ray bundle is
critical should the lens have significant spherical aberration.

[]
> But you said focus confirmation would be wider if the lens aperture is
> wider than f/2.8. Now you're saying it's narrower?
>
>> You would need to check the exact details of Canon's system for a more
>> detailed answer.
>
> I have. AFAICT you're lost in a bizarre fantasy based on a terrible
> misunderstanding. :- )

It sounds as if you may be confusing narrower and wider with f/numbers
being bigger or smaller. I would expect the "f/2.8" precision focus
sensor to have a smaller range of "in-lock" indication. I have never said
otherwise.

Cheers,
David

From: Chris Malcolm on
David J Taylor <david-taylor(a)blueyonder.delete-this-bit.and-this-part.co.uk.invalid> wrote:
> "Wilba" <usenet(a)CUTTHISimago.com.au> wrote in message
> news:00cea8a8$0$15576$c3e8da3(a)news.astraweb.com...
> []
>> But do you know why a particular AF sensor will have a range of lens
>> apertures with which it will work? It seems to me that you're assuming
>> that it's fundamentally about matching an aperture through which the
>> sensor receives light. It absolutely isn't, it's about the base of the
>> rangefinder - essentially the distance between the two elements of the
>> AF sensor.

> According to your reference:

> http://doug.kerr.home.att.net/pumpkin/Split_Prism.pdf

> it's about the exit pupil of the imaging lens.

>> Here's a photo of the AF chip fitted to my camera -
>> http://a.img-dpreview.com/reviews/canoneos450d/images/whatsnew/canon_450d_afchip.jpg
>>
>> See the four bars in a row through the centre? In discussions with
>> others, we believe that the inner two bars form the "f/5.6"
>> standard-precision sensor, and the outer two form the "f/2.8"
>> high-precision sensor.
>>
>> When talking about AF sensors, "f/5.6" and "f/2.8" refer ONLY to the
>> apertures of the lenses appropriate for each sensor. I assume that an
>> f/5.6 lens can illuminate both elements of the "f/5.6" sensor, but can't
>> illuminate both elements of the "f/2.8" sensor in the way it needs to
>> work. AFAIK, that's all there is to it.

> If your assumptions are correct, the point is that there is light from the
> f/1.8 lens which is not used for the focus measurement, and the fact that
> this unused light come from the extremes of the lens ray bundle is
> critical should the lens have significant spherical aberration.

> []
>> But you said focus confirmation would be wider if the lens aperture is
>> wider than f/2.8. Now you're saying it's narrower?
>>
>>> You would need to check the exact details of Canon's system for a more
>>> detailed answer.
>>
>> I have. AFAICT you're lost in a bizarre fantasy based on a terrible
>> misunderstanding. :- )

> It sounds as if you may be confusing narrower and wider with f/numbers
> being bigger or smaller. I would expect the "f/2.8" precision focus
> sensor to have a smaller range of "in-lock" indication. I have never said
> otherwise.

You're right that the "f2.8" AF sensor will be *capable* of higher
precision focus because of what in effect is the wider baseline of its
rangefinding phase contrasting. But can it safely *use* that higher
precision? One of the things that calibrated lenses with on chip
information can tell the camera is just how sharply they should be
focussed, i.e., what maximum size of focus error still counts as being
in focus. That's what sets the range of manual focus that gives the AF
confirmation "green light". A very sharp lens can use a small value,
which if used with a softer lens would result in it never being able
to focus.

But there's a nastier problem than that. The higher precision AF
sensors selectively use rays refracted from the edges of a wider
lens. And some lenses are sharp in the middle, but soft at the edges
(I'm referring to the middle and edges of the lens, not the image). So
the AF sensor is limited in the precision of focus it can exercise by
how good the edges of the lens are, which are often not as good
as the middle.

This problem is at its worst in the ultra wide angle rectilinear
perspective projection zoom lenses. They're very difficult optical
designs, and the edges of the lens are often a lot more optically
aberrant than the centre. So phase contrast AF sensors often have
unusual difficulty accurately focusing the shortest focal lengths of
these lenses.

--
Chris Malcolm
From: Chris Malcolm on
Wilba <usenet(a)cutthisimago.com.au> wrote:
> Chris Malcolm wrote:
>> Wilba wrote:
>>> David J Taylor wrote:
>>>> Wilba wrote:
>>>> []
>>>>> The remaining challenge is to explain why the AF system confirms focus
>>>>> over such a wide range. The idea of asymmetrical bokeh makes sense, so
>>>>> the obvious thing to do now is compare the near side and far side bokeh
>>>>> of the 50/1.8.
>>>>
>>>> Already explained, I think. The AF sensor has an f/number of 2.8 (or
>>>> whatever, depending on the camera), not 1.8, so it may report a focussed
>>>> condition over a wider range.
>>
>>> OMG! You don't think that the "f/2.8" in "f/2.8 high-precision AF sensor"
>>> is
>>> some measure of an effective aperture of the AF sensor?! It would
>>> certainly
>>> explain a lot of bizarre thinking if you did.
>>
>> Bizarre thinking? As bizarre as the diagram of how phase contrast AF
>> works in this wiki entry? :-)
>>
>> http://en.wikipedia.org/wiki/Autofocus

> LOL - you guys are seriously ****-up. :- )

> This document - http://doug.kerr.home.att.net/pumpkin/Split_Prism.pdf -
> describes your idea as an "artifice" ("Getting real", p. 7), since it
> doesn't actually exist.

> In "Aperture dependency" (p. 17) the author describes the relationship
> between the position of the two "virtual AF apertures" (note: virtual = not
> in actual fact, form, or name; existing in the mind, especially as a product
> of the imagination), and the lens aperture and exit pupil. That is the basis
> of the "f/2.8" and "f/5.6" designations for the AF different AF sensors -
> denoting only which lenses are compatible with which sensor, because of the
> distance between the elements comprising each AF sensor.

But it's a useful simplifying fiction for those learning about this
stuff for the first time, just like Newtonian physics or the earth
being a sphere. You started out this discussion not wanting to go into
more details than necessary, and not wanting to have to understand
anything not strictly necessary to answer your specific questions. You
specifically asked for explanations to be given in simple terms as
though to someone who was completely new to the topic.

But you're now criticising the explanations you were given because
they used used various "in effect" "as it were" "virtually"
etc. didactic simplifications.

You can't have it both ways. Either you must make the effort to
understand the full technical complexities, or you've got to accept
the simplfying fictions which give a good general idea of what is
going on.

--
Chris Malcolm
From: David J Taylor on

"Chris Malcolm" <cam(a)holyrood.ed.ac.uk> wrote in message
news:7qmc4fF4csU1(a)mid.individual.net...
[]
> You're right that the "f2.8" AF sensor will be *capable* of higher
> precision focus because of what in effect is the wider baseline of its
> rangefinding phase contrasting. But can it safely *use* that higher
> precision? One of the things that calibrated lenses with on chip
> information can tell the camera is just how sharply they should be
> focussed, i.e., what maximum size of focus error still counts as being
> in focus. That's what sets the range of manual focus that gives the AF
> confirmation "green light". A very sharp lens can use a small value,
> which if used with a softer lens would result in it never being able
> to focus.

Accepted.

> But there's a nastier problem than that. The higher precision AF
> sensors selectively use rays refracted from the edges of a wider
> lens. And some lenses are sharp in the middle, but soft at the edges
> (I'm referring to the middle and edges of the lens, not the image). So
> the AF sensor is limited in the precision of focus it can exercise by
> how good the edges of the lens are, which are often not as good
> as the middle.
>
> This problem is at its worst in the ultra wide angle rectilinear
> perspective projection zoom lenses. They're very difficult optical
> designs, and the edges of the lens are often a lot more optically
> aberrant than the centre. So phase contrast AF sensors often have
> unusual difficulty accurately focusing the shortest focal lengths of
> these lenses.
>
> --
> Chris Malcolm

Thanks for that further information, Chris. It would seem that, for best
results, you should have the option of excluding the wide-aperture focus
sensors under certain conditions. Do any cameras provide this option?
Mine doesn't have dual AF sensors so the issue doesn't arise.

Cheers,
David