From: Chris Malcolm on
Wilba <usenet(a)cutthisimago.com.au> wrote:
> David J Taylor wrote:
>> Wilba wrote:
>> []
>>> So you're saying that changing the aperture for the exposure can cause
>>> the focus to shift with a spherically-aberrant lens. Lovely. Wonderful.
>>> That's aperture-related focus shift as we know it
>>> (http://diglloyd.com/diglloyd/free/FocusShift/index.html).
>>
>> Correct.
>>
>>> So I ask the exact same question again - if there is no, N-O, none, nil,
>>> nought, and zero aperture change, how then can you get a focus shift due
>>> to this phenomenon?
>>
>> If the focus sensors have a different f/number than that of the lens, they
>> may focus at the incorrect position, i.e. where the convergence of the ray
>> bundle that the focus sensors can see is minimum.

> OK, again, there is no aperture _change_ in that idea, only a static
> difference, and I can't see how, if that difference were significant, it
> would not be dealt with by calibration.

Explain how you deal with it by calibration, given that there are
different kinds of designs of 50mm lenses around, some spherical, some
aspherical, some with large focus drift, some designed to minimise it,
and which are going to be used on DSLRs some of which have their
widest aperture AF sensor as wide as (approx) F2.8, and some only as wide as
(approx) f5.6?

> Again, since that idea has nothing to do with aperture-related focus shift
> as we know it (http://diglloyd.com/diglloyd/free/FocusShift/index.html), I
> need a name and references for that phenomenon.

It needs no new name since it's exactly the same phenomenon. You're
just failing to pick up on what happens when AF is brought into the
picture, which your favourite reference above doesn't consider.

--
Chris Malcolm
From: Chris Malcolm on
Wilba <usenet(a)cutthisimago.com.au> wrote:
> David J Taylor wrote:
>> Wilba wrote:
>> []
>>> Sure. So when the aperture *doesn't* change (focus and exposure both
>>> occur at the same aperture), how do you get focus shift in the centre of
>>> the image?
>>
>> I already explained that the focus sensor may only have a working aperture
>> of f/4, for example, so the aperture /is/ different.

> That's nothing to do with the only form of aperture-related focus shift that
> I'm familiar with - http://diglloyd.com/diglloyd/free/FocusShift/index.html.

> If you are talking about some other phenomenon, please provide a link to a
> good description.

> I can't see how what you're talking about would be relevant anyway, since
> the AF system is calibrated to produce a good focus under exactly those
> conditions.

How can that be, given that different lens designs have very different
amounts of this problem? How does the AF system distinguish between a
spherical and an aspherical design of 50mm lens? And if your answer is
(which is true of some DSLRs) that the camera reads the lens
calibration data from the in-lens chip, then how does the lens
calibration data deal with the different sizes of camera AF sensors
and their different algorithms?

--
Chris Malcolm
From: Wilba on
David J Taylor wrote:
> Wilba wrote:
>> David J Taylor wrote:
>>>
>>> If you are using manual focus there is still any backlash in the system.
>>
>> Irrelevant in this case.
>
> Unless the focus point is infinitely small, there is a dead-zone between
> the in-focus indication turning on and it disappearing. You might expect
> to find the the correct point of focus somewhere between those two points.

I guess you mean that there is a range of focusses over which the system
will confirm focus. Yes, with my 450D's high-precision AF sensor, that range
is about as big as the DOF near the closest focus distance. Best focus is
achieved at one end of that range, making the other end clearly OOF.

> As we seem to have drifted into things which you say are irrelevant,
> perhaps a succinct restatement of your original problem might help.

Okay. My original problem was that I couldn't figure out how to test the
"crude mechanism" theory of erratic focus with the EF 50mm f/1.8 II. Doug
McDonald gave me the idea of the beep test (like trap focussing). Using that
idea I was able to show three things - the AF system will confirm focus over
a range about as big as the DOF, where the system confirms focus depends on
where the lens was initially focussed, and that the smallest focus step the
lens can make is of the order of a few % of the DOF. Thus the "crude
mechanism" theory is disproved, and the erratic focus behaviour is
explained.

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.


From: Wilba on
Chris Malcolm wrote:
> Wilba wrote:
>> David J Taylor wrote:
>>> Wilba wrote:
>>>> []
>>>> Na, I don't buy it. I can't find any write up apart from speculation in
>>>> this thread (which seems to me to be based on a misunderstanding
>>>> of the original mention of the effective aperture of the AF sensor),
>>>> that describes what you're talking about in relation to
>>>> aperture-related
>>>> focus shift. Can you give me some references?
>>>
>>> I already pointed you to the Wikipedia article showing how lens
>>> aberrations could cause a shift of focal point as the f/number was
>>> changed:
>>>
>>> http://en.wikipedia.org/wiki/Spherical_aberration
>>>
>>> Look at this diagram:
>>>
>>> http://upload.wikimedia.org/wikipedia/commons/9/92/Spherical_aberration_2.svg
>>> and see how the rays from the edge of the lens are brought to a
>>> focus nearer the lens than those from the centre part of the lens.
>>> Of course, a good lens design will try to minimise this effect.
>>> Depending on the exact lens design, the shift of minimum circle
>>> of confusion as the aperture changes may be either towards
>>> or away from that at smaller apertures.
>>
>> So you're saying that changing the aperture for the exposure can cause
>> the
>> focus to shift with a spherically-aberrant lens. Lovely. Wonderful.
>> That's
>> aperture-related focus shift as we know it
>> (http://diglloyd.com/diglloyd/free/FocusShift/index.html).
>>
>> So I ask the exact same question again - if there is no, N-O, none, nil,
>> nought, and zero aperture change, how then can you get a focus shift due
>> to
>> this phenomenon?
>
> You keep quoting that diglloyd document. It goes into aperture related
> focus shift when focussing manually.

No, it says their examples were focussed "using Live View". There are three
ways to focus using Live View, two forms of autofocus and one manual, and
AFAICS they don't say which they used.

> It doesn't consider the focus shift problems introduced by AF
> mechanisms and the fact that they have fixed apertures which
> are usually smaller than the max aperture of the lens.

Correct. That's what I want to find out about. Where can I find documents
describing that?

> It suggests that if you want to know about that kind of thing
> you should pay money to get the fuller more detailed explanatory
> document.

I don't get any hint in that document about the phenomenon you're talking
about.

>> If you are talking about some other kind of aperture change or focus
>> shift,
>> I need a name and references to _that_. Thanks.
>
> It's exactly the same kind of focus shift, introduced by the factor
> not specifically mentioned in the diglloyd document: the difference
> which usually exists between the fixed effective aperture of an AF
> phase detection sensor, and the lens aperture which the image
> exposure in made at.

What happens because of this effect? How would the pre-exposure and exposure
focusses be different if it is and isn't present?

Does this effect change the focus at exposure-time if the exposure-time lens
aperture is the same as the focus-time lens aperture? In other words, if the
same difference is there at focussing and exposure, how is this effect
significant?

How can this effect be demonstrated? In other words, what variable can be
changed between exposures to show this effect alone?

>>> Remember that the auto-focus on some cameras can be open-loop
>>> rather than closed loop, and therefore subject to varying degrees of
>>> error. The sensor says the focus is off by so much, the firmware
>>> says move the lens by a certain amount, the lens is moved, and that's
>>> that. There's no second check to see how much error remains.
>>> In continuous-focus mode rather than single-shot-focus mode,
>>> performance may differ again.
>>
>> This is a closed-loop system, so not relevant.
>
> But the loop is closed on an AF phase detect sensor with a fixed
> aperture which is usually less than the widest aperture of the lens in
> use.

So...? I need more than just statements that there is a mysterious
inexplicable unpredictable effect. I need an explanation of how it works,
what happens because of it, and how that's relevant to the case of exposing
at the lens's widest aperture. If it has been documented, all I need is a
URL or specific name. I haven't been able to find anything about it by
searching for aperture-related focus shift.


From: Wilba on
Chris Malcolm wrote:
> Wilba wrote:
>> David J Taylor wrote:
>>> Wilba wrote:
>>> []
>>>> So you're saying that changing the aperture for the exposure can cause
>>>> the focus to shift with a spherically-aberrant lens. Lovely. Wonderful.
>>>> That's aperture-related focus shift as we know it
>>>> (http://diglloyd.com/diglloyd/free/FocusShift/index.html).
>>>
>>> Correct.
>>>
>>>> So I ask the exact same question again - if there is no, N-O, none,
>>>> nil,
>>>> nought, and zero aperture change, how then can you get a focus shift
>>>> due
>>>> to this phenomenon?
>>>
>>> If the focus sensors have a different f/number than that of the lens,
>>> they
>>> may focus at the incorrect position, i.e. where the convergence of the
>>> ray
>>> bundle that the focus sensors can see is minimum.
>>
>> OK, again, there is no aperture _change_ in that idea, only a static
>> difference, and I can't see how, if that difference were significant, it
>> would not be dealt with by calibration.
>
> Explain how you deal with it by calibration, given that there are
> different kinds of designs of 50mm lenses around, some spherical, some
> aspherical, some with large focus drift, some designed to minimise it,
> and which are going to be used on DSLRs some of which have their
> widest aperture AF sensor as wide as (approx) F2.8, and some only
> as wide as (approx) f5.6?

My 50/1.8 has been calibrated with my 450D body by the Canon service centre
in Sydney. They wouldn't tell me how they did it, but they have made it so
that that combination focusses optimally via PD AF at all apertures (as long
as I start with the plane of focus between the camera and the subject).

Where and how is your effect apparent in that?

>> Again, since that idea has nothing to do with aperture-related focus
>> shift
>> as we know it (http://diglloyd.com/diglloyd/free/FocusShift/index.html),
>> I
>> need a name and references for that phenomenon.
>
> It needs no new name since it's exactly the same phenomenon.

Let me see if I've got this right. You're saying that the focus shift you
get when the aperture stops down during exposure with a spherically-aberrant
lens (an exposure-time effect), is exactly the same shift you get because
the effective aperture of the AF sensor is different to the widest lens
aperture (a focus-time effect)?

That doesn't make sense. The way I see it, the AF sensor's effective
aperture may affect how it "sees" the rays that come through the lens, and
it will focus according to that "view" rather than what the image sensor
would "see" at the widest aperture. So if the AF system is able to focus
optimally like that (which mine does), what's the problem at exposure-time?
The only shift that can happen then is because you expose at a tighter lens
aperture, which causes the only kind of aperture-related focus shift I can
find documented (well, and in many places). What am I missing? What are you
assuming that you haven't stated?

> You're just failing to pick up on what happens when AF is brought into
> the picture, which your favourite reference above doesn't consider.

So explain it to me, or give me a reference.