But you answered a question that I didn't ask. You answered how matrix
metering works to differentiate foreground and background. I asked how
it knows the distance to various autofocus points. I'm afraid your
answer there leaves much to be desired and appears to depend on the
smoke and mirrors you mentioned.
I'll take it point by point:
>Assume just two active focus points for simplicity sake.
OK
> Point One is in focus and the distance is close to infinity.
Not OK. How did we get to this point? I haven't focused on anything
yet and besides my subject is 10 feet away.
> The camera actually knows the relative position of the focus
distance
Not OK. How does it know that? I have a simple Tokina lens which does
not have any sort of focus distance encoders nor provide any focus
distance feed back to the body.
> (That racking in and out when you turn on or mount the
lens actually identifies to the camera the full focus range) and
in some cases actually knows the specific focus distance.
Maybe you have a camera that racks the lens in and out upon mounting or
power on but I don't. My Canyon and Tokina just sit there like dumb
rocks until I push a button to tell them to do something.
> Point Two is out-of-focus. Unless the lens is a long-telephoto with
focus-limit controls, the camera immediately makes the
assumption that Point Two is closer than Point One and will
readjust focus.
Since we assumed that Point One is in focus (even though it isn't
because I haven't chosen to focus the camera yet) I'll grant you that
Point Two must be out of focus. But I said my Point One was 10 feet
away and not at your convenient notion of infinity. So, even if I give
you that Point One is in focus, short of omniscience and reading my
brain the camera has no way of knowing where Point Two is located except
that it must be somewhere ahead of *or* behind Point One.
Sorry to give you such a hard time but I think you really mumbo-jumbo'd
this one.
Chuck Norcutt
AG Schnozz wrote:
>>I'm afraid you'll have to convince the skeptic (me) on this
>>one. How can it know the distance to various autofocus
>>points without having focused on each one and gotten feedback
>>from the lens positioner?
>
>
> Each manufacturer has their own patented smoke-and-mirrors to
> accomplish this, but here goes:
>
> Assume just two active focus points for simplicity sake. Point
> One is in focus and the distance is close to infinity. The
> camera actually knows the relative position of the focus
> distance (That racking in and out when you turn on or mount the
> lens actually identifies to the camera the full focus range) and
> in some cases actually knows the specific focus distance. Point
> Two is out-of-focus. Unless the lens is a long-telephoto with
> focus-limit controls, the camera immediately makes the
> assumption that Point Two is closer than Point One and will
> readjust focus.
>
> Ok, now the camera knows the relative distance (usually as a
> percentage of the entire focus range) between Focus Point One
> and Focus Point Two. The matrix-metering now assumes that Point
> One is background and Point Two is subject. By taking a meter
> reading of these two points the camera has effectively built a
> 3D matrix of foreground/background.
>
> Depending on the camera (Minoltas and the F5 being examples),
> the camera may adjust the focus back-to-front when in "Single"
> mode. This builds a 3D profile of the entire image and knows the
> relative distance position of each sensor in relation to the one
> focused on the closest point. By applying the exposure matrix
> to this, you can see where a tremendous amount of logic can be
> applied to the data and if color information is known (like in
> the F5) additional judgements can be applied.
>
> My understanding of how this works when you have the focus set
> in "Continous" is that during the initial shutter-release press
> that the camera will typically initialize the focus at infinity
> and then lock on to the closest AF sensor point. The camera
> will then track this across adjacent AF sensor points.
>
> Sensors aren't just dumb rocks either. For example, in the
> Canon, if the subject is closing in on the camera location, and
> the AF mode is set to continous, it doesn't "hunt" the wrong
> way, but actually will track the subject as it comes closer.
> This requires the sensor to have the ability to differentiate
> between closing and departing subjects. My E-1 is a dumb rock
> in this, though. :(
>
> Anyway, that is kinda an overview of how 3D matrix metering
> works. Each manufacturer has their own specific way of making
> it happen and they all take a lot of shortcuts and assumptions.
> For example, as long as the camera knows the relative focus
> position of the lens it will assume all out-of-focus AF points
> are background.
>
> AG
>
>
>
> ____________________________________________________________________________________
> Cheap talk?
> Check out Yahoo! Messenger's low PC-to-Phone call rates.
> http://voice.yahoo.com
>
> ==============================================
> List usage info: http://www.zuikoholic.com
> List nannies: olympusadmin@xxxxxxxxxx
> ==============================================
>
>
==============================================
List usage info: http://www.zuikoholic.com
List nannies: olympusadmin@xxxxxxxxxx
==============================================
|