Maddy

First of all, I will repeat myself just a bit (for an introductory
statement):
Our circle of confusion (our eye’s “version” of depth of field)
defines our range of depth that is in clear focus. I can _see_ this
with my own eyes… consistently. (trust me… I am doing many tests
in all lighting conditions and lighting affects are not significant
until you reach the point of not being able to see anything - too dark
or too light, i.e. “blinded”)
This amount of depth that is in clear focus will increase in relative
amounts as the “far point” gets farther away, but the “deviation
amount”, which is defined as the _horizontal_ measurement between the
nearest point and the farthest point in our field of view, will stay
relatively constant (the same). What happens is when the far object
gets farther and farther away, the converging lines of sight, from our
eyes to the far point, lengthens. That point of “start of focus” (the
point in depth where the clear focus _begins_), also then moves
farther away in progressive amounts, and the angle widens, relative to
the face (eyes). It is a good idea to look at the graphic again, now,
before I say anything more, and I probably should make a new graphic
that shows this variable:
http://www.puppetkites.net/stereo3dtutgraphics/optimumdeviationeyes.gif
Next, the place where deviation is measured (see the graphic) moves
farther away from the face (it is the same point of the starting point
of focus), but when you then re-measure it, it is still the same
width!… always. When the far point is at infinity, that point will
be .89m (about 3 feet) away from your face. As the far point gets
closer, that point of “beginning of focus” or the point where
deviation is measured, gets close and closer to the face, the angle
narrows, and the deviation measurement stays exactly the same, with
the exception of ambient lighting, which is only affecting the
measurements by less than 1% or so… not much at all…. so small it
is difficult to measure a difference.
Now, as far as on screen deviation is concerned, this “theory” will
still hold true “on the screen”, but in a different way (same basic
visual results, though). The “best seat” for the image with 3.3%
deviation would be the one that has a 40 deg FOV (per Imre’s
correction). With an image with 3.3% deviation, a “correct volume of
character” would be experienced, as long as there were no “errors” in
the stereoscopic composition (again, there is a long list of possible
errors in addition to “wrong deviation”, e.g. keystoning, barrel
distortion, cardboarding). As you get progressively closer to the
screen, your FOV increases, and you start to see the z-compression
(apparent flattening). At a certain point, it will become “severe” but
the FOV can then also be defined as relatively “severe”. As you get
progressively farther away from the seat that has 40 deg FOV, your FOV
decreases, and you start to see z-exaggeration (apparent stretching).
at a certain point, it will become “severe”, but the FOV can then also
be defined as “severe”.
This scenario creates a “range”, where “optimum” deviation and FOV
are exactly in the middle (half way). As you offset this “range” in
any way, it will push the variables off to one side of the spectrum or
the other, i.e. an image with more than 3.3% deviation would
demonstrate z-compression (flattening) in a larger range and images
with less than 3.3% deviation would demonstrate more z-exaggeration in
a larger range, in progressive amounts, as you get more and more or
less and less.
0% deviation is totally flat (no depth) throughout the entire range.
6% deviation is extremely deep throughout the entire range, and
somewhere around 8% or so, stereoscopic failure occurs (the ability to
fuse or merge the images).
This is the basic idea… there are a long list of variables after
this…