Re: Twin Camera Questions: Toe-In, Hyper & Zoom

[bercov@xxxxxxxxxx (John Bercovitz)]

Mark Poole writes:
 
>      I have some questions related to Twin Camera setups, based on
>      some recent posts.
>      --------------------------------------
>      Background:
>      Pair of Minolta 7000i's
>      Triggered simultaneously by Infrared Master & 2 Slaves
>      3 foot & 6 foot Slide bar (for hyperstereo)
>      Bogen Tripod (heavy-duty) with bubble levels.
>      --------------------------------------
> 
>      [1] I have 35mm 35-75 autofocus zoom lenses on each.  From what I
>      can gather, am I supposed to use a setting that matches the lens
>      of the projection system (or viewer)?  What if that changes? I
>      use the Franka Twin 35mm Slide Viewer (does anyone know its
>      focal length?).
>      What if I am projecting the image (which I've never done), can
>      the slide projector compensate?  Isn't "normal" about 52mm on
>      35mm equipment?
 
So many questions.  Great!  Yes, the focal length of the lens on 
the camera should match the focal length of the lens on the 
viewer.  That way, when you view the pair, you "sit" the correct 
distance from the pair.  When you project, you also want to sit 
the right distance from the view but you have a variable viewer 
focal length in this situation because you can move to change your 
distance from the screen.  8-)  What you're trying to do is to 
position yourself at the perspective point, exactly the same as 
you would for flat photography only now, in 3D viewing, it's even 
more important.  So the rule is exactly the same; here's an 
example or two.  The aperture in a 2x2 mount is about 35 mm wide.  
If you shot that transparency with a 50 mm lens, then when you 
project it, you should sit at a distance from the screen of 
(50/35)*(the screen image width).  So let's say the image on the 
screen is 46" wide.  Then you should sit at a distance of 66" from 
the screen to have the correct perspective.  Same thing applies in 
stereo.  Suppose you're using a Realist which has 35 mm focal 
length lenses and a 21.4 mm aperture width in the slide mount.  If 
the projected image's width is again 46", you must sit at a 
distance from the screen of (35/21.4)(46") = 75".  Now you can see 
what's wrong with short lenses on your camera if you're going to 
project an image:  Everyone has to sit so close to the screen that 
there isn't enough space unless they all sit in a pile.  (Hey!  
That could be fun, depending.)  The synopsis would be that you 
want the image to subtend the same angle at your eye as the scene 
did at the camera's eye and so shooting for a hand held viewer is 
very different from shooting for projection.
 
You also asked what a normal lens is.  A normal lens has a focal 
length equal to the format's diagonal.  The nominal size of the 
35 mm format is 24x36 and the diagonal of 24x36 is 43 mm so 43 mm 
would be a normal lens in that format. 
 
Remember when yer using zooms, yuh gotta be careful to match the 
two lenses' focal lengths.  The eyes will tolerate a bit of 
mismatch but not a ton of it.  If the lenses are mismatched then 
the images sizes are different and it feels like you just got your 
glasses prescription changed, radically.
 
I don't know the focal length on a Franka.  There are a couple of 
solutions.  You could measure it or you could guess.  In general, 
the viewers with achromats have a focal length of 40-45 mm and the 
single lens viewers, whether plastic or glass, have a focal length 
of about 50 mm.  If it's a single lens, you can measure the 
distance from the apex of the lens to the transparency when the 
viewer is in focus and add 1/3 of the thickness of the lens.
 
What are the penalties if you view from the wrong distance or with 
the wrong focal length lenses in your viewer?  Depends on if 
you're a sensitive person or AN INSENSITIVE LOUT!  8-)  No, 
seriously, some people notice more than others but if you sit too 
close to the screen, the view will be squashed in the third 
dimension and if you sit too far, the image will be stretched in 
the third dimension proportional to the ratio of where you're 
sitting to where you should be sitting.  Same is true for lenses 
which are too short (close) or too long (far).
 
>      [2] Toe-in.  I thought this was not done (so much for book
>      learnin', John Boy).  It does make sense that the eyes can
>      converge for close-up viewing.  But what happens when you view
>      with parallel viewers?
>      Can I get away with this to compensate for the wider than normal
>      interocular distance of my camera lenses (a good 4 inches or
>      so)?  Will toeing cause eyestrain on the audience?
 
Right the first time.  Toeing in is a no-no but again you can get 
away with it if it's not excessive.  In fact, it's done all the 
time, in spite of my disapproval.  8-)  First thing to realize is 
that it's got nothing to do with whether the viewer has parallel 
optic axes or not.  That's a weak effect.  Oh, sure, you wouldn't 
want to overdo that either because excessive convergence of the 
axes makes the scene seem reduced in size, but still it's only a 
second-order sort of effect.  The primary problem with toeing in 
is that the scene scale is not the same from left to right for the 
two views.  Let's look at the left camera.  Looking down on it, it 
will be toed in if you turn it slightly clockwise.  A camera is a 
flat-to-flat imaging device, that is it images a plane in object 
space which is parallel to its film plane at some constant scale 
(magnification).  You can see that when you toe the camera, you 
are moving the planes in object space and now if you have two 
objects some horizontal distance apart and located in the same 
plane initially, they are no longer in the same plane after 
toeing.  Since they are no longer in the same plane, they are no 
longer represented at the same scale.  Here's a concrete example:  
There is an object to the left.  You toe in the left camera and 
that object's representation on the film becomes larger because 
now it is located in a closer plane.  You toe in the camera on the 
right.  That same object now becomes smaller because you toed the 
opposite direction.  Now you have a problem when you go to view 
the pair: the heights of images aren't the same (except for the 
image of an object lying in a vertical plane through the 
convergence point and perpendicular to a line between the lenses).  
Try to fuse that scene!  It's really weird.  It's one thing to 
have a scale error between images but it's another, and worse, 
thing to have scale error within images.
 
>      [3] Hyper-stereo.  I have heard (or read & forgot the source) of
>      1 to 30 or 1 to 40, ratio of the distance of lens separation to
>      nearest object in photograph.  Does this sound right?  I have an
>      aluminum bar that is 3ft wide (and thinking of adding a 6ft).  I
>      use this mainly for landscapes (views from the Blue Ridge
>      Parkway).  Related to the two questions above, what if I want to
>      include an object closer than the 1/40?  Can I toe in on this one
>      object without popping peoples eyes out?  Can I use zoom to bring
>      that object in closer?  Longer length tends to flatten 2d images,
>      does this still happen in 3d?
 
The classic 1 in 30 rule only applies to scenes which include 
infinitely-distant objects.  The real rule is that the on-film 
disparity shouldn't exceed one thirtieth of the focal length of 
the taking lens.  However, you have infinity in your scenes and so 
the rule devolves to the classic rule and your cameras should not 
be farther apart then 1/30th the distance to the nearest object in 
the scene.  
 
Toeing in doesn't affect this rule.  Toeing in is only a poor 
man's way of windowing that happens to have the deleterious side 
effect noted above.  
 
You can use zoom but then you'd flatten the image unless you move 
back when you view the transparencies.  In three dee you have not 
only the flat image perspective clues but also the binocular clues 
of perspective and so the effect is much stronger.  The two 
effects are in complete accord and so reinforce each other.
 
If you want to go closer than 1:30, then you're going to have to 
get rid of some of the distant background.  You can also 
ameliorate a little by making sure that there are no foreground 
objects partially occluding background objects.
 
>      I have tried zoom in hyper, but I failed to record exactly which
>      images I did it on and  what setting I used on each, so I have no
>      good way to quantify it. I guess I could get off my lazy rear-end
>      and do some scientific method testing, but I would appreciate if
>      anyone could provide me some rules of thumb.
 
Zoomed hypers are as easy to spot as any zoomed images if you're 
sensitive to it.  The scene will be compressed in the third 
dimension.  Mountains running from left to right will look 
impossibly steep (will exceed the angle of repose for their 
construction materials).  Hypering itself does nothing but rescale 
the scene exactly; all dimensions are reduced equally.
 
>      I am in the process of redesigning my twin camera rig, and need
>      to know if I should provide a method of allowing toe-in to the
>      sliding camera carriages.
 
Well here I have good news for you.  Don't bother to add toe.  
Just shoot on 24x36 and mount in 7-perf mounts and you will have 
all the film you need to properly window the scenes.
 
John B
 


------------------------------