P3D Re: Base Seperation Formulas

[Michael Watters <mwatters@xxxxxxxxxxxxx>]

Larry Haines brings up an older post of mine:

>message you sent to photo-3d last February relative to base separation
>for twinned stereo cameras.  You developed a simplified equation which

>B (Base Separation) = S1 (distance to nearest object) / f (focal length
>of lens).

Yup, that's the ultra-simplified version.  It makes a couple of assumptions, mostly  which  
are fairly reasonable.  Namely:

You're using 35mm film and are gonna project/view with the normal equipment.  - That's  
where giving the parallax a value of 1 (actual reasonable max is more like 1.2 I think, but 1  
is better for the calculation) comes into play.

The farthest object is located at a distance approximted by "infinity".  Realistically, this  
only has to be about 20-30 times further away than the subject is.  Basically, if you're  
outdoors, this condition is met.  Indoors?  probably not, unless there's a window to the  
outside or it's a really big room.

The lens is focused near infinity.  This is REALLY the assumption that the distance from  
the film to the node of the lens is about the same as the focal length of the lens.  This is  
gonna be true unless you're doing macros or if you're doing portraits with really long  
lenses.



>I have some questions.

Okie Dokie

>1.  I assume that in some way I do not fully understand, the equation
>gives a "good" stereo effect (however one would define "good").

Well, this is clearly a subjective thing.  What it actually DOES is tell you the base  
seperation to use if you want a 1mm parallax on the film.  Saying "1mm=good" is the  
subjective part.   Why was this chosen?  The 1.2 figure I think just comes from people's  
personal experience.  If I remember, it translates to something 6feet away IF you were  
shooting with a Realist.  (35mm lenses 2.5inch Base)

>2.  If I were using 28 mm lenses and the nearest object were magically
>28
>feet away then the cameras should have a base separation of one foot -
>correct?  Or if the distance of the nearest object was 280 feet the base

>should be ten feet - correct?  Looking for the edge of the envelope let
>me
>ask, what happens when the nearest object is about a half a mile, say
>2800 feet.  This would seem to indicate a base of 100 feet. But I guess
>there is some practical limitation for a particular focal length lens -
>is that so?
>This does not seem to be incorporated in the equation.

The practical limitations come into play because it's very very hard to find a situation  
where the nearest object is half a mile away.  Shooting off sides of cliffs, out of airplane  
windows, up into the sky (at fireworks).  That kinda thing.  For all practical use, you're  
gonna always have some foreground stuff hanging around in the picture.  That's gonna be  
especially true with the wide angle lens used in your example.

>3.  Then if I am using 135 mm lenses, the oft spoken 30/1 rule is really

People tend to mention either a 1/30 rule or a 1/50 rule.   These have been around since  
the dawn of time it seems.  Since most early cameras tended to have either 35mm or  
50mm lenses, I was mainly bringing that up to point out that under those curcumstances  
the B=S/F thing simplified to the old standard formula.

>wrong ( as you point out) since it should become rather 135/1 - right?
>So a camera base of one foot would be correct for nearest distances of
>about 135 feet - right?  This is quite a bit different than the 30 feet

Yup.  That's basically it.

>applications.  If I wanted to put clouds in stereo, I might seek even
>longer focal length lenses?

If you wanted the clouds themselves to be in stereo, yes.   Should work.  The only time I've  
really had an opportunity to test it at it's limits was when I shot some fireworks several  
years ago.  I was taking photos from about 2 miles away from display #1 and 10 miles  
away from display #2.  I was shooting with 135mm lenses and an 8-10 foot base.  I would  
have really prefered more, but that's how long my cord was.  There was clearly some  
decent stereo between the two displays, the internal stereo was somewhat lacking though.


>4.  I have spoken to a stereo photographer who takes a lot of twinned
>camera pictures.  He suggests that I carefully record what I do on
>each shot since the actual results might dictate something different
>than the
>equations.  I am not yet ready to accept this, since it does seem to me
>that these equations are pointing at defining optimum results.  I would
>like your comments on this.

I'd definatly agree with him.  Maybe not to the point of taking meticulous notes about  
everything, but on this basic point:  "Experience is more important than 'rules'".  That is  
undeniably true, no matter what aspect of photography you're talking about.  You're gonna  
know what you want your pictures to look like (a little over or under exposed, more extreme  
or more subdued stereo effect, everything behind the window or stuff poping out etc).  The  
equations are a good starting point, but I'd still bracket around them to see what I liked  
best.   The VAST majority of the time when I'm using a twin-rig, it's mounted on a  
side-by-side bar with a fixed seperation anyway.  The only thing I use a formula for is  
knowing roughly how far away my near objects should be.  If my bar allowed some  
adjustment of interoccular for example, and the formula said I should have a 6 inch  
interoccular and my cameras were sitting at 5inches (or 7 for that matter) I wouldn't change  
them.  This isn't rocket science after all.  We're not trying to hit Mars, we just want an  
acceptable photo.  "Close-enough" is a very valuable concept.
  Then again, even when a formula would dictate a MUCH larger base than I've got, I'll  
usually just change my framing and include a foreground object to put some stereo into the  
shot.  You avoid having things look too hyper that way anyway.

>5.  The matter of the perceived diminuation of the objects in the
>picture
>is of concern to me also, but this equation does not seem to address
>these
>things at all. 


It doesn't.  Not directly anyway.  Using wide-bases WILL make things look little.  That's just  
a fact of life.  Think of it this way:   For the moment, let's pretend that you had a perfectly  
orthoscopic stereo camera/display system.  Then, an object that was 3ft tall and 6ft away  
from the camera would appear (to your mind) to be 3ft tall and 6ft away when you viewed  
the image.  For the sake of argument, we'll say that object at 6ft away gives you an on-film  
parallax of 1mm.   Now, when viewed through your hypothetical ortho devise, ANYTHING  
with a 1mm parallax is gonna appear to be 6ft away.  Does that make sense?  So, if you  
alter your camera and double up the base so objects that are 12 feet away give a 1mm  
parallax, THEN view them through your usual ortho viewing devise, that object will look to  
be half the distance away (and half the size).  What you're basically doing when you play  
around with base seperations and the like is squeezing everything into that defined space.   
(about 6 feet away for the nearest object).     Same thing when you're going the other way.   
Macro stereo cameras have the effect of shoving an object further out in space than it  
really was and making it appear much larger than it really was.
    Such fiddling around can be done a little without it being too perceptable to the viewer,  
but do it too much and it'll be an obvious hyper shot.  


I should also point out that there's plenty of folks on this list who get a whole lot more  
exacting than I do when it comes to base-sep formulas.  I'm sure they could give much  
more precise versions of everything I've mentioned.   I'm very fond of the short-cut  
versions for two reasons:

1) having things be exactly ortho isn't important to me.  So having something to give me an  
approximate value is useful.

2) They're a whole hell of a lot easier to figure in your head.  With the simplified version of  
the formula, you can make a handy little slide-rule calculator.  Match the base and lens  
and it gives you the acceptable near-distance.  OR pick the near distance and read off all  
the various base/lens combinations that will work well.

mike
watters


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