[MF3D.FORUM:381] fl/30

[Tloc54452@xxxxxxx]

You guys are nothing if not timely; you ask questions 
just as memory starts to fade.  Great review for me.

Bill writes:

> Are you saying if we use your formula, that we should 
> vary the
 deviation based on the fl lens used?

In a way.  The formula assumes you will view from the
perspective point.  So if you use a focal length which
is twice normal, you will either sit twice as far from
the screen or, what is equivalent, you will put longer 
focal length lenses on your viewer.

Why this rule?  It's because tolerance for depth in a
stereo pair is an angular tolerance.  As long as the
angular difference between near and far doesn't exceed
1 in 30, the pair is easily viewed.  So just take your 
focal length and divide by 30 to find the on-film max.
linear measurement of deviation.

This really is an angle.  If you have a near object and 
a far object, and to your left eye, they are aligned, 
then your right eye shouldn't have to converge more than 
1 in 30 to acquire the near object.  This is a small angle
so it could be arcsin or arctan.  So arctan 1/30 = 1.9
degrees.  Not much, huh?  If your eyes are 65 mm apart, 
then that nearest object shouldn't be any closer than
65/tan(1.9) = 2 meters.

So to answer Bill's question, what I'm saying is that if 
you vary your focal length, you should vary your seating 
distance to match, in an ideal world.  What about the 
not-so-ideal world?  Well, that's where Mannle's way 
comes in - just keep the on-film devaition to 1/30th of 
the normal focal length of the format.  I've never 
experimented with this, never pushed the envelope, so I 
can't comment except to say that many people use it and 
they don't have trouble with viewability.

Paul writes:

> BTW, how is the 2.7 calculated?

That would be the Mannle method: format's normal focal
length divided by 30.

>> On the other hand, if you are shooting for projection, 
>> you are planning on a longer distance from observer to 
>> screen so you should use a longer focal length and can 
>> use more on-screen deviation than if your viewing 
>> distance were for 50 mm or 35 mm lenses

> Very surprising to read this!  I've always labored on 
> the impression that when shooting for projection one 
> should be conservative and aim for no more than 1mm OFD 
> (in R-format), as opposed to the 1.2mm figure usually 
> cited.  The rationale is that the image is magnified a 
> great deal in projection, and if you end up with a lot 
> more than 2.5 inches on-screen separation you can cause 
> viewing discomfort in the audience.

Yeah, you need to keep the infinity separation down to 
2.5".  So if you set infinity at 2.5" on screen, the 
near point reconstruction pops off the screen (depending 
on screen size/distance) if you have more than 1.2 mm of 
on-film deviation.  If the infinity points have extra 
separation in the mount,then the window also pops off 
the screen.  This is kind of cobbish, though.  Better to 
mask off the outer borders of the apertures in your 
standard mount and leave infinity separation in the mount 
at standard.  That way the projectionist doesn't have to 
make adjustments for infinity separation.  Too bad masks 
like this aren't offered.  Instead, the manufacturers 
opted to keep the aperture separation constant when making 
closeup masks.

Anyhow, standard mounts are extremely conservative and 
I wouldn't feel the least bit bad about masking them down.
What the heck, the window pops off the screen if you have 
a distant screen anyway.

> If I'm following correctly, 1/fl is a quickie way to keep 
> from exceeding...

OK, this is for the rule where stereobase doesn't exceed 
some fraction of the near distance.  You are trying to keep 
on-film deviation to the normal for the format.  That's
Mannle's way.  You use 50 mm lenses, near point in the scene 
is 50 times stereobase; you use 100 mm lenses, nearest point 
in the scene is 100 mm times the stereobase.

Note that you are trying to keep OFD at normal for the 
format.  So 1/fl doesn't work for MF.  There it's more 
like 2/fl because the format is about twice as big.

If you do your geometry, you'll find that 2/fl means you
are aiming for 2 mm OFD, which is ~normal for MF; and 1/fl 
means you are aiming for 1 mm OFD, which is ~normal for the 
Realist format.  If this isn't obvious, and you don't mind 
more math, I can make up a .pdf to illustrate the idea.

> In MF, 1/fl in a normal scene that includes infinity
> would also tend to keep OFD under 1.2mm, right?  And that's 
> well below the target maximum of 2.7, so 1/fl doesn't 
> directly help as a shortcut for MF.  It would need to be 
> scaled up to 8/3 * 1/fl.  With an 80mm lens, that puts us 
> back at 1/30.  Then to maintain a constant OFD shooting the 
> same scene with a 240mm lens, the new base would be (8/3 * 
> 1/240) or 1/90.  Am I warm?

You're hot!  We could quibble over the last decimal place, 
but that would be silly.

Oleg writes:

>
 1.2 mm is the most deviation you can squeeze between the 
> window and nominal
 infinity when you mount your slide.  I 
> thought this held whether the slide would
 be hand viewed 
> or projected...

Yes, that's the correct figure for Realist format (RF).
For MF, it's more like 2.7 mm.  35/30 = 1.2 mm; 80/30 = 
2.7 mm.

>> 2) base = near point/30 regardless of focal length

>
 This is what I started out with.  But I kept getting too 
> much depth in my
 telephoto shots, so I started factoring 
> in focal length.

Right, and that's because the audience was actually sitting
too close for long focal lengths; on film deviation was
also way too much for the standard mount.  Lots of problems 
here.

>> 3) on film deviation = focal length/30
>> ...Spicer's and my rule which can be back-derived from
>> Ferwerda and Waack, at least their advanced sections

>
 Lost me here, John.  Could you give an example of how this 
> rule would be
applied?

Well, let's put it in terms of the discussion above.  
Let's say you shoot with the ideal focal length for RF 
and projected viewing, which is 70 mm.  Then your MAOFD 
should be 70/30 = 2.33 mm.  Now your rule becomes (2.33/fl) 
times distance-to-nearest-point-in-scene = maximum 
stereobase.  You also have to mask your mount apertures 
down to get 2.33 mm instead of 1.2 mm on them.  (Start 
with a wider-than-normal mask.)

> I have been using 2/1 scaling between the 35mm and medium 
> formats, resulting in a 2.4mm maximum OFD.  Sounds like 
> that was way too conservative.

Not really.  The normal lens for RF is 35 mm and the normal 
lens for MF is 80 mm.  80/35 ~= 2.  You could also use 
(80/30)/fl = 2.7/fl but then to be consistent, you'd use
(35/30)/fl = 1.2/fl for RF.  Pick your poison.

=========================================================

OK, having said all that, I disagree.  8-)  I disagree with 
the idea of keeping a fixed OFD by varying FL and near point.
Also, I don't like these rules because they imply infinity 
is in the scene and often it's not.  Certainly if it's not,
you are limiting yourself by the 1 in 30 rule and no macro
shooter uses this rule.  Instead, they use charts which can
be made from the MAOFD spreadsheet or looked up in Waack or 
Ferwerda.

I would always start with the perspective point of the 
observer.  If he sits 18' from a 6' screen, why would you 
use a 35 mm taking lens for RF?  That way lies extreme 
stretch.  Instead, use a 70 mm lens (good luck finding one) 
and thereby make the observer's seat the correct seat.  
This is absolutely nothing new; read Ferwerda.

For macro, you want the same focal length.  So use a 50 mm
macro lens (RF) and rack it out to 70 mm.  Your observer's 
perspective is correct.  Then you get to choose the amount
of enlargement in the stereo reconstruction by varying the 
stereobase to whatever you like, within reason.  Don't make 
your stereobase so large that you have more than 70/30 = 
2.33 mm of OFD, though.  A lot of times you will want less
to make the enlargement of the object more.  And again, yes,
2.33 mm means masking your mount apertures down.

If anyone read this far, he is nuttier than I am, almost.  
8-)

John B