[tech-3d] Re: Stereo Base Calculation with a $20.00 Handheld

["Michael K. Davis" <zilch0@xxxxxxxxxxxx>]

Abram,

>Date: Fri, 24 Nov 2000 12:56:40 -0000
>From: "Abram Klooswyk" <abram.klooswyk@xxxxxx>
>Subject: Re: Stereo Base Calculation with a $20.00 Handheld
>
>Subject: Re: Stereo Base Calculation with a $20.00 Handheld
>
>Dear Mike Davis,
>Quite a defense you wrote (22 Nov 2000) in favor of the "General
Solution"!     
>Just a few comments. 

I encourage you to read it again Abram - it was not a defense in favor of
the General Solution vs. other stereography techniques.  It was simply a
rebuttal to your contention that newbies who attempt to use the General
Solution find themselves in an "endless regression".  The punchline of the
article to which you are responding here was:

>>   Date: Wed, 22 Nov 2000 23:31:39 -0600
>>   From: "Michael K. Davis" <zilch0@xxxxxxxxxxxx>
>>
[snip]
>>I can't imagine any newbies actually doing the math to rearrange the
>>General Solution in the first place, much less finding themselves in an
>>endless regression, solving for deviation, then for base, then for
>>deviation, then for base, etc.  You argument is simply fantastic.



>Of course the equation solves for base, but, supposing a certain 
>camera and a certain scene, some of the right hand variables are 
>more or less settled in the practical shooting: focal length, near 
>and far point distance. 

Absolutely.  These variables aren't chosen, but they can surely change from
one scene to the next.

>Then only the deviation value should be _chosen_ as input, and 
>newbies tend to find this difficult.   

It's not difficult to understand that Camera FL/30 is the Maximum Allowable
On-Film Deviation and that a little bit of experimentation would soon
reveal what percentage of MAOFD you personally find appealing.  If it's
your desire to maintain a consistent deviation from one scene to the next,
you can select some percentage of MAOFD for all your shots.  If, for some
situtations, you want to exaggerate or dimminish the deviation, you can do
that, too.  The formula enables precise control of deviation, calculating a
base value that will delver the look you choose.

>Of course test rolls will 
>give most people experience, but then, why a formula at all and 
>not just testing different bases?

For how many combinations of near and far point distances are we to test
bases?    We'll have to repeat these tests for every focal length we use.
To achieve even a maginal degree of reproducability, we would have to
evaluate, choose and a hundred or more slide pairs.   Would we measure the
on-film deviation produced with each pair or would we just eyeball each
bracketed set to select the one that looks *best*?   

Thanks, but that's a lot of work for a low probability of reproducible
results.  I'll stick with the formula.

>I wrote:
>>>When you want to emphasize tiny scratches on ancient coins, 
>>>stretch might be your purpose. 
>You answered: 
>>It is apparent that your intention here is to suggest that in a 
>>situation where stretch might be desirable, I would use the 
>>General Solution and this would somehow prohibit that goal. 
>
>I'm sorry not to have been clear enough on this point. 
>My intention was just the opposite: the so-called General 
>Solution tends to give stretch in many cases, but that should be 
>a purpose only in selected cases, like with the ancient coin. 

The General Solution does NOT tend to give stretch in many cases.  It does
NOT tend to give squash in many cases.  It TENDS to deliver precisely the
look one wants, as specified by the deviation.  Your suggestion that it
"tends to give stretch" implies you would prefer a smaller value for the
variable d, than that used to produce the stretched images you are
attributing to General Solution.  Deviation is a variable, not a constant.
It does NOT need to be set to MAOFD.  

>>The formula matches the geometry perfectly.  It is both valid 
>>and useful.  

>As said before, the matching to geometry is a matter of taste. 

No, matching to geometry is NOT a matter of taste.  The geometry is
absolute.  Choosing a deviation is a matter of taste.

>I like the concept of Viewing Space, where the difference in 
>convergence of the eye axes is about 2 degrees between 2 meter 
>(7 feet) and infinity. 
>
>Base formulas which include far point distance as well as near 
>point distance (in object space), tend to squeeze, or mostly 
>stretch, the extent of the photographed scene to _cover_ the 
>_entire_  viewing space from infinity to (mostly) 7 feet. 

Again, your contention exhibits an assumption that deviation must equal
MAOFD.  Given the right equipment, I can easily produce images using a base
calculated by the General Solution that will have too little depth for your
tastes.   All I have to do is specify an incrdibly small value for deviation.

>Remember that taking pictures is not the end, mounting, viewing and 
>perception follow. Several years ago I have put the whole process 
>from object space to viewing space in a spreadsheet
>(its gone now in the dust of MS-Dos 2 and 5 1/4'' floppies), 
>showing how a test object (cube) would be deformed by the variables 
>of the process. 
>That the "general solution" squeezes or stretches follows directly 
>from the math. That I call the Procrustean approach. 

Nonsense.  Deviation is a variable - no such conclusions can be drawn
without asssociating the conclusions to a specified value for deviation.  I
can use the General Solution to stretch or sqeeze as I see fit.

>Moreover, the effect will vary from slide to slide, when the 
>formula would be used consequently. 
>
>Of course you can use different percentages of deviation. 

Yipeee!

>>John Bercovitz and others have preceded me in communicating that 
>>one should select a DESIRED on-film deviation...
>But then sooner or later the newbie will ask for a formula :-) 
>to help him chose the deviation... My "Carrolleschery" was just a 
>pun, but such formulas have been published. 

Your Carroleshcery was definitely funny, but completely incongruous with
reality.  It's not difficult to shoot a roll of film using a deviation
equal to 80% of MAOFD and ask yourself if that's too much or too little
depth for your taste.  Again, where's the paradox?

>>It is my personal preference to use a variable base, following 
>>the General Solution, for nearly every situation, including the 
>>mid-field.
>A funny thing is that John Bercovitz, who started the original
>Tech-3d 
>list but has not posted here since long, used to call himself 
>"Orthoman", meaning that he was (is?) in favor of orthostereoscopy 
>all the time. His site is still online:
>        http://www.angelfire.com/ca2/tech3d/
>Orthostereoscopy is an orthodoxy which condemns using anything 
>other than a normal (about 6.5 cm) base for the midfield.

Your definition may be true, but I do not condemn orthostereoscopy and I
have not seen any evidence that John Bercovitz does so.    I like the
results I get with variable-base equipment, using a deviation equal to 80%
of MAOFD, for most situations where the Near:Far ratio exceeds 1:2.  Can
you permit me that perogative?

As I wrote at the top of this posting, the article to which you are
responding only made a rebuttal to your contention that newbies who attempt
to use the General Solution find themselves in an "endless regression".  

I have NEVER condemned orthostereoscopy, nor have I ever made any
statements comparing the use of the General Solution with variable base
techniques to the use of fixed-separation techniques.  Look long and hard.
You won't find anything but a comparison of the General Solution to other
approaches for calcuating base.  Even within this limited venue, I allow
others to enjoy whatever base calucation method they find appealing, but
salute those whose techniques allow them to consistently produce
pre-visualized results.  

Others have reacted to my original article, the title of which is in the
subject header of this one, with a defense of orthostereography.  Anyone
who wants to condemn the use of a base not equal to 65mm is welcome to do
so, but understand that I am NOT a variable-base biggot and I have not seen
any evidence that John Bercovitz (a.k.a. Orthoman) is either.

>I understand that you don't use the "General Solution" always, so 
>is it too ambiguous (:- )) or is it not so general after all? 

Let me quote my original article:

>>For Near:Far ratios less than 1:2, I have chosen to employ the 1/15 Rule
(modified to reduce the deviation to 80% at 1:2, instead of 100%, and to
accommodate the mismatch of my viewer focal length to my camera focal
length.)  For a 65mm camera FL and 78mm viewer FL, the rule becomes 1/15.8
for 80% of MAOFD.  Using a linear rule for Near:Far ratios less than 1:2 is
actually a good idea, in my opinion.  It automatically tapers the on-film
deviation as the Near:Far ratio shrinks toward 1:1 and that's a lot easier
than accomplishing the same thing by bumping the Percent of MAOFD variable
down incrementally.   
>>

As the last sentance suggests, the General Solution actually could be used
for Near:Far ratios less than 1:2, but I followed with this:

>>I have chosen to do this [use a 1/n rule at ratios below 1:2]] because I
believe it is more appealing than sticking with a high, fixed percentage of
MAOFD when the subject depth is so shallow.  
>>

So Abram, do you see that the General Solution is NOT ambiguous.
Personally, I have chosen to abandon it with Near:Far ratios below 1:2,
only because sticking with it would require my tapering the deviation
manually TO SUIT MY TASTES.  It is this CHOICE I have made to reduce the
deviation as the subject depth shrinks that has made the use of a linear
rule more sensible to me for ratios below 1:2.  The General Solution is a
viable, though more cumbersome solution, at these ratios - deviation is a
variable, not a constant!

>"Useful for all but macro work" you write, and "I abandon the 
>General Solution at Near:Far ratios less than 1:2". 
>(Your original post is also at:
>http://www.egroups.com/message/tech-3d/48    )
>
>The "truth embodied by the General Solution" which you defend so 
>"vigorous"ly  just has not been accepted generally as useful for 
>general use, especially not for the mid-field (which still is the 
>subject in a majority of stereopictures).

If it is true that the General Solution "has not been accepted generally as
useful for general use" it is only because too many people have treated the
deviation variable as a constant.   Every comment you've made regarding the
General Solution's "tendancies" has exhibited this very ignorance.  I'll
say it again:  Deviation is a variable.  

Even the on-film deviations produced by the protagonists of
orthostereoscopy adhere PERFECTLY to the mathematics of the General
Solution.  Examine a slide pair produce by a camera like the Stereo
Realist, entering the measured deviation, measurements of the near and far
point distances (made at the time of exposure), the focal length, etc. and
you use the General Solution to solve for a base that EQUALS the fixed
separation of the Realist.  The General Solution fits the geometry
perfectly.  There's no escaping this absolute truth.  How can it be ambigous?

>The precursors of the formula in this 20th century (we are almost 
>in its last month now) have been disputed by many well known 
>stereographers since before you and I were born. 
>Some other time I have said that "Private" Solution would be a 
>better name than "General" Solution .:-)
>
>>the Maximum Allowable On-Film Deviation is equal to f/30.
>This again is disputable. Read what Ferwerda and Krause wrote in 
>their books on "double depth",  which for the 5P system means 
>2.4 mm deviation, about 1/15 * camera focal length. 

The use of f/30 to define the Maximum Allowable On Film Deviation may be
disuputable, but your reference to Ferwerda's discussion of "double depth",
as found in sections 25.3 and 25.4 of his book, "The World of 3D", does NOT
support your contention at all.  Ferwerda discusses "double depth" as the
second of two methods one can use to deal with close-up scenes (the title
of the Ch. 25 is "Mounting close-ups with a normal base") where "the
deviation of the near-point is larger than the largest deviation
permitted."  The first method involves the use of special masks with the
5-P format and has NOTHING TO DO WITH DOUBLE DEPTH.  

How can you deduce or infer that Ferwerda believes the MAOFD for the 5-P
format is f/15 instead of f/30 ?

Can you tell me precisely where Krause supports this contention?

>>Useful for all but macro work.
>In my experience a general formula like you mean _can_ be used 
>for macro, if only (bellows) extension is included in the math, 
>accounting for "working" focal length. Also, as in any stereo 
>photography, it must be known if a more or less normal way of 
>mounting and viewing is used.
>(Surprised that I defend a "general solution" against you? :-))

No.

>The bottom line is that any formula or rule will work for the 
>knowledgable person who understands stereogeometry and 
>perception of stereo images, and that personal taste, not 
>mathematics, will be the final judge of what is acceptable.
>
>Abram Klooswyk

Incorrect.  "Any formula" includes formulas like the 1/30th rule, which
does not adhere to the geometry across the range of Near:Far ratios where
it is commonly applied (greater than 1:2), and thus, can not deliver
consistent on-film deviation.  If the inconsistent on-film deviation
delivered by the 1/30th rule actually satisfies the taste of some
individuals, how can so many of those individuals use 1/30th at Near:Far
ratio just barely exceeding 1:2, then shift to the 1/15th rule at precisely
1:2 and below?  As shown in my original article, at a Near:Far ratio of
1:2.1, these people would jump from a base of 305mm to a base of 152mm with
a tiny change in Near or Far point distance.  

No, you can't be satisfied with just "any formula", unless your taste is as
inconsistent as your technique.

Mike Davis


-------------------------- eGroups Sponsor -------------------------~-~>
eGroups eLerts
It's Easy. It's Fun. Best of All, it's Free!
http://click.egroups.com/1/9698/1/_/520353/_/975797595/
---------------------------------------------------------------------_->