[Raytrace] A useful sequence

[Peter John Smith]

Been rather busy. For those just beginning, here are some useful exercises.

In isolation, some are trivial and apparently rather stupid.  But, in
retrospect, I do think they are very useful.

Given that an optical design software package is much more than a raytracer,
it is worth useing the built in optimiser to do any trivial thing in an
attempt to become really at home with it.  The computer evaluation may be
impressive but its the optimiser that is pure magic.  Thats why few free
programs have one.

I am not sure how to do this with OSLO but would be surprised if there was
much of a problem.

As you pass from one stage to another, keep a general file handy.  Then when
you ever want to play with something similar, just modify it.

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

1/  Set up a spherical primary newtonian.  The 10 inch F:6 Newtonian
mentioned before could be a good starter.  However, since I never work in
anything but
  mm  for optical design, I would do a 250 mm diam F:6 instead.  mm is
pretty well a world standard for optics.  If I want to change to inches
later it is easy in
  Zemax.  Check oyt OSLO to see if this can also be done at the touch of a
button.
  It may be easier to modify the file already talked about by changing the
mirror to a sphere (ie conic const from -1 to 0)
  You could assess the performance - we know it will not be be very good.

2/ Now change both the conic constant and the distance to prime focus to
variables.  Leave the R of the mirror as it is. You may want to only
consider on
    axis rays since the optimiser may give slightly different results over a
wide field so only ose field rays of 0 deg.
   Using the optimiser. solve for the best focus distance and conic const.
Trivial and pointless maybe.  But since we know what the answer should be,
it is a good
   check that all is well.  And if it works, then the method can be built on
with confidence.

3/ Now lets turn the Newt into a Cassegrainian scope.   Firstly with a
spherical secondary.
    Place a mirror 1000 mm from the primary.  Set the distance from
secondary to image as 1500 mm.  (Most call this the back focal length of the
system - but
    most ATM's would call the 500 mm  back beyond the primary the Back Focal
Length.  A real point of confusion when talking about this)
    Remove all variables, then make the Radius of curvature of the sec a
variable.  Leave its surface a sphere.
    Optimise, to find the correct Radius of Curvature of the secondary.
Evaluate the design - its not very good - as expected.

4/  Lets turn this into a cass with Parabolic primary but conic secondaqry.
     Remove all variables.
     Create variables on conic const of the sec and its radius of curvature.
     Optimise.  This now tells us the shape of the sec and its curvature
fior these distances.
     At this point it may be worth introducing a field ray at 0.1 degrees
from axis.    The optimisation will be more relavant to the real world.  Now
optimise again
     and evaluate.

5/  Lets turn this into a Dall Kirkham cass - with spherical sec.
    Remove all variables.
    Make the sec conic const = 0.
    Create variables on radius of curvature of the sec and the conic const
of the primary.
    Optimise and see what eventuates.  Evaluation compares poorly with the
classical cass off axis.

6/  Of course, the RC cass is created in a similar way but the primary ,
secondary, and radius of curvature of the sec are all made variables.
     Optimisation will now create a better performer over the whole field.
Much beloved by designers but a real bitch to make.

NOTE.  Depending on your program, it may be necessary to create a new
optimisation function before each step.  Some programs are happy with minor
changes without doing this but for any major change, even a change in field
angle, it becomes necessary for best - or evrn sensible results.

Of course, most of these designs are not very sensible.  We did not bother
tailering the specs to what the customer wanted.  But if you have come this
far, you now should trust the optimiser and can investigate all sorts of
possibilities.

I am not sure whether these files should be made available in OSLO format.
The whole exercise is in creating them.

It is remarkable what we have achieved without any maths.  We did need to
know roughly where the cass sec should be - but apart from this it was all
done by the program ans some common sense.

Finally, if you have not already done so, create more and more variables in
a design and optimise and watch the design run amok.  Mathematically, it is
making sense but its not what is needed in the real world.

You have to judiciously use certain surfaces and spacinga as anchor points
when optimising to keep the design in check in conjunction with targeted
variables.


Have fun.

Peter Smith.