Back to Index
|
BENCHMARKING
POVRAY AGAINST ZEMAX
POVRAY is clearly capable of forming
real images
inside
photorealistic representations of the optics creating the
images.
Although as observed in the index page it was originally conceived as
an artistic tool, it is
worthwhile testing the accuracy of the POVRAY images using ZEMAX as a
calibration.
This first test-scene is a darkroom with a lightpipe experiment
as a test object. The lightpipe is formed by just two optical
components – a glass cylinder merged with a spherical end
forming a
lens. Light from an LED is collected over a 90deg cone angle
by
the cylinder and is guided by direct raypaths and total internal
reflection TIR to the lens. The glass material of the unit is
Schott SF6. This glass was selected, using ZEMAX as a design
tool, to have the sharpest image from a 10mm diameter 60mm long
cylinder and 10mm diameter lens at an image distance of
1.0mm.
The lens
focusses the light to a diffuse 10mm x 10mm square diffuse white
screen. The screen can also be contrived to be an image
source
glowing with the site logo.
The
assembly is mounted on a steel topped scraped texture laboratory
optical
bench. The table is drilled with countersunk holes for fixing
components. The scene is lit by a single overhead point light
source mounted
from a 30% reflective ceiling. The
camera is located
about
100mm from the pipe and has a field of view 35degrees wide.
In this first image the LED is off - the scene is lit only
by the
overhead light.
Notice how a virtual image of the web-site logo on the screen is
visible through the end of the cylinder, and the optical bench is
visible through the wall of the cylinder. But
notice also
the shadow under the cylinder. This is counter-intuitive
– any
optical
engineer would expect to see a strip patch of focussed light under the
cylinder. This was the type of image that was created by
early
versions of POVRAY - it was the introduction of 'photon
mapping'
by Nathan Kopp following its creation by Henrik Wann Jensen that
transformed the artist's toolbox into a serious
optical analysis tool.
In this image photon mapping has been applied - the
image
is now much
more realistic in that one sees that the cylinder and sphere have
formed a strip image of the overhead light source.
The
image is irregular - because the steel table is not uniformly diffuse
spatially.
The scene is rather stark because of the single lightsource - by
mounting an array of luminaires
on the ceiling the
lighting becomes much more diffuse - stark shadows are softened and
highlights appear in the countersunk holes in the bench.
Steel
tables are
rather dreary - so replace it with a shiny checkered
reflective
plastic.
The darkroom lights are then turned off and the LED
turned
on.
POVRAY was then used to calculate the image profiles on the
display screen.
POVRAY
creates a *.tga file representing the image brightness as a solid
object. Because the image is rotationally symettric it is
convenient to display the profiles through a centrally cut section. In the image
above the cross
section
has been textured with the 'jet' palette to indicate the height
This is the ZEMAX style image
in which
the ray intersections in the image plane are counted and displayed as a
graph.
These images are
clearly of
a very similar structure, so we can be confident that POVRAY is
yielding reliable results.
We now extend
the lightpipe
application to the much more complex task of a bent
lightpipe.
The lightpipe is now arranged to relay the LED through to a
screen 100mm diagonally displaced. The bend is achieved by
with a
quarter torus and two cylinders arranged at 90deg. But it's
not
known
what should be the radius of the torus to achieve the maximum
brightness on the screen. Nor is it known what will be the position of
the brightest image.
By animating the
bend radius
and observing the screen image we can answer both of these questions.
 |
The time
required to create the 32 frames in this animation was greatly reduced
by omitting the holes in the optical bench and enclosing the
lightpipe in an opaque plastic cladding.
To
view the
0.64MB animation:
pipebend1.m1v
|
 |
The image to the left is a
textured
heightfield
of the intensity at the screen for the value of the bend
radius
which maximises the intensity. The peak intensity is 0.8mm from the
axis. The values were determined from the
following 0.64MB movie:
pipebend3.m1v |
|
