PDA

View Full Version : Ray Tracing software



bananas
09-27-2011, 01:48 PM
I've tried OpticalRayTracer 3.2 which is Java and very simple. It also doesn't appear to do mirrors. I have a trial of Beam 4 which seems adequately powerful but I am still learning the interface. Anyone have a preference as to what you use for ray tracing?

wledzian
09-27-2011, 06:19 PM
Have you tried POVRay?

bananas
09-27-2011, 07:01 PM
I'm studying their documentation now. There's a lot of it.

bananas
09-28-2011, 01:14 AM
I have tried to make a correct section of mirror using a rectangle which intersects a hollow sphere in the correct way to produce a mirror of about 40* vertical and 180* horizontal. If I use the intersection command it creates not just the intersected outside part of the shell, but adds surfaces all around it to correspond with the periphery of the box as it is inside the sphere, so the concave area of interest is hidden. It seemed likely that using a hollow sphere + solid rect would create the appropriate intersection but apparently not. Obviously it is possible to create the proper surfaces, maybe using an isosurface? I have found surprisingly little on google about making convex surfaces in povray so far, maybe I just haven't been looking hard enough.

EDIT: I've decided to manually do the layout and ray trace in sketchup. This way is helping me learn the physics much more easily. What ratio between mirror and screen did you use on your build, wledzian? I am using 2:1 in my design now but I have just started learning the physics involved so I am not sure if this is correct.

Further, if I understand properly, the screen should be located approximate 1/2 radius distance from the mirror in order to best approximate a collimated image to the viewer, is that right?

wledzian
09-28-2011, 12:35 PM
the sphere command is a true sphere primitive, not just a hollow shell. A simple intersection will do what you see.

If you want to try a wysiwig interface for it, look for 'Moray'.

For an on-axis viewing position, the screen is at ~1/2 R, but a collimating mirror setup is not on-axis. Read some of Mike Powell's posts in the original collimated display thread. To determine the real 'correct' screen position, you'll need to pick an eyepoint and do some point-by-point calculations. I've written an Excel spreadsheet to do this for me, but it's not user-friendly enough to release.

If you plan on using ray-tracing to test a design, keep in mind that a pinhole camera has an infininte depth of field. It will show you image distortion, but won't really show you the focal depth effects. You can include a non-zero aperture in the render which will show blur, but won't show apparent visual depth differences.

bananas
09-28-2011, 12:44 PM
Thanks for taking the time to reply. As you can tell, I have just started learning about most of the physics behind the project. I will go through that thread again and see how many of my questions it answers.

bananas
09-28-2011, 07:38 PM
Do I have the right idea about "sets of parallel lines?" Here's a ray trace I produced in sketchup using a small circle to indicate the viewpoint with all rays tangent to it for equally spaced lines. Red dots indicate where the rays converge.5630

wledzian
10-01-2011, 01:05 AM
Yes, that's exactly the right idea. Notice how the lowest eyeline defines the bottom of the screen and the bottom of the screen limits the highest eyeline. This is one of the big technical limitations in vertical FOV, and constitutes one of your main design decisions.

bananas
10-01-2011, 02:53 PM
I understand. Now I am working on designing the frame for 90* horizontal of 4ft.r. Since I do not have a CNC mill I will have to accept a larger margin of error on the components. I haven't yet calculated what range that will lie in but it may or may not be acceptable. Certainly it will take a lot of time to produce the master. I think if extreme care is taken this method could keep distortion within acceptable limits (whatever those are). I have a variety of tools to get the job done. After finishing the design the plan is to output and tile the template from sketchup.

wledzian
10-01-2011, 03:09 PM
In all honesty, the only absolutely critical dimensions are the top and bottom radii and the separation between the two. The ShopBot is absolutely wonderful in that it saves us the trouble of drawing large arcs with beam compasses and allows us to use tabs and pockets for alignment. There's nothing inherent in the design that would stop you from building it with 'normal' shop tools.

bananas
10-03-2011, 02:43 PM
I made another ray trace using 3 parallel beams for each set instead of 2. This gives me 3 intersections per set of parallel rays. I drew arcs across the outermost sets of intersections.

5635