Sep 4, 2021

RGB Sphere (Epcot Sphere / Geodesic Polyhedron)

The RGB Sphere is an advanced version of a routine I programmed for the Amiga, released in the intro often called First by Reflect early 1992. Still keeping with the idea of recreating "oldskool demo style" real-time 3D graphics with Python ... see my bigger project Sound Vision for lots more. 

All the code can be downloaded from github.

The RGB Sphere is a morphing vector object for testing three color light sourcing. It starts from a tetrahedron, or, for better results, an icosahedron, the faces of which are equilateral triangles. All the corners (ends of edges) lay on a surface of a single sphere. Then this object can be morphed closer to a full sphere by replacing each triangle with four triangles of equal shape and size, and projecting the new corners (ends of edges), which now are inside the sphere, to the sphere surface. Repeating this a few times results in a 3D vector object looking like the Epcot Center at Disney World. All these objects are also Geodesic Polyhedra.

Three's Company


There are three modes available, and moving between the modes is controlled by the left and right cursor keys. The initial mode is wireframe, where the back side of the object can be seen through the front.


The second mode is a light sourced version of the same, where the color of each triangle depends on its angle to either one white light source or three separate light sources, which are red, green and blue (additive colors...). The light source mode can be switched with the l key. The light sources may also move about, and the speed can be selected with m and n keys. Note that unless they move, they all start at the same spot, so three light sources look the same as one.


The third mode is the same as the second, but instead of each triangle being white to begin with, they can be colored based on a image. I used a world map, but any image can be used - however projecting it to a sphere and having a limited number of triangles will result in a less accurate representation of it ... I have another project for that under way.



The changes between the modes and adding new triangles ("depth") are made in a smooth way. When new triangles are added, they are first added "in place" so that the new corners appear in the middle of the old ones, and only then are they slowly pushed to the sphere surface. This increasing or decreasing of depth is controlled by the up and down cursor keys. You can see the different modes in this video capture:



Preparations Necessary


To ensure smooth transitions, all the object data are calculated and stored before showing any 3D objects. Building new levels from the original tetrahedron or icosahedron is fully algorithmic, so even more depth could be added, but now getting six levels results in 10,242 nodes (or corners or points to rotate), 35,514 edges to connect these, and 20,480 surfaces. Unfortunately, there is persistent bug in the code when trying to figure out which edges are at the front using depth 6 in wireframe mode.

The original Amiga OCS assembly code had only 62 surfaces - 31 independent colors was the maximum one could have, and it ran in 208 x 208 pixels @ 50Hz. 


The original Amiga RGB Sphere.