Re: [stella] Real 3D Space Calculations

Subject: Re: [stella] Real 3D Space Calculations
From: "Roger Williams" <localroger@xxxxxxxxxxx>
Date: Wed, 29 Aug 2001 18:41:36 -0500
Polar- (or Spherical) coordinates seem to be one of the keys to success.

Absolutely. In order to draw the screen, you must convert everything in the game universe into ship-centered spherical coordinates. This will involve three angles, usually labelled rho, theta, and phi, as well as a radius. (The third angle is the object's rotation, which is trivial in cartesian co-ords but not in polar.) You can then translate to ship-VIEWPOINT spherical co-ordinates by doing simple addition to rho and theta. You can then locate the object on the screen by treating rho and the arctangent of rho * the radius as polar co-ords and converting back into cartesian in the screen space. Phi directly tells you the object's rotation as it needs to be plotted, and the polar radius its size. In this co-ordinate system you can also move and rotate the ship very easily before converting back to spherical.

The more I'm reading about it, the more I think it's doable on the VCS.

The main problem is that all the values are going to have to be 16-bit to get the motion smooth enough, and you will need at least 4 16-bit random multiplications to do each transform in both directions. That is a lot of uncertain cycle-count code. You might be able to distribute the math across multiple vertical blanks, though.

I have done and tested some 8086 routines that do fast integer trig
and polar-to-cart conversions, and I have an old BASIC program that
uses them to simulate a roulette wheel (in real time, on a 386DX20,
at 640x480).  They use MUL and 16-bit regs but illustrate the
principles; if you want I'll send them along.

--Roger Williams

Get your FREE download of MSN Explorer at

- Archives (includes files) at Unsub & more at

Current Thread