A while back, I made a song with this effect using single-cycle waveforms, and seem to recall that a harmonically rich waveform is much more likely to produce a convincing result.
Probably has to do with the psychoacoustic effect where each tone (there are always multiple tones) has to cover a frequency band wide enough to “cover up” the effect
The theory behind a risset/shepard tone is actually quite simple. Consider the following
/ / / / / / / / / / / / /
/ / / / / / / / / / / / /
/ / / / / / / / / / / / /
/ / / / / / / / / / / / /
This graphical pattern would give the impression of a steep angle. If we converted it into sound (going from left to right), we would actually hear an endless rising tone.
The trick is to find a sound that has the right harmonic content (make the stripes “fatter”, so to speak). Or alternatively, add more stripes.
///////////////////////
///////////////////////
///////////////////////
///////////////////////
Also, it’s important to balance the amplitude (fade sound in from the bottom, fade it out is it approaches the top), so your ears does not pick out any specific frequency at any time.
Another aspect is the tempo, how often the signal will repeat itself, If we study the previous text pattern closer, we’ll see that it repeat itself for every 4 characters
/| / /| / /| / /| / /| / /| /
/ |/ / |/ / |/ / |/ / |/ / |/
/ /| / /| / /| / /| / /| / /| /
/ |/ / |/ / |/ / |/ / |/ / |/
When time is mapped to an axis along with frequency, you could say that the tempo corresponds to the “angle” of the stripes.
It’s not possible for me to visualize a less steep angle with just text, but imagine that those stripes were repeating for every 16th character, or going down instead of up?
Once the theory is established, you’ll find that it’s possible to create endless chord sequences based on these phenomena.