Ring modulation multiplies two signals sample by sample. Feed it a 440 Hz tone and a 100 Hz carrier and neither survives: the output holds 340 Hz and 540 Hz instead.
Frequency
The carrier runs from 5 Hz to 4 kHz on a log scale, and the sound changes character completely across that range.
Below about 20 Hz the sidebands are close enough together that the ear hears a tremolo instead of two tones, though a hard one, since the carrier passes through zero on every cycle.
Between 30 and 200 Hz sits the classic metallic voice. Above 500 Hz the sidebands move far apart and the source stops being recognisable, which is where the effect is most useful for sound design.
Shape
Sine gives the cleanest result: one carrier frequency, so each input partial produces exactly two output partials.
Triangle and square add carrier harmonics, and each of those produces its own pair of sidebands. The output gets denser and harsher. The square shape is generated as a saturated sine rather than a true step, so a slow carrier changes smoothly instead of clicking on every edge.
Stereo phase
Stereo phase offsets the carrier between channels by up to 180 degrees. Since the carrier is bipolar, an offset of 180 inverts one channel against the other on every cycle, which widens the result and thins it on a mono sum.
Mix
Mix decides how much of the original survives. At 100% only the modulated signal is heard.
Speech and vocals usually need some dry signal to stay intelligible. Around 60 to 70% keeps enough of the original formants to follow the words while the metallic layer sits over them.
Example: a machine voice
Frequency 55 Hz, sine, stereo phase 0, mix 100%.
The dialogue keeps its rhythm and loses its pitch, which is the effect used for robotic voices since the 1960s.