As I understand it, the original intent was for the pair of intervals to establish all possible pitch values. Starting from 0V you go up the 1st interval, then the 2nd, then the 1st again, etc… Going down from 0V I think you first go down by the 2nd interval, then the 1st, then the 2nd, etc., though I am not positive. This would establish all possible pitches, and then for any given input, the quantizer would pick the “legal” pitch closest to the input. So the quantizer would be absolute - it would always pick the same pitch for any given input.
But the developer struggled with implementation, and ended up doing something different. The set of available notes is established based on the most recent quantized value. In other words, the last quantized pitch becomes the new 0 point. So every time you quantize a note, it is relative to the previous quantized value, and a given input may produce different notes depending on what was played previously.
For example, suppose the quantizer is set to a perfect fifth for interval 1 and a perfect fourth for interval 2. Starting from 0V you feed it value y that quantizes to the fifth plus the fourth.
But suppose instead you first feed an intermediate value of x that quantizes to the fifth, and then you feed value y. The quantizer now starts from the fifth, and then quantizes up another fifth (the 1st interval). So in this case value y yields up two fifths from 0 instead of a fifth + fourth.
The end result is some interesting key changes as the inputs jump around.
I am actually working on an a similar idea of quantizing based on intervals, but am using absolute quantization instead of relative. In addition, my quantizer will support up to 10 independent step intervals, and the minimum interval step size can be any equal division of any pseudo octave. I think it will be extremely flexible. capable of producing both beautiful and horrible results. I think anyone interested in micro tuning and/or non-octave scales will be interested.