human volume percerption roughly follows a logarithmic scale.

To compensate for this you could _try_ to use exponential steps:

(0,) 1, 3, 6, 12, 25, 50

(which follows 1.5 * 2^n with some rounding)


It will still not give the feeling of equidistant volume steps because 50% duty cycle will likely excite less overtones than f.e. 12% duty cycle (and thus will sound much softer/less alarming).

I'd start varying from the exponential steps (without really expecting to achieve consistent volume steps over a larger sample of these 2.7khz magnetic buzzers).

Greetings,
Frieder