Hello,
well I have to admit that those level were just a guess. I have now made an accurate measurement of the real volume levels by playing back white noise through the headset and recording with a microphone at the same time, while increasing the hardware volume level with 'amixer'. Then i have measured the volumes with audacity's contrast-analyzer and smoothed the resulting curve by applying a nonlinear least squares fit with a function a*log(x+d*x^2+e*x^3)+b+c*x, wich fits the data really well.
The resulting levels are:
0 invalid/unconclusive, sometimes mute, sometimes louder than 1. seems not to be working correctly.
1 -24,91 2 -22,31 3 -20,10 4 -18,15 5 -16,46 6 -15,00 7 -13,73 8 -12,61 9 -11,62 10 -10,72 11 -9,91 12 -9,17 13 -8,49 14 -7,87 15 -7,28 16 -6,74 17 -6,23 18 -5,75 19 -5,30 20 -4,88 21 -4,48 22 -4,09 23 -3,73 24 -3,38 25 -3,05 26 -2,74 27 -2,44 28 -2,15 29 -1,87 30 -1,61 31 -1,35 32 -1,10 33 -0,87 34 -0,64 35 -0,42 36 -0,21 37 0,00
Alex