Shared by every Welte system rather than declared per system, and that is a
decision about comparability rather than about accuracy. The point of running
a red issue and a green issue of one recording through two emulators is to
attribute the difference between them to the coding. If the two systems used
two velocity maps, the difference between the maps would dominate and the
comparison would measure the maps.
Nothing in either mechanism determines the map. These three anchors are
midi2exp's published velocities, joined linearly with the middle one at the
Mezzoforte pin of the half in question, and they are what the T-100's
published results were produced with. Two independent implementations
disagree with them and may well be right: Stahnke gives v_h = K_s √(w − W_s)
for vacuum to hammer velocity and V_MIDI = round(52 + 25 log₂ v_h) for
velocity to MIDI, and Phillips and Gosden's Rollmidi, in the same volume,
uses an explicitly logarithmic curve; Phillips's own bench measurement has the
suction-to-velocity relation "essentially logarithmic" (pp. 215 f.). The
default should change only once a red/green comparison has been run both ways
and the difference measured, and then for both systems at once.
How bellows travel becomes a MIDI velocity.
Shared by every Welte system rather than declared per system, and that is a decision about comparability rather than about accuracy. The point of running a red issue and a green issue of one recording through two emulators is to attribute the difference between them to the coding. If the two systems used two velocity maps, the difference between the maps would dominate and the comparison would measure the maps.
Nothing in either mechanism determines the map. These three anchors are midi2exp's published velocities, joined linearly with the middle one at the Mezzoforte pin of the half in question, and they are what the T-100's published results were produced with. Two independent implementations disagree with them and may well be right: Stahnke gives
v_h = K_s √(w − W_s)for vacuum to hammer velocity andV_MIDI = round(52 + 25 log₂ v_h)for velocity to MIDI, and Phillips and Gosden's Rollmidi, in the same volume, uses an explicitly logarithmic curve; Phillips's own bench measurement has the suction-to-velocity relation "essentially logarithmic" (pp. 215 f.). The default should change only once a red/green comparison has been run both ways and the difference measured, and then for both systems at once.