The sensors use piezoresistivity to measure stress.
By applying an external force to the sensor, the conductive material on the substrate contacts the active area, exhibiting the force versus resistance characteristics as shown:
Each Force-Sensing Resistor will have a point when increasing pressure will not affect nor change the resistance. This point is referred to as the saturation point.
Once you have that resistance measurement, you can determine velocity by comparing values over time. So if a pad is at resting values, nothings going on. If it goes over the initial threshold, the software decides what to do. If "full level" mode is on, then you immediately trigger a MIDI event with 127 velocity. If it's off, you take samples over time, and as soon as the change in value goes below some threshold you grab that value as the hit value, and divide it by the range of potential resistivity between the lowest value and hitting the saturation point to get the percentage of max, then multiply that fraction by 127. Then you trigger a MIDI event at that velocity. So if the max resistance range value were 1.6volts, and they hit with 1.5volts peak resistance, 1.5/1.6=94% of 127, so a MIDI velocity of 119.
That's just my inexperienced guess based on what you see above from their chart. But you can see why the pad gap matters then, because it takes force to bend the rubber and push the pad down to the sensor, but that does absolutely nothing until it hits! The pads themselves are dead. You are merely using the pad as an intermediary to bend the sensor beneath it. That's why if you apply a small enough force to the rubber pad that bends it and lowers it, but doesn't bend the sensor square below it enough for the software to register it as a hit, then you get nothing. It's numb until you've overcome the mechanical resistive force of the rubbery material, plus the force needed to stress the sensor below. That's why it feels so unresponsive, exactly like trying to feel with thick gloves on. When you take them off, the stress doesn't need to first bend a middle material, it's skin on skin.
Akai heard all the complaints about the 500s sensitivity when it was first out, and they added a "fix" in one of the OS updates. Not sure what they did exactly, but it felt maybe quicker to trigger the full level while sacrificing range expression. Try this, set pad sensitivity to max value of 16, tap in waves of soft to hard, and compare all the curve settings. Notice anything? I didn't really. The thing felt so numb the curves didn't even matter much.
Seems like what the software should have is a calibration button on the sensitivity page where you take a sample not touching, then hit "do it", then it says to hit it soft, then hit it medium, then max. Then it uses those detected values to make your custom curve and you can choose to need to bang it to the saturation point if you want for max level, giving more range of expression, or just use the bottom end so you have a small range of expression, but it's easier to trigger high-velocities.
But they aren't even admitting there's bugs let alone adding new stuff to make the machine work better. It seems like all their focus has been tied up with the 5000. The 2500/1000 people are being taken care of by JJ, and the 5000 is their top priority. The 500 is the runt left out in the cold begging for food scraps leftover from older sisters. I'd buy the 5000 if I thought it was any good, but the 500 is actually a better idea to me. I'd prefer a 500 over 5000 hardware-wise. Software is another story.
That's why I emailed Akai first saying it'd be a smart business move to open source the OS, then when they didn't respond I offered to purchase access to it. They lack vision.
not as weird as I was thinking they would be though.
and crying... Rolling and 
