Quick note on the motor power curve from previoys Roadster 2 discussions filtered through my 2 in the morning brain.
The power band graph is instantaneous (or at least short term) power, so thermal limiting shouldn't be in play. Think single launch vs track day.
Horsepower is torque*RPM/5252. Motor RPM generates back EMF voltage which reduces the voltage across the motor windings and thus current. Torque is proportional to current.
Assuming a standard buck (step down) type motor drive, you're limited to pack voltage. The inverter also has a current limit on its output. The pack itself has a power limit.
The first part of the power curve is inverter current limited, as RPM increases, power increases linearly. Then we hit the battery power limit. At some point, the inverter is at 100% and is putting out pack voltage. After that point, as back-EMF continues increasing power rolls off as current drops.
With the trimotor set up, each rear motor only needs to put out half the power versus driving two tires. So they can change the internal construction to reduce the amount of backEMF. This also reduces the torque per Amp, but that needed reduced anyway due to only having half the traction. They can also shift the gearing to make the motor spin faster and reduce the windings further.
End result, the backEMF per RPM value is much lower, so less roll off on the high end.
Looking at the graph:
Note how P100D, performance, and plaid all have the same inital slope, that's either the inverter current limit (in plaid, it's the sum of the three motors) or a software limit due to drivetrain torque limits.
Performance and P100D then hit power limit followed by a backEMF drop.
P85 has a lower max current and peak power due in part to the smaller pack.
The Plaid curve shows a higher max power which indicates pack improvements allowing higher pack current. Then follows a really shallow slope indicating either a very low Volt/RPM figure or possibly, V/RPM is low enough that it never hits voltage saturation of the inverter. In that case this may be loss of power due to the increased number of switching events of shorter duration vs motor inductance which could reduce the effective power output.
View attachment 672059