According to the internet, cold air must be blown upwards to displace hot air. The problem is the nozzle's limited range only allows it to blow down. Another problem is the air inlet is on the bottom of the window. The height of the effluxing hot air will have to be studied.
Getting the nozzle to point up requires rotating the nozzle's inner ring more than 1 revolution. It can't be tracked with a boundary sensor.
The leading idea is now a 2 step manual reset process. 1st, the outer 2 rings are reset automatically. Then the user uses the remote control to align a fiducial on the inner ring. The inner ring just has to get within a 180 deg range. Then the program resets the inner ring as it did before.
Another theory has an accelerometer on the middle ring giving extra information about the inner ring's position. This would make the reset automatic. This involves a lot more wiring. When the motors switch to quadrature encoders, it might be lighter to put a control board & H bridge on every ring, in which case 1 accelerometer makes sense.
Another theory says accelerometers can replace all the encoders. The accelerometers could eliminate all reset procedures.
Whether or not accelerometers are used, there's still a case where more than 1 revolution of the inner ring could lead to the same accelerometer position. It would need either a deadband or a manual reset procedure. The easiest next step is just a manual reset procedure.
Quadrature encoding may still never happen. Binary encoding with debouncing has been good enough.
The current nozzle has 6 wires for each motor. 2 motor power, 2 sensor power, & 2 sensor outputs. For quadrature encoding, it's finally practical to go with a multiplexing board for each motor with just 4 wires: 1 GND, 1 logic power, 1 motor power & 1 data. That would multiplex 1 boundary & 2 rotation sensors. A way to limit current in the motor could get it to 3 wires.
As problematic as the nozzle is, never forget the 20 years of torture which the nozzle fixed.
The decision was made to eliminate all limits on the nozzle angle. The user has to make sure the motors are all facing down before commanding a reset. It's a half manual reset. There's always been an abort functionality during reset. Press the reset button during the reset procedure to stop the motors.
The kinematic solver is brutal. There are 2 sets of limits. 1 set of limits is for the lookup table which translates nozzle pitch to motor positions. Another set of limits is for the raw encoder values & is tied to the 1st set of limits. A 3rd set of limits would be required to limit the nozzle angle to a range beyond a single revolution. The nozzle pitch should be independent of the nozzle angle, but it was written before the kinematics were fully solved.
The user has to make sure the cables don't run out of room. Since most usage is loading a preset & tweeking 1 or 2 stops, it's not normally a problem.