ArduCopter-Overly sensitive swashplate response to cyclic inputs for Dual heli frame

[bnsgeyer]

Issue details

A dual heli frame was added to Copter 3.5 to support tandem rotor helicopters. It was noted in @frizensami post (https://discuss.ardupilot.org/t/tandem-helicopter-pid-tuning/17849) that tuning parameters for PID in pitch and roll seemed very low (1 order of magnitude smaller) when compared to single main rotor /tail rotor configurations. Recent, @PittRBM posted (https://discuss.ardupilot.org/t/dual-heli-frame-frame-class-11-problem/20972) an issue when setting up his transverse configuration using the 3.5 version dual heli frame that the swashplate was overly sensitive using the tuning parameters that he used in tuning his personal version of a copter 3.4 dual heli frame derived from the same user @fhedberg.
In comparing the motorsheli_single.cpp and the motorsheli_dual.cpp, I noticed an inconsistency in the code between the two files that I believe is causing this increased sensitivity.
In these lines of code there is a term that is divided by 0.45

ardupilot/libraries/AP_Motors/AP_MotorsHeli_Dual.cpp

Lines 530 to 535 in 528d4d8

	float servo1_out = (_rollFactor[CH_1] * roll_out + _pitchFactor[CH_1] * pitch_out + _yawFactor[CH_1] * yaw_out)/0.45f + _collectiveFactor[CH_1] * collective_out_scaled;
	float servo2_out = (_rollFactor[CH_2] * roll_out + _pitchFactor[CH_2] * pitch_out + _yawFactor[CH_2] * yaw_out)/0.45f + _collectiveFactor[CH_2] * collective_out_scaled;
	float servo3_out = (_rollFactor[CH_3] * roll_out + _pitchFactor[CH_3] * pitch_out + _yawFactor[CH_3] * yaw_out)/0.45f + _collectiveFactor[CH_3] * collective_out_scaled;
	float servo4_out = (_rollFactor[CH_4] * roll_out + _pitchFactor[CH_4] * pitch_out + _yawFactor[CH_4] * yaw_out)/0.45f + _collectiveFactor[CH_4] * collective2_out_scaled;
	float servo5_out = (_rollFactor[CH_5] * roll_out + _pitchFactor[CH_5] * pitch_out + _yawFactor[CH_5] * yaw_out)/0.45f + _collectiveFactor[CH_5] * collective2_out_scaled;
	float servo6_out = (_rollFactor[CH_6] * roll_out + _pitchFactor[CH_6] * pitch_out + _yawFactor[CH_6] * yaw_out)/0.45f + _collectiveFactor[CH_6] * collective2_out_scaled;

where in the Motorsheli_Single.cpp file the same expression is multiplied by 0.45.

ardupilot/libraries/AP_Motors/AP_MotorsHeli_Single.cpp

Lines 444 to 445 in 528d4d8

	float servo1_out = ((_rollFactor[CH_1] * roll_out) + (_pitchFactor[CH_1] * pitch_out))*0.45f + _collectiveFactor[CH_1] * coll_out_scaled;
	float servo2_out = ((_rollFactor[CH_2] * roll_out) + (_pitchFactor[CH_2] * pitch_out))*0.45f + _collectiveFactor[CH_2] * coll_out_scaled;

I believe changing the expression to have it multiplied by 0.45 as is done in the heli_single frame will fix the oversensitivity of the cyclic in the heli_dual frame. By the way, I checked @PittRBM's repository for his version of Copter 3.4 with the Dual-Heli library which he is running on his synchropter and it also has the expression where it is multiplied by 0.45 like the Heli_single frame.

Version

latest Copter 3.5.X

Platform

[ ] All
[ ] AntennaTracker
[X] Copter
[ ] Plane
[ ] Rover
[ ] Submarine

Airframe type

helicoter tandem and transverse

Hardware type

Pixhawk

Logs

No logs

@

[tridge]

you're absolutely right, thanks! I've created a PR for it here:
#7026

[auturgy]

Can this be closed?

[OXINARF]

Yep, fixed by #7026.