Duty cycle vs pwm

[nilay994]

avitra_esp/test/rpm_tuning/main/motor.c

Line 20 in 488bf4f

motor->duty_cycle += motor->del_duty_cycle;

Do you really need += ? Can't the operator be only pure =?

[shanks-d]

@nilay994 No in our case we don't know the optimum value of duty_cycle that corresponds to our desired rpm. Thus we cannot use the form: motor_pwm = optimum_pwm + error.
So what we are doing that we are directly feeding the error calculated to the motors by constantly adding the errors continuously. Now if we replace += by only =, the motor will never attain desired rpm since when they are running at desired rpm the error becomes zero which is given to the motor and it'll slow down the motor speed.
So either we have to find a way to calculate the optimum value of PWM which is required for a particular rpm or we have to change the implementation of PID itself.

[sm1393]

It is not possible to calculate the optimum value of PWM. We should go for the second option.

[nilay994]

https://github.com/uzh-rpg/rpg_quadrotor_control/blob/master/documents/theory_and_math/theory_and_math.pdf, page 4

Create a linear/polynomial fit by plotting for various values of PWM and motor RPM. The slope will give you your so-called optimal PWM. The y-intercept will tell you the minimum PWM required to start the motor (similar to Integral gain term).

a simple equation should look like:

static float rpm_prev_error = 0; 
static float rpm_curr_error  = rpm_d - rpm_c;
pwm_out_to_motor = slope * rpm_d + kp * rpm_curr_error + kd * (rpm_curr_error - rpm_prev_error) * freq_hz;
rpm_prev_error = rpm_curr_error;

where rpm_c is current rpm read out from encoders in correct units, and rpm_d is the desired rpm is set to a constant value during your tests. freq_hz = 1/(time take per execution of this sequential thread).