We are building a 1 meter humanoid with MX-106. We have tried tuning the PID in Protocol 1 along with punch, as well as in protocol 2.0 . We are struggling with removing around 1-2 degrees of movement around the goal position even with only 5-10% torques. We have seen better control on cheap chinese servos since they are hard coded for maintaining position and not for compliance. Could you please help us with our current situation as we feel we are probably not using the features on your motors correctly.
In order to help you, please specify your current setting(such as operating mode, protocol version, PWM/Current Limits, PID gains) of MX-106
and how much of actual torque is applied when operating the joint.
2018-01-10 20:09:10 Aadarsh MehndirattaFirmware Version 41, Model number 321, Joint Mode, Position Control, Protocol 2.0, Current Limit 2047(default), PWM Limit 885(default), actual torque applied ~ 10kgcm Deflection observed in present position ~ +-(3) degrees We have tried ranging the P values throughout the range in r+ manager 2.0, This only reduces the deflection to 1.5-2 degrees. Changing I and D values manually have shown no improvement. We would like to know an approximate value for PID that we should test it on for such conditions.
The mechanical backlash of the normal MX-106 is 20 arcmin(about 0.33degree).
If reduction gears are worn out, the gap could get bigger.
In order to acquire better position traction near the goal position, please try bigger Position P gain(such as 1,600) and I gain(such as 800).
However, these values should be adjusted based on your system as lager values could cause unwanted vibration.