What role does each encoder play in a dual-loop encoder axis?

A dual-loop encoder axis utilizes two encoders to control a single motor. Typically there is a rotary encoder mounted on the motor (the "minor encoder") and a linear encoder mounted on the axis driven by the motor ("the major encoder"). This arrangement is beneficial if you wish high positioning accuracy and there is backlash or slippage or other sources of position errors in the motor drive system.

In this configuration, the two encoders play the following roles:

The minor (motor) encoder produces the commutation angle. This angle is used to generate the 3 phases of the motor current output. It is not critical that this angle be accurate to more than a couple of degrees but it is vital that the relationship between the commutation angle and the motor phase not change over time in order to guarantee that the motor torque can be efficiently produced.
For the PID loop, the position and the integrator feedback loops are always closed around the major (linear) encoder. So, the final positioning accuracy of the system is a function of the resolution of the major (linear) encoder. The velocity feedback loop can be specified via the "Velocity feedback blending factor" (DataID 10305) to be generated exclusively from the minor (motor) encoder or exclusively from the major (linear) encoder or from any blended combination of these two encoders. The selection of the blending factor is application dependent and should be based upon factors such as the relative resolutions of the two encoders and which encoder provides better stability for the axis.

If the blending factor is set to 1, all of the PID loop is produced based upon the major (linear) encoder and the current loop commutation angle is solely a function of the motor encoder.