hello Matteo
both schemes are OK, they only differ for tracking performance
assuming :
- plant transfer is P
- controller transfer is C
- your input is t (before the summing point) and the output is y
for controller in the main path : tracking performance = y/t = C*P / (1 + C*P) , so you see the controller transfer is also in the numerator of the closed loop
for controller in the feedback path : tracking performance = y/t = P / (1 + C*P) , so you see the controller transfer is only in the denominator of the closed loop
for output perturbation rejection (that is assume y is corrupted by perturbation p ) , both schemes give exactly the same performance :
y/p = 1 / (1 + C*P)
one typical example of feedback is PID; usually we put the P and I action in the main path , but it is not recommended to put the D term here because we want to avoid to derive the input data, so the D term is located only in the feedback path - so damping applies only on the measured output and not on the input.
hope it helps !
