LP filtering


Thread Starter


I am just wondering is it true that PID fails control when:
1. Noisy signal comes
2. Long nonlinearity (dead time) within the process

Ref 1. is solved usually by LP filtering. Is it really good way to do so?

Firstly filter makes some phase shift, secondly we cannot differenciate between PV raise and signal hardware failure - both have the same raise time.
And this is why we don't use D term in PID with noise.

Would it be better to do this way:
de = e1 - e2
if (de(normalized) > (system raise time)*dt)) then
de = de (old)

as a kind of acceleration filter?
Do I do domething wrong?