KDER – Derivation and filtering of the input signal

Block SymbolLicensing group: ADVANCED

Function Description
The KDER block is a Kalman-type filter of the norder-th order aimed at estimation of derivatives of locally polynomial signals corrupted by noise. The order of derivatives ranges from $0$ to $\mathtt{\text{norder}}-1$. The block can be used for derivation of almost arbitrary input signal $\mathtt{\text{u}}=u_0\left(t\right)+v\left(t\right)$, assuming that the frequency spectrums of the signal and noise differ.

The block is configured by only two parameters pbeta and norder. The pbeta parameter depends on the sampling period $T_S$, frequency properties of the input signal u and also the noise to signal ratio. An approximate formula $\mathtt{\text{pbeta}}\approx T_S{\omega }_0$ can be used. The frequency spectrum of the input signal u should be located deep down below the cutoff frequency ${\omega }_0$. But at the same time, the frequency spectrum of the noise should be as far away from the cutoff frequency ${\omega }_0$ as possible. The cutoff frequency ${\omega }_0$ and thus also the pbeta parameter must be lowered for strengthening the noise rejection.

The other parameter norder must be chosen with respect to the order of the estimated derivations. In most cases the 2nd or 3rd order filter is sufficient. Higher orders of the filter produce better derivation estimates for non-polynomial signals at the cost of slower tracking and higher computational cost.

2020 © REX Controls s.r.o., www.rexygen.com