Inverse Frequency Response
Identify filter parameters from frequency response data.
invfreq(
h,
w,
nb,
na,
wt = rep(1, length(w)),
plane = c("z", "s"),
method = c("ols", "tls", "qr"),
norm = TRUE
)
invfreqs(
h,
w,
nb,
na,
wt = rep(1, length(w)),
method = c("ols", "tls", "qr"),
norm = TRUE
)
invfreqz(
h,
w,
nb,
na,
wt = rep(1, length(w)),
method = c("ols", "tls", "qr"),
norm = TRUE
)h |
Frequency response, specified as a vector |
w |
Angular frequencies at which |
nb, na |
Desired order of the numerator and denominator polynomials, specified as positive integers. |
wt |
Weighting factors, specified as a vector of the same length as
|
plane |
|
method |
minimization method used to solve the normal equations, one of:
|
norm |
logical indicating whether frequencies must be normalized to avoid matrices with rank deficiency. Default: TRUE |
Given a desired (one-sided, complex) spectrum h(w) at equally spaced
angular frequencies w = (2 π k) / N, k = 0, ... N-1, this function
finds the filter B(z)/A(z) or B(s)/A(s) with nb zeroes
and na poles. Optionally, the fit-errors can be weighted with respect
to frequency according to the weights wt.
A list of class 'Arma' with the following list elements:
moving average (MA) polynomial coefficients
autoregressive (AR) polynomial coefficients
Julius O. Smith III, Rolf Schirmacher, Andrew Fitting, Pascal
Dupuis.
Conversion to R by Geert van Boxtel,
G.J.M.vanBoxtel@gmail.com.
order <- 6 # order of test filter
fc <- 1/2 # sampling rate / 4
n <- 128 # frequency grid size
ba <- butter(order, fc)
hw <- freqz(ba, n)
BA = invfreq(hw$h, hw$w, order, order)
HW = freqz(BA, n)
plot(hw$w, abs(hw$h), type = "l", xlab = "Frequency (rad/sample)",
ylab = "Magnitude")
lines(HW$w, abs(HW$h), col = "red")
legend("topright", legend = c("Original", "Measured"), lty = 1, col = 1:2)
err <- norm(hw$h - HW$h, type = "2")
title(paste('L2 norm of frequency response error =', err))Please choose more modern alternatives, such as Google Chrome or Mozilla Firefox.