A matrix where each column contains a transfer function - tf()$H would be appropriate.
n
An integer indicating the half-length of the impulse response to return, L = 2*n+1.
Normally this would be the blockSize2 agrument used in tf().
realPart
A logical indicating whether to return only the real part of the impulse response
(default = TRUE) or return the complex-valued impulse response (FALSE)