The Hammerstein model is used as the representation of
the PA nonlinear behaviors, with a memoryless distortion
block followed by a linear subsystem h(n) which represents
the memory effect. The associated predistorder is the Wiener
model formed by a linear subsystem followed by a predistortion
memoryless system.
For the PA memoryless distortion, the predistortion can be
implemented using a general power series or with a look-up
table (LUT). For the polynomial model, the principle relies on
the identification of a polynomial function which represents
the inverse characteristic of the PA. The LUT predistorder
consists in addressing a table which contains a quantified
version of this inverse characteristic. The first solution suffers
from curve fitting limitations whereas the second suffers
from quantization aspects [16]. In both cases, as for memory
polynomial model, the LMS algorithm is used. Concerning
the nonlinear subsystem characterized by its impulse response,
[17] proposes a frequency domain method for the identification
of the Hammerstein