Behavioral model and predistorter for modeling and reducing nonlinear effects in power amplifiers
View Patent ↗The behavioral model and predistorter for modeling and reducing nonlinear effects in power amplifiers addresses the model size estimation problem. The GMP model is replaced by the hybrid memory polynomial/envelope memory polynomial (HMEM) model within a twin nonlinear two-box structure to reduce the number of variables involved in the model size estimation problem, without compromising model accuracy and digital predistorter performance. A sequential approach is presented to efficiently estimate the model size. Experimental validation is carried out to evaluate the performance of the size estimation and the accuracy of the HMEM-based twin-nonlinear two-box model with respect to that of the GMP-based twin-nonlinear two-box model.
1. A behavioral model of a nonlinear power amplifier, comprising:
a first module implementing a highly nonlinear static behavior of the power amplifier, the first module having an input and an output;
a second module implementing dynamic distortion behaviors of the power amplifier using a hybrid memory polynomial envelope memory polynomial function defined by a combination of a memory polynomial and an envelope memory polynomial, the second module having an input and an output, the first module being coupled to the second module; and
a behavioral model size estimation module sequentially computing an estimated behavioral model size using a formula characterized by S HMEM =(N×M)+(N E ×M E ), where N and M represent the nonlinearity order and memory depth of the memory polynomial sub-function, respectively, and N E and M E represent the nonlinearity order and memory depth of the envelope memory polynomial sub-function, respectively.
2. The behavioral model according to claim 1 , wherein the behavioral model size estimation module further comprises:
means for setting a first dynamic distortions function to a memory polynomial model;
means for sweeping N and M to evaluate the memory polynomial model size;
means for setting N=N MP and setting M=M MP ;
means for setting a second dynamic distortions function to the hybrid memory polynomial model;
means for sweeping N E and M E to evaluate the envelope memory polynomial model size; and
means for setting model parameters to N=N MP and M=M MP , N E =N EMP and M E =M EMP .
3. The behavioral model according to claim 2 , further comprising means for extracting the values of the memory polynomial and the envelope memory polynomial sub-function parameters leading to accurate modeling of the power amplifier distortions.
4. The behavioral model according to claim 3 , further comprising means for performing a computation of an output x out _ DD (n) of the dynamic distortion behaviors, the computation being characterized by:
x
out
DD
(
n
)
=
∑
j
=
1
M
∑
i
=
1
N
a
ji
·
x
in
DD
(
n
+
1
-
j
)
·
x
in
DD
(
n
+
1
-
j
)
i
-
1
+
∑
j
=
1
M
E
∑
i
=
1
N
E
b
ji
·
x
in
_
DD
(
n
)
·
x
in
_
DD
(
n
+
1
-
j
)
i
-
1
,
where a ji are time aligned model coefficients, and b ji are lagging cross-terms model coefficients.
5. The behavioral model according to claim 3 , wherein the means for extracting further comprises for each total number of coefficients, means for calculating the combination of nonlinearity order (N E ) and memory depth (M E ) leading to the best possible NMSE between the model's predicted output and a desired output of the power amplifier, wherein NMSE is characterized by:
NMSE
d
B
=
10
log
10
∑
l
=
1
L
x
out
_
meas
(
l
)
-
x
out
_
DD
(
l
)
2
∑
l
=
1
L
x
out
_
meas
(
l
)
2
,
where x out _ meas and x out _ DD are measured and estimated baseband output waveforms at the output of the power amplifier, respectively, and L is the number of samples in the output waveform.
6. A predistorter for reducing nonlinearity in a power amplifier, the predistorter comprising:
a first circuit for implementing a highly nonlinear static predistortion function of the power amplifier, the first circuit having an input and an output; and
a second circuit for implementing a dynamic distortion predistortion function of the power amplifier using a hybrid memory polynomial envelope memory polynomial function defined by a combination of a memory polynomial and an envelope memory polynomial, the second circuit having an input and an output, the first circuit being connected to the second circuit in cascade for producing a predistortion signal input to the power amplifier compensating for nonlinear distortion behavior of the power amplifier and producing a linear output from the power amplifier; and
means for sequentially computing an estimated predistortion function size using a formula characterized by S HMEM =(N×M)+(N E ×M E ), where N and M represent nonlinearity order and memory depth of a sub-function of the memory polynomial, respectively, and N E and M E represent nonlinearity order and memory depth of a sub-function of the envelope memory polynomial, respectively.
7. The predistorter according to claim 6 , wherein said means for sequentially computing further comprises:
means for setting a first dynamic distortions function to a memory polynomial model;
means for sweeping N and M to evaluate the memory polynomial model size;
means for setting N=N MP and setting M=M MP ;
means for setting a second dynamic distortions function to a hybrid memory polynomial model;
means for sweeping N E and M E to evaluate the envelope memory polynomial model size; and
means for setting model parameters to N=N MP and M=M MP , N E =N EMP and M E =M EMP .
8. The predistorter according to claim 7 , further comprising means for extracting values of the memory polynomial and the envelope memory polynomial sub-function parameters leading to linear performance of the power amplifier.
9. The predistorter according to claim 8 , further comprising means for performing a computation of an output x out _ DD (n) of the dynamic distortion function, the computation being characterized by:
x
out
DD
(
n
)
=
∑
j
=
1
M
∑
i
=
1
N
a
ji
·
x
in
DD
(
n
+
1
-
j
)
·
x
in
DD
(
n
+
1
-
j
)
i
-
1
+
∑
j
=
1
M
E
∑
i
=
1
N
E
b
ji
·
x
in
_
DD
(
n
)
·
x
in
_
DD
(
n
+
1
-
j
)
i
-
1
,
where a ji are time aligned dynamic distortion function coefficients, and b ji are lagging cross-terms dynamic distortion function coefficients.
10. The predistorter according to claim 8 , wherein the means for extracting further comprises for each total number of coefficients, means for calculating the combination of nonlinearity order (N E ) and memory depth (M E ) leading to the best possible NMSE between the predistorter's predicted output and a desired output of the predistorter, the NMSE being characterized by:
NMSE
d
B
=
10
log
10
∑
l
=
1
L
x
out
_
meas
(
l
)
-
x
out
_
DD
(
l
)
2
∑
l
=
1
L
x
out
_
meas
(
l
)
2
,
where x out _ meas and x out _ DD are desired and estimated baseband output waveforms, respectively, at the output of the predistorter and L is the number of samples in the output waveform.