Decorrelated power amplifier linearizers
View Patent ↗Procedures for decorrelating the branch signals of a signal adjuster of an amplifier linearizer are presented herein. The decorrelation procedures can be performed with or without self-calibration.
1. A method of decorrelating M control signals in a multibranch feedforward linearizer having M monitor signals and a first signal, said method comprising the steps of:
performing bandpass correlations pairwise between the M monitor signals to form a signal correlation matrix, each pairwise bandpass correlation a component of the signal correlation matrix;
inverting the signal correlation matrix;
performing bandpass correlation between the first signal and each of the M monitor signals to form a correlation vector, each bandpass correlation being a component of the correlation vector; and
computing the M control signals using the inverted signal correlation matrix and the correlation vector.
2. A method according to claim 1 , wherein the steps are iteratively repeated.
3. A method according to claim 1 , wherein the computing step also uses a scalar step size parameter.
4. A method according to claim 1 , wherein a is a control signal vector of M length, R a is an M×M signal correlation matrix, R a −1 is the inverse of the signal correlation matrix, r ae is a correlation vector of M length, s is a scalar step size parameter, and n is an iteration, and the M control signals of the n+1 iteration are computed as follows:
a ( n+ 1)= a ( n )+ sR a −1 r ae ( n ).
5. A method according to claim 1 , wherein the first signal is an error signal of the linearizer.
6. A method according to claim 1 , wherein the first signal is an output signal of the linearizer.
7. A method of decorrelating M control signals in a multibranch feedforward linearizer having M monitor signals and a first signal, said method comprising the steps of:
performing partial correlations pairwise between the M monitor signals at N frequencies;
for each monitor signal, summing the pairwise partial correlations over N frequencies to form a signal correlation matrix, each sum being a component of the signal correlation matrix;
inverting the signal correlation matrix;
performing partial correlations between the first signal and each of the M monitor signals over N frequencies;
for each monitor signal, summing the partial correlations over N frequencies to form a correlation vector, each sum being a component of the correlation vector; and
computing the M control signals using the inverted signal correlation matrix and the correlation vector.
8. A method according to claim 7 , wherein the steps are iteratively repeated.
9. A method according to claim 7 , wherein the computing step also uses a scalar step size parameter.
10. A method according to claim 7 , wherein a is a control signal vector of M length, R a is an M×M signal correlation matrix, R a −1 is the inverse of the signal correlation matrix, r ae is a correlation vector of M length, s is a scalar step size parameter, and n is an iteration, and the M control signals of the n+1 iteration are computed as follows:
a ( n+ 1)= a ( n )+ sR a −1 r ae ( n ).
11. A method according to claim 7 , wherein the first signal is an error signal of the linearizer.
12. A method according to claim 7 , wherein the first signal is an output signal of the linearizer.
13. A method for generating M control signals in a M branch signal adjuster for a linearizer, where M is greater than 1, the signal adjuster having M branch signals and a corresponding M monitor signals, and M observation filters between the respective M branch and monitor signals, the method comprising the steps of:
estimating the gains of the M observation filters; and
decorrelating the M control signals using the estimated gains of the M observation filters.
14. A method of computing M control signals in a M branch signal adjuster for a linearizer, where M is greater than 1, the signal adjuster having M branch signals and a corresponding M monitor signals, a first signal, and M observation filters between the M branch and monitor signals, said method comprising the steps of:
estimating the gains of M observation filters;
performing bandpass correlations pairwise between the M monitor signals to form a signal correlation matrix, each pairwise bandpass correlation being a component of the signal correlation matrix;
adjusting the components of the signal correlation matrix using the corresponding estimated gains of the M observation filters;
inverting the signal correlation matrix;
performing bandpass correlation between the first signal and each of the M monitor signals to form a correlation vector, each bandpass correlation being a component of the correlation vector;
adjusting the components of the correlation vector using the corresponding estimated gains of the M observation filters; and
computing the M control signals using the inverted signal correlation matrix and the correlation vector.
15. A method of computing M control signals in a M branch signal adjuster for a linearizer, where M is greater than 1, the signal adjuster having M branch signals and a corresponding M monitor signals, a first signal, and M observation filters between the M branch and monitor signals, said method comprising the steps of:
determining the gains of M observation filters;
performing partial correlations pairwise between the M monitor signals at N frequencies;
for each monitor signal, summing the pairwise partial correlations over N frequencies to form a signal correlation matrix, each sum being a component of the signal correlation matrix;
adjusting the components of the signal correlation matrix using the corresponding estimated gains of the M observation filters;
inverting the signal correlation matrix;
performing partial correlations between the first signal and each of the M monitor signals over N frequencies;
for each monitor signal, summing the partial correlations over N frequencies to form a correlation vector, each sum being a component of the correlation vector;
adjusting the components of the correlation vector using the corresponding estimated gains of the M observation filters; and
computing the M control signals using the inverted signal correlation matrix and the correlation vector.
16. A linearizer for an amplifier comprising:
an FIR signal adjuster having two signal branches, wherein the power of the signals on each branch are unequal; and
an adaptation controller for decorrelating a plurality of control signals for said FIR signal adjuster.
17. A linearizer for an amplifier comprising:
a signal adjuster having three or more signal branches; and
an adaptation controller for decorrelating a plurality control signals for said signal adjuster.
18. A linearizer for an amplifier comprising:
a non-FIR signal adjuster having two or more signal branches; and
an adaptation controller for decorrelating a plurality of control signals for said non-FIR signal adjuster.
19. A method according to claim 1 , wherein a is a control signal vector of M length, R a is an M×M signal correlation matrix computed as the weighted sum of measured signal correlation matrices R a (n) at successive iteration steps n=1, 2, 3, . . . , R a −1 is the inverse of the signal correlation matrix, r ae is a correlation vector of M length computed as the weighted sum of measured correlation vectors r ae (n) at successive iteration steps, and a is computed by least squares as a=R a −1 r ae .
20. A method according to claim 1 , wherein a is a control signal vector of M length, R a is an M×M signal correlation matrix, R a −1 is the inverse of the signal correlation matrix, and a and R a −1 are computed iteratively according to a recursuve least squares method.
21. A method according to claim 7 , wherein a is a control signal vector of M length, R a is an M×M signal correlation matrix computed as the weighted sum of measured signal correlation matrices R a (n) at successive iteration steps n=1, 2, 3, . . . , R a −1 is the inverse of the signal correlation matrix, r ae is a correlation vector of M length computed as the weighted sum of measured correlation vectors r ae (n) at successive iteration steps, and a is computed by least squares as a=R a −1 r ae .
22. A method according to claim 7 , wherein a is a control signal vector of M length, R a is an M×M signal correlation matrix, R a −1 is the inverse of the signal correlation matrix, and a and R a −1 are computed iteratively according to a recursuve least squares method.
23. A method for generating a plurality of control signals for a FIR signal adjuster of an amplifier linearizer having two branches, each branch having unequal power, comprising the steps of:
decorrelating a plurality of monitor signal of the signal adjuster; and
computing said plurality of control signals accounting for the decorrelated monitor signals.
24. A method according to claim 23 , in which the decorrelating step comprises:
correlating the monitor signals between themselves to form a signal correlation matrix;
inverting the signal correlation matrix; and
correlating an error signal of the linearizer and the monitor signals to form a correlation vector.
25. A method according to claim 24 , wherein the computing step uses the inverted signal correlation matrix and the correlation vector to generate the control signals.
26. A method for generating a plurality of control signals for a signal adjuster of an amplifier linearizer having three or more branches, comprising the steps of:
decorrelating a plurality of monitor signal of the signal adjuster; and
computing said plurality of control signals accounting for the decorrelated monitor signals.
27. A method according to claim 26 , in which the decorrelating step comprises:
correlating the monitor signals between themselves to form a signal correlation matrix;
inverting the signal correlation matrix; and
correlating an error signal of the linearizer and the monitor signals to form a correlation vector.
28. A method according to claim 27 , wherein the computing step uses the inverted signal correlation matrix and the correlation vector to generate the control signals.
29. A method for generating a plurality of control signals for a non-FIR signal adjuster of an amplifier linearizer having two or more branches, comprising the steps of:
decorrelating a plurality of monitor signal of the signal adjuster; and
computing said plurality of control signals accounting for the decorrelated monitor signals.
30. A method according to claim 29 , in which the decorrelating step comprises:
correlating the monitor signals between themselves to form a signal correlation matrix;
inverting the signal correlation matrix; and
correlating an error signal of the linearizer and the monitor signals to form a correlation vector.
31. A method according to claim 30 , wherein the computing step uses the inverted signal correlation matrix and the correlation vector to generate the control signals.
32. A method for an amplifier linearizer having a signal adjuster with two or more branches, comprising the steps of:
self-calibrating the signal adjuster; and
decorrelating the signal adjuster.
33. A method according to claim 32 , wherein the self-calibrating and decorrelating steps comprise the substeps of:
computing an observation filter gain for each branch of the signal adjuster;
correlating monitor signals of the signal adjuster between themselves to form a signal correlation matrix; and
adjusting the signal correlation matrix using the observation filter gains.
34. A method according to claim 33 , wherein the self-calibrating and decorrelating steps further comprise the substeps of:
inverting the adjusted signal correlation matrix; and
correlating an error signal of the linearizer and the monitor signals to form a correlation vector; and
computing said plurality of control signals using the adjusted inverted signal correlation matrix and the correlation vector to generate the control signals.
35. A linearizer for an amplifier comprising:
a signal adjuster having two or more signal branches; and
an adaptation controller for self-calibrating and decorrelating a plurality of control signals for said signal adjuster.