# US 2018/0026587 A1 · Stabilization of Direct Learning Algorithm for Wideband Signals

> Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt

- Number: US 2018/0026587 A1
- Document ID: US20180026587A1
- Status: Published application
- Kind: A1
- Application number: 15/549,750
- Filed: 2015-02-10
- Published: 2018-01-25
- Earliest claimed: 2015-02-10
- Assignees: Nokia Solutions and Networks Oy
- Applicants: Nokia Solutions and Networks GmbH and Co. KG
- Inventors: Christian REICHL; Shavantha KULARATNA; Rene ROESCHKE; Bjoern JELONNEK
- Web page: https://patentgrep.com/patent/US20180026587A1

## CPC

- H03F 1/3247: {using feedback acting on predistortion circuits (H03F1/3264 takes precedence)}
- H03F 1/3258: {based on polynomial terms}
- H03F 3/189: High-frequency amplifiers, e.g. radio frequency amplifiers
- H03F 3/195: in integrated circuits
- H03F 3/24: of transmitter output stages
- H04L 1/1887: {Scheduling and prioritising arrangements}
- H04L 27/368: {adaptive predistortion}
- H03F 2201/3212: Using a control circuit to adjust amplitude and phase of a signal in a signal path
- H03F 2201/3233: Adaptive predistortion using lookup table, e.g. memory, RAM, ROM, LUT, to generate the predistortion
- H03F: AMPLIFIERS
- H04L: TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION (arrangements common to telegraphic and telephonic communication H04M)

## Abstract

The present invention addresses method, apparatus and computer program product for stabilization of the direct learning algorithm for wideband signals. Thereby, a signal to be amplified is input to a pre-distorter provided for compensating for non-linearity of the power amplifier, and the pre-distorted output signal from the pre-distorter is forwarded to the power amplifier. Parameters of the pre-distorter are adapted based on an error between the linearized signal output from the power amplifier and the signal to be amplified using an adaptive direct learning algorithm, and the linear system of equations formed by the direct learning algorithm are solved using a conjugate gradient algorithm, wherein, once per direct learning algorithm adaptation, at least one of the initial residual and the initial direction of the conjugate gradient algorithm are set based on the result of the previous adaptation.

## Claims

### Claim 1

A method for controlling a power amplifier operating in a non-linear state, comprising:
inputting a signal to be amplified to a pre-distorter provided for compensating for non-linearity of the power amplifier;
forwarding the pre-distorted output signal from the pre-distorter to the power amplifier;
adapting parameters of the pre-distorter based on an error between the linearized signal output from the power amplifier and the signal to be amplified using an adaptive direct learning algorithm; and
solving the linear system of equations formed by the direct learning algorithm using a conjugate gradient algorithm,
wherein, once per direct learning algorithm adaptation, at least one of the initial residual and the initial direction of the conjugate gradient algorithm are set based on the result of the previous adaptation.

### Claim 2 (depends on 1)

The method according to claim 1, wherein the initial residual and/or the initial direction of the conjugate gradient algorithm is/are set at the beginning of the adaptation.

### Claim 3 (depends on 1)

The method according to claim 1, wherein the initial residual and/or the initial direction of the conjugate gradient algorithm is/are set so as to fulfill the following equations:
rinit=b−A*xn−1
dinit=rinit
wherein rinit, is the initial residual vector, dinit, is the initial direction vector, b is the cross-correlated vector of the error, A is the autocorrelation matrix, and xn−1 is the solution vector at the (n−1)th direct learning algorithm adaptation.

### Claim 4 (depends on 1)

The method according to claim 1, wherein loops of the conjugant gradient algorithm are iterated until reaching a preset number of loops.

### Claim 5 (depends on 1)

The method according to claim 1, wherein the initial residual and/or the initial direction of the conjugate gradient algorithm is/are computed based on the partially converged solution of the previous adaptation.

### Claim 6 (depends on 1)

The method according to claim 1, wherein an error scaling factor is introduced in the adaptive direct learning algorithm, so as to fulfill the following equation:
y(n)=x(n)−μ*e(n)
wherein y is the pre-distorted signal vector, x is a vector of the signal to be amplified, e is the error vector, μ is the error scaling factor, and indicates the adaptation number.

### Claim 7 (depends on 6)

The method according to claim 6, wherein the error scaling factor is set to a value smaller than 1.0

### Claim 8

An apparatus for controlling a power amplifier, comprising:
a pre-distorter;
at least one processor, and
at least one memory for storing instructions to be executed by the processor, wherein
the at least one memory and the instructions are configured to, with the at least one processor, cause the apparatus at least to perform
inputting a signal to be amplified to a pre-distorter provided for compensating for non-linearity of the power amplifier;
forwarding the pre-distorted output signal from the pre-distorter to the power amplifier;
adapting parameters of the pre-distorter based on an error between the linearized signal output from the power amplifier and the signal to be amplified using an adaptive direct learning algorithm; and
solving the linear system of equations formed by the direct learning algorithm using a conjugate gradient algorithm,
wherein, once per direct learning algorithm adaptation, at least one of the initial residual and the initial direction of the conjugate gradient algorithm are set based on the result of the previous adaptation.

### Claim 9 (depends on 8)

The apparatus according to claim 8, wherein the initial residual and/or the initial direction of the conjugate gradient algorithm is/are set at the beginning of the adaptation.

### Claim 10 (depends on 8)

The apparatus according to claim 8, wherein the initial residual and/or the initial direction of the conjugate gradient algorithm is/are set so as to fulfill the following equations
rinit=b−A*xn−1
dinit=rinit
wherein rinit, is the initial residual vector, dinit, is the initial direction vector, b is the cross-correlated vector of the error, A is the autocorrelation matrix, and xn−1 is the solution vector at the (n−1)th direct learning algorithm adaptation.

### Claim 11 (depends on 8)

The apparatus according to claim 8, wherein loops of the conjugant gradient algorithm are iterated until reaching a preset number of loops.

### Claim 12 (depends on 8)

The apparatus according to claim 8, wherein the initial residual and/or the initial direction of the conjugate gradient algorithm is/are computed based on the partially converged solution of the previous adaptation.

### Claim 13 (depends on 8)

The apparatus according to claim 8, wherein an error scaling factor is introduced in the adaptive direct learning algorithm, so as to fulfill the following equation:
y(n)=x(n)−μ*e(n)
wherein y is the pre-distorted signal vector, x is a vector of the signal to be amplified, e is the error vector, μ is the error scaling factor, and indicates the adaptation number.

### Claim 14 (depends on 13)

The apparatus according to claim 13, wherein the error scaling factor is set to a value smaller than 1.0

### Claim 15 (depends on 1)

A computer program product for a computer, comprising software code portions for performing the steps of claim 1 when said product is run on the computer.

### Claim 16 (depends on 15)

The computer program product according to claim 15, wherein
the computer program product comprises a computer-readable medium on which said software code portions are stored, and/or
the computer program product is directly loadable into the internal memory of the computer and/or transmittable via a network by means of at least one of upload, download and push procedures.
