Granted Patent B1

US 12,425,971 · App. 19/077,638 · Granted Sep 23, 2025

Inventors: Rosanah Murugesu (Murray Hill, NJ); Hungkei Chow (Murray Hill, NJ)

Assignee: Nokia Solutions and Networks Oy

H04W52/0235H04B7/0456H04B2001/0425

Abstract

Claims (52)

1. An apparatus, comprising:

at least one processor;

and at least one memory storing instructions that, when executed by the at least one processor, cause the apparatus at least to:

receive a signal to be transmitted via an array of antennas from a linear precoder, wherein the signal is processed by the linear precoder based on one or more input power criteria for power amplifiers of the array of antennas;

apply a distortion reducing matrix to the processed signal, wherein the distortion reducing matrix is trained to reduce the distortion at each output of the power amplifiers based on minimizing an error signal corresponding to a difference between measured and calculated outputs of the power amplifiers after a coupling effect between the power amplifiers based on the applied matrices,

and wherein the distortion reducing matrix is further trained based on minimizing an error signal corresponding to a difference between a measured total input power to the power amplifiers and a total input power provided to the power amplifiers at startup of the apparatus; and

provide an output of the distortion reducing matrix to be used in digital pre-distortion processing of an input signal for the power amplifiers.

2. The apparatus of claim 1 , wherein the apparatus further comprises the linear precoder configured to receive the signal to be processed and output the processed signal to the distortion reducing matrix; and

wherein the one or more criteria comprise at least one of equal or fixed input power for the power amplifiers.

3. The apparatus of claim 1 , wherein the signal is processed by applying an equal power zero-forcing precoding matrix for equal power input to the distortion reducing matrix.

4. The apparatus of claim 1 , wherein the apparatus is further caused to:

process the output of the distortion reducing matrix based on digital pre-distortion models trained to compensate nonlinear distortion of the power amplifiers; and

provide an output of the digital pre-distortion models as the input signal to the power amplifiers.

5. The apparatus of claim 1 , wherein the apparatus further comprises at least one of the array of antennas, wherein the number of the antennas in the array is greater than a number of receivers served by the antennas, or the power amplifiers of the array of antennas.

6. The apparatus of claim 1 , wherein the distortion reducing matrix is one of a frequency independent matrix, a frequency independent vector, a frequency dependent matrix per a subcarrier or a group of subcarriers or a frequency dependent vector per a subcarrier or a group of subcarriers.

7. The apparatus of claim 1 , further caused to:

train the distortion reducing matrix at a startup of the apparatus for all power amplifiers of the array of antennas using random signals and random channel matrices.

8. The apparatus of claim 1 , wherein the apparatus is further caused to:

update coefficients of the distortion reducing matrix based on minimizing a weighted sum of the error signal of the outputs of the power amplifiers and the error signal of the total input power to the power amplifiers with a least mean square algorithm.

9. The apparatus of claim 8 , further caused to:

detect a change in at least one of behaviour of the plurality of power amplifiers, the coupling effect between the plurality of power amplifiers or an ambient environment; and

update the coefficients of the distortion reducing matrix in response to the detected change.

10. The apparatus of claim 8 , wherein the distortion reducing matrix is iteratively trained and the apparatus is caused to:

set the distortion reducing matrix initially to an identity matrix; and

for every iteration round:

generate a random channel matrix;

calculate the linear precoder based on a channel response determined based on the random channel matrix, a matrix model of the coupling effect and the distortion reducing matrix;

generate a random test signal;

calculate an output of the distortion reducing matrix based on the calculated linear precoder and the test signal, and apply the output to the power amplifiers;

obtain measured outputs of the power amplifiers after the coupling effect; and

update coefficients of the distortion reducing matrix based on the measured and calculated values of the iteration round.

11. The apparatus of claim 8 , wherein the distortion reducing matrix is iteratively trained and the apparatus is caused to:

set the distortion reducing matrix initially to an identity matrix;

determine that a signal-to-noise ratio at outputs of the power amplifiers is below a threshold; and

perform iteration rounds while the signal-to-noise ratio is below the threshold, wherein for every iteration round the apparatus is caused to:

generate a random channel matrix;

calculate the linear precoder based on a channel response determined based on the random channel matrix, a matrix model of the coupling effect and the distortion reducing matrix;

generate a random test signal;

calculate an output of the distortion reducing matrix based on the calculated linear precoder and the test signal;

calculate an output of the digital pre-distortion models based on the output of the distortion reducing matrix, and apply the output of the digital pre-distortion models to the power amplifiers;

obtain measured outputs of the power amplifiers after the coupling effect;

determine if a current iteration round count is divisible by a predetermined positive integer;

if the current iteration round count is divisible by the predetermined positive integer, update model coefficients of one of the digital pre-distortion models or the distortion reducing matrix based on the measured and calculated values of the iteration round while maintaining coefficients of one of the distortion reducing matrix or the digital pre-distortion models; and

if the current iteration round count is not divisible by the predetermined positive integer, update coefficients of one of the distortion reducing matrix or the digital pre-distortion models based on the measured and calculated values of the iteration round while model coefficients of one of the digital pre-distortion models or the distortion reducing matrix are maintained.

12. The apparatus of claim 11 , wherein the coefficients of the digital pre-distortion models are updated per power amplifier with a direct learning algorithm based on the measured outputs of the power amplifiers and the output of the distortion reducing matrix with the matrix model for the coupling effect applied.

13. The apparatus of claim 11 , wherein the coefficients of the digital pre-distortion models are updated per power amplifier with an indirect learning algorithm based on the measured outputs and the output of the digital pre-distortion models with the matrix model for the coupling effect applied.

14. The apparatus of claim 1 , wherein the apparatus is a network device.

15. A method, comprising:

receiving a signal to be transmitted via an array of antennas from a zero-forcing precoder, wherein the signal is processed by the linear precoder based on one or more input power criteria for power amplifiers of the array of antennas;

applying a distortion reducing matrix to the processed signal, wherein the distortion reducing matrix is trained to reduce the distortion at each output of the power amplifiers based on minimizing an error signal corresponding to a difference between measured and calculated outputs of the power amplifiers after a coupling effect between the power amplifiers based on the applied matrices,

and wherein the distortion reducing matrix is further trained based on minimizing an error signal corresponding to a difference between a measured total input power to the power amplifiers and a total input power provided to the power amplifiers at startup of the apparatus; and

providing an output of the distortion reducing matrix to be used in digital pre-distortion processing of an input signal for the power amplifiers.

Assignments (3)

Priority Claims (1)

References Cited (13)