Non-linear model with tap output normalization

1. A method of compensating an input signal for distortion introduced by an electronic device operating on the input signal to produce an output signal, the method comprising:

generating, for each of a plurality of sampling time instances, first and second signal samples representing said input and output signals;

generating, for one or more of said sampling time instances, a corresponding set of data samples from said first signal samples and a set of basis functions in a non-linear distortion model of the electronic device or predistorter;

normalizing the data samples so that the normalized data samples from each basis function have a pre-determined variance;

calculating model weighting coefficients to fit the normalized data samples to the second signal samples according to the non-linear distortion model;

determining predistortion weights from the model weighting coefficients; and

applying the predistortion weights to the input signal to compensate for the distortion introduced by the electronic device.

2. The method of claim 1 , wherein generating the corresponding set of data samples for a given sampling time instance comprises computing one or more data samples based on first signal samples corresponding to two or more sampling time instances.

3. The method of claim 2 wherein computing one or more data samples based on first signal samples corresponding to two or more sampling time instances comprises:

computing basis function output signals corresponding to two or more sampling time instances; and

generating said one or more data samples from the basis function output signals.

4. The method of claim 3 , wherein normalizing the data samples comprises normalizing said basis function output signals and generating normalized data samples from said normalized basis function output signals.

5. The method of claim 3 wherein normalizing the data samples comprises generating the one or more data samples from said basis function output signals as non-normalized data samples and normalizing said non-normalized data samples.

6. The method of claim 1 , wherein normalizing the data samples comprises multiplying the data samples generated from each basis function by a corresponding normalization factor proportional to the reciprocal of the square root of the variance of the data samples generated by the basis function.

7. The method of claim 1 wherein the non-linear distortion model comprises a predistorter model and wherein determining predistortion weights from the model weighting coefficients comprises computing predistortion weights from the model weighting coefficients of the predistortion model.

8. The method of claim 7 wherein computing predistortion weights from the model weighting coefficients comprises computing predistortion weights as the products of the model weighting coefficients and respective normalization factors.

9. The method of claim 1 wherein the non-linear distortion model comprises a distortion model of the electronic device and wherein determining predistortion weights from the model weighting coefficients comprises:

computing power amplifier weighting coefficients from the model weighting coefficients, and

computing predistorter weighting coefficients as the inverse of the power amplifier weighting coefficients.

10. The method of claim 1 , wherein the data samples for a given sampling time instance comprises computing data samples as branch output signals from said first signal samples and a set of basis functions corresponding to said respective branches.

11. The method of claim 10 wherein normalizing the data samples comprises normalizing the branch output signals.

12. The method of claim 10 wherein said branch output signals are computed as a weighted sum of basis function output signals corresponding to two or more sampling time instances.

13. A predistortion circuit for predistorting an input signal to compensate for distortion introduced by an electronic device operating on the input signal to produce an output signal, the predistortion circuit comprising:

an input circuit configured to generate, for each of a plurality of sampling time instances, first and second signal samples representing said input and output signals;

a distortion modeling circuit to model distortion of the electronic device or predistorter and to compute predistortion weights, said distortion modeling circuit configured to:

generate, for one or more of said sampling time instances, a corresponding set of data samples from said first signal samples and a set of basis functions in a non-linear distortion model of the electronic device or predistorter;

normalize the data samples so that the normalized data samples from each basis function have a pre-determined variance;

calculate model weighting coefficients to fit the normalized data samples to the second signal samples according to the non-linear distortion model;

determine predistortion weights from the model weighting coefficients;

a predistorter to apply the predistortion weights to the input signal to compensate for the distortion introduced by the electronic device

14. The predistortion circuit of claim 13 wherein the distortion modeling circuit is configured to generate the corresponding set of data samples for a given sampling time instance by computing one or more data samples based on first signal samples corresponding to two or more sampling time instances.

15. The predistortion circuit of claim 14 wherein the distortion modeling circuit is configured to compute the one or more data samples by computing basis function output signals corresponding to two or more sampling time instances, and generating said one or more data samples from the basis function output signals.

16. The predistortion circuit of claim 15 wherein the distortion modeling circuit is configured to normalize the data samples by normalizing said basis function output signals and generating normalized data samples from said normalized basis function output signals.

17. The predistortion circuit of claim 15 wherein the distortion modeling circuit is configured to normalize the data samples by generating the one or more data samples from said basis function output signals as non-normalized data samples and normalizing said non-normalized data samples.

18. The predistortion circuit of claim 15 wherein the distortion modeling circuit is configured to normalize the data samples by multiplying the data samples generated from each basis function by a corresponding normalization factor proportional to the reciprocal of the square root of the variance of the data samples generated by the basis function.

19. The predistortion circuit of claim 13 wherein the non-linear model comprises a predistorter model and wherein the distortion modeling circuit is configured to compute predistortion weights from the model weighting coefficients of the predistortion model.

20. The predistortion circuit of claim 19 wherein the distortion modeling circuit is configured to compute the predistortion weights as the products of the model weighting coefficients and respective normalization factors.

21. The predistortion circuit of claim 13 wherein the non-linear model comprises a distortion model of the electronic device and wherein the distortion modeling circuit is configured to:

compute power amplifier weighting coefficients from the model weighting coefficients of the electronic device distortion model; and

compute predistortion weighting coefficients as the inverse of the power amplifier weighting coefficients.

22. The predistortion circuit of claim 13 wherein the distortion modeling circuit is configured to compute data samples as branch output signals from said first signal samples and a set of basis functions corresponding to said respective branches.

23. The predistortion circuit of claim 22 wherein the distortion modeling circuit is configured to normalize the data samples by normalizing the branch output signals.

24. The predistortion circuit of claim 22 wherein the distortion modeling circuit is configured to compute the branch output signals as the weighted sum of basis function output signals corresponding to two or more sampling time instances.