System and method for fast polynomial pre-distortion
View Patent ↗A system and method of pre-distorting a signal applied to a power amplifier to compensate for distortion introduced by the power amplifier at different power settings, comprising pre-distorting the signal according to a pre-distortion polynomial, prior to amplification of the signal by the power amplifier; adaptively estimating and applying pre-distortion polynomial coefficients to the pre-distorter by saving successive iterations of adaptive estimation of the pre-distortion polynomial coefficients; and regularizing estimation of the pre-distortion polynomial coefficients over successive iterations according to a modified objective function that initially constrains growth in the magnitude of the pre-distortion polynomial coefficients and then relaxes the constraint in growth over successive iterations until estimation is reduced to a steady-state least squares estimation.
1. A system for pre-distorting a signal applied to a power amplifier to compensate for distortion introduced by the power amplifier at different power settings, comprising:
a pre-distorter for pre-distorting the signal according to a pre-distortion polynomial, prior to amplification of the signal by the power amplifier;
an adaptive estimator for adaptively estimating and applying pre-distortion polynomial coefficients to the pre-distorter, wherein the adaptive estimator includes
a first estimation loop for saving successive iterations of the adaptive estimator; and
a second estimation loop including an estimation regularizer for regularizing estimation of the pre-distortion polynomial coefficients over successive iterations according to a modified objective function that initially constrains growth in a magnitude of the pre-distortion polynomial coefficients and then relaxes the constraint in growth over successive iterations until an estimation is reduced to a steady-state least squares estimation.
2. The system as in claim 1 , wherein the adaptive estimator includes a complex-to-magnitude (|u|) converter and a complex-to-phase (<u) converter for establishing parallel estimation paths for magnitude and phase pre-distortion coefficients, respectively, each path comprising a multichannel vector FIR filter and a linear equation solver for generating least squares estimates of magnitude and phase pre-distortion coefficients, respectively.
3. The system as in claim 2 , wherein the adaptive estimator includes a polynomial generator connected to the complex-to-magnitude (|u|) converter for generating a vector of different polynomial powers that forms an outer product, and a two-dimensional matrix FIR filter for averaging the outer product over time.
4. The system as in claim 3 , wherein the first estimation loop comprises a feedback loop between the output and input of each multichannel vector FIR filter and the two-dimensional matrix FIR filter, each feedback loop including a delay (τ) and a multiplicative weighing coefficient.
5. The system as in claim 4 , wherein the adaptive estimator includes a Matrix Formation block and feed forward parameters α 1 and α 2 whose outputs are respectively applied to a respective first input of each linear equation solver, a respective second input of each linear equation solver being connected to the respective output of each multichannel vector FIR filter.
6. The system as claimed in claim 5 , wherein the parameters α 1 and α 2 are iterative adjusted following a power step from higher values for increased constraint on estimation error to progressively lower values for increasingly relaxed constraint on the estimation error.
7. The system as claimed in claim 6 , wherein the parameters α 1 and α 2 are progressively reduced to zero.
8. The system as claimed in claim 6 , wherein the parameters α 1 and α 2 are progressively reduced to a non-zero value for improved estimation robustness in response to a parametric variation.
9. A method of pre-distorting a signal applied to a power amplifier to compensate for distortion introduced by the power amplifier at different power settings, comprising:
pre-distorting the signal according to a pre-distortion polynomial, prior to amplification of the signal by the power amplifier;
adaptively estimating and applying pre-distortion polynomial coefficients to the pre-distorter by saving successive iterations of adaptive estimation of the pre-distortion polynomial coefficients; and
regularizing estimation of the pre-distortion polynomial coefficients over successive iterations according to a modified objective function that initially constrains growth in a magnitude of the pre-distortion polynomial coefficients and then relaxes the constraint in growth over successive iterations until an estimation is reduced to a steady-state least squares estimation.
10. The method as in claim 9 , wherein regularizing estimation of the pre-distortion polynomial coefficients further includes formulating a constrained optimization problem using a modified objective function that minimizes estimation error of the pre-distortion polynomial coefficients and adds a weighted squared norm of the pre-distortion polynomial coefficients to the estimation error to regularize the estimation in the event that the magnitude of the pre-distortion polynomial coefficients exhibits unstable growth.
11. The method as claimed in claim 10 , wherein the squared norm is weighted by a tunable parameter α.
12. The method as claimed in claim 11 , wherein the tunable parameter α is iterative adjusted following a power step from a higher value that increases the weight of the weighted squared norm for constraining the estimation error to progressively lower values for reducing the weight of the weighted squared norm for relaxing the constraint on the estimation error.
13. The method as claimed in claim 12 , wherein the tunable parameter α is progressively reduced to zero.
14. The method as claimed in claim 12 , wherein the tunable parameter α is progressively reduced to a non-zero value for improved estimation robustness in response to parametric variation.