13results about "Memory effect compensation" patented technology

A digital pre-distortion circuit and technique for a signal being amplified. Digital pre-distortion is performed using a model-based pre-distortion function. Pre-distortion is further corrected using a data-driven pre-distortion function. The model-based pre-distortion function and the data-driven pre-distortion function are trained using training data including a digital loopback signal generated by sampling a loopback signal of an amplifier, the digital pre-distorted signal input to the amplifier for a time period, and an intermediate signal output from the model-based pre-distortion function driven by the digital loopback signal or an output of the data-driven pre-distortion function. Parameters may be continually updated in order to account for changes in characteristics of the amplifier.

A pre-compensation method for a pre-compensation circuit coupled to a power amplifier to compensate for nonlinearity of the power amplifier includes performing pre-distortion according to at least one parameter or at least one hyperparameter to convert a first input signal received by the pre-compensation circuit into a first pre-distortion output signal, updating the at least one parameter or the at least one hyperparameter according to Bayesian Optimization, Causal Bayesian Optimization, or Dynamic Causal Bayesian Optimization, and performing pre-distortion according to the at least one parameter updated or the at least one hyperparameter updated to convert a second input signal received by the pre-compensation circuit into a second pre-distortion output signal.

A device may include a first model implemented using a first processing circuit configured to implement a transformation function to generate an initial predistortion signal based on an input signal corresponding to an output of a power amplifier. The device may include a second model implemented using a second processing circuit configured to implement a modification function to modify the initial predistortion signal. A device may include a configurable multiplier configured to weight an output of the second model based at least in part on a weighting value to obtain a weighted output. A device may include a combiner configured to combine the initial predistortion signal that is output by the first model with the weighted output to generate a digital predistortion signal for the power amplifier in communication with the digital predistortion system.

Systems, devices, and methods are provided for using model architecture search for hardware configurations. The method includes receiving, by a computer-implemented system, information associated with a pool of processing units, receiving, by the computer-implemented system, a dataset associated with a data transformation operation, training a parameterized model associated with the data transformation operation based on the dataset and the information associated with the pool of processing units, where the training includes updating at least one parameter of the parameterized model associated with configuring at least a subset of the processing units in the pool, and outputting one or more configurations for at least a subset of the processing units in the pool based on the training.

This invention relates to distortion compensation devices and methods, computer-readable storage media, and communication devices. One provided is a distortion compensation device that uses a distortion compensation model to compensate for distortion in a signal to be amplified by an amplifier, wherein internal states affecting distortion characteristics in the amplifier change, and the distortion compensation model includes: a plurality of computational models having corresponding distortion compensation characteristics for the amplifier in different internal states; and a combiner that combines the plurality of computational models with a combination ratio corresponding to the changed internal states.

A device may include a first model implemented using a first processing circuit configured to implement a transformation function to generate an initial predistortion signal based on an input signal corresponding to an output of a power amplifier. The device may include a second model implemented using a second processing circuit configured to implement a modification function to modify the initial predistortion signal. A device may include a configurable multiplier configured to weight an output of the second model based at least in part on a weighting value to obtain a weighted output. A device may include a combiner configured to combine the initial predistortion signal that is output by the first model with the weighted output to generate a digital predistortion signal for the power amplifier in communication with the digital predistortion system.

According to embodiments, a method performed by an electronic device includes: identifying a channel bandwidth; identifying one or more active DPD units among a plurality of DPD units of a digital pre-distortion (DPD) circuit based on the channel bandwidth and amount of resource allocated within the channel bandwidth; generating a DPD output signal of the DPD circuit based on the one or more active DPD units; generating an amplifier output signal based on the DPD output signal of the DPD circuit, a digital-to-analog converter (DAC), and a power amplifier (PA). While the DPD output signal is generated, at least one DPD unit different from the one or more active DPD units among the plurality of DPD units of the DPD circuit is deactivated.

A pre-compensation method for a pre-compensation circuit coupled to a power amplifier to compensate for nonlinearity of the power amplifier includes performing pre-distortion according to at least one parameter or at least one hyperparameter to convert a first input signal received by the pre-compensation circuit into a first pre-distortion output signal, updating the at least one parameter or the at least one hyperparameter according to Bayesian Optimization, Causal Bayesian Optimization, or Dynamic Causal Bayesian Optimization, and performing pre-distortion according to the at least one parameter updated or the at least one hyperparameter updated to convert a second input signal received by the pre-compensation circuit into a second pre-distortion output signal.

A digital pre-distortion achieving near ideal spectral regrowth suppression with a bandlimited loopback. To eliminate mismatch between the bandwidth of the pre-distortion model terms and the bandwidth of the signal used to determine the model a band limiting operation may be applied to both the loopback signal and the input to the power amplifier used for estimating the pre-distortion coefficients. In addition, oversampling may be performed in the digital domain to avoid aliasing in the nonlinear terms of the pre-distortion model.

A digital pre-distortion achieving near ideal spectral regrowth suppression with a bandlimited loopback. To eliminate mismatch between the bandwidth of the pre-distortion model terms and the bandwidth of the signal used to determine the model a band limiting operation may be applied to both the loopback signal and the input to the power amplifier used for estimating the pre-distortion coefficients. In addition, oversampling may be performed in the digital domain to avoid aliasing in the nonlinear terms of the pre-distortion model.

A digital pre-distortion circuit and technique for a signal being amplified. Digital pre-distortion is performed using a model-based pre-distortion function. Pre-distortion is further corrected using a data-driven pre-distortion function. The model-based pre-distortion function and the data-driven pre-distortion function are trained using training data including a digital loopback signal generated by sampling a loopback signal of an amplifier, the digital pre-distorted signal input to the amplifier for a time period, and an intermediate signal output from the model-based pre-distortion function driven by the digital loopback signal or an output of the data-driven pre-distortion function. Parameters may be continually updated in order to account for changes in characteristics of the amplifier.

There is provided mechanisms for enabling linearization of a non-linear electronic device. A method is performed by a linearizer device. The method comprises receiving an input signal destined to be input to the non-linear electronic device. Input-output characteristics of the non-linear electronic device is in the linearizer device represented by a linearization function defined by a LUT based model of base functions. The linearizer device is configured to in a greedy pursuit framework select the base functions according to a signal reconstruction criterion. The method comprises obtaining an output signal by subjecting the input signal to the linearization function. The method comprises providing the output signal, instead of the input signal, as input to the non-linear electronic device, thereby enabling linearization of the non-linear electronic device.

This disclosure provides an apparatus and method for compensating for nonlinearity in a power amplifier. An apparatus configured to perform wireless communication to accelerate a neural network includes: a predistortion circuit configured to generate a predistorted input signal by performing predistortion on an input signal based on a parameter set including multiple coefficients; a power amplifier configured to generate an output signal by amplifying an RF signal based on the predistorted input signal; and a parameter acquisition circuit configured to obtain second memory polynomial modeling information corresponding to an operating frequency band based on first memory polynomial modeling information corresponding to each of a plurality of frequency bands, and to use the second memory polynomial modeling information to obtain a parameter set according to an indirect learning structure.