System and method for computing parameters for a digital predistorter

A computerized method comprising:

• generating feedback data from input data with a non-linear circuit;
• initializing a software model of the non-linear circuit by using the input data and the feedback data to generate initial software model parameters,
• wherein the software model includes a linear block and a non-linear block in cascade,
• wherein the step of initializing further comprises:
• determining initial linear block parameters,
• determining initial non-linear block parameters, comprising:
• convolving the input data by the linear block to generate convolution data, wherein the linear block uses the initial linear block parameters;
• and
• generating initial non-linear block parameters from the convolution data;
• and
• refining the initial software model parameters following the step of initializing using the feedback data and input data, wherein the step of refining includes:
• applying the input data to the linear block;
• applying output data from the linear block to the non-linear block to generate filtered data;
• determining the difference between the filtered data and the feedback data to generate error data;
• and
• determining refined software model parameters that minimize a cost function for the error data;
• validating the refined software model parameters;
• and
• storing the refined software model parameters.

A computerized method comprising:

• processing a baseband signal to generate input data;
• capturing the input data with a first capture buffer;
• applying the input data to a power amplifier to generate feedback data;
• capturing the feedback data with a second capture buffer;
• initializing a software model of the power amplifier using the input data and feedback data from the first and second capture buffers, wherein the software model includes a linear block and a non-linear block in cascade, and wherein the software model is initialized by:
• applying the input data to the linear block;
• applying output data from the linear block to the non-linear block when the non-linear block is in a pass through state;
• determining the difference between output data from the non-linear when the non-linear block is in a pass through state and the feedback data to generate error data when the non-linear block is in a pass through state;
• minimizing a cost function for the error data when the non-linear block is in a pass through state to determine initial linear block parameters;
• and
• convolving the input data by the linear block to generate convolution data, wherein the linear block uses the initial linear block parameters;
• and
• generating initial non-linear block parameters from the convolution data;
• refining the initial linear block parameters and the initial non-linear block parameters by:
• applying the input data to the linear block using the initial linear block parameters;
• applying output data from the linear block to the non-linear block to generate filtered data, wherein the non-linear block uses the initial non-linear block parameters;
• determining the difference between the filtered data and the feedback data to generate error data when the initial linear block parameters and the initial non-linear block parameters are being refined;
• and
• determining refined linear block and non-linear block parameters that minimize the cost function for the error data when the initial linear block parameters and the initial non-linear block parameters are being refined;
• validating the refined linear block and non-linear block parameters;
• and
• storing the refined linear block and non-linear block parameters.