PA Modelling with Volterra series

Hello to everybody,

since some time I am stuck with a modelling problem, maybe someone here has already done this before and could give me a tipp on how to procede.

Basically I have some data, representing the input and the output signal of a power amplifier. What I would like to do is model the answer of the device by black box modelling. I have alread tried different models (polynomials, polynomials with memory, FIR, Hammerstein), but I do believe that in my case would be better to use a Volterra serie in order to model such non linearity.

I have read many papers and books about the topic but still I find it difficult to implement in matlab. Building the first order kernel is no problematic but things start to get really tricky while increasing the order of the model (so the memory).

Does anybody have already tried this in MATLAB and can give me a tipp on how to build the problem of calculating the kernels?

At the moment I cannot figure out a way to recursively generate the regression matrix to build up my solution.

Best Regards, Giovanni

Answer 1

[sxy]

Hi Giovanni, Actually, Volterra Series is really very complex and computation consuming. However its expression form is somehow straightforward. In my opinion, it is not that complex to code when you specify limited and certain memory order and nonlinear order.

I have code a GMP model which is the simplified and degraded form of volterra series. Could you please post the code out here so that we can be more specific on the difficulities.

BR// Rycsxy

[Frank Underwood]

Could you post the GMP model code, thank U

[Abdelwahab Afifi]

can you please share GMP code. thanks in advance

[behnoosh meskoob]

I also need it, would you please send it. Thanks in advance.

[Michal Kmiecik]

Can you please share me your code?

Thanks in advance.