How to avoid Hidden State in VerilogA model for SpectreRF

Hi Andrew,

 I will like to know while this Normalised LMS algorithm could not be supported in PSS analysis due to the hidden state variable "WtsReg".

Can you please help proffer solution or suggestion to the code in verilogA.

// VerilogA for IM2_Cancellation, LMS, veriloga

`include "constants.vams"
`include "disciplines.vams"

module LMS(clk,Inp_Sig,Error, Wts);

input clk, Error,Inp_Sig;
electrical clk, Error,Inp_Sig;
output Wts;
electrical Wts;

parameter real mu=0.2 ; //varies between 0 to 2

integer  vth;
integer Clk;
real ErrorReg;
real WtsReg;
integer Vin;
real Quot;

analog begin
        Clk = (V(clk)> 0.5)? 1:0;

        @(initial_step)
        begin

                WtsReg = 0;
                V(Wts) <+ 0;
        end

        if (V(Inp_Sig)) begin

                ErrorReg =  V(Error);
                V(Wts) <+ WtsReg;

        end

        @ (cross(Clk - 0.5, +1)) // rising edge of clock signal
        begin
                ErrorReg =  V(Error);
                Quot = V(Inp_Sig)/(V(Inp_Sig)*V(Inp_Sig));

                WtsReg = WtsReg + mu * ErrorReg * Quot;

        end
end
endmodule

Did you read the http://www.designers-guide.org/Analysis/hidden-state.pdf paper I pointed at earlier in this thread? The problem is that you are relying on sampling the signal at the clock edge into the variable WtsReg which is then a hidden state variable. You need to employ one of the strategies discussed in that paper so that this state is represented as an electrical node rather than a variable; variables have to be updated on every iteration.

It's also not a good idea to have the contribution statement (the V(Wts)<+WtsReg;) inside a conditional statement - in fact I'm not sure that's really doing anything useful (so remove the surrounding if() statement).

This thread seems to have become a bit of a dumping ground for questions on hidden state; the forum guidelines ask forum users not to post on the end of old threads to avoid muddying the answers and making it harder for people to find specific answers to specific questions.

Kind Regards,

Andrew.