I have implemented (-1,+1) boxed-constraint algorithm into FPGA desgin. The algorithm is to solve the equation Rh+n=b in telecommunication application. For my project case, all the elements of size NxN Matrix R in the region [-1,1], and n is the Gauss noise. h is a size Nx1vector . For example: R =
1.0000 0.5000 0.5000 0.5000 0.5000 1.0000 0 0 0.5000 0 1.0000 1.0000 0.5000 0 1.0000 1.0000b =
-1.2672 -0.5291 2.1703 1.2486
R1=round(2*R); b1=round(2*b); It is easily seen that all elements can be presented by 4 bit. But for the transmission to FPGA board, I extend sign bit to 8 bit . Then I write Matrix R1 and vector b1 into FPGA board.
In Matlab simulation, I used constraint H=2^0 as the biggest step size .Then is d=2^-1, d=2^-2....to approach the expected h value. So my question is: Shall I use H=00000001 to presnted 1? Some said I used 4 bit to present input data. I should use 2^4 to present H. I don't agree with it. And now I am a little confused about it. Does anybody can help me figure me out? I have tried my best to explain it clearly. Thank you very much.
Zhi