Further Study

One of the problems of this experiment is that the output in frequency domain diverges to very large values when there are more than 250 layers. It's not simply caused by round-off errors or bad choice of units. It's because of the term $\sqrt{\rho/s}$ in equation (6). As long as it is not equal to 1, the matrix ${\bf A}$ has a large element resulting from the successive multiplication of $\sqrt{\rho/s}$.Does it mean that it's wrong to use the equation (6) for this experiment? I don't know yet. After figuring out this problem, more examples should be studied.