Roundoff error in computing derivatives using the Chebyshev differentiation matrix (Q1346548): Difference between revisions

The authors propose a simple procedure to compute diagonal elements of the Chebyshev differentiation matrix, so that the constant null vector is preserved, and a dramatic reduction in roundoff error in the computation of the high-order derivatives is achieved. The procedure is used to the differentiation of a function \(f\) using the Chebyshev pseudo-spectral method. Two examples are presented using 3 methods. For these cases the authors find that the accuracy obtained from the matrix multiply approach is comparable to the accuracy obtained from transform techniques.