I am estimating the dissipation rate from (du'/dt), estimated at each time using the central differencing scheme. Some authors especially Prof. Mi (ref: Meas.Sci.Technol. 22 (2011)) have used spectral like stencil to calculate (du'/dt) from (u',t) data.Can anyone give me an idea of setting the matrix for this procedure. I tried forming a pentadiagonal matrix (A) (with the coefficients of gradient terms) in the right hand side, [A] [Grad]=[u']; This matrix is very sensitive to the boundary conditions and shows very large values compared to central differencing. How can I resolve this?