Erbes [9] proposed a semi-Lagrangian method of characteristics (MOC) with quadratic interpolation to solve the shallow water equations with a large CFL number, but there was
dispersion error produced.
Model of empirical distribution function of the number of defect-free reference lengths in a normal distribution function with the expectation that varies linearly over the manufacturing process and stable
dispersion error control: 1--empirical distribution function [F.sup.*]{n(er<4); 2--distribution function at the beginning of the observation period; 3--distribution function at the end of the observation period; 4--model distribution function
Using this modified AH-FDTD method, the
dispersion error induced by the space grid can be reduced, which makes it possible to set coarser grid and generally reduces the total number of grids.
For a Runge-Kutta method the
dispersion error (phase-lag error) and the dissipation error (amplification error) are given, respectively, by
Although it remains second-order accurate, RD-FD is able to half the
dispersion error. It is identical to our LFE-FC-7 stencil in Section 4.1 derived from using a spherical Fourier-Bessel series (SFBS).
In other words, for a given numerical
dispersion error, QI-ADI-FDTD can take a time step larger than that of ADI-FDTD, leading to a saving of computation time.
The various types of errors, namely, dissipation error and
dispersion error, are tabulated for some values of cfl numbers in Table 8.
It is well known that the ADI-FDTD method will result in numerical
dispersion error for larger step size.
Analysis of Stability and Numerical
Dispersion ErrorHadley's is that we analytically derived both the local LFE error and the global CLF
dispersion error equations and proved that this LFE-9 equation is of the sixth-order accuracy, which is theoretically the highest order of accuracy on a nine-point compact stencil [12] for the 2D discrete Helmholtz operator.
Nevertheless, it presents large numerical
dispersion error with large time steps.