


 Implicit finite difference in timespace domain with the helix transform  

Next: Nonseparability of velocity from
Up: Barak: Implicit helical finite
Previous: Validity tests of derived
Implicit finitedifference propagation of a wavefield is the solution of the system shown in Equation 17. I tested wavefield propagation with constant velocity using a standard linear equation solver, and compared it to the wavefield propagated by repeated deconvolutions with spectrally factorized coefficients of the same finitedifference weights (Equation 14). Figure 7 shows this comparison  standard linear system solver is on the left, deconvolutions with spectrally factorized coefficients are on the right. The bottom figures were created with a larger time steps than the top ones. For the same time step size, the methods produce similar images. Both of the bottom images show greater dispersion than the top ones. This suggests that the dispersion (visible also in Figure 6) is an intrinsic property of the finitedifference weights derived by the formulation in Equation 9, and is not the result of the spectral factorization method.
Propagation with a linear equation solver is time consuming, which is why the wavefields in Figure 7 are rather small (only 100x100 elements). To test whether this method is stable over longer periods, I compared explicit wavefield propagation to the helical implicit propagation. These are shown in Figure 8. The figure is after 4 seconds of propagation, with a time step almost equal to the stability limit of the 2nd order explicit finitedifference scheme. The explicit propagation exhibits less dispersion. However, it diverges for greater time step sizes, whereas the implicit solution does not. The dispersion problem however does get worse for the implicit solver as the time step increases.
Subsections



 Implicit finite difference in timespace domain with the helix transform  

Next: Nonseparability of velocity from
Up: Barak: Implicit helical finite
Previous: Validity tests of derived
20100519