next up previous print clean
Next: SUMMARY Up: Jedlicka: Cascaded normal moveout Previous: PROOF FOR 15-DEGREE PARTIAL

RATE OF CONVERGENCE

The rate of convergence may be estimated from equation (31). It can be rewritten
\begin{displaymath}
\vert T_N-\tilde{T}_N \vert < {x^4 \over 8Nv^4 T^2_N (T_N+\tilde{T}_N)}.\end{displaymath} (32)
Because
\begin{displaymath}
T_N+\tilde{T}_N \gt 2T_N\end{displaymath} (33)
we can obtain this estimate of the number of iterations:  
 \begin{displaymath}
N \gt {x^4/v^4 \over 8 (t^2-x^2/v^2)^{3/2}\epsilon},\end{displaymath} (34)
where $\epsilon$ is the required precision.

The number of iterations estimated from equation (34) is higher than the real number of iterations needed to achieve the desired precision. For a low number of iterations (up to about 10), inequality (34) gives a good estimate of the actual number of iterations. Curves of the numbers of iterations required to achieve particular precisions are shown on Figure [*]c. These curves are not computed from equation (34), but they are the numbers of iterations really needed to achieve the required precision.


next up previous print clean
Next: SUMMARY Up: Jedlicka: Cascaded normal moveout Previous: PROOF FOR 15-DEGREE PARTIAL
Stanford Exploration Project
1/13/1998