Convergence
As mentioned in Briggs (1974), and strongly recommended in Smith and Wessel (1990), Surfer uses a "multiple lattice strategy." It starts with a coarse grid and then incrementally refines the grid until the final density is achieved.
The relaxation approach is a local smoothing process and, consequently, short-wavelength components of Z are found quickly. On the other hand, the relaxation process does not propagate the effects of the data constraints to longer wavelengths efficiently.
As recommended by Briggs (1974) and Smith and Wessel (1990), this routine determines convergence by comparing the largest magnitude nodal change in one iteration to the specified tolerance (Maximum Residual).
The status of the algorithm is reflected on the status line. For example: pass 2 of 4, iteration 360 (0.1234 > 0.08)
This says that there are four levels of grids considered (the fourth is the final grid), and the algorithm is currently working on the second. The algorithm is currently on iteration 360. If the iteration number exceeds the Maximum Iterations parameter, the algorithm terminates without generating the grid and provides a failure-to-converge warning. The numbers in the parentheses are the current largest residual and the largest allowed residual. The largest allowed residual equals the Maximum Residual parameter on the final pass.
The Minimum Curvature method requires at least four data points.
Minimum Curvature References
Briggs, I. C. (1974), Machine Contouring Using Minimum Curvature, Geophysics, v. 39, n. 1, p. 39-48.
Press, W.H., , Numerical Recipes in C , Cambridge University Press.
Smith, W. H.