Next: Higher-order Accuracy Revisited Up: Incorporating the DWN into Previous: Alternative MDKC for the

## Alternative MDKC for (2+1)D Parallel-plate System

The same idea extends simply to the (2+1)D parallel-plate equations (3.67). In this case, we again look at the lossless source-free system, scaled by a factor of , and can develop an MDKC along the lines of Figure 4.47(b), with now treated as a voltage instead of a current as in Figure 3.17. We remind the reader that we use the notation , in circuit diagrams to indicate directional derivatives in the directions defined by the coordinate transformation (3.22). Also, we have written and for and , in keeping with the WDF literature [62]. The alternative MDKC is shown in Figure 4.49, and the element values are given by

is, as before, a free parameter which can be scaled to achieve an optimal space step/time step ratio.

The two two-ports with terminals , , , and , , , are both linear and shift-invariant, and their right-hand pairs of terminals are both connected in parallel with the capacitor. The hybrid matrices for these two-ports are

and under the spectral mappings

with , we again have a decomposition of the discrete hybrid two-ports into pairs of multidimensional unit elements (impedance ) connected in series/parallel, i.e.

The MDWD network, corresponding to checkerboard''-type sampling, is as shown in Figure 4.50, and where are the shifts in the coordinate directions , and the port resistances are given by

For the one-port inductors and the capacitor, we have again applied the trapezoid rule, with a step-size of (implying a delay of , written as in Figure 4.50).

When spatial dependence is expanded out, the signal flow graph is identical to the interleaved type III DWN for the parallel-plate equations shown in Figure 4.21, without the loss/source ports. As in (1+1)D, loss and source elements can be reintroduced into the MDKC of Figure 4.49 without difficulty.

It should also be possible to derive DWNs from MDKCs for the same system on alternative grids, such as the hexagonal and triangular grids mentioned in §4.6.1, though we have not investigated this in any detail. In this case, one would presumably begin from the MDKC under the appropriate coordinate transformation. One of these coordinate transformations (which generates a hexagonal grid under uniform sampling in the new coordinates), was discussed briefly in §3.3.3. The full MDKC for the parallel-plate equations in these coordinates is given in [62].

Next: Higher-order Accuracy Revisited Up: Incorporating the DWN into Previous: Alternative MDKC for the
Stefan Bilbao 2002-01-22