for |

Thus the current density component tangential to the boundary is uncoupled from the other dependent variables. It is convenient to assume, then, that is initially zero, so that it will remain so permanently. In this case, we can drop the series junctions corresponding to on the southern boundary from the network. Otherwise, we may allow the junctions to remain, as lumped damped (by a factor ) elements, still uncoupled from the rest of the network. In either case, the parallel junctions at the grey points in Figure 4.23(a) may be short-circuited as in the (1+1)D case in order to realize boundary condition (4.68). The waveguide mesh termination corresponding to (4.68) is shown in Figure 4.24(a).

To deal with the boundary condition
, we may proceed as in the (1+1)D case, and write down a difference approximation to (4.58c), where we use the one-sided difference approximation

and centered differences in the time and directions,

where and are as given in (4.60).

Here, the voltages on the boundary are related to the tangential currents, which, from Figure 4.23(a), are also calculated on the boundary. This implies that the corresponding junctions will be connected to one another by waveguides which lie directly on the boundary. Also notice the doubled weighting of the grid function at the boundary; this requires special care in the DWN implementation, though it also follows from a structurally passive termination, provided we make use of *transformers* along the boundary waveguides. Though we have not discussed transformers in the DWN context [166], they are identical to wave digital transformers, which were mentioned in §2.3.4. In effect, we may introduce multiplies by and , for any real in the two signal paths of any waveguide without affecting losslessness, provide we scale impedances at both ends of the waveguide accordingly. The DWN termination corresponding to at a southern boundary is shown in Figure 4.24(b). We have used transformers of turns ratio 2, implying that the immittances on the boundary satisfy

(4.82) |

The corresponding difference equation at a parallel boundary junction is then

where

(4.83) |

The junction updating will be equavlent to the centered difference scheme if we choose

(4.84) |

Given that the northward immittances at the boundary junctions must be set as interior values, the settings for the remaining immittances for the type I and II meshes discussed in §4.4.2 will be

Type I:

Type II:

The passivity conditions which follow from the positivity of the boundary self-loop immittances will be

Type I | |||

Type II |

and clearly are