As already discussed above, by using the standard single-mode AC a general complex-valued amplitude can be applied to an arbitrary number of terminals of the device. However, under these circumstances the calculation and assembly of the complete admittance matrix is a cumbersome task. For that reason the simulator has been extended to provide a feature for the automatic calculation of the complete matrix.
The intrinsic scattering matrix can be calculated by the following analytical formula [244]:
(3.20) |
(3.21) | |
(3.22) |
(3.23) |
(3.24) | |
(3.25) | |
(3.26) | |
(3.27) |
The capacitances are calculated according to the charge-based capacitance model [9]: the terminal charges are in general a function of the terminal voltages. Each terminal has a capacitance with respect to the remaining terminals. For that reason a four terminal device has 16 capacitances.
All capacitances form the so-called indefinite admittance matrix. Each element of this matrix describes the dependence of the charge at the terminal with respect to the voltage applied at the terminal with all other voltages held constant. In general,
(3.28) |
The signs are chosen to keep the capacitances positive for all devices for which the node charge is directly proportional to the voltage at the same node, but indirectly proportional to the voltage of any other node. Thus, for a four terminal MOS transistor the 16 capacitances of the matrix are defined as follows:
(3.29) |
Each row and column sum must be zero in order to fulfill Kirchhoff's laws. For that reason, the capacitances are not independent from each other, but one of the four can be calculated with the remaining three. The gate capacitance is therefore:
(3.30) |
The extended small-signal features have been evaluated by a comparison with results of the commercial simulator DESSIS [111]. The structure, which has been designed with the program MDRAW, was converted by using ISE2PIF and is shown in Figure 3.11. For MHz, , and the simulator calculates the admittance matrix shown in Table 3.1.
The last row of Table 3.1 contains the column sums, and the last column the row sums. Due to numerical reasons, zero can hardly be obtained, but the error is significantly lower than the entries in the matrix. After the calculation of the steady-state operating point, the solution of the complex-valued linear equation system requires 4.2s on a 2.4GHz Intel Pentium IV with 1GB memory running under Suse Linux 8.2. The dimensions of the complete and inner equation system are 6,610 and 4,805, respectively. In Figure 3.12, the gate drain capacitance as calculated by MINIMOS-NT is compared with results of DESSIS. Note, that the sign of the DESSIS result had to be inverted.
|