6.3.1 Output Stage of an OpAmp

A simplified schematic of the $\mu$A709 operational amplifier output stage is shown in Fig. 6.8. Transistor T₃ acts as a common-emitter driver stage for the complementary output devices T₁ and T₂. Problems are caused by the large bias voltages of $\pm$15 V and by the sensitivity of the circuit to the state of transistor T₃. The operating point is calculated for $\varphi_{\mathit{in}}^{}$ = - 14.5 V and from this point $\varphi_{\mathit{in}}^{}$ is stepped until -14.1I>V. Within this interval the internal state of the circuit changes completely as $\varphi_{1}^{}$ moves from $\approx$ 13.6 V down to $\approx$ - 15 V. With $\varphi_{1}^{}$ the states of the two output transistors change.

Figure 6.8: Simplified output stage of the $\mu$A709 operational amplifier.

Figure 6.9: Evolution of the node voltages during DC operating point calculation for the OpAmp output stage with $\kappa$ = 4.

Figure 6.10: Comparison of the iteration counters for a DC transfer characteristic for the OpAmp output stage with $\kappa$ = 4.

Figure 6.11: Evolution of the node voltages during DC operating point calculation for the OpAmp output stage with constant shunt conductance.

The evolution of the node voltages during DC operating point calculation is shown in Fig. 6.9. Best results were obtained with $\kappa$ = 4. With ABC 34 iterations were needed and K^A = 34/30 = 1.13 whereas with DBC 38 iterations were needed and K^D = 38/31 = 1.226.

For the DC transfer characteristic the required number of iterations is shown in Fig. 6.10. In addition $\varphi_{1}^{}$ is shown to represent the internal state of the circuit. In this case DBC is superior compared to ABC.

To demonstrate the importance of G_S^k the evolution of the node voltages for a constant G_S^k = 1/G_min is shown in Fig. 6.11. To obtain convergence at all, the global damping parameter $\delta$ had to be increased by a factor of 100 and 170 iterations were necessary.

Although ABC required less iterations than DBC this type of boundary condition seems to have a negative impact on the condition of the system matrix as each contact voltage depends on the node voltages of all other contacts. Since an iterative solver is used which is very sensitive to the condition of the matrix the total simulation time is unfortunately larger than for DBC.