7.1.2 Sensitivity Study: Pseudomorphic HEMT on GaAs at 77 GHz

Especially for W-band applications at 77 GHz, the understanding of the process variations is desirable as a function of as many parameters as possible due to the gate length ${\it l}_{\mathrm{g}}$ required. The absolute changes of transistors are relatively more important due to the scaled geometries and frequencies than for larger gate-lengths necessary e.g. for 2 GHz applications. This is especially true for the small- and large-signal performance. As was shown in Fig. 7.2, the output characteristics including self-heating gives the necessary agreement to allow for such a study. The ultimate verification of a HEMT simulation model assuming a specific equivalent circuit such as given in Fig. 4.1, is the comparison with bias dependent S-parameter measurements. Except from thermal and frequency dispersive effects, the bias dependence of the S-parameters describes large-signal operation completely. Simulation allows to extract the physical sources of the bias dependence of the extracted circuit elements. Fig. 7.7 and Fig. 7.8 show simulated and measured S-parameters for two different biases and temperatures.

Figure 7.7: Simulated (-) and measured (+) S-parameters at $T_L$= 373 K and $V_{DS}$= 1.5 V and $V_{GS}$ for $g_{mmax}$.

Figure 7.8: Simulated (-) and measured (+) S-parameters at $T_L$= 300 K and $V_{DS}$= 2 V and $V_{GS}$= -0.5 V.

Fig. 7.7 presents S-parameters between 0.5 GHz and 50 GHz for ${\it V}_{\mathrm{GS}}$ for ${\it g}_{\mathrm{m,max}}$ and ${\it V}_{\mathrm{DS}}$= 1.5 V. An overall good agreement is found between simulation and measurements at ${\it T}_{\mathrm{sub}}$= 373 K. The agreement is similar to the agreement presented in Fig. 4.5 for ${\it T}_{\mathrm{sub}}$= 300 K. Fig. 7.8 shows the simulated and measured S-parameters for the transistor at pinch-off for ${\it V}_{\mathrm{DS}}$= 2 V and ${\it V}_{\mathrm{GS}}$= -0.5 V, measured and simulated between 1 GHz and 120 GHz for ${\it T}_{\mathrm{sub}}$= 300 K. Also the behavior of the pinched-off device is predicted correctly. $S_{\mathrm {21}}$ and $S_{\mathrm {21}}$ merge to the symmetrical passive behavior. For $S_{\mathrm {11}}$ and $S_{\mathrm {22}}$ an asymmetry is seen in the simulation, which is due to a slight deviation of the simulated and measured currents for subthreshold bias.