Solid-state switches can be fabricated with a variety of different semiconductor technologies, usually selected on the basis of performance requirements. For this study, single-pole, single-throw (SPST) switches were evaluated based on three different high-electron-mobility-transistor (HEMT) devices: InP, GaAs, and GaN. Although the least-mature process, the GaN HEMT switches exhibited high breakdown voltages compared to the other two processes, with better power handling (to +15 dBm without degradation in RF isolation and insertion loss).

Such solid-state switches have been used successfully as control devices for transmitter (Tx) and receiver (Rx) switching functions in a variety of digital communications systems. The choice of semiconductor material for these devices is generally dictated by the material's critical breakdown field, its saturation drift velocity, and its thermal conductivity. GaAs- and InP-based HEMT switches offer high isolation, low insertion loss, and high switching speed1 but are limited in power-handling capabilities due to modest critical breakdown fields. On the other hand, due to its wide bandgap structure, AlGaN/GaN HEMTs are ideal for realizing microwave control and amplification functions.2

GaN is a wide bandgap semiconductor material, with a high critical breakdown field, high saturation drift velocity, and good thermal conductivity compared to GaAs and InP (Table 1). If offers more than twice the bandgap energy of InP and GaAs materials, with more than three times the thermal conductivity of GaAs (implying better power-handling capability).

To evaluate the three semiconductor materials, a HEMT device structure was selected as optimum for high-frequency switching and control functions. The resistances and capacitances forming the HEMT equivalent circuit can be subdivided into configurations representing the two switch states (on and off). For most control applications, at frequencies of most interest, the on-state impedance is mostly resistive while the off-state impedance is mostly capacitive. Figure 1(a) shows the physical origin of the various important resistances and capacitances in the conducting state. The resistances present in this state are the channel resistance, Rch, and the parasitic source resistance, Rs and parasitic drain resistance, Rd. The capacitances present in the on-state configuration are the gate capacitance, Cg, the extrinsic drain-gate capacitance, Cdgext, and the extrinsic source-gate capacitance, Csgext. These resistances and capacitances are estimated using the same techniques used in refs. 3 and 4.

The nonconducting-state resistances and capacitances are shown in Fig. 1b. In addition to the resistances present in the conducting-state, there is also a resistance Rmax introduced to characterize the maximal channel resistance beyond the gate-threshold voltage when the 2DEG electron gas is suppressed. In addition to the parasitic capacitances Csgext and Cdgext, the capacitances in the nonconducting state consist of three components. The first component is the intrinsic capacitance, Cigs(gd) which reflects the coupling between the gate and the inner side of the n+ diffusion regions. This capacitance was computed by the relationship of ref. 5.

The second component is the intrinsic capacitance Csdint that couples through the GaAs substrate. The third component is the extrinsic capacitance Csdext that couples through the air, namely the metallic source-to-drain coupling capacitance. Determining the values of these last two elements is based on the relationships of ref. 3. The model used in this article includes an additional capacitance corresponding to the InGaAs, AlGaAs, or AlGaN neutralization.

Finally, the modeled capacitances and resistances can be combined to form an equivalent circuit that represents the two states of the HEMT control device. It is well known that the on-state resistance, RON, and the off-state capacitance, COFF, are the key equivalent-circuit elements used in characterizing broadband HEMT switches. The broadband cutoff frequency figure of merit, fc, can be defined as:

Because RSD and CSD are key equivalent-circuit parameters affecting HEMT device and switch performance over gate voltage and input power, this analysis will focus on these two parameters.

Figures 2 and 3 show computed characteristics of the normalized source-drain resistance and capacitance at a frequency of 1 GHz for three different technologies. A significant dependence of the extracted total source-drain resistance on Vg can be observed. When the gate voltage decreases toward the threshold voltage (−0.1 V for InGaAs/InP, −0.46 V for AlGaAs/GaAs, and −1.04 V for AlGaN/GaN), RSD increases rapidly. This is in good agreement with the results obtained by Caverly5 on modeling the AlGaN/GaN HEMT. In the gate-voltage range greater than the threshold voltage, RSD is rather flat, which is a good evaluation of the on-state resistance, RON. Due to the lower bandgap energy, the InGaAs/InP HEMT offers a lower on-state resistance compared to the other two devices, which leads to improved high-speed switching.

Figure 7 shows simulated source-drain capacitance for three devices. Under off-state biasing conditions, the variation of CSD is dominated by the extrinsic capacitances, which depend upon the FET layout. The GaN-based HEMT presents the lower COFF, because of its low relative dielectric constant. When the on-state HEMT was biased, the total capacitance presented approximately the same variations.

Page Title

Insertion loss and isolation are shown in Figs. 4 and 5 as a function of gate voltage for various technologies, for devices with a gate periphery of 200 µm. For the shunt connected switch, the InGaAs/InP HEMT switch presents a small variation on the level of insertion loss compared to the other technologies, but demonstrated an improvement of isolation of 4 dB than GaAs-based HEMT and 9 dB than the GaN-based HEMT switches due to lower on-state resistance. The converse is true in the series case, where the isolation is approximately constant. The insertion losses of the three technologies are low, due the low on-state resistances of the InGaAs/InP, AlGaAs/GaAs, and AlGaN/GaN HEMT switches.

The power-handling characteristics of each technology in microwave control applications were also investigated. To define the total resistance and the total capacitance in large-signal environment, a simple equivalent circuit of the shunt HEMT switch was used (Fig. 6). For a broadband HEMT switch, power handling is dependent on the gate-bias circuitry. For improved isolation of the gate-control voltage, the gate is usually biased with a large gate-bias resistance, RGG, or a resistor incorporated monolithically within the gate.

Figures 7 and 8 show the variations of RSD and CSD, respectively, with input power for the three types of HEMT switches at 1 GHz. When the on-state HEMT was biased, the on-state resistance increases with increasing input power above +5 dBm for both the InP- and GaAs-based HEMT switches and above +15 dBm for the GaN-based HEMT switch. Figure 8 shows that the on-state capacitance varies with increasing input power above +15 dBm for the GaN-based HEMT switch, and no degradation of this capacitance was observed for inputs to +25 dBm for the two other technologies.

Figures 9 and 10 show the variations of shunt-connected HEMT isolation and series-connected HEMT insertion loss, respectively, with input power for the three technologies at 1 GHz. The effect of input power isolation and insertion loss was evaluated under on-state biasing conditions of Vg = +1.0 VDC). For both the InP- and GaAs-based HEMT switch technologies, there are strong degradations of isolation and insertion loss with increasing level of input power above 0 and +3.5 dBm, respectively. Compared with the AlGaN/GaN HEMT switch, the same phenomenon occurs at power levels above +15 dBm due the wide bandgap of the GaN material. This result shows the feasibility of using AlGaN/GaN HEMTs for high-power microwave control.

Table 2 summarizes cutoff frequencies and breakdown voltages calculated by the relationship from ref. 6. These cutoff-frequency results are based on a gate periphery W = 200 µm and a gate bias voltage of −2 VDC (COFF) and +2 VDC (RON).

REFERENCES

  1. J.L. Cazaux, D. Pavlidis, G.I. Ng, and M. Tutt, "A HEMT monolithic double channel attenuator with broadband characteristics and wide dynamic range," 18th European Microwave Conference, September 1988, pp. 999-1004.
  2. E. Alekseev, Shawn S.H. Hsu, and D. Pavlidis, "Broadband AlGaN/GaN HEMT MMIC Attenuators with High Dynamic Range," the 30th European Microwave Conference.
  3. N. Jain and R. Gutmann, "Modeling and Design of GaAs MESFET control devices for Broad-Band Applications," IEEE Transactions on Microwave Theory & Techniques, Vol. 38, February 1990, pp. 109-117.
  4. N.V. Drozdovski, R.H. Caverly, and M.J. Quinn, "Large-Signal Modeling of Microwave Gallium Nitride-Based HFETs," 2001 AP MC, Taipei, Taiwan, December 3-6, 2001, Proceedings of APMC'01, Vol. .1, pp. 248-251.
  5. R.H. Caverly, N.V. Drozdovski, and M.J. Quinn, "Gallium Nitride-Based Microwave and RF Control Devices," Microwave Journal, Vol. 44, No. 2, February 2001, pp. 112-124.
  6. S.H. Wemple, W.C. Niehaus, H.M. Cox, J.V. Dilorenzo, and W.O. Schlosser, "Control of gate-drain avalanche in GaAs MESFETs," IEEE Transactions on Electron Devices, Vol. ED-27, 1980, pp. 1013-1018.