Broadband 1.2- and 2.4-mm Gallium Nitride (GaN) Power Amplifier Designs

The US Army Research Laboratory (ARL) has been working with Raytheon to design efficient, broadband, linear, high-power amplifiers and robust, broadband, low-noise amplifiers for future adaptive, multimodal radar systems. Raytheon has a high-performance, W-band, gallium nitride (GaN) fabrication process and a process design kit (PDK) that ARL used to design low-noise amplifiers, power amplifiers, and other circuits for future radar, communications, and sensor systems. After the first set of ARL and Raytheon designs was submitted for fabrication, test designs of broadband Class A/B power amplifiers were developed. While these designs did not get fabricated in the initial effort, they serve to demonstrate the performance, bandwidth, and capability of this GaN process and could potentially be fabricated in the future.

Broadband Power Amplifier

The preliminary design of a single, high-electron mobility transistor (HEMT) and 2-way parallel combined HEMT power amplifier was performed. These initial broadband power amplifiers are based on a 12 × 100μm HEMT at a nominal recommended DC bias. This size HEMT had an optimal match provided by Raytheon as “RLoad” ohms in parallel with a negative “CDS” pF in capacitance. Since a negative reactance can only be matched over a limited band, an initial design was performed of an ideal, double, tuned, Q bandpass match for broadband operation centered around 4.5 GHz, with a goal of achieving 2 to 10 GHz. A schematic of the ideal load as a resistor in parallel with a capacitor and the ideal, double, tuned output matching circuit is shown in Figure 1.

Figure 1. Microwave Office (MWO) schematic for the ideal power load and match (12 × 100-μm HEMT-nom-inal DC bias).

After designing a lumped-element output match, the capacitors and inductors were replaced with monolithic microwave integrated circuit (MMIC) elements from Raytheon and retuned to achieve a broadband match. Then microstrip bends, tees, and decoupling elements for the DC bias were added to complete a layout of the MMIC output match. A simulation of the output match showed a better than 20-dB return loss from 2.3 GHz to above 8.7 GHz versus the ideal lumped-element, double tuned match with slightly less bandwidth but an excellent match midband.

Comparing the impedance match of the ideal lumped-element output match to the lossy MMIC output match over frequency to the ideal RLoad normalized impedance and negative “CDS” normalized capacitance showed good broadband performance. The ideal output match is close to the ideal RLoad load line of a 1.2-mm HEMT from 3.5 to 6 GHz, while the MMIC output match undershoots the real part of the impedance but stays very close to 95% of RLoad over a broader range of 3 to 7 GHz. Since an ideal reactance equivalent to a negative CDS capacitance can only be maintained over a finite bandwidth, the output matching circuits match well over the band, diverging at the low end of the frequency range (2 to 3 GHz).

After designing the output match for the broadband power amplifier, the S-parameters of the 12 × 100μm (1.2-mm) HEMT are generated at the nominal DC bias. Initially, these S-parameters were exported from Advanced Design System (ADS) and imported into Microwave Office (MWO) to perform an initial amplifier design. Small-signal stability was analyzed and established with a shunt resistor and a parallel series resistor and capacitor on the gate of the HEMT.

After stabilizing the 1.2-mm HEMT, the input impedance at midband (4.5 GHz) was simulated, resulting in a higher Q matching impedance (Q = 2.4) than the output, making it more difficult to broadband match the power amplifier input. An initial ideal input match provided better than 10-dB return loss from 3.5 to 6 GHz but was limiting the amplifier bandwidth compared to the output matching circuit. An ideal, coupled line provided a broader frequency range for the input match, while sacrificing additional loss.

Figure 2. Unfinished layout of the broadband (2- to 8-GHz) 1.2-mm GaN HEMT power amplifier.

A preliminary ideal transmission line input matching circuit provided good performance from 2 to 7 GHz. The ideal input matching elements were replaced with MMIC components resulting in two relatively large inductors. Next, the folded, spiral, coupled line required electromagnetic (EM) simulation to verify its performance. A pseudo layout of the full one-stage, 1.2-mm, 2- to 8-GHz power amplifier is shown in Figure 2; note the large area required for the broadband input match.

Performing Simulations

With a preliminary layout and MWO simulations for a stable, broadband power amplifier from 2 to 8 GHz based on a 1.2-mm GaN HEMT, the next step was to perform nonlinear simulations using the design kit and ADS. The nonlinear HEMT model within the ADS Raytheon design kit is needed to do performance simulations. MWO schematics for the MMIC input and output circuits were translated into ADS schematics. Ideal bias tees were added to provide the DC bias as a convenience to simulating the ADS schematics, though the matching circuits have the appropriate components for DC and RF decoupling. This power amplifier design would still need design rule checks (DRCs), layout versus schematic (LVS), and final EM simulations.

Figure 3. ADS dynamic load line simulation of the broadband (4.5-Ghz) 1.2-mm HEMT power amplifier.

A dynamic load line simulation at the center frequency of 4.5 GHz for the one-stage power amplifier at nominal DC bias is shown in Figure 3. As an additional verification, ADS was used to repeat the small-signal S-parameter simulations but with the nonlinear HEMT model at the nominal DC bias. The gain seems a little higher than the previous simulations in MWO but the return loss and gain with frequency has a similar shape, as expected.

To get a measure of the losses due to the physical MMIC output, input, and matching circuits, an ADS schematic of the power amplifier using the original lossless element input and output matching circuits was simulated. Output power and efficiency is slightly higher in comparison for the broadband, 1.2-mm HEMT power amplifier with lossless matching elements.

In addition to the 1.2-mm broadband power amplifier, a 2.4-mm power amplifier was implemented using two parallel combined 1.2-mm HEMTs. First, the ideal output match for a single 1.2-mm HEMT was transformed from a 50-Ω output match to 100 Ω so that two devices could be easily paralleled into a 50-Ω load. This simple lossless combiner circuit would need to be modified to supply DC bias, and there are a several easy ways to modify it. The 2-way combiner output matching circuit has the same broadband return loss, with a better than 20-dB return loss match to the ideal load from 2.7 to 7.6 GHz.

ADS was used to simulate the performance of the broadband power amplifier as a 2-way combined (2.4 mm) HEMT power amplifier using the ideal output matching circuit. The input of the 2-way combined amplifier was simulated as two of the coupled line ideal input matching circuits into a 25-Ω source. Output power would be expected to double (+3 dB), with similar efficiency and bandwidth in comparison to the single 1.2-mm HEMT power amplifier. Performance simulation at the frequency of 4.5 GHz, with output power equal to that expected and PAE of 55% for a lossless matched broadband, 2.4-mm HEMT single-stage power amplifier was performed.

Losses for the MMIC output match were calculated to be a reasonable 0.3 dB over most of the band, with up to a 0.5-dB loss at the low end of the band — 2.5 to 3 GHz. Additional losses on the MMIC input match would similarly affect small signal gain and PAE.

The performance data were typically 3 to 4 dB compressed for the Class A/B, biased power amplifier. For the ideal 2.4mm power amplifier, the input power level is 3 dB higher, corresponding to a 3-dB higher output power, with the same large signal gain as the ideal 1.2-mm power amplifier. Nominal performance for the MMIC 1.2-mm HEMT amplifier was within 85% (0.6 dB) of expected output power with 50% PAE at 4.5 GHz. In comparison, the ideal version of the 1.2mm power amplifier was 100% (0 dB) of expected output power with 57% PAE. As expected, the 2-way combined ideal amplifier has double the output power with similar bandwidth and efficiency, showing double the power of a single 1.2-mm HEMT with 55% PAE at a comparable gain compression level.

Conclusion

A preliminary design of a broadband, 1.2-mm HEMT power amplifier and a 2.4-mm HEMT power amplifier using Raytheon's GaN process was performed. The intent was to explore the bandwidth and performance of a Class A/B biased 1.2-mm HEMT power amplifier designed to maximize bandwidth, output power, and PAE over the 2- to 8 GHz band. Trying to increase the band to 2 to 10 GHz would certainly require more matching losses to extend the bandwidth.

A similar 2-way combined, 2.4-mm HEMT power amplifier should achieve comparable performance based on a preliminary design using ideal, lossless matching elements. For the one-stage, 1.2-mm HEMT design, a preliminary layout was implemented, including EM simulations of critical elements such as the folded coupled line for the broadband input match.

These designs illustrate broadband, Class A/B power amplifiers using a 1.2-mm HEMT cell, which should provide good efficiency with matching network losses within 0.6 dB of ideal at these frequencies at the recommended DC bias. To get these designs ready for fabrication would require additional steps to perform full EM simulations, simulate process variation effects, and perform normalized determinant function stability analyses. The Raytheon process is very capable for high-power RF amplifiers and robust low-noise amplifiers for receivers.

This article was written by John E. Penn of the Sensors and Electron Devices Directorate at the Army Research Laboratory, Adelphi, MD. For more information, visit ARL here .