The US Army Combat Capabilities Development Command Army Research Laboratory (ARL) has been evaluating and designing efficient broadband high-power amplifiers for use in sensors, communications, networking, and electronic warfare (EW). ARL submitted designs of Ka-band low-noise amplifiers (LNAs), power amplifiers (PAs), and transmit/receive (T/R) switches using Qorvo Inc.’s high-performance 0.15-μm gallium nitride (GaN) fabrication process. These amplifiers were fabricated as one- and two-stage designs, as well as integrated T/R modules for bidirectional transceivers as part of a recent ARL Qorvo Prototype Wafer Option (PWO), which yields many different designs from two full 4-inch GaN wafers. This research documents testing and analysis of these designs, as well as lessons learned for improvements to future design efforts.
The key component for a Ka-band transceiver is the LNA, which, when implemented in GaN, has the added advantages of high dynamic range and robust survivability to high-power interference signals. These LNAs were designed with a goal of several gigahertz bandwidth centered around 28 GHz.
Various matching topologies, stabilizing approaches, and tradeoffs of gain versus noise figure were explored for two high-electron-mobility transistor (HEMT) sizes, using the limited devices in the process design kit (PDK) that had noise data, for 5V or 10V biases. These designs were intended for 10V operation, so while they work over biases of 5V to 28V, the targeted optimal performance is at 10V with a typical 100-mA/mm drain current.
The LNAs were designed as two-stage amplifiers, with the first stage optimized for low noise figure. Even though the first stages were not designed for optimal use as a standalone amplifier, they were fabricated as test circuits for testing and analysis of the two-stage LNAs. There were two designs based on a 4- × 25-μm and a 6- × 25-μm HEMT, each trading off stability, noise figure, return loss, and gain. Initially, the larger 6- × 25-μm LNA design seemed a narrower band stable design compared with a potentially broader band gain with the smaller HEMT size but with a riskier tradeoff of stability versus stability.
Figure 1 shows a plot of measurements (solid) versus simulations (dash) of the small signal s-parameters of the first-stage 6- × 25-μm LNA, at the nominal 10V DC bias. While the shapes are similar, the actual gain is higher but slightly narrower band. A similar comparison plot is shown in Figure 2 for the same LNA measured at 5V. The shift to a lower frequency, both simulated and measured, is noted at the lower 5V bias. Recall that the design was intended for 28 GHz at 10V operation but could be used at the lower 5V operation for a slightly lower frequency band operation, or conversely, over slightly higher frequency bands for DC voltages higher than 10V.
An electromagnetic (EM) resimulation of the full one-stage LNA layout was repeated to eliminate the possibility of unsimulated parasitic interaction among the input match, source inductance, and output match of this very compact layout. But the full EM layout result was similar to the original simulation where those three EM layouts were independent sections. The higher gain peak could be explained by lower-than-expected source inductance, or could be due to typical process variation in fabrication.
This work was done by John E Penn and Ali Darwish for the Army Research Laboratory.For more information, download the Technical Support Package (free white paper) below. ARL-0240
This Brief includes a Technical Support Package (TSP).
Ka-Band Front-End Monolithic Microwave Integrated Circuits (MMICs) and Transmit/Receive (T/R) Modules Testing
(reference ARL-0240) is currently available for download from the TSP library.
Don't have an account? Sign up here.