InP HEMTs were obtained from the Microwave and Lidar Technology Section at the Jet Propulsion Laboratory, and their scattering parameters were measured on a microwave probe station with a network analyzer. The SuperMix optimizer was used to fit Marian Pospieszalski's FET model to the measured data to determine the model parameters. The noise prediction from the model, given a best-guess value of the noisy drain-resistor temperature, was used without any direct measurement of the transistor noise correlation matrix.
This model was then used with SuperMix to simulate and evaluate different tuning circuits for the amplifier. The SuperMix optimizer was used to optimize the designs for flat gain above 30 dB, low noise, input reflections below about -10 dB, output reflections below -20 dB, and unconditional stability at all frequencies. During this initial design phase, efforts were made to keep total power consumption low by minimizing resistive elements in the drain current path.
The final designs chosen for layout use inductive tuning at the transistor gates, some source inductance for stability, gates biased through large resistors (large enough to not contribute significant noise) and resistively biased drains.
After the ideal circuit designs were completed, physical resistors, capacitors, and inductors were laid out. The substrate includes air bridges and a resistive layer, but for simplicity does not include via holes. Each component, including five spiral inductors, was simulated with Agilent's EM simulator, Momentum. Results of the EM simulations were saved as scattering parameter files, which were then imported back into SuperMix for the final simulation. At this point, the predicted results including field simulations differed enough from the original design that the amplifier was re-optimized, and the component layouts tweaked accordingly.
The predicted gain and noise of the final design are plotted below.