Analysis of Process Variations in W-Band GaN MMIC PAs Using Nonparametric Statistics

This article presents an analysis of the effects of amplifier process-related performance variation on the radiation patterns of a ${W}$ -band active transmit antenna array. Based on measured on-wafer $S$ -parameters for about 80 amplifier chips designed in an experimental 90-nm gallium nitride (GaN) on SiC process, a statistical analysis of the array patterns is performed. Several types of probability density functions (pdfs) are generated from measured data and compared to determine which statistical approach is most relevant. We find that a joint distribution that maintains the correlation structure between $|S_{21}|$ and $\angle S_{21}$ is important for an accurate analysis. The monolithic microwave integrated circuits (MMICs) are connected to the waveguide-fed horn antenna elements in the array via microstrip-to-waveguide transitions fabricated on alumina. The transitions are analyzed in full-wave simulations with fabrication tolerances included in the analysis. Finally, the MMIC and transition statistical variations are cascaded, resulting in a quantitative evaluation of spatial power combining. Given a random choice of power amplifier chips in a $4\times 4$ array, the EIRP is shown to vary by ±3 dB at 94 GHz.