PhD Defense: "Characterization, modeling, and design of N-polar GaN devices for mm-wave power"

Matthew Guidry

May 26th (Friday), 10:30am
Phelps Hall, Room 1410 (TMP Executive Learning Center)

The millimeter-wave frequency spectrum is in increasing demand for wireless applications including communication and radar. High transmitter powers are needed to increase range and data rates, but conventional transistor technologies have been limited in output power density. Nitrogen-polar GaN devices developed at UCSB have now shown transformative performance with record-breaking power density and efficiency in the W-band (75-110 GHz) frequency range. The work presented here has assisted these device developments through new characterizing and modeling methods which have been used to understand and improve device design, and in the future will be applied to circuit design.

On-wafer loadpull is invaluable for guiding device development because it provides a rapid and unambiguous evaluation of device performance. All published work and vendor experience at this frequency had been for low-power device loadpull (GaAs, SiGe, Si). Different tradeoffs exist for high-power GaN loadpull which were necessary to analyze before committing large monetary resources to system construction. A new analysis of the multivariable tradeoffs between loadpull system design and semiconductor device design will be shown which quantitatively evaluates a loadpull system’s utility for a particular device, replacing inaccurate rules of thumb. This was used for successful design and construction of UCSB’s W-band loadpull system which is the highest power system of its kind in the literature.

While loadpull is invaluable it gives little insight into the effect of specific device parameters on overall performance. Equivalent circuit models can fill that gap, and guide device improvement and optimization. Achieving physically meaningful linear models is difficult for small-area high-speed devices. Some conventional methods in the literature cannot be applied, which is discussed. The methods used to obtain excellent model fits over the full 0.25 – 67 GHz measurement range are presented including extrinsic parasitics, real parts of intrinsic feedback, and substrate modes. This model was extrapolated to 94 GHz and used to analyze all of the factors limiting the gain, and thus efficiency. Several methods were proposed to improve the gain (and efficiency), which were successfully employed in subsequent device runs. In conjunction to intrinsic device improvements made by the device team this helped lead to record GaN W-band efficiency of 34% at 88 GHz and 28% at 94 GHz.

Finally an initial evaluation of large signal modeling methods applied to N-polar GaN devices will be presented. These models are important for aiding circuit designs in constructing power amplifiers or other circuits. For the most part a conventional GaN model can be employed, but areas where conventional models can be improved for application to N-polar devices will be presented.

About Matthew Guidry:

Matthew Guidry obtained his B.S. in Electrical Engineering from UCSB in 2008. He joined UCSB’s MS/PhD program in Electrical Engineering later that year where he was initially co-advised by Professors Patrick Yue and Robert York. During that time he has also worked at Toyon Research Corporation and HRL Laboratories with projects focused on RF and microwave circuits, sensors, and solid state device development and characterization. In the summer of 2012 he joined Professor Umesh Mishra’s research group to work on characterization and modeling of N-polar GaN devices and to manage Professor Mishra’s and York’s measurement laboratory. His research interests include characterization, metrology, and device modeling and their application to novel devices, and also development of new semiconductor devices.

Hosted by: Umesh Mishra