|Ph.D Student||Wainstein Nicolas|
|Subject||Reconfigurable Radio Frequency Circuits|
based on Emerging Memory Technologies
|Department||Department of Electrical and Computer Engineering||Supervisors||ASSOCIATE PROF. Shahar Kvatinsky|
|DR. Eilam Yalon|
|Full Thesis text|
Adaptive radiofrequency integrated circuits (RFIC) have become critical in wireless communication systems to effectively use the RF spectrum and to sustain the always-increasing performance demand of modern mobile devices. Currently, more radios are being added to smartphones to fulfill the incremental frequency bands and data rates of new standards (e.g., 5G and millimeter wave (mmW), and eventually 6G and THz communications). Adaptive RFIC allow for redundant hardware removal, which results in smaller area, lower complexity, and reduction in the number of modules in radio chips. The core of these systems is the RF switch, controlling the flow of the RF signal and providing tunable capabilities to different blocks. Recently, resistive memory technologies such as resistive random-access memory (RRAM), phase change memory (PCM) and conductive-bridge RAM (CBRAM), have emerged as great candidates for RF switches due to their superior performance, small footprint, non-volatility and back-end-of-line compatibility. They achieve state-of-the-art cutoff frequency (fCO), a figure-of-merit of RF switches, making them great contenders to the best-in-class RF switches (FETs, pHEMT, and MEMS).
This research explores the novel capabilities of resistive memories for adaptive RF circuits. This research is multidisciplinary and covers a fundamental study in device, circuit, and architecture levels to achieve our vision of reconfigurable RFICs. At the device level, I have developed the fabrication process of inline phase-change RF switches (IPCS) using germanium telluride (GeTe) as the phase-change material (PCM). These devices achieve state-of-the-art fCO (>5 THz), while reducing the area and switching energy of RF switches. As part of this work, I have developed the first physics-based compact model of the IPCS, experimentally validated by 1) nanosecond electrical thermometry of the heater during device operation, 2) finite element method simulations, and 3) RF measurements. The model enables rapid and accurate (>92%) device optimization, and the design and simulation of large-scale circuits composed of IPCS with varying substrates, dielectrics, and PCM. Furthermore, this dissertation presents the use of the IPCS device structure as a platform for electro-thermal characterization of phase transition properties of different PCM in nanoscale films at nanosecond temporal resolution. With this platform, the phase transition, as well as temperature-dependent properties up to temperatures of ~1100 K, can be measured during heating pulse application with nanosecond resolution. Additionally, RF modeling of nanoscale CBRAM RF switches and RRAM-based RF switches is carried out in this research. These models predict the device behavior with RMS error of 10% on average.
At the circuit level, I have explored the use of CBRAM for tunable inductors which show a tunability ratio of 300% with no area overhead. The fabrication process of tunable inductors based on the IPCS is currently under development. The proposed tunable inductor is used in a multi-band low-noise amplifier, which achieves a noise figure below 2.3 dB. Furthermore, I have worked on the design and fabrication of several IPCS-based reconfigurable RF front-end blocks (e.g., multiplexers and phase-shifters), for the future implementation of a multi-standard radio transceiver. In the system level, this research explored the design spectrum to find the best system topology for a resistive memory-based reconfigurable radio. It is envisioned that the small size and low power consumption of these devices will portray a major advantage in sub-6 GHz systems. For high-frequency systems (5G, radar, and 6G), the low-loss of resistive memory-based RF switches will allow for adaptable beamforming systems and in general, the resistive memory-based RF switches will eventually enable the long-desired FPGA RF systems.