Design and optimisation of THz and mm-wave components for communications

Abstract

This thesis proposes millimetre-wave (mm-wave) and terahertz (THz) components that can be used in highly integrated systems. These components have been designed for low cost and ease of manufacture using traditional manufacturing techniques, such as printed circuit board (PCB) printing and microlithography. The proposed frequency regions pose unique challenges, including high water-vapor attenuation [1] and the need for micro-scale components. This can limit the achievable power transfer, thereby lowering the signal-to-noise ratio (SNR) associated with components. Components have therefore been highly optimised for power efficiency to compensate for this.
A genetic algortihm (GA), Agenoria, written and revised by the author throughout this work, is presented that can optimise a variety of passive mm-wave and THz components. This uses pre-written FDTD software, Lucifer, which was previously designed in-house, to simulate component performance. A GA is chosen, as compared to other numerical methods, they are more likely to scale well to THz frequencies due to being able to optimise within many constraints at once. The challenge of manufacturing at the micrometre scale can be avoided by specifying constraints such as minimum feature size.
Agenoria is tested on a series of planar patch antennas with varying frequency responses. The first is a broadband antenna, optimised between 100 GHz and 5.0 THz, where 79% of the spectrum shows some improvement. The second antenna is optimised for a single peak at 1.0 THz. The evolved antenna’s simulated power output is considerably higher than that of the reference case, with an improvement at 1 THz of 267%. The third antenna is a dual peak antenna, with peaks optimised at 1.0 THz and 1.2 THz showing a 4,510% and 250% improvement on the reference case from the first generation.
One candidate antenna design, the broadband antenna, was then fabricated and experimentally verified, both in-house by the author. Due to the extremely broadband nature of the antenna, it was not possible to validate the simulated performance across the entire frequency range. Therefore, the verification was completed at a frequency band of 0.8 THz to 1.0 THz, within the optimised range. A vector network analyser (VNA) with two bidirectional VDI WR 1.0 frequency extenders with high-gain horn antennas operating between 0.75 THz and 1.10 THz was used for testing. The antenna transmission and angular dependence was measured and found to concur with simulated results for both the evolved antennas and a reference plain patch antenna. Feedline losses were investigated using the experimental set-up shown in Figure 61. The calculated loss of the microstrip line, connecting the antenna pairs, is 0.79 ± 0.06 dB/mm, which shows promisingly low losses. The combination of these antenna measurements verifies the simulation output within the region 0.8 THz to 1.0 THz. This provides strong evidence that the simulations would match experimental results across the entire simulated range.
In addition, a simple modulation platform for THz communications has been developed to be used in conjunction with WR1.0-VNAX frequency extenders. This consists of an on-off switching board, designed and simulated for a 12.3 GHz carrier wave, corresponding to the RF input of the extender modules, and a power detector board, designed for the LO output of the extenders. These components were designed using traditional techniques, such as transmission line theory, and optimised using Keysight Advanced Design System (ADS) Method of Moments (Momentum) electromagnetic simulator.
The power detector was successfully designed and fabricated in house. It was characterised, and a clear peak seen between 310 and 320 MHz, of 1.38 V at 314 MHz. This PCB was then used in the characterisation of the on-off keying capabilities of the VNA and extender heads. The switch was found to work best at 10 kHz but can modulate the VNA 12.3 GHz input at frequencies between 1 Hz and 1 MHz.
A series of frequency multipliers and mixers were also produced. The frequency multipliers were designed and simulated in the 10 to 30 GHz frequency range. Simulations consistently showed a strong third harmonic while others were successfully filtered by on-chip microstrip filters. However, the preliminary experimental results show a stronger effect of the filter output due to tolerances than had been predicted.
The frequency mixers were designed for two consecutive frequency ranges. The low frequency mixers, combining 2 GHz and 1.685 GHz signals, had a measured output power of -42.3 dBm at 0.315 GHz, where the mixer is optimised for. A mid-frequency mixer, designed to combine frequencies of 20 GHz and 18 GHz to produce a 2 GHz output signal, produced an output power of -45.8 dBm at the optimised frequency.