This chapter began with a overview of the parameters optimised for an aperture coupled microstrip patch antenna. This was followed by a dissertation on the optimisation algorithms employed and the results obtained when optimising the antenna to 50Ω and transistor source reflection coefficient (Γs) at L1 and L2 frequency points.
The Random search, with either Least Squares or Minimax error function, has shown to be the most robust when matching the antenna to 50Ω. A suitable antenna design with a return loss of greater than 10 dB in less than an hour using these optimisers has been achieved. The Genetic optimiser tended to converge to a local minimum in most trials, however the search can evade this saddle point and an excellent result could be obtained. Reasonable results could be obtained using the Least Pth optimiser, however this search needed to be restarted to avoid converging to a local minimum. Searches involving derivative evaluation for direction finding did not perform well, such as the Gradient and Quasi-Newton optimisers. The Discrete optimiser, which implements a direct search technique, also proved to perform badly.
The aperture coupled patch antenna can be designed to have a specific complex impedance. As expected, the results indicate that it is easier to converge to Γs at L1 compared to a 50Ω impedance match at the same frequency. The best searches have been the Random, Least Pth, Discrete and Genetic algorithm. Even though the Discrete search has been successful in obtaining a suitable solution, it is very time consuming compared with the other searches. The work was further developed to investigate the viability of matching this type of antenna to two complex frequency points rather than the conventional 50Ω impedance.
Optimising to Γs at both L1 and L2 posed a considerable challenge, however the results prove that it is possible to match these impedance points using a single layered aperture coupled patch antenna. From the results obtained, it can be clearly seen that the best optimisation algorithms for optimising aperture coupled patch antennas to dual-frequency points are Random searches. The Genetic search, and to a lesser extent the Least Pth, have the potential to perform well when optimising to dual-frequency points. However, both of these algorithms have a tendency to converge to a local minimum and were unable to evade this saddle confinement in most trials. It was necessary to stop the optimisation and perturb the dimensions slightly before continuing. Overall, in dual-frequency optimisation it has been observed that it has been easier to optimise to the L2 frequency point compared with the L1 frequency. The impedance phase has been relatively simple to optimise, however the magnitude proved to be the most difficult. From these results obtained, the Genetic and Least Pth optimisers are best suited for the initial phase in a dual-frequency optimisation when the error function is relatively large. When the optimiser becomes lodged in a saddle point, random based optimisers are robust enough to converge to a potential global minimum. Searches involving determining derivatives for direction finding performed poorly.
optimiser involves a decision-making component that has made it a more effective search technique. Al- though a suitable solution exists, surprisingly, the Genetic search did not converge to a suitable solution in every trial and became locked in a local minimum. However, this technique of restarting the optimisation could take a longer period of time to obtain a suitable solution compared with implementing a Random search technique. Perhaps using a Genetic algorithm where priori knowledge is required could produce improved results or another breeding technique could enable the search to perform better. Although the Genetic algorithm technique is a directed random search, it is not as robust as a purely Random search.
The unconstrained searches, which involve determining the derivatives of the objective function, have performed inadequately in reducing the error function for the dual-frequency problem presented. Either starting with a large error or a significantly smaller error function did not aid the search. These algorithms are unsuitable for optimising an antenna as they converged to a local minimum that exists at a significant distance from the goal or are not able to progress from the initial starting point. Similarly, even if initiating the search for a small error still did not reach a better solution.
In each of the three search objectives mentioned in this chapter, the optimisation searches that did not converge to a suitable solution indicates that the way in which the objective function was calculated appears to be inadequate in converging to a solution for that particular objective. The achievements in this section will be implemented in creating an active antenna, which will be continued in the next chapter.
Chapter 4
Active Antenna Design
This chapter describes the process involved in designing a low noise amplifier connected to the aperture coupled microstrip patch antenna as part of the front-end section of a GPS receiver. A dual-band active antenna design is presented, including simulated and measured results.
4.1
Introduction
The increasing demand for low cost, compact and conformal design and manufacturing ease has spurred advances in both substrate and surface mount technology. Without such major achievements in microwave technology over the recent decades, the enormous benefits of printed antennas could not have been exploited.
An active antenna or active integrated antenna has one or more active devices directly connected to the antenna [63]-[76]. Active antennas can be broadly characterised into two categories, transmitting and receiving [66], depending on which port of the active device is coupled to the antenna. Active antennas with both functions are characterised into different groups such as repeaters, transponders and transceivers. The benefits of using active antennas include saving power, size, weight and cost.
This concept emerged as early as 1928 [66] and directly integrating a solid-state device to a dipole was proposed in 1966 [65]. More recently, printed antennas connected to low noise amplifiers without the use of a matching network have been investigated [1], [77]-[79]. A novel active dual-band antenna design is presented in this work that implements a single layered aperture coupled microstrip patch antenna without a matching network between both entities.
This chapter describes the process involved in designing a low noise amplifier connected to the aperture coupled microstrip patch antenna as part of the front-end of a GPS receiver. An Agilent ATF-35143 Gallium Arsenide (GaAs) pseudomorphic High Electron Mobility Transistor (pHEMT) was employed because of its low noise and high gain properties at L-band frequencies. This feature is paramount in
low noise amplifier design as any noise generated in this first stage in a receiver will multiply through in proceeding receiver stages [80]-[82]. In addition, the ATF-35143 pHEMT is compact in construction which makes it ideal for connectionless receiver design. Surface mount components were used in the amplifier design since realising the output matching network with microstrip line sections would consume a considerable amount of circuit board real estate at the desired frequencies.