D.M. Kokotoff* and R.B. Waterhouse

Microstrip patch antennas are used in a variety of communication systems due to their many inherent advantages. The printed annular ring has received attention recently because it is usually smaller than other forms of patches for a given operating frequency.

A coaxial connector is a robust and efficient technique to isolate the feeding network from the radiating patches. Unfortunately, probe-fed patch antennas suffer from inherently narrow impedance bandwidths, typically 2-3%. One common means to improve the bandwidth is to use an electrically thick substrate. However, as shown in [1], a special attachment mode must be used to accurately predict the RF performance of a probe-fed patch antenna utilising electrically thick substrate.

Here, we present an attachment mode for probe-fed printed annular ring antennas. The attachment mode is incorporated into a full-wave spectral domain analysis. Several design examples are presented. Insight into the design philosophy of probe-fed printed annular ring antennas is also addressed.

Consider an annular probe-fed microstrip patch antenna on a infinite grounded dielectric slab, as shown in Figure 1. The printed element, the probe feed and the ground plane are all assumed as perfect conductors. The formulation considered here uses a full-wave spectral domain Green’s function/moment method approach. If a poor choice of expansion functions is made, the number of functions required to approximate the exact solution to an acceptable level may become intractable in terms of CPU time and memory requirements.

Special care must be taken to consider both the continuity and the singular nature of the current at the point where the probe is connected to the patch. These conditions can be ensured by the proper choice of a singular expansion function in the solution, commonly referred to as an "attachment mode." As in [1], the attachment mode considered here is derived from the corresponding cavity model. The attachment mode be written in terms of a series of ordinary bessel functions.

Figure 2(a) illustrates the expansion function in vector form. The current is singular at the attachment point (Figure 2(b)), but is well behaved over the rest of the structure. The "negative" current in Figure 2(b) indicates the vector direction of J_r and J_f. In this case, the mode looks similar to the TM₁₁ mode on the patch.

Since the attachment mode is a good approximation to the original problem, only a small number of additional entire domain expansion functions, typically less than 15, are necessary in the full-wave solution. These additional expansion functions are the eigenmodes of the magnetic wall cavity model of an annular ring geometry [2].

When operated in their fundamental mode (TM₁₁), probe-fed printed ring antennas have a high input impedance at resonance. Thus, there is a strong coupling between the feed and the patch, that is, the radiator is over-coupled. This is why proximity coupling a microstrip line to a printed ring is the preferred option [3], whereby moving the feedline away from the patch reduces the coupling, and hence the input impedance. However, strong coupling between a probe feed and a printed annular ring is extremely advantageous when considering stacked patch configurations. Typically, for any stacked configuration, the lower resonator is deliberately over-coupled and the top patch is used to effectively impedance match the entire configuration. Thus the high impedance observed for a probe-fed ring is well suited to a stacked version. Importantly, the high impedance level is relatively independent of substrate thickness. Thus, thick dielectric layers below the lower resonator can be used without the penalty of poorly behaved input impedance due to the inductive nature of the feed. This directly contrasts with probe-fed circular and rectangular stacked patch configurations where the lower substrate cannot be made too thick [4].

Using the above design philosophy, a probe-fed stacked annular ring antenna was designed to operate at 1.9 GHz. Figure 3(a) shows the geometry of a stacked annular ring, with its relative dimensions. The predicted and measured input impedance plot of the antenna is shown in Figure 3(b). Very good agreement between theory and experiment is achieved. The predicted and measured 10 dB return loss bandwidth are 20% and 21%, respectively. The CPU time per frequency point using 15 entire domain expansion functions on each patch and the attachment mode was 18 seconds using a 200 MHz AMD K6 running LINUX. The E- and H-plane co-polar and cross-polar far-field radiation patterns measured are very similar to a conventional patch radiation characteristics. The gain of the stacked configuration is measured as 7.2 dBi, compared to a theoretical value of 7.5 dBi.