Ultrawideband (UWB) technology offers several advantages over conventional communications methods. For example, UWB systems feature information bandwidths of 1 GHz and more while being able to share spectrum with other applications without causing interference. UWB systems use narrow pulses to transmit data. Since no carrier frequencies are involved, the transmitter (Tx) and receiver (Rx) hardware can be made very simple. Still, a challenge lies in the development of an antenna capable of handling these high-speed pulse trains. UWB antennas must cover multiple-octave bandwidths in order to transmit pulses that are of the order of a nanosecond in duration. Since data may be contained in the shape of the UWB pulse, antenna pulse distortion must be kept to a minimum.

The Vivaldi antenna offers great promise for UWB applications. This planar design is suitable both for radar-like and communications applications. First conceptualized in 1979 as a wideband antenna,¹ the initial designs were balanced structures and therefore had to be fed by a wideband balun transformer. The primary disadvantage of this architecture is that the balun must provide good performance over the entire bandwidth of the transmitted signal, significantly increasing the implementation costs. The Vivaldi antenna is still in use today in its original form for broadband microwave and electronic-countermeasure (ECM) applications. Newer designs² involve the use of double-sided and stripline versions employing innovative techniques to eliminate the balun.

The Vivaldi antenna can preserve the shape of transmitted UWB pulses, ensuring error-free, high-data-rate communications. To understand why, Fig. 1a compares the time-domain S₂₁ response for a Gaussian monocycle input when fed to a log-periodic dipole antenna (a classical broadband antenna structure) and a Vivaldi antenna. In the log-periodic antenna, the smallest antenna element radiates the highest-frequency component while the largest antenna element radiates the lowest-frequency component after the pulse has had time to propagate to the far end of the antenna. The use of these resonant elements, however, results in an antenna, which is dispersive in the time domain. The dispersion results in difficulties distinguishing individual multipath signals—a well-known advantage of UWB—at the Rx, due to broadening of the pulses resulting in significant overlap. Figure 1b shows the time-domain S₂₁ response of a Vivaldi antenna. This antenna produces a near-perfect Gaussian doublet in response to the Gaussian monocycle input, (i.e., the first-order derivative³), and has a greater efficiency than the log-periodic dipole antenna.

What follows is meant to guide the reader through an engineering analysis of the Vivaldi antenna, attempting to explain how the different aspects of the antenna affect its performance, as well as provide a prototyping methodology to design and fabricate the antenna. Vivaldi antennas have been designed for diverse purposes from ranging to communications within the FCC approved limits.⁴ The performance of several antenna designs were evaluated in an anechoic chamber, and the effects of several different designs will be reviewed. The Vivaldi antenna is composed primarily of three different structures: a microstrip feed, a paired-strip middle section, and the radiating section. The design of the microstrip and radiating sections have a critical impact on the antenna performance, while the paired strip serves primarily as a transition region. Adjustments to the designs presented here can be made without causing a substantial loss in overall performance. Figure 2 shows an image of the overall Vivaldi antenna.

Early Vivaldi designs⁵ used longitudinal tapered slots on semirigid coaxial cable for the antenna feed. Although covering several octaves in bandwidth, this method is difficult to implement since the taper (100:1) which would have to be cut into the shield of the coaxial cable. A more elegant feed structure is a simple microstrip transmission line. The width of the microstrip is designed for a 50-Ω characteristic impedance for the type and thickness of dielectric material, using standard formulae.⁶

In the current design, the microstrip transmission line gradually tapers to a paired-strip transmission line. The transition region is responsible for connecting the highly capacitive feed structure to the inductive radiating section, and decouples the microstrip structure from the radiating portion of the antenna. It was empirically discovered that the transition region should be three to five wavelengths long to prevent a sharp discontinuity (and the resulting pulse distortion) between the feed and radiating regions. In addition, a properly designed transition region will convert the unbalanced feed into a balanced structure that can then be connected to the radiating region of the antenna. A detailed analysis and discussion of the paired strip structure can be found in ref. 7. The equation helps calculate the characteristic impedance of the paired strip as a function of width, dielectric constant, and thickness of substrate:

SEE EQUATION BELOW

where:

Z₀ = the characteristic impedance,
e_r = the dielectric constant of the substrate,
e₀ = the characteristic impedance of free space (377 Ω),
a = the width of the paired line × 0.5, and
b = the thickness of the dielectric substrate × 0.5.

The paired-strip section is slowly tapered away on each side to develop into the radiating sections of the antenna. The taper into the radiating section of the antenna is the most critical aspect of the design. The taper should be as gradual and as smooth as possible to avoid significant discontinuities at higher frequencies, which will cause reflections resulting in a distorted pulse-shape.