Beamforming is a technique used to identify the source (or sources) of signals. Using multiple sensors arranged in a grid, it is possible using beamforming techniques to tell the directions from which signals are coming from. It does not matter whether the signals are low frequency acoustic (marine sonar), ultrasonic (medical imaging) or radio frequency. The same principles apply.
| The conventional delay-and-sum beamformer is a spatial filter that operates on the output of an array of sensors in order to enhance the amplitude of a coherent signal relative to background noise and directional interference, i.e. to improve the signal to noise interference ratio (SNIR). The figure provided shows a curved array of sensors, where each sensor (black circle) is located at an (x,y) coordinate as shown. These sensors are pointed in known directions (red arrows), and we wish to form beams which point in chosen directions (purple arrows). |
|
The response of a single element is plotted on a polar graph, where the angle is offset from the beam directions, and the radius is the magnitude response (dB) in that direction. Element responses, determined generally by means of the 3 dB down point, are very wide - in this instance the width is about 90°.
The goal of conventional beamforming is thus to delay and sum the outputs from the individual sensors in the array so as to increase the SNIR and in so doing to achieve a narrower response in a desired direction or set of directions - the pointing direction is referred to as the Maximum Response Axis, or MRA, and can be chosen arbitrarily for the beams. In this way, when a source is detected in a given beam or sector of beams, the approximate direction from which it came is known.
Design Limitations
The design of conventional beamformers is generally optimised according to the principle of minimising the associated implementational efficiency, in order to reduce costs, with simplifying assumptions often being made about the sensor array geometry in order to facilitate the use of computationally efficient line array or circular array processing techniques. It is also generally assumed that the sensor data will always be "good" enough to form beams with, which in operational environments is not always the case, with problems routinely arising through the presence of malfunctioning sensors and/or data glitches. Once bad data has been allowed to enter the beamforming system, however, the damage cannot be undone.
As a result of such assumptions the associated beamformer performance is often compromised. It is essential therefore that if performance is to be maintained that such problems be catered for within the design of the beamformer. This means that the solution needs to be independent of sensor array geometry and needs to be able to detect and correct for bad sensors and/or bad data, in real time, within the processing.