Формирование луча в сенсорных решетках: принципы и применение.

Signal processing technique used in sensor arrays for directional signal transmission or reception

Beamforming or spatial filtering is a signal processing technique used in sensor arrays for directional signal transmission or reception. This is achieved by combining elements in an antenna array in such a way that signals at particular angles experience constructive interference while others experience destructive interference. Beamforming can be used at both the transmitting and receiving ends in order to achieve spatial selectivity. The improvement compared with omnidirectional reception/transmission is known as the directivity of the array. Beamforming can be used for radio or sound waves. It has found numerous applications in radar, sonar, seismology, wireless communications, radio astronomy, acoustics and biomedicine. Adaptive beamforming is used to detect and estimate the signal of interest at the output of a sensor array by means of optimal (e. g. least squares) spatial filtering and interference rejection.

Sonar beamforming utilizes a similar technique to electromagnetic beamforming, but varies considerably in implementation details. Sonar applications vary from 1 Hz to as high as 2 MHz, and array elements may be few and large, or number in the hundreds yet very small. This will shift sonar beamforming design efforts significantly between demands of such system components as the "front end" (transducers, pre amplifiers and digitizers) and the actual beamformer computational hardware downstream. High frequency, focused beam, multi element imaging search sonars and acoustic cameras often implement fifth order spatial processing that places strains equivalent to Aegis radar demands on the processors. Many sonar systems, such as on torpedoes, are made up of arrays of up to 100 elements that must accomplish beam steering over a 100 degree field of view and work in both active and passive modes. Sonar arrays are used both actively and passively in 1 , 2 , and 3 dimensional arrays. 1 dimensional "line" arrays are usually in multi element passive systems towed behind ships and in single or multi element side scan sonar. 2 dimensional "planar" arrays are common in active/passive ship hull mounted sonars and some side scan sonar. 3 dimensional spherical and cylindrical arrays are used in 'sonar domes' in the modern submarine and ships. Sonar differs from radar in that in some applications such as wide area search all directions often need to be listened to, and in some applications broadcast to, simultaneously. Thus a multibeam system is needed. In a narrowband sonar receiver, the phases for each beam can be manipulated entirely by signal processing software, as compared to present radar systems that use hardware to 'listen' in a single direction at a time. Sonar also uses beamforming to compensate for the significant problem of the slower propagation speed of sound as compared to that of electromagnetic radiation. In side look sonars, the speed of the towing system or vehicle carrying the sonar is moving at sufficient speed to move the sonar out of the field of the returning sound "ping". In addition to focusing algorithms intended to improve reception, many side scan sonars also employ beam steering to look forward and backward to "catch" incoming pulses that would have been missed by a single sidelooking beam.

A conventional beamformer can be a simple beamformer also known as delay and sum beamformer. All the weights of the antenna elements can have equal magnitudes. The beamformer is steered to a specified direction only by selecting appropriate phases for each antenna. If the noise is uncorrelated and there are no directional interferences, the signal to noise ratio of a beamformer with antennas receiving a signal of power , (where is Noise variance or Noise power), is:
A null steering beamformer is optimized to have zero response in the direction of one or more interferers. A frequency domain beamformer treats each frequency bin as a narrowband signal, for which the filters are complex coefficients (that is, gains and phase shifts), separately optimized for each frequency.

The delay and sum beamforming technique uses multiple microphones to localize sound sources. One disadvantage of this technique is that adjustments of the position or of the number of microphones changes the performance of the beamformer nonlinearly. Additionally, due to the number of combinations possible, it is computationally hard to find the best configuration. One of the techniques to solve this problem is the use of genetic algorithms. Such algorithm searches for the microphone array configuration that provides the highest signal to noise ratio for each steered orientation. Experiments showed that such algorithm could find the best configuration of a constrained search space comprising ~33 million solutions in a matter of seconds instead of days.

To receive (but not transmit), there is a distinction between analog and digital beamforming. For example, if there are 100 sensor elements, the "digital beamforming" approach entails that each of the 100 signals passes through an analog to digital converter to create 100 digital data streams. Then these data streams are added up digitally, with appropriate scale factors or phase shifts, to get the composite signals. By contrast, the "analog beamforming" approach entails taking the 100 analog signals, scaling or phase shifting them using analog methods, summing them, and then usually digitizing the single output data stream. Digital beamforming has the advantage that the digital data streams (100 in this example) can be manipulated and combined in many possible ways in parallel, to get many different output signals in parallel. The signals from every direction can be measured simultaneously, and the signals can be integrated for a longer time when studying far off objects and simultaneously integrated for a shorter time to study fast moving close objects, and so on. This cannot be done as effectively for analog beamforming, not only because each parallel signal combination requires its own circuitry, but more fundamentally because digital data can be copied perfectly but analog data cannot. (There is only so much analog power available, and amplification adds noise.) Therefore, if the received analog signal is split up and sent into a large number of different signal combination circuits, it can reduce the signal to noise ratio of each. In MIMO communication systems with large number of antennas, so called massive MIMO systems, the beamforming algorithms executed at the digital baseband can get very complex. In addition, if all beamforming is done at baseband, each antenna needs its own RF feed. At high frequencies and with large number of antenna elements, this can be very costly, and increase loss and complexity in the system. To remedy these issues, hybrid beamforming has been suggested where some of the beamforming is done using analog components and not digital. There are many possible different functions that can be performed using analog components instead of at the digital baseband. Beamforming, whether done digitally, or by means of analog architecture, has recently been applied in integrated sensing and communication technology. For instance, a beamformer was suggested, in imperfect channel state information situations to perform communication tasks, while at the same time performing target detection to sense targets in the scene.

Beamforming can be used to try to extract sound sources in a room, such as multiple speakers in the cocktail party problem. This requires the locations of the speakers to be known in advance, for example by using the time of arrival from the sources to mics in the array, and inferring the locations from the distances. Compared to carrier wave telecommunications, natural audio contains a variety of frequencies. It is advantageous to separate frequency bands prior to beamforming because different frequencies have different optimal beamform filters (and hence can be treated as separate problems, in parallel, and then recombined afterward). Properly isolating these bands involves specialized non standard filter banks. In contrast, for example, the standard fast Fourier transform (FFT) band filters implicitly assume that the only frequencies present in the signal are exact harmonics; frequencies which lie between these harmonics will typically activate all of the FFT channels (which is not what is wanted in a beamform analysis). Instead, filters can be designed in which only local frequencies are detected by each channel (while retaining the recombination property to be able to reconstruct the original signal), and these are typically non orthogonal unlike the FFT basis.