Fullscreen HTML5Med Res (??MB)HTML4

ACOUSTIC IMAGING Image source: https://unsplash.com/photos/empty-chairs-in-a-room-3rW1HAakg8g By Dr David Maddison, VK3DSM Those of us lucky enough to still have good hearing in both ears can instinctively tell where sound is coming from. However, some sounds can be difficult to locate; sometimes, doing so is a matter of life and death! That is where technology comes to the rescue, with Acoustic Imaging Systems. W ouldn’t it be nice to locate the source of a sound that we can hear but can’t see or locate precisely? Depending on their level, frequency and spectra, sounds are not as easy to locate as certain other phenomena, such as light leaking into a darkened room. Seeing sounds as an image is not altogether unusual. Animals such as bats and dolphins use sound to ‘see’ (see Fig.1). The same can be said for medical ultrasounds and submarine sonar. With active sonar, a sound wave is emitted and its reflection from the target is analysed to form an image. Alternatively, for passive sonar, no sound is emitted by the sonar; instead, it listens to sound waves emitted or reflected by objects being surveilled. Directional or stereo microphones, or our ears, can give some cues as to the location of a sound based on differential timing, frequency shaping (due to the shape of the ear and head) and so on. However, it can be difficult to 12 Silicon Chip locate a sound precisely; sometimes we only know the general area. At times, sounds can appear to come from one place but are really coming from another, perhaps due to reflections, refraction, standing waves or other phenomena. However, there is a way to visualise the source of sounds precisely, making them visible to us in the same way as we can see the source of light leaking into a darkened room. The source of the sound can be rendered visible by a device called an acoustic imaging camera. In contrast with the active sonar mentioned above, where acoustic signals are reflected back to form an image, in acoustic imaging, signals are only received from an external source. Like passive sonar, acoustic imaging relies on detecting sounds directly from the source, but it visualises sound fields for applications like industrial monitoring, setting it apart from passive sonar’s underwater tracking role. With an acoustic imaging camera, Australia's electronics magazine sound waves are detected and pinpointed using a microphone array for precise location. The sounds are overlaid in real time (or sometimes later) onto a digital camera image of the scene of interest. Acoustic imaging can also detect sounds inaudible to the human ear (eg, infrasound or ultrasound). It should be noted that, confusingly, there are other devices also called acoustic cameras that emit acoustic signals for tracking like active sonar. In this article, unless stated, we are only describing the passive devices. The core of acoustic imaging lies in beamforming, a technique that electronically shapes received sound (or radio) signals into focused beams by adjusting their timing (phase) and strength (amplitude) to enhance sounds from specific directions while reducing others. We previously mentioned beamforming in our September 2020 article about 5G Networks (siliconchip. au/Article/14572). siliconchip.com.au Visualising sounds as acoustic imaging is just one application of this technology. Others include acoustic microscopy, ultrasound imaging, photoacoustic imaging and thermoacoustic imaging, as well as sonar, which will not be discussed in this article. Sonar was already described in some detail in our June 2019 article on that topic (siliconchip.au/Article/11664). How it is used Examples of the use of acoustic cameras include locating the source of an unwanted sound to rectify it, such as reducing noise in prototype motor vehicles, aircraft, trains or other vehicles. It can also be used to locate a gas leak in a chemical plant, which often can be hard to detect otherwise (eg, if it’s a clear gas escaping). Alternatively, we might want to analyse the frequency spectrum of sounds emanating from certain locations for various diagnostic or suppression purposes. We can also map traffic noises or locate the origins of noises from wildlife. It could also be used to analyse the source of noise entering a building from outside, so that soundproofing can be installed. In fact, just about anywhere there is a sound that needs to be eliminated, located or analysed, there is an application for the acoustic camera. We previously published a review of the CAE SoundCam (October 2020; siliconchip.au/Article/14610). It was one of the first commercial devices on the market and took ~15 years to develop. In this article, we will go into more detail about the theory of operation of such devices and the latest developments. 17th century CE Sir Isaac Newton attempted to measure the speed of sound and understood sound to be a wave like a water wave. 1626 Sir Francis Bacon emphasised the importance of investigating “the nature of sounds in general” which he called “acoustica”. His observations and experiments on sounds were published posthumously in 1627, in Sylva Sylvarum (siliconchip.au/link/ac95). He observed “frisk and sprinkle” when he rubbed the rim of a glass of water. 1671 Robert Hooke saw patterns on a flour-covered plate along which a violin bow was drawn. 1680 Ernst Chladni repeated and enhanced Hooke’s work and developed a method to show the various modes of vibration of rigid plates. 1877-1878 Lord Rayleigh laid the foundations for the theory of the behaviour of sound waves in his treatise, “The Theory of Sound”. 19th century Hermann von Helmholtz made substantial contributions to acoustics. 20th century Microphones and oscilloscopes greatly facilitated the study of acoustics. 1910s to 1920s Sonar was developed for imaging underwater. 1917 Nobel Prize winner Jean-­ Baptiste Perrin invented the télé- sitemètre for the French military, for the acoustic detection of enemy aircraft. In 1917, it was said to be able to detect aircraft 7-8km away with an angular error of 2-3°. It used two sets of a number of sub-arrays of listening horns grouped together and combined via an acoustic waveguide to a listening point at each of the observer’s ears. It was a type of acoustic beamforming before its modern implementation with computers and signal processing. A version appeared on the cover of 1930 Popular Mechanics (Fig.2). According to the magazine, that version “automatically registers their flying speed, altitude and distance from the finder”. 1930s to 1940s Directional microphone arrays emerged for sound ranging during World War II, advancing multi-microphone techniques. Phased array antennas were used similarly for radar. 1940s to 1950s Phased arrays of hydrophones were used for sonar. Sonar principles were applied in the development of medical ultrasound. 1960s to 1970s acoustic methods were developed for non-destructively testing materials, eg, looking for cracks in aircraft parts or other critical components. Beamforming techniques were used in medical ultrasound. History of acoustic imaging Developments leading up to acoustic imaging included the following discoveries regarding the behaviour of sound and developments in beam-forming: 6th century BCE Pythagoras studied musical sounds from vibrating strings. 4th century BCE Aristotle suggested that sound propagates as motion through air. 1st century BCE Vitruvius contributed to the acoustic design of theatres and determined the correct mechanism of sound wave transmission. 6th century CE Boethius documented a link between pitch and frequency. siliconchip.com.au Fig.1: an image of a man as seen by a dolphin’s natural sonar. Source: www.speakdolphin. com/pressRelease/ Press_Release_what_the_ dolphin_saw.pdf Fig.2: the cover of Popular Mechanics from 1930 shows a version of Jean Baptiste Perrin’s télésitemètre. Australia's electronics magazine January 2026  13 Fourier Transforms for Dummies Fourier transforms let us view signals in terms of their frequencies rather than time; a bit like turning a recording of a song into its individual notes. Fourier theory says that any waveform can be represented as the sum of sinewaves of different frequencies, phases and amplitudes. If you are not familiar with a Fourier transform, it may seem like a complex and exotic mathematical concept that you are unlikely to ever fully understand. However, it actually turns out to be relatively simple when you think about it the right way. One way to approach it is to consider the inverse Fourier transform first. If a Fourier transform turns regularly sampled time-domain amplitude data into frequency/phase data (as a complex number, but don’t worry about that now), the inverse Fourier transform turns frequency/phase data back into a set of points sampled at fixed intervals in time. Its output is exactly the input of the original Fourier transform. The frequencies that we’re breaking the signal down into are at fixed intervals (eg, DC, 100Hz, 200Hz, 300Hz etc), so the output of the Fourier transform is simply a series of amplitudes and phases, with each frequency ‘bin’ allocated a scaling factor and phase offset. We can easily visualise how to reverse the Fourier transform. You take a sinewave at each frequency, scale it by the corresponding amplitude value, shift it by the phase shift, and add the lot together. Voilà, you have your original waveform back. Mathematically, this is just a linear operation – a kind of matrix multiplication – where each row represents one sinewave at a different frequency. After all, a sinewave of a specific frequency sampled at specific intervals is simply a set of numbers between -1 and +1 calculated using the sin(ωt) function. If we expand that function to Asin(ωt + φ), where A is the amplitude scaling factor and φ is the phase shift, we get our original sinewave back. Then we just need to add them up, giving us the final formula: In this formula: xn is the nth input sample; N is the total number of samples in the transform; k is the frequency bin index; and Xk is the result for a given k. If you haven’t studied high-level maths, that may look like gobbledegook, but it’s essentially just performing the sum-of-scaled-and-phase-shifted-sinewaves mentioned above, with some normalisation applied so the magnitude of our result matches the original scale. Now, through the lovely properties of linear algebra, it turns out that the forward Fourier transform has almost exactly the same formula, with just a sign change and the removal of the scaling factor (as per convention). It is: How can our sum-of-sinewaves algorithm break down a time-domain signal into its constituent sinewaves? It makes sense if you think of it this way: what a Fourier transform is essentially doing is calculating the correlation between the input signal and each sinewave at a different frequency. A correlation is a statistical calculation that tells you how similar two sets of data are, with a larger result meaning they are more similar. Its formula is quite simple: In other words, the correlation between two sets of discrete data is simply the sum of the products of corresponding data points. If you think about it, if your data rises and falls at a similar rate to the sinewave you’re correlating it with, you’re going to get a large resulting sum. If they are not synchronised, the products are going to essentially be random and cancel out when you sum them. So, the scary-looking Fourier transform formula above is basically just doing this correlation with a set of sinewaves at different frequencies, and out pop the correlated sinewave amplitudes. By using complex numbers, the transform simultaneously captures both amplitude and phase; the magnitude of the complex number gives the amplitude, while its angle gives the phase. Finally, to resolve any confusion over the use of complex numbers giving us the phase shift; there is a simpler, geometric way to think of what we’re doing. Effectively, we are correlating the input signal with each sinewave along with its corresponding cosine wave, ie, the same sinewave phase shifted by 90°. The cosine component (the real part) measures how much the input aligns with a zero-phase reference wave. The sine component (the imaginary part) measures how much it aligns with a 90°-shifted version of the same frequency. Together, these two numbers form a 2D vector: one axis for cosine, one for sine. That vector’s angle gives you the phase of that frequency in the signal, ie, how far along the cycle your signal’s version of that frequency is compared to the reference cosine. The length (magnitude) of that vector gives you the amplitude, or how strongly that frequency appears in your signal. In summary, the Fourier transform is a set of two orthogonal correlations, with sine and cosine waves, at various frequencies, producing vectors where the angle represents phase shift and the length, amplitude. So while it’s advanced mathematics, it’s also incredibly elegant once you understand what’s going on. 14 Silicon Chip Australia's electronics magazine Fig.3: the concept of beamforming. The beam is electronically scanned to capture the signal from various parts of a soundscape, producing a sound map. 1970s the first experimental acoustic imaging systems emerged, using arrays to map sound sources, influenced by sonar and ultrasound. In 1974, John Billingsley invented the first “acoustic telescope”, a precursor to the acoustic camera. 1976 Billingsley and Roger Kinns develop a full-scale acoustic microscope system to analyse sounds from the Rolls Royce Olympus engine used in the Concorde. It used 14 condenser microphones, with signals digitised with 8-bit resolution at a sampling rate of 20kHz. The computer used had a memory of 48kiB and data was stored on floppy disks with a capacity of 300kiB. The processed data was displayed on a colour TV. This was the basis of modern systems, and in the following decades, improvements were made in the sampling rate, number of microphones, digitisation resolution, software and size and portability of the equipment. This was also the first time a real engineering problem, the determination of noise sources from the engine, had been analysed with acoustic imaging techniques. 1980s to present digital signal processing methods were developed, and high-speed computers enabled realtime beam-forming. 1997 a reporter coined the term “acoustic camera”. 2001 the first commercial acoustic camera was introduced by GFaI tech GmbH (www.gfaitech.com). The introduction of commercial devices marked the transition from research to practical tools, integrating digital signal processing (DSP) and array technology. siliconchip.com.au Fig.4: beamforming in the time domain using the delay-and-sum technique. Original source: www.gfaitech.com/knowledge/faq/delayand-sum-beamforming-in-the-time-domain Fig.5: how a Fourier transform converts data between the time and frequency domains. Original source: https://visualizingmathsandphysics. blogspot.com/2015/06/fouriertransforms-intuitively.html 2000s to present advances in array design and software have refined acoustic imaging for industrial and environmental use. How they work An acoustic imaging camera uses an array of multiple microphones to detect the source of a sound. One microphone cannot locate the source of a sound; two microphones can to a certain extent, like our ears, but even that does not give precise locations. For example, the shape of our ears combined with our brain is how we determine where sound is coming from. If you were to change the shape of your ears, it would take some time before your brain could readjust, and therefore you wouldn’t be able to precisely pinpoint where sound was coming from. An array of microphones, often 64 or more, is necessary so that triangulation and advanced mathematical techniques can be used to locate the source of the sound very precisely, while also filtering them by frequency. The microphones may be sensitive to frequencies from around 2kHz to 100kHz (well above what we can hear, ie, ultrasound). The precise method used to locate sounds is called beamforming, a signal processing technique also used for radio waves. It is how a mobile phone tower focuses its radio lobe directly at your phone to maximise the signal it receives while using minimal power and not interfering with other devices. In acoustic imaging, beamforming works differently. The camera, acting siliconchip.com.au as a receiver, focuses on acoustic energy naturally emitted by a sound source, enhancing sounds from specific directions while ignoring others. Essentially, it is the reverse process used for transmitting signals. Acoustic beam-forming The microphone array of an acoustic imaging camera is in the form of a geometric array. Sound waves reaching individual microphones are processed in such a way that some sounds from particular directions are selectively reinforced while others from different directions are attenuated by adjusting their relative amplitudes and phases. The ‘sound field’ is scanned either sequentially or digitally all at once, similar to how a spectrum analyser can be swept or a ‘snapshot’ processed using a Fourier transform. This amplifies and reinforces sounds from particular directions while attenuating others, thus building up an image showing intensity and frequency of sounds from particular areas – see Fig.3. Methods of acoustic beamforming using microphone arrays to produce directional images include: Delay-and-sum technique This is one of the simplest and most common methods of acoustic beamforming. Consider a microphone array that is picking up sound waves from multiple directions. Because sound waves travel at a more-or-less constant, finite speed (about 343m/s in air at sea level with average pressure, temperature and humidity), the sound waves from a Australia's electronics magazine specific direction will arrive at each microphone at a slightly different time. That time difference is determined by the distance between the microphone and the sound source. Delayand-sum adjusts for these time differences in software by delaying the signal from each microphone so that waves from the desired direction align exactly when it adds them together. If a desired sound wave comes from straight ahead, the closest microphone will receive it first; others will be slightly delayed. The software of the signal processor will delay the signal of the first (closest) microphone the most, and the others less so. When the signals are summed, the desired signal from straight ahead is reinforced, while others from undesired directions are attenuated or cancelled. Since this technique focuses on one direction at a time, it is repeated across the entire sound field, thus building an image. It is computationally straightforward, making it suitable for realtime imaging. This is less effective than other techniques in noisy environments or in complex sound fields, though. It generates a sound intensity map only, and does not separate individual sound frequencies. The beamforming and acoustic map generation process seems complicated, but it is simple in principle (although more complex in practice). Fig.4 shows an example with two sound sources, Source 1 (red) and Source 2 (blue), and four microphones (yellow circles). The steps are: 1. Signal acquisition: microphones record the sounds from a sound field January 2026  15 Fig.6: delay-and-sum beamforming in the frequency domain. of interest; four waveforms recorded are shown at the bottom. The plots show sound pressure (vertical axis) vs time (horizontal axis). The relative positions (in time) of the red and blue waveforms vary for each microphone based on its relative proximity to the sound source. 2. A time delay is added: each waveform has a distance along the time axis (horizontal) relative to its position from the source. The actual distances can be worked out by knowing the distance between the microphones and sound sources, and the speed of sound. We are interested in mapping Source 1 (Source 2 can be mapped at another time on another part of the sound field scan). A variable time delay indicated by ∆tx is added to each microphone waveform so the signals from Source 1 (red) for each microphone are aligned. 3. Signal summing: the signals with the time delays ∆t1, ∆t2, ∆t3 and ∆t4 are summed, resulting in a combined waveform where the signals from Source 1 are strengthened and those from Source 2 are not. 4. Signal normalisation: the signals are then normalised based on the number of microphones. The time delay to the largest peak is a measure of the position of the sound source in the sound field. 5. Mapping: the process is repeated over the entire sound field to create an acoustic map, showing the sound 16 Silicon Chip intensity at different locations. Frequency-domain beamforming This technique processes sound in the frequency domain rather than the time domain. Thus, the frequency spectrum of each sound source can be analysed. It allows the determination of which frequencies come from which directions so that acoustic maps of both sound intensity and frequency can be created. It uses beamforming techniques on each frequency band. It is computationally intensive and is often performed by post-processing data rather than in real time. Frequency domain beamforming is shown in Fig.6. In the approach described here, it is based on delayand-sum beamforming. The steps are as follows: 1. Signal acquisition: identical to the delay-and-sum technique. 2. Fourier transformation: the ‘Fourier transform’ is a powerful mathematical tool that converts a signal such as sound pressure over time, known as the time domain, into its underlying frequency components and their amplitudes, represented in the frequency domain (see panel). It decomposes a signal into a combination of sinewaves that represent both the amplitude and phase angle for each frequency component in the signal. Plots of amplitude vs frequency and phase angle vs frequency can be made from this information. This offers two views of the same data, revealing, for example, which frequencies dominate (see Fig.5). For instance, just as a piano chord can be separated into individual notes, the transform can break down the hum of machinery into its distinct frequency parts, aiding acoustic imaging analysis. 3. Phase vs frequency determination (Fig.6): Fourier analysis is applied to the amplitude vs time signal from each microphone to give a spectrum showing phase vs frequency representing the signals received at each of the four microphones. Each of the four signals from each microphone can be seen to have a different phase angle as a function of the frequency. 4. Phase adjustment: a time delay correction aligns the phases for Source 1, making its red signals in phase, Fig.8: adaptive beam-forming; the reception pattern of the lobes of the microphone array is shown. Undesired signals coming from directions other than the main beam are nulled in the signal processing. Original source: www.researchgate.net/publication/283639759 Australia's electronics magazine siliconchip.com.au Fig.7: phased-array beam-forming. The signals from each microphone (p1, p2 & p3) are phaseshifted into alignment and summed for each look direction to maximise signal strength. Source: https://dspace.mit.edu/ handle/1721.1/154270 while Source 2’s blue signals remain out of phase. This is evident in the lower middle graphs of Fig.6, where red signals align at the same phase angle, and blue ones diverge. 5. Summing: the adjusted signals are summed and normalised by the number of microphones. The in-phase red signals of Source 1 strengthen (overlapping as a single peak), while the out-of-phase blue signals interfere destructively, reducing their strength. 6. Mapping: the summed values for each frequency can be plotted on an acoustic map, with the positions of the sources of each frequency being determined from the time delay and phase angle information, resulting in a “heat map” of sound intensity and frequency. Phased-array technique The phased-array technique is a beam-forming method that uses precise control of the phase, the position of each acoustic signal’s sinewave cycle received by microphones, to electronically steer the listening beam across the sound field (see Fig.7). Unlike delay-and-sum, it adjusts the phase of each microphone’s signal, causing acoustic wavefronts to interfere constructively and reinforce sounds from the target direction while destructively cancelling others. This offers excellent directional precision, ideal for imaging dynamic sources, but demands computationally intensive processing and careful equipment calibration. Adaptive beam-forming Adaptive beamforming (Fig.8) adjusts to challenging sound environments by modifying delays and microphone weightings (amplification) in real time to suppress noise or interference, such as from a specific direction. This dynamic approach requires significant processing power, although it is ideal for complex acoustic imaging tasks. Acoustic imaging system configurations Acoustic imaging cameras come either as fully integrated all-in-one units (handheld) or as separate micro- phone and camera arrays, data acquisition units and a laptop computer (see Fig.9). The sound map being recorded and processed here is shown in Fig.10. Handheld acoustic imaging cameras For industrial inspection purposes, it is often more convenient to use an all-in-one handheld acoustic camera rather than separate system components. The SoundCam Ultra is a handheld unit that images audible sound and ultrasound (see siliconchip.au/link/ ac97). It is used for compressed air/gas leak localisation, vacuum leak localisation, partial discharge localisation, condition-based monitoring, animal studies and non-­destructive testing. Another example is the GFaI tech Mikado. It uses an array of 96 digital MEMS microphones and a Microsoft Surface Pro tablet as its data processing and display unit – see Fig.11. Acoustic microphone arrays Separate microphone arrays are also available for use with the separate Fig.9: a GFaI tech acoustic imaging camera system with separate components (microphone array, data recorder and computer) recording sounds from a sewing machine. Source: www.gfai. de/fileadmin/user_upload/GFaI_product_ sheet_acoustic_camera_en.pdf Fig.10: a sound map from the sewing machine being recorded in Fig.9. Source: www.gfai.de/fileadmin/ user_upload/GFaI_product_sheet_ acoustic_camera_en.pdf siliconchip.com.au Australia's electronics magazine January 2026  17 cameras, data recording units and a computer with the appropriate software. The spacing and relative location of microphones in an acoustic imaging array are crucial, carefully designed to optimise goals like resolution (clarity of sound sources), side-lobe suppression (reducing unwanted beams) and spatial aliasing reduction (avoiding imaging artefacts). These microphone arrays can be 2D linear (square or rectangular), circular, random, or even follow a Fibonacci pattern, similar to a sunflower. Various 3D arrangements are also possible. A key design rule is that the microphone spacing should be less than half the wavelength of the highest frequency to prevent aliasing (derived from the Nyquist-Shannon sampling theorem). The relevant equation is d = v ÷ 2fmax, where d is the spacing in metres, v is the speed of sound in air (343m/s), and fmax is the maximum frequency to be imaged. For example, to image up to 5kHz (a wavelength of 68.6mm), the spacing should be about 34mm; for up to 20kHz (a wavelength of 17.15mm), it should be around 8.6mm. One example of a 2D microphone array is the SoundCam Octagon (Fig.12), which has 192 MEMS microphones along with an integrated camera, data recorder and notebook computer running suitable software. The large number of microphones allows very high resolution imaging and acoustic holography (more on that later). Another example of a 3D microphone array is GFaI tech’s Sphere48 AC Pro48 channel system for acoustic measurements in 2D and 3D with 48 electret condenser microphones (see siliconchip.au/link/ac96). It has a frequency response from 20Hz to 20kHz. It is designed for sound localisation in confined spaces such as a motor vehicle. It is used with NoiseImage software that allows sound sources to be isolated, localised and analysed with respect to both frequency and time response. It also allows a 3D acoustic map to be produced, and imagery is provided by an integrated Intel RealSense Depth Camera to record depth information. Suggested uses include noise, vibration and harshness (NVH) analysis in cars, trains and aeroplanes; location of squeaks and rattles in vehicles; leakage detection; and sound design and analysis of building acoustics. An additional example of a 2D array is the Fibonacci120 AC Pro (Fig.13), a 120-element microphone array in the form of a Fibonacci pattern. It allows near-field and far-field measurements and, according to the manufacturer, the spiral pattern gives the “highest possible spatial resolution and the best possible map dynamics”. A further example of a microphone array is the GFaI tech Star48 AC Pro (siliconchip.au/link/ac9c). It is optimised for mid-range frequency measurements of outdoor objects like aircraft flyovers or the observation of large wildlife, like elephants. Applications In this section, we will discuss various applications of acoustic imaging. Acoustic detection of drones Hostile drones pose risks to military and civilian people and infrastructure; therefore, their detection is extremely important. Drones can be flown autonomously, without RF communications (or via fibre-optic cables), making their detection even more difficult. Their small size can also make radar detection difficult. Airspeed Electronics Ltd (www. airspeed-electronics.com) has developed passive acoustic imaging arrays to detect drones (Fig.14), which can each detect small quadcopters at a range of 200-300m. Each sensor can be integrated into a network to make a fully scalable array connected by wireless mesh radio. Multiple sensors enable accurate target location via triangulation. A drone’s acoustic signature also provides valuable information such as the number of rotors, pitch imbalances and rapid pitch variations, which allow the drone class to be detected, an estimate of its payload mass (weight can affect the rotor pitch) and whether the drone is manually or autonomously controlled. Airspeed’s microphone arrays use phased-array signal processing to help separate drone sounds from other background noises. Electret condenser microphones are used in Airspeed’s microphone arrays as they have superior performance to MEMS Fig.12: the SoundCam Octagon has an integrated camera and data recorder. Source: www.gfaitech.com/products/ acoustic-camera/all-in-one-soundcam-octagon Fig.13: the GFaI tech Fibonacci120 AC Pro. Source: www.gfaitech.com/fileadmin/gfaitech/documents/ datasheets/acoustic-camera-fibonacci-array-120datasheet-20.pdf Fig.11: the GFaI tech Mikado. The object behind the device is the microphone array (the video camera is not visible). Source: www. gfaitech.com/products/ acoustic-camera/handheldsoundcam-mikado 18 Silicon Chip siliconchip.com.au microphones, according to the company. Airspeed performs its own in-house modelling and performance evaluation of microphones; a simulation of a microphone array is shown inset in Fig.14. Fig.15 shows the dashboard from a sensor array tracking a small drone. Aircraft An example of the acoustic analysis of a business jet is shown in Fig.16, The image shown represents a spectral analysis for the third octave band of 315Hz at 53dBA (“A-weighted decibel”, a sound measurement weighted to reflect human hearing). The hardware setup is the same as described below for the car measurements. The software used was Photo 3D and Spectral Analysis 3D for precalculated narrowband analysis to create acoustic photos from a spectrum. Building acoustics Acoustic imaging can be used in concert halls and other large interior spaces to optimise acoustics. It can diagnose and correct acoustic problems such as undesirable echoes (reflections), absorption of sounds, or differential absorption or reflections of sounds of different frequencies. Acoustic imaging can be used to optimise ‘acoustic comfort’ in buildings by detecting the source of sound leaks or the effectiveness of various acoustic treatments. For example this video (https://youtu.be/ykchSQX-sfg) shows a Sorama CAM iV64 being used to detect sound leaks around a window frame. Fig.14: a network of Airspeed’s TS-16 acoustic remote sensors at the British Army’s AWE-24 exercise, Salisbury Plain, UK. Inset: a simulated beam pattern from a microphone array at 1.2kHz. Source: www.airspeed-electronics.com/ technology Fig.15: drone tracking by Airspeed using an acoustic image array. The image at upper left shows the target drone location by azimuth and elevation. At upper right is a polar plot, while the lower left shows a view from the target drone; at lower middle is a spectrogram of the target, and the lower right shows the predicted target type based on spectral information. Source: www.airspeedelectronics.com/technology Cars Automotive and other engineers strive to minimise NVH (noise, vibration & harshness) in vehicles (or other machines). For cars, NVH can be perceived as unwanted and unpleasant for passengers and drivers. These sounds may originate from the engine, drivetrain, suspension, tyres, road, air conditioning, wind noise etc. One way to locate the source of these noises is through the use of acoustic imaging cameras. Some examples of locating such noises are shown in Figs.17 & 18. The experimental setup to obtain those images comprised the GFaI tech Sphere48 AC Pro microphone array mapping frequencies from 291Hz to 20kHz. Fig.16: acoustic measurement and location-finding in a Bombardier BD-700 - 1A10 business jet. Source: www.gfaitech.com/ applications/aircraft-interior siliconchip.com.au Australia's electronics magazine January 2026  19 Figs.17 & 18: analysing and locating noise sources in VW interiors with a microphone array. Source: www.gfaitech. com/applications/ vehicle-interior Also used were an mcdRec data recorder with a sampling rate of 192kHz and a depth of 32 bits, and Noise­Image software with the Acoustic Photo 2D and Acoustic Photo 3D modules for mapping the sound sources onto a common interior or exterior CAD model. Other software modules used include the Record Module, Spectral Analysis, Advanced Algorithms and Project Manager. Cooling fans Acoustic imaging technology can be used to develop quieter cooling fans in electronic equipment. For example, PC fan manufacturer Cooler Master uses this technology, as shown in Fig.19. Notua also use similar technology to develop their fans, this includes acoustic imaging to map noise (siliconchip. au/link/ac99). Drone-based acoustic imaging Acoustic imaging cameras can be mounted on drones (see Fig.21) for various purposes such as industrial inspection, natural disaster response or security. The obvious problem of self-induced drone noise can be reduced by spectral (Fig.20) and other methods, such as making sure the beam-forming direction ignores any part of the drone’s airframe. Fig.19: Cooler Master computer fans are developed with a Sorama acoustic camera. Source: https://youtu.be/0UFli2BUCL4 Fig.20: a block diagram of a spectral ‘denoising’ scheme to remove selfgenerated noise from a drone-mounted acoustic camera. Original source: https://doi.org/10.3390/drones5030075 Fig.21: the Crysound (www.crysound.com) CRY2626G is the first dronemounted acoustic camera designed for detecting pressurised system leaks and electrical partial discharge. Source: https://sdtultrasound.com/ products/crysound/cry2626g/ 20 Silicon Chip Australia's electronics magazine Echoes in rooms The sampling rates for acoustic imaging can be as high as 200kHz. Thus, it is possible to watch echoes bounce around a room, as shown in Fig.22. The picture shows all the bounces, but in reality they happen sequentially. Electrical discharge inspection Detecting high-voltage partial discharges from insulation and corona discharges is a necessary task to prevent dangerous or expensive problems in high-voltage installations. Techniques such as infrared thermography are not always reliable for detecting them because certain types of discharges might not cause a significant temperature rise, or not pinpoint the exact location of the problem. In addition, in a high-voltage installation, heat may be generated for other reasons. It is also often difficult to detect the sounds that these discharges make using the ear or microphones. Thus, acoustic imaging can be a good tool to detect such problems. siliconchip.com.au Fig.22: echoes bouncing around a room. Sources: https:// petapixel.com/2023/03/23/how-acoustic-cameras-can-seesound/ & https://youtu.be/QtMTvsi-4Hw Acoustic imaging is also potentially safer in the hazardous environment of high-voltage installations, as it can be used from further away than some other techniques. An example of discharge detection is shown in Fig.23. The instrument used is the Fluke ii915. Research has shown that the frequency of sound emissions from electrical discharges is mostly in the range of 20-110kHz, with 95% of the acoustic energy in the range of 48kHz to 100kHz, with a peak frequency of 68.3kHz. Thus, this instrument is optimised for detection at those frequencies. Fixed or mobile applications The Sorama L642 (https://sorama. eu/products/l642-acoustic-monitor) can be permanently mounted on a pole or placed on a mobile robot for continuous monitoring or inspections. It can be used indoors or outdoors, in a Fig.24: detecting a noisy vehicle exhaust with a Sorama L642. Source: https://sorama.eu/solutions/vehicledetection-system siliconchip.com.au Fig.23: detecting high-voltage electrical discharges using the Fluke ii915. Source: www.seesound.com.au/partialdischarge factory environment or even an urban environment to monitor noises and their sources. One application is to detect noisy vehicles, as shown in Fig.24 and https://youtu.be/fQEkkFGPbU8 Gas leak detection Acoustic imaging can be used for gas leak detection and is able to detect leaks that people cannot even hear. This method of gas leak detection is considered superior to, or at least supplemental to, gas detectors, because acoustic imaging can detect a small leak before there is a substantial buildup of gas (see Fig.25). fireworks, noisy vehicles or alarms going off. Hydrogen leak inspection Finding hydrogen leaks is difficult, as hydrogen can escape from the smallest openings. Acoustic cameras such as those from Sorama have been designed specifically to be able to detect hydrogen leaks from tanks, pipes and valves – see Fig.26. Mechanical inspection Acoustic imaging can discover defective parts of machinery, such as a defective robot joint that has developed a squeak. General environmental monitoring The Sorama L642 series can be used for noise measurements and anomaly detection in urban environmental monitoring, such as identifying the location of inappropriately lit Mining equipment Sounds from mining equipment can be identified and appropriate action taken. These sounds can indicate a possible occupational safety concern. One example is abnormal noise from Fig.25: gas leak detection using a Sorama acoustic imager. Source: https://sorama.eu/solutions/gas-leakinspection Fig.26: detecting a hydrogen leak from a valve using a Sorama camera. Source: https://sorama.eu/solutions/ hydrogen-leak-inspection Australia's electronics magazine January 2026  21 the conveyor bridge of an excavator (see siliconchip.au/link/ac99). Road noise management We already mentioned the Sorama L642, but other companies make devices for monitoring noisy vehicles. Noisy vehicle detection technologies are already on trial in Australia: siliconchip.au/link/ac91 siliconchip.au/link/ac92 Editor’s note – there are several large boxes in the middle of Foreshore Road near Port Botany in Sydney, powered by solar panels, that appear to be used to monitor noise from the many trucks on that road. Apart from Sorama, companies that make noisy vehicle detection systems include SoundVue (https://soundvue. com – used in Australia), General Noise (www.generalnoise.co.uk) and acoem (www.acoem.com/en). Fig.27: an overhead view of Philips Stadion with acoustic camera data overlaid. Source: https://sorama.eu/fan-behavior-analytics-with-acoustic-data-engaginginsights-for-sports Fig.28: an acoustic image of a high-speed train. Source: www.gfaitech.com/ knowledge/faq/passby-2d-integration-time Fig.29: studying elephant vocalisations in Nepal. Source: https://youtu.be/ Xl7LnAob2T8 22 Silicon Chip Australia's electronics magazine In stadiums Acoustic imaging is used to analyse, map and localise cheers from fans in stadiums. Competitions can be organised to enhance fan engagement so that the loudest and proudest fans win. The winner for the noisiest fans or real-time noise production by fans can be determined with a “SoundSurface map” display on the large screen being shared in real time at the stadium and on social media – see Fig.27. The noise level changes second by second and corresponds to events happening within the game being observed, such as scoring a goal. There are two different Sorama acoustic camera systems installed at the Philips Stadion in the Netherlands. One is the Sorama L642XL, which is equipped with 64 microphones arranged in a sunflower pattern to provide seat-level accuracy, right down to individual fan reactions. The other system uses 30 Sorama L642 cameras, covering all seats, to observe crowd behaviour at a higher level. The system can also detect unwanted chanting, shouting, slurs or breaking glass. Trains Investigating noises emanating from trains was one of the first commercial usages of acoustic imaging. An acoustic image of a high-speed train is shown in Fig.28. Not surprisingly, the wheels seem to be the main source of noise, but there was also noise from siliconchip.com.au Figs.30 & 31: examples of vibration analysis using the GFaI tech WaveCam software on large structures such as a wind turbine and tower, and smaller structures such as a car engine. Source: www.gfaitech.com/products/structural-dynamics/ vibration-analysis-with-wavecam the pantograph. This discovery led to design efforts to minimise noise from that source. Vibration analysis Vibration analysis can be used as a supplementary technique to acoustic imaging. It is performed optically, using a camera and software to detect small variations in an image due to vibrations. GFaI tech offers the WaveCam software for this purpose. Figs.30 & 31 show some examples of such vibration analysis. A combination of both vibration analysis and acoustic imaging can be used to give a deeper understanding of a vibration and noise problem, as shown in the video at https://youtu. be/0Z7E5Ql7Xiw Vacuum cleaner development Perhaps one of the noisiest domestic appliances is the vacuum cleaner, so it is not surprising that considerable efforts are made to quieten these machines. Figs.32 & 33 show frame grabs from Steve Mould’s video at https://youtu.be/QtMTvsi-4Hw showing sources of sound from a vacuum cleaner; one at 400Hz, the other at 7000Hz. Wildlife Acoustic imaging cameras have been used to study wildlife vocalisations, such as elephant sounds, including infrasound – see Fig.29. A better understanding can thus be made of how animals communicate and the parts of the body involved in generating various sounds. Acoustic holography Acoustic holography is a specialised technique that reconstructs the entire sound field (a 3D representation of the distribution of sound waves), including amplitude and phase over a surface or volume, based on measurements taken at a limited set of points. It uses wave propagation principles to create a ‘holographic’ representation, akin to optical holography, but with sound waves. It uses some of the same techniques as acoustic imaging, such as acoustic wave analysis, microphone arrays and signal processing, and can be seen as an extension of acoustic imaging. It has niche applications in research, requiring extremely advanced mathematical models. Acoustic imaging maps sound sources using beamforming, while acoustic holography extends this by reconstructing the full sound field, including phase, for a detailed analysis. Acoustic imaging can be seen as a ‘snapshot’, while acoustic holography is a complete 3D model of sound. The future of acoustic imaging Over the last few years, the cost of acoustic imaging has gone down, and the capabilities have gone up. Possible or likely developments in the future include higher-resolution microphone arrays, integration with AI for automated source detection, plus cheaper and more portable designs. Challenges include improving low-­ frequency detection, reducing setup complexity (although existing handheld units are virtually ‘plug & play’), and handling reverberation, where the sound reflects off multiple surfaces even after the source has stopped. Research trends include advanced signal processing, wearable sound cameras (possibly with military applications) and multi-modal imaging (say, measuring vibration and sound at the same time by the same device). Future applications include the use in robotic imaging, smart cities and SC consumer applications. Figs.32 & 33: frame grabs from https://youtu.be/QtMTvsi-4Hw showing noise from a vacuum cleaner. On the left, it shows the 400Hz noise from the tube, while on the right, the 7kHz noise is coming exclusively from the motor. siliconchip.com.au Australia's electronics magazine January 2026  23