Acoustic beamforming arrays, commonly known as acoustic cameras, enable the user to visualise different sound sources at different frequencies and source strengths. The resolution and ability to resolve sound sources spaced closely apart, and at lower frequencies, is mainly decided by overall size and number of microphones of the equipment being used. Although image manipulation and deconvolution techniques on the beamformed results might give added resolution, in practise the properties of the array still influence the results. This size versus resolution criteria is the crux of the acoustic camera market. Users want something that is small, light weight, and portable, while at the same time having excellent resolution, and the ability to go low in frequency. This has been an impossible demand for a single system – until now.
The Norsonic Hextile is a module based approach to acoustic camera that gives the user both portability and great resolution for a wide range of measurement situations. The array dish is based on a hexagon shape, given it both its name, and the ability to combine several tiles into larger systems.
With a single Hextile, the user has a small, portable and lightweight acoustic camera that can be used for a wide range of measurement situations. The Hextile is a USB based acoustic camera, with a single USB cable for both power and data transfer – no extra battery cable needed. The array is made from robust and lightweight aluminium, has 128 MEMS microphones, and is less than 3 kg in weight while having a maximum diameter of 46 cm. The low frequency limit for the Hextile is 410 Hz.
For users that require better resolution both in lower frequencies and overall, three single Hextiles can be combined to a larger Multitile system, consisting of 384 microphones with a maximum diameter of 96 cm. The low frequency limit for the Multitile is 220 Hz.
For special low frequency applications below 1 kHz, it is also possible to utilise the Multitile in the low frequency configuration as the Multitile (LF mode). By placing the individual Hextiles further away, the maximum diameter of the complete array system is increased to 1.46 m, making it ideal for low frequency measurements. The Multitile (LF mode) is for low frequency measurements below 1 kHz, with a lowest frequency limit of 120 Hz.
Connection: USB
Microphones: 128 MEMS microphones
Max sound level: 120 dB
Min sound level (system): 9 dBA
SNR per microphone: 65 dBA
SNR array (system): 82 dBA
Audio sampling rate: 44.1 kHz
Camera resolution: 2592 x 1944
Opening angle: 105°
Frame rate: 15 FPS
Operating temperature range: -40 to +85
Per microphone (flat): 100 Hz – 20 kHz
Per microphone: -26 +/-3dBFS/Pa @1 kHz 94 dB
Spatial sensitivity Hextile: 410 Hz – 20 kHz
Spatial sensititivy Multitile: 220 Hz – 20 kHz
Spatial sensititivy Multitile (LF mode): 120 Hz – 1 kHz
Dimension Hextile: 41 cm x 48 cm, Ø 48 cm
Dimension Multitile: 83 cm x 84 cm, Ø 96 cm
Dimension Multitile (LF mode): 126 cm x 121 cm, Ø 146 cm
Weight Hextile: < 3 kg
Weight Multitile: < 10 kg
Material: Aluminium
Power consumption: < 3 W
The biggest improvement when going from a single Hextile to the two different Multitile configurations is best demonstrated on a low frequency source. Seen below are the results from recordings on a single omnidirectional noise source emitting pink noise, with the colour plotting being done when the input signals are filtered at 500 Hz. This should give a direct comparison of the low frequency capability of the different arrays.
At the top are the different array configurations used for the recordings, with a 128 element Hextile, a 384 element Multitile, and a 384 element Multitile (LF mode). The diameters of the array configurations are 46 cm, 96 cm, and 1.46 m respectively.
The second rows show the beampattern for the different array configurations at 500 Hz and 3 dB dynamic range. As can be seen the beampattern gets more narrow, thus giving better resolution, as the overall array size increases.
Lastly the plotting results from the three different array configurations recorded on a real noise source are shown with 3 dB dynamic. The improvement in terms of resolution and pin-pointing the source is clearly visible when using bigger equipment.
The software design strategy has always had user friendliness and ease of use in mind. We want the user to be able to get results quickly, and start analysing recordings easily, thus spending time on the analysis, rather than the measurement set up or configuration of parameters. Live view of measurements combined with an intuitive software interface enables users without prior experience to make measurements within the first five minutes after powering the device.
The one feature that really sets the software apart is the virtual microphone. The virtual microphone enables the capability to only get audio signals from the chosen listening point, and listen to sounds coming from specific directions of the video image, while suppressing noise and sounds emitting from other positions than what is selected. With this tool the user has the power of super hearing, and may gain more insight in addition to regular colour plotting of sources. Such super hearing may be especially useful in noisy and complex sound environments, where for instance different noise sources greatly impair the ability to distinguish which machinery is producing a faulty noise.
In addition to live plotting and directive listening, it is also possible to record measurements and do the analysis at a later time. The raw signal from all microphones are then saved, and all parameters such as frequency selection, time selection and so on can be changed in post-processing. This means that a recording can be done without selecting the optimal parameters during the measurement, since these can be changed when analysing the recording. This also means that anybody can do the actual recordings themselves since it is then basically a matter of pointing the array roughly towards the area of interest and pressing record. All analysis and changes of parameters can be done in post-processing such as directive listening, graphical overlay of sources, spectrogram, FFT analysis and so on.
Sometimes sources may be closely spaced apart, or a strong noise source in the area of interest is interfering with the recording and impairing the image quality. Often this will be seen as either a single large source, or the source of interest will be completely shadowed by the stronger source. Seen in the image below is a situation where two equally strong sources are positioned close to one another, and the resulting colour plot will display a single large source. In such situations the acoustic eraser feature may prove valuable. This function will add a red circle to the screen that can be dragged to any point, and remove the source from that point. This is highly effective when several noise sources are present. As seen on the pictures the acoustic eraser completely removes the source where the suppress point button is positioned. The virtual microphone can further be positioned on the source of interest.
Especially in automotive applications RPM measurements may give vital information. The acoustic camera software has the possibility to display frequency content as a function of RPM by using the order analysis function.
In the spectrogram window, frequency as a function of RPM is plotted. It is further possible to select a square in the spectrogram window to isolate interesting events. By pressing the “apply” button on the selection, the RPM and frequency limits in the main view window automatically change to the limits set by the selection in the spectrogram. The user may then find and interesting sound event in the spectrogram, and automatically get the corresponding colour plotting of the event chosen.
Measurements in office complex
Lier, Norway, April 2016
These recordings were made with the Nor848A-10 1.0 m acoustic camera system with 256 microphones, now replaced with Nor848B acoustic camera system.
Problem
A newly built office complex is designed with glass facades between the offices and the hallway. The glass facades include a glass door. Although the glass structures themselves have a sufficient sound reduction value, the sound insulation between office and hallway was measured at 19 dB, which was far below the sound insulation criteria for offices given in the regulations. It was therefore important to find out where any weaknesses were introduced in the overall structure.
Measurements
A common way to detect cracks and gaps in barriers is by placing an omnidirectional loudspeaker emitting white noise in the sending room, and use the acoustic camera in the receiving room pointing at the structure of interest. For this situation the Norsonic Nor848A-10 1.0m acoustic camera with 256 microphones was placed on the outside of the glass facade filming directly at it. Gaps and cracks in the structure should then be detectable by being visualised as small noise sources in the structure.
Results
The first results from the initial recordings proved disappointing. Although it was possible to hear clear differences by using the virtual microphone, which enables the user to listen to specific points in the image, the coloring of sources was only seen on a glass facade standing perpendicular to the wall of interest, as seen on the left side in the image below. Clearly this was a strong acoustic reflection and not the main source itself.
Since the recording environment was highly reflective, a good approach was to try to dampen the influence from surrounding structures. By using a piece of cloth to cover parts of the reflective wall, the acoustic reflection was absorbed enough so that the sources of interest could be visualised as seen in the image below.
Now various weak points in connection with the door frame became apparent. Sources were seen between the door and the floor, and especially around one of the door hinges. By studying the hinge in detail, one could easily see how the rubber seals weren’t completely sealed off around the hinge, but left a small gap as seen in the image below. This gap allowed noise to leak through.
In addition to the main weaknesses being the door hinge and the seal between the door and the floor, also the top right corner of the door wasn’t closed tightly enough, so that noise leaked through here as well. This weakness could be clearly heard with the virtual microphone. Since it was weaker than the two main sources, it was better visualised in the image by using the acoustic eraser to try to remove the main sources from the visualisation as seen in the images below.
By closer inspection one could see how the rubber strip around the door edge wasn’t completely sealed shut in this position as seen from the image below.
Measurements in bar and bistro
Oslo, Norway, March 2016
These recordings were made with the Nor848A-10 1.0 m acoustic camera system with 256 microphones, now replaced with Nor848B acoustic camera system.
Problem
A combined bar, bistro and concert venue in the city center has been renovated with a great emphasis on acoustic noise dampening. Nevertheless, the venue is still getting complaints from neighbours close by due to breakout noise from the location, especially during late night concerts. The establishment consists of a bar and bistro on the ground floor, with the concert venue on the floor above. The concert venue has several windows facing the outside street and neighbourhood buildings, and it was desirable to pin point any acoustic weaknesses in these windows. Also it was of interest to see if the wall itself needed additional measures, or if the main source contribution came from the windows alone.
Measurements
The 1.0 m Nor848A-10 with 256 microphones was used for the recordings. The camera was plugged into an external battery pack for easy transportation and mobility. In addition to measuring the wall and windows of the concert venue, a wall between the café and a patio area on the ground floor was also of interest.
The camera was placed outside pointing at the facade of interest, with the audio system inside of the music venue playing white noise at volumes up to 100 dBA. The room inside would then act as a sending room, and the outside as the receiving room. In order to get close enough to the windows of the concert venue, a truck mounted crane was hired, with both camera and operator around 7 m above ground during measurements.
Results
Since the measurements were conducted around noon during a normal weekday, the city traffic was a constant source of background noise, especially during the recordings on the concert venue facade. Cars and trams driving past at regular intervals made it challenging to get a window of opportunity with relative quiet measuring conditions. However since all analysis can be done in post processing, you only needed a window of 5-10 seconds of proper measurement conditions to get the recording needed.
Seen in the pictures below are some of the measurements of the concert venue facade. As can be seen one could quickly rule out that any additional measures had to be taken on the wall itself, as the only sound leakage that could be seen came from the windows only. It was also possible to zoom in on different areas of interest, either at the site during a measurement, or at a later time in post-processing analysis to further locate weak spots.
The second measurements were conducted on the patio outside the café on the ground floor. Here the measurement conditions were substantially easier, as the patio was shielded from the city traffic noise. Again the music system of the café emitting white noise was used as source. The wall between the café leading to the patio consists of both a door and several windows. The first step consisted in seeing what made the biggest noise contribution. As seen in the image below, where the dynamic range is set to 10 dB, the door had a noise contribution that was approximately 10 dB higher than the nearest window.
Measurements in apartment
Oslo, Norway, February 2016
These recordings were made with the Nor848A-10 1.0 m acoustic camera system with 256 microphones, now replaced with Nor848B acoustic camera system.
Problem
An apartment on the ground floor of a two story house in the city lies close to a busy road. The living room in the apartment is facing the road, and the inhabitants are bothered by traffic noise, especially in the morning and the afternoon, when the traffic is the heaviest.
Traffic noise could be clearly heard when standing in the living room. The facade facing the city street consists of a large window and a porch door. It was thought that the main contribution of noise came from these two parts, but it was difficult to verify if those assumptions were true, or exactly where any weaknesses in the structure might be located.
Measurements
The Norsonic Nor848A-10 1.0m acoustic camera with 256 microphones was used for the recordings. The camera was placed inside the living room pointing at the facade facing the street. The living room would hence act as the receiving room, and the outside as the sending room, much in the same sense as the procedure for sound insulation measurements. Weaknesses in the facade would then be possible to be seen as small noise sources in the structure. It was possible to use regular traffic as sound source, regardless whether the traffic was steady, or just a single vehicle from time to time.
In addition to using traffic as noise source, measurements were made by placing a omnidirectional loudspeaker emitting white noise on the outside of the facade. This created a more stationary sound field on the outside of the structure, and made detection of small cracks and gaps in the structure even easier.
Results
Initial recordings when using traffic as noise source displayed a single strongest facade weakness at the top left of the living room wall as seen from the video below. This strongest source position was also confirmed when using the omnidirectional loudspeaker as noise source. At this position a ventilation valve was installed, and most of the traffic noise came from this location.
Having determined that the intake valve was the main noise contributor, this spot was covered up with a pillow to remove it from the overall noise field, and try to locate secondary sources. By filtering on a relatively high frequency band around 3 – 4 kHz, it was possible to filter out only the noise being emitted by small gaps and cracks. This produced two new possible weaknesses, one on the porch door, and another on the air valve above the window as seen in the image below.
The next step was to look at those positions in more detail. The acoustic camera was then moved, first to cover the porch door, and secondly to look only at the large living room window. Again the frequency filtering stayed at the approximate same frequency limits. By looking at the coloring, and also listening to the points with the virtual microphone that enables the user to listen to sounds emitted from a single position, these two new noise positions were confirmed as seen from the image and video below.
This same procedure with omnidirectional loudspeaker on the outside of the facade emitting white noise, and the acoustic camera on the inside filming an area of interest, could be used on other walls and windows as well. As seen below on a different window and wall, again the main source contribution is visualized as being the air valve above the window.
Measurements in caravan car
Lier, Norway, October 2015
These recordings were made with the Nor848A-10 1.0 m acoustic camera system with 256 microphones, now replaced with Nor848B acoustic camera system.
Problem
When driving a caravan car, the driver could hear annoying noises from behind the driver’s cab in the living compartment. Even though the noises could be heard clearly, it was very difficult to discover the exact location and cause of the various annoying sounds due to overwhelming background noise.
When the car was standing still with the engine running, no extra noises could be heard, but especially two distinctive sounds that were different from ordinary engine and wheel noise were heard when the caravan was driving. The first sound had the characteristics of a loose screw sliding back and forth on a hard surface, so whenever the vehicle was turning, what sounded like a small piece of metal rolling from one side to the other on a hard surface could be heard. This was only heard when the caravan was in a turn, and not when driving straight. The second noise sounded like squeaking and creaking of wood, which made sense since a lot of the interior consisted of wood. This sound also appeared when the caravan was driving straight, and not only when turning. Since there were several pieces of the interior that consisted of wood, including the floor, it was not know what specific piece of interior made the noise. Based on previous listening experiments when the car was driving, it was thought that the indoor fridge which was positioned next to the entrance door might be root of of both noises.
Measurements
The Norsonic Nor848A-10 1.0m acoustic camera with 256 microphones was used for the recordings. The camera was plugged into an external battery pack for easy transportation and mobility. Although the 40 cm Nor848A-0.4 acoustic camera would have been more mobile, at the time of the measurements it was unavailable. Nevertheless there was more than enough room in the caravan car to use the Nor848A-10.
The measurements were done with one person driving the caravan, and another placing the camera dish on a table and performing the measurements. As the array dish itself weighs in at only 11 kg, it was easy to control it with one arm, and starting and stopping a measurement in the software on the MacBook with the other arm. No tripod was used or needed for the measurements.
Results
Initially the camera was pointed towards the fridge in the center of the living compartment of the caravan and measurements were made as this was thought to be the origin of the sources, In addition to looking at the coloring of the sources, it was also very useful to use the virtual microphone to listen to the sound field from a specific direction. In addition it worked very well to enable the band pass filter when listening in order to filter out background noise such as engine noise and wheel noise.
Seen in the image below is the initial result when filming at the fridge in the living compartment of the caravan with the caravan driving and turning, and when the distinct metal rolling sound was heard. As seen from the image there was no indication whatsoever that the sound originated from the fridge in the middle of the picture. Instead the coloring indicated that the sound originated from a different location. Also listening with the virtual microphone could more or less rule out the possibility that the fridge was the source origin. Instead the camera was positioned to aim to the left of the fridge where the coloring indicated a source.
The position of the true source was seen more or less instantly to be inside a cabinet positioned on the wall of the car. This true source location was also confirmed when listening with the virtual microphone on a recording, and even when holding the ear into the cabinet when driving.
Finding the second noise source proved to be a trivial case after the fridge had already been ruled out as the source. Since most of the interior floor was wooden, it was thought that the creaking might originate from here, and this was the first position the camera was aimed at. Both coloring and listening with the virtual microphone confirmed this.
Normally using acoustic cameras in the interior of cars will often provide poor results, due to high background noise, and a very reverbant noise field. This means that sounds and signals will arrive multiple times at the array at different times, thus making it very difficult to determine the true source location. In this case study it is demonstrated how ordinary beamforming can provide accurate results even in such a situation. This could be thanks to the added space of the interior compartment of the caravan van, in addition to various soft furniture and so on that may help in reducing reverberation effects. In addition the Nor848A-10 1.0 m camera was used that has higher resolution than smaller cameras, in addition to excellent side lobe level suppression.
In addition to being able to locate the various sources correctly by looking at the coloring, also using the virtual microphone in combination with the band pass filter to filter out background noise proved valuable.
Measurements in apartment building
Oslo, Norway, September 2015
These recordings were made with the Nor848A-10 1.0 m acoustic camera system with 256 microphones, now replaced with Nor848B acoustic camera system.
Problem
An apartment complex consists of five floors, with several apartments over four floors, and an attic on the top floor. In the attic an air circulation system is installed to provide circulation in the bathroms of all the apartments in the building. The circulation system is driven by an air fan distributing air through pipes going to all apartments. The air pipes are cemented in to the structure of the building itself, and in some apartments a low frequency structure born noise with the same frequency content as the frequency of the air fan can be heard.
Measurements
The Norsonic Nor848A-10 1.0m acoustic camera with 256 microphones was used for the recordings. The camera was plugged into an external battery pack for easy transportation and mobility. Measurements were made both in the attic at the source location, and also in the bedroom and bathroom of one of the apartments three floors below. The Nor848A-10 was chosen for the recordings over the more compact and mobile 40 cm and 128 microphone Nor848A-4, mainly due to the low-frequency nature of the noise. An array that is larger in size will have better resolution for all frequencies, and will also be able to go lower in frequency content. Even though the Nor848A-10 has a diameter of 1.0 meter, it weighs in at only 11 kg with tripod mounting brackets, and could easily be mounted on a tripod for inspection of the air fan in the attic, or laid down on the floor, or a bed or similar for inspection of the roof in the bedroom and bathroom of the apartment.
Results
Looking at the recordings from the attic, it was clear that the intake fan was the main culprit having a dominating fundamental frequency at 200 Hz. Also several harmonics of the fundamental frequency could be seen in the frequency spectrum.
Since the distance between the attic and the measurement apartment was several floors, and the sound pressure level of the fan noise in the attic was around 40-50 dB, it would have been impossible for the noise in the apartment to be anything other than structure born noise. By inspecting the pipes it was seen that they were cemented in place to the building structure itself without any form of vibration damping measures in place.
The air flows through the pipes and enters the bathroom in the measurement apartment through an air valve in the roof. By positioning the camera so that the measurement direction is straight up at the roof, it was possible to film the structure born noise and get images as seen below. Also by looking at the frequency spectrum one could see how the frequency content of the obtained noise in the bathroom had the same characteristics as the frequency spectrum in the attic. However now more sub harmonics below 200 Hz were seen in the spectrum as seen from the image below.
Recordings were also made in the bedroom of the apartment. The camera was placed on the bed with measurement direction up towards the roof. As was the case in the bathroom, the frequency spectrum also showed tonal tendencies, however here the sub harmonic at 100 Hz was the most dominant frequency. For both bathroom and bedroom the measured sound pressure level was around 30 dB.
Case study: Structure born noise
Measurements in apartment building
Oslo, Norway, April 2015
These recordings were made with the Nor848A-0.4 40 cm acoustic camera with 128 microphones, now replaced with Nor848B acoustic camera system.
Problem
A newly built apartment building consists of several floors with multiple apartments on each floor. One of the ground floor apartments is disturbed by sanitary noise from the apartment above, which is heard whenever the toilets in the top apartment are being flushed. The sanitary noise is heard in several rooms, and the level was measured to be around 30 to 35 dB, which is above the noise criteria set in the regulations. The culprit was thought to be embedded pipelines in one of the corners of the living room, and several measures were made on this area. Although increased insulation improved the noise dampening capabilities of the embedded pipelines, the noise was still heard. Also the improvements could not explain the fact that the noise was also heard in two bedrooms, one of which was not adjacent to the living room.
After the construction of the apartment complex was finished, the contractor in charge of piping had gone bankrupt, and also there existed no documentation or blueprints on the actual position of the piping system. Since the first measures did not produce the desired outcome and had little overall effect, it was quite certain that the problem was also located elsewhere. Given that the measured sound pressure levels were so low, it was very difficult to detect any real change in sound pressure level from different measurement positions. Also trying to locate the origin by hearing in different positions proved to be futile when trying to determine source position. In addition to hearing the noise in several rooms, it was also heard in the hallway and it was uncertainty whether there existed a single source location or several.
Measurements
The 40 cm Nor848A-0.4 acoustic camera with 128 microphones was used for the measurements. Although the external battery pack could have been used for extra mobility, the easy access to power outlets meant that the camera could be plugged in directly in the wall outlet in the different measurement rooms.
The light weight and small size of the array ensured easy portability and also great flexibility in measurement positions. The camera could be set up on the tripod, but also positioned on the floor, in a bed, on a couch or just hold it by hand all based on what part of the room one would like to listen to and get the acoustic image from. The 40 cm camera proved to be very handy in a measurement situation where it was not given beforehand exactly where the source of interest was positioned. In that way, the camera could easily be used to scan around different parts of different rooms.
To actually get a recording of the sanitary noise, one person was positioned in the apartment above and told when to flush via cellphone from the apartment below. This was repeated several times, with recording time around 20 seconds each time to cover the entire event duration.
Results
The first recording of the flushing event did not produce the desired result. The source colouring was on a wall adjacent to the embedded pipes, but no pipes were in that part of the building, and it was impossible that the noise could originate from that position. By looking at the recording and listening to the center of the colouring on the video, one could quickly realise what was happening. Because of the very low noise levels, the camera was picking up the strongest source in the room, which in this case was a reflection on the wall from the noise of the cooling fan of the Macbook that was being used for the recordings. This was solved by placing the Macbook in a different room and closing the door. The ethernet cable from Macbook to the acoustic camera was small enough in diameter that the person doing the recordings could sit virtually anywhere he or she pleased, as long as the cable was long enough. On the second try one could clearly hear the sanitary noise for a period of around 8 to 10 seconds. When filming at the location of the encased pipes in the corner of the living room, the acoustic camera did not pick up any energy at that position, but rather the colouring was upwards towards the roof and outside the field of view of the camera at that measurement position. This was a very strong indication that the origin of the noise was in fact located somewhere else, and the camera was directed accordingly for the subsequent measurements.
When pointing the camera towards the ceiling it became apparent that the source was at this position. No other spots, either from the embedded pipes in the corner of the living room, or from the hallway, had any visible colouring, and could be excluded as the likely position of the origin of the source. By using the virtual microphone, which enables the user to listen to a specific position in both real time and in a recording, one could clearly hear the sanitary noise from the recorded position at the roof. Also by enabling the bandpass filter one could further be able to filter out background noise and very clearly hear the sanitary noise from the piping.
Based on these measurements one could deduct that the water pipes in were in fact not positioned as previously thought. In addition to measurements in the living room subsequent measurements were also made in the two bedrooms. These measurements displayed the same result, the energy of the noise did not come from the walls, but were confined to the roof. All in all, this was a very strong indication that the problem was piping located in separating floors between apartments.
Measurements in car factory
Korea, July 2015
These recordings were made with the Nor848A-0.4 40 cm acoustic camera with 128 microphones, now replaced with Nor848B acoustic camera system.
Problem
A well-known car manufacturer in Korea were testing an automatic car door window intended for one of their new car models. The window is driven by an electric engine positioned in the middle of the car door that drives the window up and down. When the window was driven up by the electric motor, a short timed high pitch squeak noise could be heard as seen from the level versus time plot below. The squeak noise was obviously connected to the window and the car door, but the localisation of the source of the problem proved difficult.
Measurements
The Nor848A-0.4 40 cm camera with 128 microphones was used for the recordings. The camera was positioned at a distance of 2.0 m from the car door, with the front-end of the camera pointed straight at the door. The recording consisted of an event of the car window going up whilst being driven by the electric engine.
Results
In addition to the high pitch squeak noise, sound from the electric engine could also be heard which was around 670 Hz as seen below. Since this sound was present during the entire recording, it could be easy to misinterpret the results and pinpoint the source of the high pitch noise to the wrong location without doing proper analysis. As seen in the image below, the location of the electric engine was in the middle of the door.
As opposed to the stationary sound from the electric engine, the high pitch squeak noise was only happening for a brief period of time, around 300 ms. By defining a scene in the sound level indicator pane of the acoustic camera software, it was then possible to analyse events happening only within the time frame of the extent of the scene. The scene is made by simply dragging the cursor over the region of interest in the software as seen below.
Now by calculating the average spectral density over that part of the recording, it was easy to see that in addition to the noise from the electric motor around 670 Hz, a high pitch tonal noise with a fundamental frequency around 1890 Hz could also be seen in the frequency spectrum. In addition, harmonics of the fundamental frequency up to the 7th harmonic was visible in the spectrum.
Now by filtering the frequency content of interest around the fundamental frequency, it was possible to get a clearer indication of the position of the source of the high pitch squeak noise as seen below.
In the software, the so called virtual microphone can be used, which makes it possible to listen to and analyse sounds from a given point in the image. By positioning the virtual microphone on top of the indicated position of the squeak source, and looping the scene over and over again, it was also possible to look at the spectrogram for that scene and that position. Clearly in addition to the stationary sound coming from the electric engine, bursts of energy at higher frequencies made by the source of the squeak noise were also seen.
Lastly, due to the tonal nature of the noise it was possible to switch from the ordinary, robust wideband algorithm, to the adaptive algorithm for narrowband applications. This further helped in pinpointing the exact location of the problem. Higher resolution could also be achieved by zooming further in in the image.
Measurements in conference hotel
Trondheim, Norway, November 2014
These recordings were made with the Nor848A-0.4 40 cm acoustic camera with 128 microphones, now replaced with Nor848B acoustic camera system.
Problem
A conference hotel is using modular walls to divide large halls into several smaller conference rooms. The rooms are divided by modular walls that provide several different opportunities for subdivision and multipurpose use of the large area spaces.
When measuring the sound insulation between adjacent rooms through the modular walls, the resulting value was found to be too low, and noise from one conference room could possibly disturb listeners at adjacent rooms. The dividing modular walls cover large areas, and are as high as 7 meters from bottom to top, which makes intensity measurements with hand held sound level meters difficult. The room dividers could have several weak points, which were not easily identifiable. It was thought that identifying and fixing the weak points in the individual modular walls would help increase the overall sound insulation capabilities of the entire wall element.
Measurements
The measurements were conducted with the Nor848A-0.4 40 cm and 128 element acoustic camera. The camera was plugged into an external battery pack for easy transportation and mobility. The measurement procedure consisted of choosing two adjacent rooms divided by a modular wall of interest. One of the rooms was chosen to act as receiving room, where the acoustic camera was positioned. A noise source and omnidirectional loudspeaker generating white noise at high volume was positioned in the source room. The speaker was placed in one of the corners of the room furthest away from the dividing wall, in order to achieve as diffuse source noise field as possible.
Due to the large size of the modular walls, the camera was pointed to different areas of the walls, and several measurements were made. The individual measurements could then be examined further in post-processing analysis. Due to the source being used at high volume in the sending room, cracks and gaps in the modular walls would appear as small noise sources at specific location on the walls when recording with the acoustic camera in the receiving room.
Results
The acoustic camera was able to locate several weak spots on the walls, even though the range where differences could be discovered were for certain areas below 0.05 dB.
The measurement system’s virtual microphone feature was also very helpful during live measurements. With this function you can scan and listen to the desired spots in the image, and also filter the listening function to desired frequency range. This made it possible to scan along edges and hear differences in frequency from different points. A change in frequency may indicate a sound leakage. Also by using the spectrogram function to get a visual representation of the spectrum of frequencies as they varied with time, one could further indicate a leakage at various parts of the walls.
A very useful function is the so called acoustic eraser, which is a functionality that enables source suppression in order to find interesting plotting points. Seen in the video below is a recording of two walls meeting at a corner. The coloring is smeared over a larger area than usual, which may indicate the presence of several sources of roughly equal strength located at close proximity to one another. Or in this case, a weakness or small gap in the walls located so closely that they initially may be interpreted as a single source. When the acoustic eraser is enabled it is seen as a red circle with a white x and placed on a point in the image to suppress a source. By enabling the acoustic eraser, and dragging the point suppressor to the desired location, it was easy to locate the two individual points of interest. Further analysis could be conducted by placing the virtual microphone on the point of interest.
Measurements on LNG gas terminal
Stavanger, Norway, September 2014
These recordings were made with the Nor848A-10 1.0 m acoustic camera system with 256 microphones, now replaced with Nor848B acoustic camera system.
Problem
A large LNG gas facility (approximately 300m x 150m) producing 300 000 tons of LNG annually is situated in a terminal area with the nearest populated area at a distance of around 1 km. Within the gas production facility, a low-frequency tonal noise at around 500 Hz is generated causing complaints from nearby neighbours. The tone imposes a more stringent noise requirement on the facility, forcing noise reducing actions being made on the source.
In addition to the tonal noise, the entire LNG gas facility is rich in noise emitting sources, including lossing and loading of maritime vessels, which further complicates the source location of the single tonal noise source. Also the location of the facility at the coastal regions of the western part of Norway, ensures that windy conditions are frequent, with wind noise further impeding the quality of acoustic recordings.
Based on measurements with hand held sound level meters, the problem area was narrowed down to be a large pipe in the midst of the facility. However it could not be determined if the emitted tonal noise was from the entire pipe itself, or if it originated at a specific part of the pipe. There was also uncertainty whether there existed multiple sources within the pipe, for instance at both the base and top layer. In the worst case the noise insulation would have to be performed over the entire pipe length, which could have been a very expensive solution.
Measurements
The measurements were conducted over two subsequent days with the Nor848A-10 1.0m and 256 element acoustic camera. The camera was plugged into an external battery pack for easy transportation and mobility. The entire measurement system could easily be moved around to different positions to get a noise mapping of different sides of the pipe. Different positions would also ensure that noise sources being different from the source of interest would not inflict too much on the measurements. The primary measurements were conducted at a distance of approximately 25-30 meters from the pipe. In addition measurements were made close to the source from 2-5 m distance by climbing up onto the pipe with the camera. Since the flight of stairs were too narrow to get the 1.0 m camera through the stair’s safety rails, this was solved by hoisting the camera up and down by rope.
Results
By positioning the center of the array towards the pipe and adjusting the frequency to display only coloring within the 500 Hz 1/3-band, the noise source was located within seconds, and the source producing the tonal part from the pipe was detected. Measurements from different measurement positions also confirmed the source location.
By placing the virtual microphone on the localised source and using the spectrogram function, it was easy to verify the position of the source emitting a tone at 460 Hz.
Although the measurement location had quite windy conditions, the wind noise did not affect the measurement results at all. Wind noise can be viewed as spatially white, which means that wind noise sampled at different places in space, as is done with the Nor848A, is not correlated from position to position. When many different signals from many microphones are added in the beamforming algorithm, the wind noise will be added out of phase and attenuated proportional with the number of microphones being used.
With the acoustic camera it was possible to detect the tonal sound of the most crucial parts of the turbine. This meant that the facility could focus on and implement noise reduction actions in the right places. After pin pointing the location of the noise source, further analysis could be made with measurements performed closer to the source of interest in order to further determine the position and cause of the generated tonal noise.
Another useful function is the so called acoustic eraser, which is a functionality that enables source suppression in order to find interesting plotting points. Seen on the images from the acoustic camera software on the next page is a recording of the pipe without and with point suppression enabled. Seen in the bottom image, the acoustic eraser is seen as a red circle with a white x and placed on the tonal source in the image to suppress it. By enabling the acoustic eraser, and dragging the point suppressor to the desired location, one could further identify if the pipe had other locations that generated tonal noise. As seen in the bottom image, no such additional tonal sources were found.
- Case study: Office noise leakage
- Case study: Breakout noise from café
- Case study: Traffic noise in apartment
- Case study: Wall leakage test in lab
- Case study: Interior caravan car noise
- Case study: Structure born noise
- Case study: Car window squeak
- Case study: Apartment sanitary noise
- Case study: Modular walls
- Case study: Industry tonal noise
- Case study: Office noise leakage – 中文
- Case study: Breakout noise from café – 中文
- Case study: Traffic noise in apartment – 中文
- Case study: Wall leakage test in lab – 中文
- Case study: Interior caravan car noise – 中文
- Case study: Structure born noise – 中文
- Case study: Car window squeak – 中文
- Case study: Apartment sanitary noise – 中文
- Case study: Modular walls – 中文
- Case study: Industry tonal noise – 中文