Blog

During last week I conducted the measurements inside the anechoic chamber of the AudioLab of the University of York. The measurements started on Thursday the 18th and ended on Thursday the 25th of July 2019. With the help of Andrew Chadwick, the tree trunk was screwed on two wooden planks in order to be stabilised, allowing it to stand vertically. He also built the rotating mechanism that would hold the microphone for the measurements and placed the tree on top of it in the centre of the anechoic chamber. The loudspeaker was then placed in the appropriate position, all distances were measured and the angles of rotation were found and marked with the help of lazer pointers. Using a Fireface interface, the equipment were connected to Cubase 5.2 which was responsible for playing back and recording the sine sweeps. The experimental procedure was ready be carried out.

For each measurement session, 3 sine sweeps were recorded for each receiving position, and 12 receiving positions were obtained from each side of the trunk ranging from 180 to 15 degrees, split in 15-degree-intervals. The measurements were repeated for 4 different sides of the trunk by rotating it by 90 degrees each time. After completing each measurement session, MATLAB was used to split each individual sine sweep, deconvolve it into an impulse response (IR) and save it under the appropriate name. The IRs obtained underwent some basic time domain analysis by plotting and comparing their graphs.

The measurements were repeated and analysed multiple times each time improving the set up in an attempt to eliminate unwanted reflections detected in the plots. The final measurements were conducted on Thursday the 25th of July.

Figure: Preparing the setup for the measurements inside the anechoic chamber

Searching for the appropriate tree for the experiment was a long process. Various tree sergeants across North Yorkshire were contacted. Finally, a cylindrical and smooth Maple trunk was found in Scarborough at the beginning of July. The tree was cut on Monday the 8th of July by Ali Nunn from “Mundaka Tree Care”. The trunk was handled by professionals, cut to the appropriate size and delivered to the department of Electronic Engineering of the University of York on Thursday the 18th of July. Its dimensions were measured using measuring tape. Its height was found to be between 120cm and 125cm at different sides, and its average circumference 102cm. The tree was almost cylindrical with some of its sides being rougher than others.

Figure: Picture of the Maple trunk taken as soon as the tree was cut down.

The plan for the experimental procedure was to obtain measurements for the sound reflection from the surface of a cylindrical trunk at different angles around it. The tree will be placed on the centre of the anechoic chamber with a loudspeaker at a fixed position directly facing towards it. Then a microphone will pick up the sound reflections at different angles around the tree. This can be achieved by building a metallic mechanism to hold the microphone at a fixed distance and allowing it to rotate around the maple trunk.

The experimental procedure was planned thanks to the help of Andrew Chadwick [1], the Technical Assistant of the Audio Lab of the University of York. With his design, the angles will ranging from 180 to 15 degrees. A 0-degree-angle will not be required as the source and receiver cannot exist in the same position. Laser pointers will be used to mark and track the exact positions for the measurements.

Andrew will be responsible  for designing and building the mechanism. In the meantime, I will need to find a cylindrical tree with hard and smooth surface for the experiment. If such a tree is not found, the procedure will be carried out using a normal cylinder instead, and the project’s focus would shift significantly.

Figure: A diagram of the set up of the anechoic chamber with the exact position of the tree, loudspeaker and microphone as well as the angle of rotation around it in a bird’s eye view.

[1] Andrew J. Chadwick, “University of york directory :: People directory :: Mr AJ chadwick,” https://www.york.ac.uk/directory/user/searchdetail.cfm?scope=staffref=M95.

As shown in previous studies, the acoustics can vary significantly depending on the type of forest. The shape of the trees, together with the roughness and the density of their bark result into different values for the resultant scattering and energy absorption. When searching for the appropriate tree trunk for this project I contacted Sam Gilchrist, a Tree Surgeon from Manchester (UK), to find out more information about the types of trees, their texture and other characteristics. In summary he stated the following:

“The trunks can have a hard or soft surface, with the roughness of the bark coming in many variations. The texture ranges from very soft or having craters, all the way to consisting of thin layers that can often peel off. The hardness and roughness of the bark can change significantly with the age of the tree. After cutting down a tree, water slowly starts to escape making the bark softer and causing it to peel off.”

To narrow down the options, in order for the experimental procedure to be possible,
realistic and produce useful results, the following parameters were taken into  consideration when searching for an appropriate trunk:

• Be accessible in the UK
• Have a cylindrical shape
• Have a hard but smooth surface
• Have a maximum diameter of 0.5m
• Have a maximum height of 2m

Figure: Chopped trunks in a forest, taken from Pixabay.

During the past week I revisited Forest Acoustics, focusing on the acoustics and sound propagation in such an environment. My aim was to investigate the effect a forest has on sound waves.

Noise Reduction in Forests

It has been noticed that trees and other vegetation are capable of significantly reducing urban noise. A study investigating how vegetation can reduce noise caused by a train, showed that 50 m of forest can result into an extra 8-9 dB noise reduction when compared to a grass-covered country [1]. The absorption of a surface can be determined using the absorption coefficient, which varies between different materials and textures. In a forest, three main categories of surfaces have an effect on the absorption and reflection of sound:

• Tree trunks
• Brunches and foliage
• The forest floor

Sound Absorption in a Forest

When considering a tree on its own, its trunk appears to be the most neutral surfaces as it has a very low absorption coefficient which results into maximum reflection of sound. While brunches behave in a similar way, they tend to be covered by leaves and other vegetation; therefore, the two can be considered as a single medium. Leaves and foliage absorb high frequencies, usually greater than 2kHz [1], and sometimes slightly amplify middle ones. The range of frequencies being absorbed increases with the size of the leaves.

Trees and bushes tend to have a very small absorption coefficient, but noise in forests significantly attenuates. This is mainly because of the forest floor, which has greater noise reduction capabilities, and absorbs some of the lower and middle frequencies [2]. Reethof in 1977 used a Standing-Wave Tube to measure the absorption coefficients of different trees and of the forest’s ground at different frequencies [3]. As part of his study he examined how the texture of the ground affects the absorption of the ground. The results show that the absorption increases with moisture, roughness and vegetation covering the floor.

Conclusion

Overall, while the surfaces in forest tend to have small absorption coefficients, multiple reflections and sound scattering taking place by the tree trunks causes sound do diffuse and attenuate. Next, I will be investigating the scattering of sound in a forest together with Morse book “Vibration and Sound” [4], which studies the scattering of sound in the surface cylinder.

Figure: Sound reflections in different parts of a tree taken by Dobson [2]

[1] Bullen, R., and F. Fricke. 1982. “Sound Propagation through Vegetation.” Journal of Sound and Vibration 80 (1): 11–23.

[2] Dobson, Martin, and Jo Ryan. 2000. Trees and Shrubs for Noise Control. Arboricultural Advisory & Information Service.

[3] Reethof, G., O. H. McDaniel, and G. M. Heisler. 1977. “Sound Absorption Characteristics of Tree Bark and Forest Floor.” In In: Heisler, Gordon M.; Herrington, Lee P., Eds. Proceedings of the Conference on Metropolitan Physical Environment; Gen. Tech. Rep. NE-25. Upper Darby, PA: US Department of Agriculture, Forest Service, Northeastern Forest Experiment Station. 206-217. Vol. 25. https://www.fs.usda.gov/treesearch/pubs/11547.

[4] Morse, Philip Mccord, Acoustical Society of America, and American Institute of Physics. 1948. Vibration and Sound. Vol. 2. McGraw-Hill New York.

For the beginning of my literature review I started by presenting the fundamentals of acoustics and the way sound waves behave and interact with their surroundings. The aim is to establish different terms which are going to be used considered and used throughout the report.

The Behaviour of Sound in Space

Acoustics can be defined as the science of sound [1]. Sound waves, like all mechanical waves, as they interact with their surroundings they can experience different effects such as absorption, reflection, scattering, diffusion, diffraction and more. When writing this section of the literature review, the aim was to understand how a forest would act as a system and hence, how sound would behave within such an environment.

The absorption coefficient of a tree can determine the energy lost when a wave bounces on its surface, and its effect on the harmonic content of the reflected ray. The scattering indicates how the angle of reflection changes depending on the roughness of the surface and how sound diffuses across the space. Diffraction allows sound to travel around the tree trunk making the source audible to the receiver even when the direct path is blocked. In a forest, low frequencies will diffract due to their longer wavelength, while higher ones will scatter towards different directions [2]. Several of these phenomenons can be noticed through studying, the direct path, early reflections and reverberations of Impulse Responses (IRs) in a known system.

Conclusion

After obtaining an understanding of the way sound behaves in space, it is time to investigate more specifically how certain effects apply in a forest. Next I am going to look into literature that studies the way absorption and scattering takes place in forests, Trees and other vegetation.

Figure: Direct sound and early reflections in an enclosed space taken by IKoustic [3]

[1] Kuttruff, Heinrich. 2006. Acoustics: An Introduction. CRC Press.

[2] Reethof, G., O. H. McDaniel, and G. M. Heisler. 1977. “Sound Absorption Characteristics of Tree Bark and Forest Floor.” In In: Heisler, Gordon M.; Herrington, Lee P., Eds. Proceedings of the Conference on Metropolitan Physical Environment; Gen. Tech. Rep. NE-25. Upper Darby, PA: US Department of Agriculture, Forest Service, Northeastern Forest Experiment Station. 206-217. Vol. 25. https://www.fs.usda.gov/treesearch/pubs/11547.

[3] IKoustic. Acoustic glossary, 2019. URL https://www.ikoustic.co.uk/acoustic-glossary

After obtaining a better understanding on Forest Acoustics and reviewing previous research on the field, I decided to investigate different ways of modelling a virtual forest for my project. This directed me towards Physical Modelling Synthesis and Digital Waveguides, as it has been used twice before for the same purpose.

What is Physical Modelling

Physical Modelling Synthesis makes use of mathematical models to mimic the physical characteristics of an instrument [1]. Digital Waveguide models are physical models made up using delay lines, digital filters and other components which have an effect on the signal chain, targeted towards a realistic virtual representation of a musical instrument [2]. The wave propagation and hence the delay lines can be considered in 1D (string), 2D (membrane) or 3D (space). For the modelling of a virtual forest, a 2D representation would be sufficient with the tree trunks acting as circles on a horizontal plane, but a 3D representation would result into significantly more accurate results due to more realistic scattering and sound attenuation.

Modelling a Virtual Forest

In a 2D system, as sound moves across the membrane, its intensity decreases over time. When the signal reaches a tree trunk, absorption and scattering are taking place due to the medium, texture and shape of the surface. The absorption coefficient will act as a digital filter, diminishing the harmonic spectrum of the signal and attenuating the sound. Scattering can be modelled by applying a different resistance on the two mediums (from air to the tree trunk), which varies depending on the wavelength of the wave. The process is repeated depending on the number of times a sound wave reflects on surfaces before reaching the receiving position. After a certain number of reflections, some sound waves completely lose their energy and disappear.

Conclusion

I discussed with Professor Damien Murphy and Dr Amelia Gully about the use of physical modelling for my own project, and they suggested that I should prioritise other things due to the complexity of the topic. They still directed me towards further literature to approach and understand Digital Waveguide Synthesis. This made me change my project aims and reconsider my project approach, targeted more towards improving already existing models rather than inventing my own.

Figure: Digital Waveguides with different resistances on the delay lines resulting into Signal Scattering, as shown by J. Smith [2].

[1] Välimäki, V., Pakarinen, J., Erkut, C. and Karjalainen, M. (2005). Discrete-time modelling of musical instruments. Reports on Progress in Physics, [online] 69(1), pp.1-78. Available at: https://iopscience.iop.org/article/10.1088/0034-4885/69/1/R01/meta

[2] Smith, J. (2006). A Basic Introduction to Digital Waveguide Synthesis (for the Technically Inclined). [online] Center for Computer Research in Music and Acoustics. Available at: https://ccrma.stanford.edu/~jos/swgt/

Ever since the beginning of this project I have been thinking of different ways to approach it. After attending two different meetings with the project’s supervisor Dr Frank Stevens and conducting individual research, the following aims were set:

Project Aims

Set up an experiment to study the scattering of sound in the surface of a cylinder at different angles using Impulse Response (IR) measurements. Compare the readings to the theoretical values obtained by Mores on his book Vibration and Sound [1]. Repeat the experiment replacing the cylinder by a real tree trunk and use the readings to compare the two.  Analyse the IRs obtained to estimate the scattering and absorption of the tree trunk; use them to implement a digital filter. Create a simulation of the tree trunk in the anechoic environment virtually based on the real one. Model a forest based on the virtual tree trunk created, and study its acoustic characteristics.

Planning and Structuring the Literature Review

The aims set are challenging, but can be achieved, while taking into account that new ideas and problem will arise. While studying the background of forest acoustics, the topic was split into different components aiming to conduct individual research on each before the final project initiation. The areas aimed to be researched as part of the literature review are the following:

1. Acoustics and Sound waves
2. The Background of Forest Acoustics
3. Acoustic Measurements, and Impulse Responses
4. Artificial Reverberation
5. Physical Modelling and Ray Tracing
6. Virtual Simulation of a Real Acoustic Space

Figure: Picture of a forest taken from Miscellaneoushi.

[1] Morse, Philip Mccord, Acoustical Society of America, and American Institute of Physics. 1948. Vibration and Sound. Vol. 2. McGraw-Hill New York.

I started my research by looking at existing papers which investigate different aspects of Forests Acoustics. This was chosen to be the starting point of this project in order to familiarise with the field, and break it down into different components.

Forest Acoustics

The topic was primarily addressed by Carl F. Eyring in 1946 with his study of “Jungle Acoustics”, focusing on the different physical properties of the jungle such as temperature, wind velocity and humidity, and their effect on the timbre and sense of direction of sound [1]. His publication approached the forest as a whole, showing some of its interesting acoustic characteristics, which inspired several researchers towards investigating the way in which sound behaves within such an environment.

In 2001, a different approach was taken by Sakai, who investigated the acoustics of a bamboo forest [2]. After noticing some interesting characteristics on the projection and reverberation of sound in such an environment, he conducted acoustic measurements targeted towards obtaining and analysing Impulse Responses. Through performing auralisations, his study showed that music sources with higher frequency components are more appropriate in such an environment.

Virtual Models of Forests

The next two papers approached the topic using virtual representations of a forest to obtain Impulse Response Measurements. The first one used a network of Digital Waveguides named “Treeverb”, to simulate the acoustics of a forest [3]. This paper mainly focused on the effect of the shape of the tree trunk towards the scattering and reflection of sound, by modeling all trees as perfect cylinders distributed randomly in space, with no energy absorption.

Based on that idea, a second paper was published focusing on simulating and studying the acoustics of open-air spaces including forests, using a “Waveguide Web” [4]. This network made use of the Digital Waveguides used in Treeverb, as well as Scattering Delay Networks, resulting into more realistic results. Instead of approaching the trees as perfect reflecting surfaces, the scattering junctions used filtered off the reflected signal to create the effect of sound attenuation.

Conclusion

I chose these papers as my starting point, as the first two papers notice some of the unique characteristics of forests and perform different types measurements, while the next two focus on simplifying the system enough in order to model a certain characteristic of the forest. Taking these into account, I was directed towards Physical Modelling and Digital Waveguide Synthesis. This made me want to obtain a better understanding of artificial reverberation and sound attenuation in a forest. In addition to that, I am currently looking into additional literature to obtain a better understanding on Physical Modelling Synthesis and Digital Waveguides.

Figure: The different paths of sound between a source S and a receiving position R in a forest consisting of 3 Trees, as expressed in “Treeverb Digital Waveguide” [4]

References

[1] Eyring, C. (1946). Jungle Acoustics. The Journal of the Acoustical Society of America, [online] 18(1), pp.245-245. Available at: https://asa.scitation.org/doi/abs/10.1121/1.1916362

[2] Sakai, H., Shibata, S. and Ando, Y. (2001). Orthogonal acoustical factors of a sound field in a bamboo forest. The Journal of the Acoustical Society of America, [online] 109(6), pp.2824-2830. Available at: https://asa.scitation.org/doi/10.1121/1.424564

[3] Spratt, K. and Abel, J. (2008). A Digital Reverberator Modeled After the Scattering of Acoustic Waves by Trees in a Forest. In: AES 125th Convention. [online] San Francisco, CA, USA: AES. Available at: http://www.aes.org/e-lib/browse.cfm?elib=14801

[4] Stevens, F., Murphy, D., Savioja, L. and Valimaki, V. (2017). Modeling Sparsely Reflecting Outdoor Acoustic Scenes using the Waveguide Web. In: IEEE/ACM Transactions of Audio, Speech and Language Processing. [online] IEEE. Available at: http://eprints.whiterose.ac.uk/115692/1/modeling_sparsely_reflecting.pdf

After the end of the Easter break and the completion of the last two assignments, the taught side of the Master’s degree has come to an end, indicating the initiation of the End of Year Research Project, which will be my main area of focus until the beginning of September.

The Project

The project is focused on “The Measurement and Analysis of Forest Acoustics”, supervised by Dr Frank Stevens and Dr Jude Brereton. I chose this topic because of the following reasons:

• I want to find out how forests and vegetation can be used to minimise urban noise.
• I find interesting the process of isolating and simplifying components of a complex system in order to study each individually. Outdoor acoustic measurements are complex and unpredictable due to the asymmetry of nature and the various parameters affecting the system, making forests very hard to predict and simulate.
• Minimum research has taken place in the field providing plenty of space for new ideas and opportunities for novel work.

The first deadline of this project is the Literature Review which needs to be submitted on the 13th of June. As a result, the next few weeks will be focused on determining the approach and aims of the project. This will be achieved through reviewing existing literature, in order to obtain a better understanding of the field. Then, the objectives for the final project will be created in order to meet the aims set.

Starting the Project

When approaching a project of such magnitude, it is important to understand the current state of the field. A small amount of research has been conducted on Forest Acoustics; therefore a good starting point would be the studying and reviewing of existing papers to identify the different components of the field. This on its turn, would help determine the final project’s approach as well as the next steps for the literature review.

Figure: Research Project Initiation