k-Wave

1. nsaquib
   

   Member
   Joined: Dec '13
   Posts: 2

   Hi,

   Thank you for this great open source effort and the forum. This is really impressive!

   I am very new to the field of acoustic wave simulation, and I apologize if this is a trivial question. For a research project, I will have to simulate a city environment with vehicles moving through streets. As part of the simulation, I will have to incorporate wave propagation of sounds created from these vehicles (say, a horn sound). The simulation does not have to be very accurate through the buildings, I only need to see the approximate amplitude vs. distance traveled over time across the environment for each sound source.

   I am reading the documentation and I am assuming that I can model the city as a heterogeneous medium (two types of regions - air and building blocks). However, I am wondering how I can model the moving vehicles as a sound source. Should I think of each source as a time varying pressure field defined over a path in the grid?
   Also, how can I create a realistic time-varying pressure field for, say, a truck or a car horn sound? In the examples, the magnitudes are defined in [au]. What unit is that exactly? Any suggestion and ideas would really help me. Thank you.

   Posted 10 years ago #

2. Bradley Treeby
   

   Developer
   Joined: Mar '10
   Posts: 1,640

   Hi nsaquib,

   As a first approximation, you could certainly try to model your environment as a combination of air and building blocks. This will capture the bulk behaviour of the medium, i.e., free-space propagation and reflections from buildings. You will also need to think about how to model the ground.

   For moving sources, you could try following the Doppler Effect example (you can access this example and the associated m-file from the k-Wave entry in the MATLAB help browser). You could either extend this to include several moving sources, or you could model them separately and add the output fields together at the end.

   The source you define can be anything, provided that the frequency content is supported by the spatial grid (see Sections 3.2 and 3.4 in the k-Wave Manual). For example, you might try to find a recording of a truck or car horn, import this into MATLAB, and then assign this to source.p (with the appropriate resampling if needed).

   The pressure is specified in Pascals. The units [au] (meaning arbitrary units) are used in some examples as there is no significance to the absolute values themselves, only their relative values.

   Hope that helps,

   Brad.

   Posted 10 years ago #

3. nsaquib
   

   Member
   Joined: Dec '13
   Posts: 2

   Hi Bradley,

   Thank you very much for your response. In fact, I followed the Doppler Effect example immediately after posting here, and got some nice results to start my investigation. Good to get the assurance that I am following the right example.

   A question about the model: is there a reference that states the alpha coefficient values for air and building blocks? Is the power law of absorption a good model when it comes to sound propagation in air and buildings?

   Also, I tried the example with a terrain where the top and bottom are mostly building blocks (with only small gaps between them) and the left and right sides are kept open. The simulation seems to blow up in that case (using a real car sound with the sensor, resampled, 8 kHz). Any idea what could be going wrong here? The simulation also blows up (or nothing happens in the visualization) if I try realistic density and sound speed values, for example, 1.18 kg/m^3 for air and 1400 kg/m^3 for bricks, and 330 m/s for air and 3200 m/s for bricks.

   Thank you for your time.

   Posted 10 years ago #

4. Bradley Treeby
   

   Developer
   Joined: Mar '10
   Posts: 1,640

   Hi nsaquib,

   Unfortunately I don't have a reference for the absorption in air to hand, but there's bound to be something in the room acoustics and outdoor acoustics literature.

   For your first example (with the real horn), does it also blow up if you manually define your own source, for example, using toneBurst?

   Regarding the instability, there's a few things you could try. First, set 'PlotScale' to 'auto' just to make sure it's not a visualisation problem. Second, try making the material properties of bricks much closer to air, and then gradually increasing the impedance to see if there's a point at which things go wrong. Third, try reducing the value of the CFL (for example, as an input to makeTime).

   Hope you manage to get it working,

   Brad.

   Posted 10 years ago #

RSS feed for this topic

Reply

You must log in to post.