Multiple Sources in 3-Dimensional Space

k-Wave User Forum » General Acoustics » Acoustic Simulations

Multiple Sources in 3-Dimensional Space

(12 posts) (3 voices)

Started 6 years ago by char7114
Latest reply from lbruju

char7114

Member
Joined: Mar '18
Posts: 2

First and foremost, I would like to express my appreciation for the release of this package and its continued efforts. Most impressive.

I am still wrapping my head around its functionality however I had a question in regards to creating a 3-Dimensional space with multiple di-pole and mono-pole sources. I have managed to incorporate multiple di-poles into the same space however I cannot seperate them into different locations in the grid and I cannot find a way to effectively combine mono-poles in the same space. Is there a simple way to add sources of different characteristics into different locations in the grid?

My goal is to model the acoustics generated by a helicopter, which I believe will require velocity sources to represent the rotors and monopoles for the power plant and transmission. Each of these will be of different frequency and magnitude. I understand that the sources can be located by defining the location in the grid as a 1 in the matrix but am unsure how to specify an individual source to its own location.

Many thanks.

Posted 6 years ago #
Bradley Treeby

Developer
Joined: Mar '10
Posts: 1,643

Hi char7114,

There are some details on the ordering of the inputs in the Monopole Point Source In A Homogeneous Propagation Medium Example (this is accessible from the MATLAB help browser), and also in the k-Wave manual.

Hope that helps,

Brad.

Posted 6 years ago #
char7114

Member
Joined: Mar '18
Posts: 2

Brad,

I have worked through your user manual as well as the available examples and have managed to combine pressure and velocity sources in the same space whilst utilising the column wise indexing. Thank you for your help in this regard. I am however currently unable to combine velocity sources of different direction (.ux, .uy and .uz) into this space. Is this something which should be possible? I can only get the individual sources to all propagate in the same direction through specification of the source.u matrix which I am using to reference the sources to the desired mask location. Could this possibly be achieved by creating a second mask and incorporating it into the same space? Perhaps I am just defining my source matrix incorrectly.

Again, thank you for your aid.

Posted 6 years ago #
Bradley Treeby

Developer
Joined: Mar '10
Posts: 1,643

Hi char7114,

It's not possible to define a different source mask for the different components of the velocity. A work around is to use a single source.u_mask that covers all of the sources. Then, for each component of the velocity, for the grid points that are not transmitting, set the source term to zero.

Hope that helps,

Brad.

Posted 6 years ago #
lbruju

Member
Joined: Mar '18
Posts: 19

Hi,

I am trying to simulate the pressure field generated by 8 rings defined on a spherical cap. All the points on a same ring have the same p_source_input signal, but each ring has a different signal.
I made the simulation by two different ways.
1) I simulate the pressure field generated by one ring after the other. I put p_source_many=0. kspaceFirstOrder3DGC-OMP.exe is executed for each ring and I sum the results (with a sensor_mask which contains all the points of the FOV).
2) I simulate the pressure field generated by all the rings at the same time. I put p_source_many=1. kspaceFirstOrder3DGC-OMP.exe is executed only once.

I don't understand why I don't obtain the same result with both simulations.
The only difference is that all the inputs (index values and input signals) are separated in 16 different matrices (8 matrices for p_source_index and 8 for p_source_input ) in the first case. Whereas the inputs are separated in only 2 matrices (1 for p_source_index and one for p_source_input ) in the second case.

Do you have an idea of where the error could come from?

Thanks a lot,
lbruju

Posted 6 years ago #
Bradley Treeby

Developer
Joined: Mar '10
Posts: 1,643

Hi lbruju,

Nope, provided your simulation is linear, they should give the same results. Perhaps it's an error in the indexing in the second case?

Brad.

Posted 6 years ago #
lbruju

Member
Joined: Mar '18
Posts: 19

Hi Bradley,

Thanks for your answer.
I have checked the values, and they seem to be the same.

In the first case I have:

P_SOURCE_INDEX:
p_source_index0=np.array[[3363674]] #only one point for the top of the spherical cap
p_source_index1=np.array[[3273700,3274150,3274749,...]] #index of points on the 1st ring
ETC... until
p_source_index7=np.array[[2891235,2891684,2892284,...]] #index of points on the 7th ring

P_SOURCE_INPUT:
p_source_input0=sig0 #where sig0 is a list of size (1,Nt)
p_source_input1=sig1 #where sig1 is a list of size (1,Nt)
ETC... until
p_source_input7=sig7 #where sig7 is a list of size (1,Nt)

And in the second case I have:

P_SOURCE_INDEX:
p_source_index=np.array[[3363674,3273700,3274150,3274749,...,2891235,2891684,2892284,...]] #index of all the points

P_SOURCE_INPUT:
p_source_input is a matrix of size np.size(p_source_index) x Nt where:
p_source_input[0,:]=sig0
p_source_input[1:np.size(p_source_index1),:]=sig1
ETC... until the last one.

Did I well understand that the the first line in p_source_input will be the signal attributed to the first index in p_source_index and so on?

Thanks a lot for your help and for taking time to answer.
lbruju

Posted 6 years ago #
lbruju

Member
Joined: Mar '18
Posts: 19

Hi,

I have finally found the error! The matrix p_source_input should be transposed. The correct shape is Nt x np.size(p_source_index) and not the opposite as I generated before.

Thanks,
lbruju

Posted 6 years ago #
Bradley Treeby

Developer
Joined: Mar '10
Posts: 1,643

Great! One other point, if you're manually creating the input file, you will need to apply to source scaling term. See Line 208 in the example_cpp_io_in_parts.m (you don't need MATLAB to see the code).

Brad.

Posted 6 years ago #
lbruju

Member
Joined: Mar '18
Posts: 19

Hi Bradley,

Thanks a lot, I haven't noticed that.
But I am confused regarding the final pressure values obtained when using this source scaling term.
I add: source_scaling_term=(2*dt)/(3*1500*dx) which is around 7.4E-8.
The p_source_input is a sinus with an amplitude of 10Pa, so the simulation should give the pressure field in Pa too. Without the source scaling term, I obtained a pressure of 6.33E7 at the focal point, which seems to be correct. But with the source scaling term, I have only 4.8 (which corresponds to 6.33E7*source_scaling_term ).

Thanks,
lbruju

Posted 6 years ago #
Bradley Treeby

Developer
Joined: Mar '10
Posts: 1,643

Hi lbruju,

I'm not sure of your input conditions, but getting a focal pressure of 60 MPa with an input pressure of 10 Pa implies a focal gain of 6 million, which seems much too high. The value after scaling seems more plausible, at least to me.

Brad.

Posted 6 years ago #
lbruju

Member
Joined: Mar '18
Posts: 19

Hi,
I am not sure of my simulation yet, but thanks for the information.

lbruju

Posted 6 years ago #

RSS feed for this topic

Reply

You must log in to post.

k-Wave

A MATLAB toolbox for the time-domain simulation of acoustic wave fields

Multiple Sources in 3-Dimensional Space

Reply

A MATLAB toolbox for the time-domain
simulation of acoustic wave fields