I've begun to use your toolbox, and am very impressed with it. Thanks much for providing it as freeware. Most of my current interest is in modeling propagation and scattering from spherically focused acoustic beams. Have you considered allowing one to work in cylindrical coordinates, so that one could model focusing of a spherical beam with the speed of a 2-D simulation? That is, assume no phi dependence in a cylindrical system? I haven't tried it, but I suspect the cartesian 3-D modeling is prohibitively slow (I'm working in the 10MHz regime, with distances on the order of 2.5 cm). If you've done anything with a cylindrical coordinate system, I would be very interested.
Thanks again,
Rick Stearns
k-Wave
A MATLAB toolbox for the time-domain
simulation of acoustic wave fields
2D - cylindrical coords?
(14 posts) (6 voices)-
Posted 13 years ago #
-
Hi Rick,
Thanks for your feedback. We have discussed including a cylindrically symmetric (r, z) implementation (including some of the idiosyncrasies that arise when switching to this coordinate system) but have not progressed any further than that. It is on our "to-do" list but we do not currently have anyone working on this.
In the meantime, if you have access to a recent NVIDIA GPU, you may wish to try running some simulations in 3D to see how you get on. I regularly run simulations using a 256 x 256 x 256 voxel grid on an NVIDIA TESLA C2070. This takes less than ten minutes for simulations with around 1000 time steps. For a 25 x 25 x 25 mm domain that would enable you simulate frequencies up to 7.5 MHz (for a sound speed of 1500 m/s and two grid points per wavelength).
Brad.
Posted 13 years ago # -
Hello Brad:
Thank you for your response. I'm actually scoping out some processors and graphics cards, and I have a question: what processor do you use, along with the Tesla C2070. For example, is there a huge difference between using a Xeon vs I7 processor? Any experience you have regarding hardware choices would be most appreciated.
Best Regards,
Rick StearnsPosted 13 years ago # -
Hi Rick,
My understanding is that the Xeon and i7 are essentially the same chip aside from some minor differences, for example the amount of power they consume (the Xeon is more efficient and thus marketed for computer servers). The desktop machines we use have i7 processors, while the server boxes have Xeon processors. I do not have any concrete benchmarks for comparison, but we use both with the latest generation of NVIDIA GPUs (Telsa C2070 and GTX580).
My only other suggestion is that if you are planning on installing multiple GPUs on one machine, try and go with an architecture that has an x16 PCI Express connection to each card (rather than the links being shared between cards).
Unfortunately my knowledge of computer architectures and hardware is fairly limited, so it would be wise to run my musings past a more reliable source.
Brad.
Posted 13 years ago # -
Hi Bradley,
Just to bring back to life the original post: I just came across one of your recent paper ("Full-wave Nonlinear Ultrasound Simulation in an Axisymmetric Coordinate System Using the Discrete Sine and Cosine Transforms") in which you write: "The model was implemented in MATLAB [...] as an extension to the MATLAB k-Wave toolbox"...
However, I don't see the feature in the topic dedicated to the release 1.1 of k-wave: do you still plan to make this feature available?
Best regards,
AnthonyPosted 11 years ago # -
Hi Anthony,
We definitely intend to release this code. Unfortunately doing so is slightly trickier than the Cartesian codes, as it relies on using the fast discrete cosine and sine transforms (in place of the FFT). As these don't exist in MATLAB, we have written our own mex wrapper to the FFTW library. We still need to make sure this runs and compiles across multiple platforms, MATLAB releases etc, which may take some time. This will dictate whether it makes it in for V1.1 or V1.2!
Brad.
Posted 11 years ago # -
Hi Bradley,
Many thanks for all what you are doing!
Anthony
PS : By the way, I have just found this matlab script that perform a KZK simulation in axisymmetric configuration for HIFU:
http://www.fda.gov/AboutFDA/CentersOffices/OfficeofMedicalProductsandTobacco/CDRH/CDRHOffices/ucm301529.htm
It also solves BHTE and therefore provides a rather useful tool for a quick estimation of HIFU lesions. Maybe you could add it to the list of free acoustics software.Posted 11 years ago # -
Thanks Anthony - I've added it to the software list.
Brad.
Posted 11 years ago # -
Are there still plans to include axisymmetric coordinates in k-Wave? I've run a few 3D simulations and been pleased with the results. However, I would like to run simulations for a larger field that is currently possible in 3D.
Thanks,
DennisPosted 5 years ago # -
Hi Dennis,
Yes, the code is essentially production ready, bar a few small idiosyncrasies that still need ironing out. It will be in the next release.
Brad.
Posted 5 years ago # -
Hi Brad,
Thanks for the update. I look forward to using the new capabilities!
DennisPosted 5 years ago # -
Hello,
Bringing this post back to life. Any news about the availability of the axisymmetric code ?
Best,
Moad.Posted 5 years ago # -
Hi Moad,
It's finished, and will be in the next release. We've had a few hold ups for other reasons, but I will be out before the end of the summer.
Brad.
Posted 5 years ago # -
Hi Bradley,
Coming back to the above, has the 2D cylindrical coordinate model already been released?
Many thanks!
FlorisPosted 4 years ago #
Reply
You must log in to post.