a list of acoustic FEM SW k-Wave: A MATLAB toolbox for the time domain simulation of acoustic wave fields
no experience, just looking - several want Matlab despite being "free" - user forum reports some usability of k-wave with Octave
no experience, just looking - several want Matlab despite being "free" - user forum reports some usability of k-wave with Octave
Comsol pricing surprisingly reasonable for this sort of thing
don't know if this is special academic pricing?
http://collaborate.bu.edu/engit/Comsol?action=AttachFile&do=get&target=comsol-pricelist.pdf
https://www.comsol.com/models/acoustics-module/page/4
don't know if this is special academic pricing?
http://collaborate.bu.edu/engit/Comsol?action=AttachFile&do=get&target=comsol-pricelist.pdf
https://www.comsol.com/models/acoustics-module/page/4
You can buy it like that, but do not expect results just out of the box! I needed enormerous amounts of time to get even a simple simulation up and running and even that still has some flaws on the electrical part of things. And when it runs you still have to do the post processing, which I switched over to Matlab, which adds to the price because you need an extra module.
It is a powerfull software, but needs lots of getting in to. ABEC is not as straight foward in terms of the user interface, but due to its specialisation in audio easyer in getting started with. The documentation is really good to! Also, it takes most part of the post processing of your shoulders, which takes some getting used to in Comsol. A downside for me in ABEC is that you can not so simply import 3D models.
Also, ABEC simulates the Frequency domain only, Comsol can do both frequency and time domain, but time domain simulations are more fore things like distortion analysis of Hornchambers due to compression effects, which should be possible, but is really hard to do as far as my experiments go...
MATLAB offers a lot in terms of FEM, but as far as I know it has some flaws in recource manegement in terms of the hardware it is running on, at least it does within normal use. Extra libarys might work around that. ABEC will get a GPU update in the future which looks promising. Comsol does not, nor will it get GPU support in the future but is rather well done in terms of recourcemanegement as it is. Even runs on servers in a network just fine.
Be aware that ALL of those programms use tremendeus amounts of Computer power. Even a simple waveguide can take 30min+ to solve in Comsol on a quad core CPU with 8Gb Ram... I recommand doung several simulations in a row and reading a good book while waiting, sports did it for me, whatever floats your goat...
One the usefullnes of BEM/FEM, no matter which Programm you use, they all do the job, some make it easyer to set up then others, some offere more features or comfort, but no matter what eyecandy or whiplashes for setting them up you have to take, it is still s*it in, s*it out. Meaning even Comsol, which I dare call the most powerfull tool depending on which modules you have, will give false results if not set up properly and only your common sense will notice it, if at all. So if they are helpfull depends on your abillety to interpret the results.
Other then that, what do you even want to simulate? For simple ratationsymmetric structures VACS would be a way to go. Keep also in mind that Programms like Comsol need a 3D Model that has to fulfill certain criteria to not generate errors. A project I helped a other member here with it had some problems grasping what those where even as the guy profassionally worked with CNC machines and made the models, which where really good, but not right for simulations.... we got it sorted out eventually^^
I hope that helped some, if you have questions, ask away ! =)
Many thanks for the overview.
I suspect a steep learning curve to generate the 3D mesh models and get a BEM sim running. The shapes are irregular for most speakers and not likely to be in a std library. I'm looking for suggestions and possibly experience to avoid dead ends. I don't mind point tools and post processing images as long as the docs are good.
I'll trying to make a OmniDirectional speaker, I have another thread on this forum http://www.diyaudio.com/forums/multi-way/303941-omnidirectional-work-progress.html that describes it. I'm currently trying to improve the HF performance and will be directing a circular waveguide into a reflector (shape TBD). There are other issues as well and I thought time domain BEM would be the best bet as I'm looking for reflections and interference patterns. Ideally the entire speaker is simulated so GPU-CUDA would be a real boost, and I don't mind overnight CPU runs.
I suspect a steep learning curve to generate the 3D mesh models and get a BEM sim running. The shapes are irregular for most speakers and not likely to be in a std library. I'm looking for suggestions and possibly experience to avoid dead ends. I don't mind point tools and post processing images as long as the docs are good.
I'll trying to make a OmniDirectional speaker, I have another thread on this forum http://www.diyaudio.com/forums/multi-way/303941-omnidirectional-work-progress.html that describes it. I'm currently trying to improve the HF performance and will be directing a circular waveguide into a reflector (shape TBD). There are other issues as well and I thought time domain BEM would be the best bet as I'm looking for reflections and interference patterns. Ideally the entire speaker is simulated so GPU-CUDA would be a real boost, and I don't mind overnight CPU runs.
Oh that one, I have seen it, interesting concept indeed! For such a system VACS would be the best choice. It is free to a certain degree and your project is rotationally symmetric, except for the sub cabinet, but that can be ignored I think. The Doc is also good as it is made by the ABEC guys.
A user over at the German diy Hifi forum wrote about some processing times on his rig with ABEC via CPU and GPU, I try if I can find it again.
How long will it take?
I'd thought I'd close this thread off, as I can finally use ABEC. I modeled a simple ABC cabinet to get both the frequency response (LE engine) and the sound fields (BEM engine). It took me 3 full days to get to "student driver" level but it is an amazing design tool. It was not a straight forward process (for me anyways), and required lots of trial and error to "infer" some operations by using the examples and help files. This seriously,.. I can't emphasis that enough,.. needs at least few tutorials written up explaining how to use it.
Initial results and others at http://www.diyaudio.com/forums/multi-way/303941-omnidirectional-work-progress-7.html#post5040158
I'd thought I'd close this thread off, as I can finally use ABEC. I modeled a simple ABC cabinet to get both the frequency response (LE engine) and the sound fields (BEM engine). It took me 3 full days to get to "student driver" level but it is an amazing design tool. It was not a straight forward process (for me anyways), and required lots of trial and error to "infer" some operations by using the examples and help files. This seriously,.. I can't emphasis that enough,.. needs at least few tutorials written up explaining how to use it.
Initial results and others at http://www.diyaudio.com/forums/multi-way/303941-omnidirectional-work-progress-7.html#post5040158
Attachments
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.