Note: This discussion is about an older version of the COMSOL Multiphysics® software. The information provided may be out of date.
Discussion Closed This discussion was created more than 6 months ago and has been closed. To start a new discussion with a link back to this one, click here.
acoustic pressure, frequency response on modal basis
Posted Nov 20, 2013, 9:52 a.m. EST Acoustics & Vibrations, Studies & Solvers Version 4.3a, Version 4.3b 7 Replies
Please login with a confirmed email address before reporting spam
Hi all,
I am using Comsol 4.3a and I can't find what's wrong with my use of the frequency response study using reduced model. This is my second post on the subject, as the first one didn't attract much answers...
Attached is a very simple example, where we first compute the first ~10 eigenmodes, then use it as a reduced basis for a frequency sweep.
The model is composed of a rectangular box with a volume flow point source near a corner, and a point were we would like to measure the pressure near another corner. The walls are slightly lossy, with a finite, real impedance.
When running "Study 1", the eigen modes are properly computed, but the output of frequency sweep on modal basis is the null solution for every frequency. No error are notified. The log seems clean too...
Any idea ?
Thx,
Sami
I am using Comsol 4.3a and I can't find what's wrong with my use of the frequency response study using reduced model. This is my second post on the subject, as the first one didn't attract much answers...
Attached is a very simple example, where we first compute the first ~10 eigenmodes, then use it as a reduced basis for a frequency sweep.
The model is composed of a rectangular box with a volume flow point source near a corner, and a point were we would like to measure the pressure near another corner. The walls are slightly lossy, with a finite, real impedance.
When running "Study 1", the eigen modes are properly computed, but the output of frequency sweep on modal basis is the null solution for every frequency. No error are notified. The log seems clean too...
Any idea ?
Thx,
Sami
Attachments:
7 Replies Last Post Nov 22, 2013, 9:08 a.m. EST
Please login with a confirmed email address before reporting spam
Posted:
1 decade ago
By the way, I checked the generated "System matrices" of the reduced model : the load vector is null, despite the Point Flow Source.
Please login with a confirmed email address before reporting spam
Posted:
1 decade ago
Ok, I think I found the reason so I'll post here the solution : the default behaviour of the modal solver, concerning linearity, is to assume a perturbation around a initial configuration. In my case the initial configuration is zero, and the problem is perfectly linear.
If, in the solver options I set "Linearity" to "Linear" (instead of the default "Linear perturbation"), it suddenly works !
Still, linearizing a problem that is already linear, around the zero solution, should give the same solution, but for some reason it doesn't work... Is that a Bug from Comsol, or did I miss something here ?
Sami
If, in the solver options I set "Linearity" to "Linear" (instead of the default "Linear perturbation"), it suddenly works !
Still, linearizing a problem that is already linear, around the zero solution, should give the same solution, but for some reason it doesn't work... Is that a Bug from Comsol, or did I miss something here ?
Sami
Please login with a confirmed email address before reporting spam
Posted:
1 decade ago
Hi Sami,
This is a design choice in COMSOL, as you usually want to be able to apply
two different loads to problem such as this. (Such as a static preload and
then a smaller superimposed perturbation load).
There are two ways to make this work in the linear case. One is the
solution you have found, the second is expressing the point source as
linper(1e-3) to tell the solver that this should be treated as the
perturbation.
Best regards,
Sven Friedel
This is a design choice in COMSOL, as you usually want to be able to apply
two different loads to problem such as this. (Such as a static preload and
then a smaller superimposed perturbation load).
There are two ways to make this work in the linear case. One is the
solution you have found, the second is expressing the point source as
linper(1e-3) to tell the solver that this should be treated as the
perturbation.
Best regards,
Sven Friedel
Please login with a confirmed email address before reporting spam
Posted:
1 decade ago
Hi Sven,
If I may : I don't get why my load is considered as "static preload". We are in a frequency domain, so the constant coefficient of my load is supposed to be a coefficient that will be put in front of a exp(-jwt) term, no ? It is the case if I solve the problem with the "frequency domain" classical solver, without having to change anything... So why would the meaning of a term in the "physics" interface change, depending on the solver used ?
Shouldn't there be two boxes to fill in the "source" interface : one for static preload, the other for the oscillating term ?
Thanks for the tip anyway. Precious piece of information.
It would be nice to have it mentionned somewhere in the documentation, especially considering that there is rarely a static preload in classical acoustics...
I didn't read the documentation of all other modules, so maybe it was mentionned in the structural mechanics one, or another one (is it?), but still, I think it would be a great help to new users to have it in the general doc (user reference guide / user manual), in the section dealing with the modal solver.
Sami
If I may : I don't get why my load is considered as "static preload". We are in a frequency domain, so the constant coefficient of my load is supposed to be a coefficient that will be put in front of a exp(-jwt) term, no ? It is the case if I solve the problem with the "frequency domain" classical solver, without having to change anything... So why would the meaning of a term in the "physics" interface change, depending on the solver used ?
Shouldn't there be two boxes to fill in the "source" interface : one for static preload, the other for the oscillating term ?
Thanks for the tip anyway. Precious piece of information.
It would be nice to have it mentionned somewhere in the documentation, especially considering that there is rarely a static preload in classical acoustics...
I didn't read the documentation of all other modules, so maybe it was mentionned in the structural mechanics one, or another one (is it?), but still, I think it would be a great help to new users to have it in the general doc (user reference guide / user manual), in the section dealing with the modal solver.
Sami
Please login with a confirmed email address before reporting spam
Posted:
1 decade ago
Hi Sami
The default setting for the modal solver lineaity is "Linear Perturbatio" this means that in order for sources to be used you need to add the linper(..) operator. So try and use it in your source definition, for example, linper(1) for the "volume flow rate out of the source" in stead of just 1. The other option is to change linearity to "Linear" in the solver setting. Either one should make the model solve.
Remember that as a good rule of thumb you should find eigenvalues up to twice the frequency you want to model in the frequency response.
We will look into if this should be better documented or if the behavior should be different for acoustics, for example, having Linear as the default setting.
Best reads
Mads
Technical Product Manager, Acoustics
COMSOL
The default setting for the modal solver lineaity is "Linear Perturbatio" this means that in order for sources to be used you need to add the linper(..) operator. So try and use it in your source definition, for example, linper(1) for the "volume flow rate out of the source" in stead of just 1. The other option is to change linearity to "Linear" in the solver setting. Either one should make the model solve.
Remember that as a good rule of thumb you should find eigenvalues up to twice the frequency you want to model in the frequency response.
We will look into if this should be better documented or if the behavior should be different for acoustics, for example, having Linear as the default setting.
Best reads
Mads
Technical Product Manager, Acoustics
COMSOL
Hi all,
I am using Comsol 4.3a and I can't find what's wrong with my use of the frequency response study using reduced model. This is my second post on the subject, as the first one didn't attract much answers...
Attached is a very simple example, where we first compute the first ~10 eigenmodes, then use it as a reduced basis for a frequency sweep.
The model is composed of a rectangular box with a volume flow point source near a corner, and a point were we would like to measure the pressure near another corner. The walls are slightly lossy, with a finite, real impedance.
When running "Study 1", the eigen modes are properly computed, but the output of frequency sweep on modal basis is the null solution for every frequency. No error are notified. The log seems clean too...
Any idea ?
Thx,
Sami
Please login with a confirmed email address before reporting spam
Posted:
1 decade ago
Hi Sami
Ups, only saw the rest of the thread after answering. Saw you already figured it out and that my colleague gave you an explanation. Anyways, I will make sure to document this behavior better in the future.
Thanks!
Mads
Ups, only saw the rest of the thread after answering. Saw you already figured it out and that my colleague gave you an explanation. Anyways, I will make sure to document this behavior better in the future.
Thanks!
Mads
Please login with a confirmed email address before reporting spam
Posted:
1 decade ago
Thanks anyway, Mads.
I like it better to have twice the answer than none.
(As with my first post on the subject...)
Bests,
Sami
I like it better to have twice the answer than none.
(As with my first post on the subject...)
Bests,
Sami