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 The information provided may be out of date.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.

Harmonic simulation of high-Q resonators

Please login with a confirmed email address before reporting spam

Sometimes I need to perform frequency domain analysis of high-Q resonators. Those resonators are always operated at their resonance and have Q-factor of tens of thousands which is set as such it corresponds to measurement (isotropic or anisotropic material damping in the simulation) of different modes. The problem I have been always having is that at certain level of damping (Q-factor >1000's) the harmonic simulation stops converging (the relative error is larger than relative tolerance). I have tried various things including changing of direct and indirect solvers, different nonlinear methods with various damping of solvers, different scaling but I was not able to find a way how to make the solver converge. I was partially successful by ramping the load, however, the higher the Q-factor is, the larger the ramp must be and sometimes even very small boundary loads do not converge.

I know that when I use frequency-domain modal, there is no issue with converging. Essentially, it is possible to run simulations with Q-factor of millions. The interesting thing is that if I set initial conditions of the harmonic solver to the solution of frequency-domain modal I am able to increase the Q-factor for which the frequency domain simulations converge, however, when Q-factor reaches certain level (say Q-factor of 20 000), the simulations stop converging again. I also know that both simulations give very close results so what I don’t understand is when I give initial conditions to the solver which almost correspond to the solution, why the solver fails to converge. I have tried frequency-domain modal simulations with various number of eignemodes considered with very little effect on the convergence of subsequent frequency domain simulations. Due to the very low damping and simulating at exact resonance I suspect some numerical issues. However, why frequency domain does not work whereas frequency-domain does is something which I don’t understand.

I could always use frequency domain with low Q-factor or use frequency-domain modal with high Q-factor in case I don’t need any prestressing of the structure. Nonetheless, not being able to simulate what I want to has been bothering me for a while so I would be happy if anyone may have some advice or solution for me.

Mikulas

-------------------
Mikulas

1 Reply Last Post 18 dic 2017, 01:29 GMT-5
Henrik Sönnerlind COMSOL Employee

Please login with a confirmed email address before reporting spam

Posted: 6 years ago 18 dic 2017, 01:29 GMT-5

Hi Mikulas,

The underlying problem here is that the system matrix for high Q (low damping) tends to be ill-conditioned at resonance. Without damping, it is actually singular.

I have two different suggestions:

The first one is to use prestressed mode superposition. Add a Stationary study above the Eigenfrequency study. Then, select geometric nonlinearity in the Eigenfrequency study.

The other is to use the ordinary Prestressed Frequency Domain study, but explicitly force the solver to assume linearity.

Regards,

Henrik

-------------------
Henrik Sönnerlind
COMSOL
Hi Mikulas, The underlying problem here is that the system matrix for high Q (low damping) tends to be ill-conditioned at resonance. Without damping, it is actually singular. I have two different suggestions: The first one is to use prestressed mode superposition. Add a Stationary study above the Eigenfrequency study. Then, select geometric nonlinearity in the Eigenfrequency study. The other is to use the ordinary Prestressed Frequency Domain study, but explicitly force the solver to assume linearity. Regards, Henrik

Note that while COMSOL employees may participate in the discussion forum, COMSOL® software users who are on-subscription should submit their questions via the Support Center for a more comprehensive response from the Technical Support team.