Discussion Forum

Hi

The finer the better (mostly ;)
In general "my way" is to say I need 3-5 or more elements of 2nd order per wavelength or gradient slope, to resolve the fields. For the wavelength, one should not forget the indexof refraction or er that makes the waves shorter than in ar/vacuum, hence denser mesh. One can use a formula for the largest mesh size based on the material data.

Another point to look closer to are all shap corners, can you define a continuous field gradient around these ?
if not consider to add some small fillets, this will increase the local mesh density and improve the results (and avoid singularities)

In general read the COMSOL PDF doc ! For me the tousand of pages is becoming the best multi-physics reference books you can find, with many tips and tricks, but it's dense, so read it several times and in smaller chunks ;)

--
Good luck
Ivar

Thanks Ivar

Could you please explain more about the idea of 3-5 or more elements of 2nd order per wavelength ?

Could you please give an example ?

And Is there any difference if I did that for a specific one frequency? or sweep of frequencies ? for example, from 2 GHz to 3 GHz ?

Regards

Nayef

Hi

you can read about all this in the doc too, for the mesh in RF: some 10 elements (or more) across a wavelength, but as COMSOL uses 2nd order discrtization polynomial it's enough with 5 mesh elements across a wavelength. But the wavelength varies in vacuum and in material, depending on the index of refraction or on e_r (whatever applies for your model) so do not forget to correct for this.

No real reason to change the mesh for your frequency change, so take the highest frequency, which corresponds to the shortest wavelength.

lambda * 2*pi * f = c_const / epsilon_r

and max mesh size should be < lambda / 5

--
Good luck
Ivar

Many thanks Ivar :-)