frequency domain using reduced basis (modal)

Hi all,

I'm trying to make a frequency sweep on a (big) model of an acoustic cavity, with specified surface impedance as boundary conditions.
If I use the "Frequency domain" study, everything goes well, except that it is really, really slow.
Considering that 1) few eigenmodes participate to the vibrations at these frequencies (20-200Hz), that 2) I would like to have a good frequency resolution in my sweep, and 3) I need to test a lot of combinations of parameters for the boundary conditions, I figured that using the "Modal frequency domain" study type is perfect for me.

However, it does not seem to work : I always get a null solution, for all frequencies, when I use this type of study.
The predefined study contains two step : one for finding the eigenfrequencies and eigenmodes, and the second to use these modes as a reduced basis for the frequency sweep. The first step seems to work well (I can visualize the modes, they seem ok to me), but the second gives mysteriously a null result everytime (without any error message).

If anyone knows that problem and how to overcome it, I'd be really glad to know!

Regards,
Sami

PS : I'm using Comsol 4.3a.
PS2 : I use a point volume flow source, which works as expected with the traditional "frequency domain" study.