Discussion Forum

Model it in time domain.

Thank you, I moduled it in the time domain however it doesn't seem to converge. Is there something I'm missing or not understanding? Here is my excitation voltage

Vo=Ac * sin(2 * pi * fc * t) + Ac * m * 0.5 * cos(2 * pi * f_LSB * t) + Ac * m * 0.5 * cos(2 * pi * f_USB * t)

my time steps are range(0,0.1,1)

But does seem to converge. My model works in the frequency domain just fine.

My suggestion to you would be to start with the circuit interface alone and build up a very simple transient simulation using a time dependent source with one freq and a load resistor. Get this to work. Then add more of your text commands to the source, watch the display, and verify it looks correct. You will eventually need to link up the 2D simulation with the circuit "front end" driving it. There are a couple sample files to check out on the comsol website.

Thank you for the responses do you think if i switch from mf physics to mef physics and run the time-dependent study I'd have better luck? I'm not very familiar with the circuit interface but that seems like a good idea. Thank you

I am not certain if you will get differences between those two ac/dc physics as i have not used mef all that much. I was only suggesting using the cir interface as this is simpler if you have not done a transient before in comsol. It would be a way to verify your syntax before trying it in another physics. You would then take that syntax and use it in the intended physics. It is easier to break the problem up and get small parts to work on their own so when you combine them you have a better idea of what results you should get.

You can easily define a function with this expression and check if the resulting waveform looks as expected. The main issue regarding convergence seems to be the time stepping. For a quasi-periodic waveform like this you need manual time stepping. I suggest to try the Generalized Alpha time stepper. Set a manual time step that resolves the waveform with 4 - 6 steps per period and and choose a time range that is long enough to represent the complete waveform. Make sure the waveform starts at zero for t = 0, otherwise ramp it up in suitable manner.

I am wondering what you expect to find out by modulating the waveform. It may be sufficient to do a frequency scan covering the modulation bandwidth.

Cheers Edgar