Dear all,

I've been modeling an LNA including non-linearity (Taylor series) + bandwidth limitation (bandpass filter). All is fine in the time-domain (transient HSPICE simulation).

However, when I try to run Harmonic Balance analysis (HSPICERF) it turns out that the simulator won't return accurate results due to some code constructs.

In particulator, I found out that HSPICERF does not handle well the laplace_nd() function. I used this to model the bandpass characteristic of the LNA. Based on the hidden-state discussion/papers on this site, I can see why an RF simulator would not accept the laplace_nd() construct since a filter has state variables which are hidden.

But what is than best practice to model finite bandwidth effects if no laplace_nd() can be used? Of course, we could mimic any filter using passives but I find this not so elegant.

Any feedback welcome!

Peter

Dear Ken,

Ok, that's good to know.

The symtom is as follows: I model the LNA with/without the aforementioned laplace_nd()  and with a 3rd order Taylor series.

In my harmonic balance simulation, I use an input signal with a certain power. Whenever I exclude the filtering from my model, I can see 2nd and 3rd order harmonics at the exact, theoretical power level. Whenever I do include the filtering, the output power levels are all at -300dBm (very low).

I know there is nothing wrong with the filter statement itself, because when I run a conventional SPICE AC analysis, I see the correct bandpass characteristic.

I guess this is an issue I have to take up with the tool vendor (since there does not seem to be a fundamental incompatibility between these code constructs and an RF simulator).

Regards

Peter