I have a very basic question: in ADE -> Simulation -> Convergence Aids, what is the difference between Node Set and Initial Condition? When should I use the formar and when the latter?
In Spectre Circuit Simulator Reference, it is stated: "The nodeset statement is used to provide an initial guess for nodes in any DC analysis or the initial condition calculation for the transient analysis."
This confuses me, because I don't see the difference with Initial condition, then..
Nodesets are less strong than initial conditions. If you specify a nodeset, it applies that voltage as a current in parallel with a small resistor (by default 1 Ohm), as a Norton equivalent to a voltage in series with a resistor, connected to the node in question. This is held there during DC convergence, and is then removed once the circuit has converged. This nodeset is then removed, and the convergence algorithm carries on from where it left off, reaching final convergence.
With initial conditions, the "hold" is left through DC convergence and is only removed once the transient starts.
So a nodeset steers the convergence in a particular direction - useful to speed up DC convergence when you've had a tricky convergence problem, or useful to steer it towards a desired solution when maybe there is more than one solution. An initial condition is useful when you want to force the circuit to start a transient in a particular condition; it's useful for (say) oscillators, which might naturally DC converge to a metastable state and not oscillator, whereas with an initial condition you can force them to start away from the metastable state.
By default initial conditions are honoured in transient (and other time-domain analyses), but not in AC or DC. The "ic" parameter on a transient analysis allows you to turn off initial conditions in transient, and the "force" parameter on ac or dc allows you to enable them for AC and DC analysis. The main use however is for transient or transient-type analyses.
You should use initial conditions with caution, because you can get misleading results if you've forced the circuit into a non-realistic initial condition. With nodesets, the effect of an incorrect initial nodeset tends to be reduced by the completion of the DC operating point anyway, so it's less critical.
Hope that helps!