Simulators and convergence

[rtarbell]

Regarding simulators (Pspice, TopSpice, SwitcherCad, etc.),

Let us say I have a circuit that will not converge when simulated on an older computer.

==> Is there a chance that, using the exact same simulation program, the same circuit WILL converge on a more powerful computer (i.e. faster, more RAM, etc.)?

 

[HFGuy]

nope, the solution sets arnt closing to a single value (converging to a single answer). It has nothing to do with your computer.

 

[jcx]

LtSpice is supposed to be pretty good on convergence - the more recent versions have added a gshunt hack that lets you add a conductance from every node to gnd - sometimes helpful

did you try the ps ramping hack? - replace your supplies with pwl sources that start a 0 - surely it can find an operating point - then ramp the sources to the operating V, this can still get to non-physical states but sometimes works

ideal current sources can run off to KV, volatage sources can supply KA - clamping diodes on ccs to ps helps me occasionally

 

[HFGuy]

Or switch to Cadence, ive yet to have it not converge on me

 

[I_Forgot]

Or switch to Cadence, ive yet to have it not converge on me

SwitcherCAD (aka LTSpice) is free. Most of us could not afford the annual maintenance contract, let alone the cost of a license, for Cadence.

I_F

 

[TwoSpoons]

Simetrix always converges for me. You can also change the integration method - fast and loose or slow and accurate. You may also find using a smaller timestep on transient sims will work better. Also setting up your initial conditions better will help. Sometimes adding a small resistor in series with reactive elements (l and C) can help too.

 

[Curmudgeon]

On earlier pspice variants that I used, the above suggestions were helpful; also changing the starting conditions could cause convergence to a different answer, sometimes a better one.