Spice Sucks?

Swapnil

I am using hspice right now and I have to say that it is not the most intuitive thing in the world. What other softwares like spice are out there? Which ones do you guys use for simulating a circuit that you made?

chroot

With power comes complexity.

If you're using Windows, I like Linear Technologies' LTSPICE (SwitcherCAD III). It's free, and is about the most intuitive simulator around, assuming you understand Spice syntax and analysis directives.

- Warren

Maxwell

Take the time to know HSPICE. It's very useful. With AvanWaves, it's a very nice program.

LeBrad

What specifically do you not like about hspice?

Swapnil

The fact that you can't drag and drop the circuit elements in some sort of a virtual breadboard. Also, the that you have to use hspice in combination with AWaves or Csope. It is just too much of a hassel. Why can't they just integrate everything into one?

budala

guys, what do you think about PSpice?

chroot

PSpice is not really used by anyone in industry, but it's a good for students. It's a little hard to use, though.

- Warren

budala

Warren, thank you, I am using it in school and you are right it's not easy to use it.

Swapnil

I concur. :)

Ibaun

There's also Electronics Workbench if you'd like an alternative, albeit not as advanced (windows only I believe).

jbusc

hspice error messages are not too intuitive. Also, in some (entirely valid) analog circuits with feedback, it can have convergence problems, and even if the equations do converge, you can get weird things like signals rising _before_ its trigger comes along.

(this is about cadence hspice - the recommendation by cadence is to use spectre, which isn't really helpful :/ )

es1

Wow! What entirely valid analog circuit failed to converge for every timestep?

I am really curious to know.

It must not have been a linear circuit. I have been thinking and the only thing I can come up with is that the simulation started with an invalid initial condition (a la bias point problems), or some feedback loop that is so fast, or some component value that is so extreme, the precision of the native floating point type of the machine matters.

I do occasionally have problems with letting the computer figure out dc bias points because things like leakeage into a cap confuse it, but just using an .ic statement fixes that.

If the problem was some esoteric math issue with the model you were using couldn't you just map out that part to a thevinin or norton equivalent via a dependant source statement? Or to state the same thing another way, just substitute in the equations that you mentioned that do converge, and presumably give the desired accuracy, via a dependant source statement. Kinda cheating, sure, but it does fix the problem.

Did you mock up the circuit on a bench? Are you sure there wasn't a positive feedback loop or resonance hiding in there?