On Monday 22 May 2006 15:56, Stuart Brorson wrote: > Another way to check your results is to read your netlist > into LTSpice and/or Gnucap and run your sim using it. > Indeed, since Gnucap is a completely different simulator, > comparing ngspice to Gnucap is a good robustness test. If > you get same result (likely), then the problem is your > circuit, not your sim.
I tried it on gnucap .. It takes the names with + and -, but it tripped on another thing ... On BJT's the substrate node is optional. When you specify an area and no substrate, or is that a substrate and no area .. The parse is ambiguous. It got the wrong one. The fix is to put the node list in parentheses, which is a good idea anyway for readability. It gave me answers. Usually, checking on another simulator is a good robustness test. Sometimes it can trick you. On the help-gnucap list, there was recently a report of a convergence error. I quickly sent a patch then realized that it still wrong. It converges, but to the wrong result, the same wrong result as ng-spice.
