Re: [Qucs-devel] error when simulating a simple push pull

[Stefan J. <st...@gr...>]

Am Mo, 5.02.2007, 16:57, schrieb wernher korff:

Hi Wernher,

> Yup.. the amplitude is indeed 1uV. I haven't learnt much about capacitive
> coupling.. so that is ignorance on my part.. :-p
>
> i just did the simulation again.. this time i changed the integration
> method
> from Eular to Trapezoidal(i'll be honest , i have no idea what this is
> for)..
> after changing it i simulated ithe same circuit i sent you the last time,
> 100mV AC input, i got warning readouts again. I then changed the input to
> 3V
> and simulated and it completed the simulation, but now when i changed it
> back
> to 100mV i could do a complete simulation without any errors. Closing the
> schematic and reopening it and going through the same steps reproduces the
> same results.

This problem is weird, as transient simulation is as well. :-)

I think it occurs due to changing the order of appearance of the voltage
source in the netlist.  This order also determines how the circuit matrix
is build.  With badly conditioned matrices this can happen from time to
time.  In this case it seems to be order between the DC source (3V) and
the AC source.  You can switch order and see the phenomena by change
position of the sources forth and back abit, then save schematic, then
re-simulate.  There is currently nothing I can do about it.

Also each of the transient analysis parameters influences how such
a circuit behaves during transient analysis.  It is suggested to use
Trapezoidal (order 2) as integration method, then Gear order 2.  Then Euler
probably.  As I said it all depends sometimes.

Hope this helps,
  Stefan.