MiniMoog VCO

[harry bissell]

Good Tips JH:  Here's more...

1) on some Osc.  (like the TomG LFO 1A mentioned earlier) the integrators are
"perfect" and won't start. Bias the integrator input with a ridiculous high
value resistor (like 100M) and it will work fine...

2) Watch out for faulty models... In the Minimoog there's a LM393 comparator
with an open collector output... Though the model appears to exist in the
libraries... its in with 3 (or 5) terminal OpAmps. If it shares the model with
the 741 you KNOW its not open collector...

3) A switch (like JH said) is a good kick in the pants for oscillator start. If
the simulation is interactive (like EWB) you can kick it on at will...

:^) Harry

jhaible wrote:

> > Oscillators always seem to be a bit of a problem. Iterative loop errors
> > and all that. And getting the initial states right can save a lot of
> > simulation time. And fast rising edges...
>
> A few hints for VCO simulation:
>
> (1) Connect a switch across the integration cap which will open at t=0.
> Keep the Ron and Roff of the switch within reasonable ratio. Infinite (off)
> and zero (on) is the worst you can do. You need a high resistance for
> off, though, so start with 10GOhm (off) and 1kOhm (on).
>
> (2) While the variable timestep was one of the great advantages that made
> Spice the standard it is, there is only one choice for accurate VCO
> simulation:
> Turn it off ! That is, set a maximum timestep, like 1us if you're simulating
> a VCO at 1kHz. This will slow the simulation down, but it's the only cure
> against the simulator running over an immediate incident like a sawtooth
> capacitor discharge, and then hardly finding back to a stable point again.
>
> (3) Separate the problem.
> Waveshapers can be optimized in Spice to 99% with DC voltage sweeps.
> If you have AC coupling in your waveshapers, get the DC operating point
> and then replace the capacitors with voltage sources.
> Simulating core and waveshapers separately will help you save the time
> which you have to spend for fixed timestep method as in (2).
>
> JH.