[sdiy] An Improved Sine Shaper Circuit

Donald Tillman don at till.com
Thu Apr 23 00:16:27 CEST 2020

On Apr 22, 2020, at 2:34 PM, Tom Wiltshire <tom at electricdruid.net> wrote:
> Doing this in simulation makes me wonder how accurate the simulation is. I know we use tanh as the equation to model the behaviour of the transistor, but how accurate is that model in reality? Do real transistors behave like that? To what degree?

Usually pretty good.  I mean, SPICE has been around for almost 50 years, whole industries depend on it, and the issues have been worked out.

On the other hand, SPICE has its quirks, and situations where it doesn't converge, and there are various hacks to get around them.

And there are some horribly incorrect models out there.  (Grrr.)

But if we're talking specifically about the tanh() curve, the tanh() curve is the mathematical result of two exponential curves sharing a common current source.

(And that kind of makes sense... the constant current source is split between the two exponentials, and then one is flipped over with the current mirror, so now the difference is constant, just like a hyperbola.  If you visualize it just right.) 

And we know from our (electronic music) experience with exponential current sources, that exponential curve can be sooper accurate.  0.3% or so over a few decades.

  -- Don
Donald Tillman, Palo Alto, California

More information about the Synth-diy mailing list