You and @crutschow are on the same page...
See post 13

Actually he mentioned a well known and well understood method.

The method I mentioned (note I didn't say described) is something I've seen in project schematics but never really delved into how it actually works, to the best of my knowledge the diode chain is immune to frequency sweep as a function generator does, and integrating with CR is pretty much tuned to one frequency - if you deviate from that, the sinewave gets pretty distorted.

 

The biased-diode method, either in a feedback loop or as a series-pass circuit, is called a breakpoint shaper. It uses the logarithmic relationship between Vf and If to distort the incoming ramp, a variation of what goes on inside an analog multiplier. It is immune to frequency (within reason) but very dependent on peak to peak amplitude, source impedance, and load impedance. That is why it usually is buried inside a function generator chip or circuit rather than as a part of an output amplifier.

ak

 

I thought Ian meant something like this http://sound.westhost.com/articles/sinewave.htm#s8

 

I thought Ian meant something like this http://sound.westhost.com/articles/sinewave.htm#s8

Searching; breakpoint shaper, got mostly irrelevant hits.

Something I'll have to try searching when its not so late in the day.

 

This is a National Semiconductor app note reworked by TI. Page 11:

ak

 

and this is 2 days with no response from TS.
Beware the drive-post.