Speaking of Sine Shapers

[Terry Michaels]

Message text written by Haible Juergen
>That's what I expected.
I only wonder why anybody would bother to use an external feedback loop
in combination with a *discrete* pair, where emitters are freely acessible
to add local feedback. 

JH.<

Well, it's more complicated than that.  The article had a lot of
mathmatical analysis and computer simulations, the authors found a certain
magic combination that produced really low distortion, it required both
emitter degeneration and feedback.  Either one  by itself is an improvement
over a normal differential pair, though.  

In order to get extremely low sine distortion, you need to start with an
extremely precise triangle wave.  This is not easy, for example, a 3080
driving an integrator is not precise enough to get sine distortion much
below about 0.25%.   Because the 3080 uses a different current mirror
arrangement for the current source and current sink, it outputs slightly
different current levels when sourcing or sinking current, resulting in a
triangle wave that does not precisely have a 50/50 duty cycle.  A descrete
version of the 3080 will have the same problem, unless it is carefully
designed and built, with lots of trims, etc.   Getting it to work well over
a wide frequency range is even harder.  

I built a signal generator for my bench using this technique, but even with
the above design and adjustments, the best I could do was about 0.04%,
which is far short of what the article said was possible.   

I would guess you won't get anything even close to that number for the real
world example of a synthesizer VCO, especially when the triangle wave is
derived from the sawtooth by the saturating transistor trick, or full wave
rectification.
 
Another minor problem with deriving a sine wave from a triangle wave with
the above technique is, the sine wave is derived from the highly amplified
the output of a saturating differential transistor pair, this process adds
a noticable amount of white noise and shot noise to the sine, compared to
the triangle wave you start with.  I could clearly see it on a scope when I
ran the sine wave through a high pass filter in tandem with a notch filter,
which is the way I measured the residual distortion.

Terry Michaels