Oh dear, I seem to have caused some confusion!

> Interesting...I was assuming the poor-quality sine was a result of
> using diodes rather then the more common differential transistor
> pair circuit.

Basically it is: the existing A-110 sine-shaping circuit is a pair of
back-to-back diodes in parallel with a resistor, and a second
resistor acting as a sort of potential divider; the differential pair
method is another common way of doing it, is evidently better, and
is often implemented using an OTA or two; the diodes+JFET way is a
third (good) way of doing it, which from the books/circuits I have,
seems at least as popular as the differentail pair way.

> But by adding the trimpots to improve the diode sine,
> it's now looking more like a bug in the original design – the two
> operating points may be set incorrectly

The trimpots I 'added' to the circuit on my website apply to the
proposed, *new* diodes+JFET way, and *not* the existing two
diodes+resistors way: the trimpots were always a part of the proposed
circuit, it's just that I originally didn't show them (whilst they
are essential to its operation, they are not the 'key' components
*of* the circuit's operation, which are the diodes and JFET
themselves).

> If the triangle level is consistent, maybe all we need to do is
> tweak the two resistor values to optimum points as you suggest.

This is what I was alluding to here:

> > Incidentally I also wondered whether the component values for the
> > existing set-up could be better 'optimised': after a handful of
> > simulations, and then discovering the other circuit, I was
> convinced
> > that that path wasn't worth pursuing further!

It may be that it *is* possible to optimize the resistors used in the
existing A-110 two diodes+resistors method: however, after running
lots of simulations with different values for the resistors, it was
far from immediately clear to me just *what* could be tweaked in
order to get a better sine shape out of this existing circuit. Which
is not to say that it cannot be done: if it wasn't for the fact that
it would take me many hours to re-discover my knowledge of non-linear
numerical approximation and get it up to speed, I would probably give
it a shot. (Knowing what the original start-point for the circuit was
might help, but it was almost certainly practically/empirically
based, rather than theoretical!)

Hopefully the above might remove some of the confusion ( !).

Tim