thru-zero-FM VCO without matched pnp pair

[Haible Juergen]

Hi!

I wanted to build a Thru-zero-FM VCO - I have a 3340-based one, but I 
wanted to
build a discrete one.
In theory, it's easy: you need a triangle type VCO, with the 
hysteretic switch replaced
by a D-flipflop and two comparators. Upper threshold-> set, lower 
threshold->reset,
and the clock input is used for phase reversal.
Usually, a 3080 is used in triangle VCOs, to form a current switch to 
charge and
discharge the capacitor.
Problem: the 3080 needs a current *source* for its control current, 
rather than the
usual exponential current sink. So you either build an expo converter 
with pnp's,
or use a current mirror (again with a pnp pair). This I wanted to 
avoid.
Instead, I tried the following:
I used the Oberheim-SEM expo converter (with a 3046 npn array).
The exponential current goes to another npn pair (to the common 
emitter).
The base of one of these transistor is fixed at +4.5V. The base of the 
other transistor
is switched between +4V and +5V by the output of the flipflop (4013).
So the exponential current is switched between the two collectors 
(col_1 and col_2) -
but of course, we still have two current *sinks*, not sources.
Now  instead of simple integrator, a differential integrator is used. 
It looks like this:
One opamp (1/4 TL074) has a 1nF cap from output to its negative input. 
A second
1nF cap goes from the positive input to a +10V reference voltage. Two 
100k resistors
are connected between the 10V reference and the two opamp inputs. the 
collectors
of the switching pair (col_1 and col_2) are connected directly to the 
opamp inputs.
The circuit looks a little phantastic, but I tried it and it really 
works. The 074 contains
4 opamps. One forms the differential integrator, one buffers the 10V, 
and two are used
as comparators. The 074 is supplied with +15 and 0V, so it can drive 
the 4013 directly.
The triangle swings between +5V and +10V.
All this is based on an idea I found in EN (so don't think I invented 
it (;->) ), but
honestly I didn't expect it to work so well. I also included the 
typical Thru zero
stuff (Rectifier, edge detector etc.), and I can get all the typical 
thru zero FM waveforms.

Unfortunately, the circuit does *not* show all the imperfections and 
quirks I was used to
have on my 3340-based circuit. This was something I hoped for: That I 
could reproduce these
distorted-guitar like dirty transistions between two notes. Sorry, no. 
Too perfect ...
Anyway, it was an interesting experiment.

JH.