Korg MS 20 f->Volt Converter

[Bissell, Harry]

There are some good f/v converters. I traced out a old 360 Systems
"Slavedriver" an early guitar to synth adapter. After the filtering to
remove harmonics, the comparator to extract the zero-cross (square wave)
information (same same) the Square wave (fundamental) was used with a ramp
generator and a sample/hold. The ramp increases (or decreases) in voltage at
a linear rate. You start the ramp up (down) at the first zero cross, and
then sample the ramp value at the second zero cross. This gives a rather
flat voltage which can respond after the first cycle and updates every
cycle. There is no traditional "ripple" but the dc output will vary from
cycle to cycle, resulting in some stairstepping or dithering. But the
response time is always as good as the input frequency, there is no
slew-limited response like in tachometer style circuits. It could be made a
"full wave" detector by using two ramps and two sample/holds, but ramp
matching would be critical.  Bob Moog published a similar circuit for use
with his "Etherwave" theremin, I don't know if it is on the web (I bought
one).  If your interested I'll try to find the 360 Systems circuit. I had to
trace it out by hand (with all chip numbers dusted off). Maybe some errors
but I was able to identify all thr chips from their pinouts (4000 CMOS).
Harry Bissell

> -----Original Message-----
> From:	Martin Czech [SMTP:martin.czech at intermetall.de]
> Sent:	Wednesday, February 10, 1999 3:39 AM
> To:	synth-diy at mailhost.bpa.nl
> Subject:	Korg MS 20 f->Volt Converter
> 
> Yesterday I took a look at the Korg MS20 frequency to voltage converter.
> 
> I think this is how it works:
> 
> First the input signal is conditioned, ie. amplified and  bandpass
> filtered. A second order highpass A and a 4th order lowpass B do the job.
> B cuts away strong harmonics that could fool the detector, and the A
> limits the lower end of frequency range, it reduces unnecessary bandwidth
> below the expected fundamental of the input signal thus reducing noise.
> Unfortunately 4x pots are required for both filters.  The signal then goes
> into two 339 comparators, the first working as zero crossing detector,
> the second simply as inverter. The rectangular waves of these comparators
> are about 180 deg out of phase, they feed a capacitor & diode network,
> which forms a positive spike for every negative and positive zero
> crossing. These spikes discharge a one-shot via TR2, the one-shot made
> out of one 339 and C2. The last comparator does almost nothing, it is
> allways pulling a low level and seems to provide a reference potential
> for the one-shot. I think this is some kind of temperature compensation.
> Now we have a train of constant lenght pulses and the frequency of the
> pulses is double input frequency. The 4096 inverter chain does perfect
> wave forming of the pulse train (maybe some glitch suppresion also). The
> second lowpass filter C (2nd order) follows A (4x pot) but with some
> lower frequency, it is there to integrate the pulse train into an average
> level. The following enshures low ripple for every cutoff value of A. The
> one-shot lenght is critical, because it goes directly into the average
> level, temp compensation is therefore logical.
> 
> Even if the system reaches a steady state at the output, it suffers from
> ripple. So the detection should be quite good provided that A and B are
> proprely adjusted, but the output voltage still has ripple, which should
> lead to a certain roughness off following vco processing.
> 
> I think Tom G. build the f->Volt converter, and maybe some others too.
> 
> Any experiences ?
> Comments?
> 
> 
> 
> m.c.
>