square to ramp

[WeAreAs1 at aol.com]

Re:  square to ramp conversion:

The Square-to-Ramp conversion circuits used in the Roland Juno 60, JX3P, and 
JX8P (as well as a few other DCO-based synths) seem to work pretty well.  
They don't output *absolutely perfect* sawtooth waves, but they're very close 
- at least to the ear.  These circuits require a control voltage that is 
proportional to the pitch of the square wave (a 1v/octave CV will work, if 
you already have it).  The CV is used to control the charge rate of a ramp 
integrator whose discharge is triggered by the rising edge of the square 
(actually, you would just differentiate the square into a tiny spike).  

Interestingly, the Juno and JX's use these "fake" sawtooth waves to generate 
their variable-with pulse waves (by running them through a typical 
voltage-controlled PWM comparator circuit).  Evidently, the sawteeth must be 
reasonably stable to be useful in this context.

By varying the aforementioned CV proportionally with the pitch of the square, 
you can keep the sawtooth amplitude relatively constant.  With careful 
trimming and some temperature compensation, you can get a sawtooth that's 
almost the same as one coming from a purely analog ramp-generating VCO.  In 
fact, that's what these DCO's essentially are, except that their reset pulse 
is not self-generated.  Rather, it is derived from the externally-generated 
square wave.  BTW, the amplitude-controlling CV need not be quite as 
perfectly scaled as a CV you would need for controlling a VCO's pitch, since 
our ears are not as sensitive to subtle amplitude differences as they are to 
pitch differences.

Also, if your square wave doesn't already have a handy pitch-controlling CV, 
it would be possible to directly convert the square's frequency to a 
proportional voltage, using traditional F-to-V technique such as an F-to-V 
tachometer.  This voltage would be inversely proportional to the needed 
amplitude-controlling CV, so it would need to be run through some kind of 
trimmable anti-log circuit to compensate.  Also, the F-to-V circuit might 
have a bit of lag in response to rapid changes of incoming pitch, which could 
manifest as momentary fluctuations in the sawtooth amplitude.

I know I'm not explaining this as well as I could, but a look at the JUNO 60 
schematics would probably best clear things up.  Also, the Kawai SX-240 uses 
a similar system for converting square waves into sawteeth, and theirs is 
more tightly scaled and better temperature-compensated than the one used by 
Roland.

Sawteeth generated by these means would be best when used in the audio 
frequency range, and probably would be problematic if used in the sub-audio 
(LFO) range, where sawtooth amplitude fluctuations or shape irregularities 
might render the wave useless.

Michael Bacich