SV: Re: [sdiy] Digital VCO

[Eric Brombaugh]

Michael Bacich wrote:
>
> In most cases, they usually also have an analog voltage that rises in 
> proportion to the frequency of the programmable divider.  They use 
> that varying voltage to charge the ramp, resulting in a ramp that 
> stays fairly uniform in amplitude.  It's not perfect, but pretty good 
> enough.  Typically, they also use the output of that ramp to feed a 
> comparator to make the pulse wave.  They give the comparator a 
> CV-variable triggering threshold that allows the DCO pulse wave to 
> have PWM.  Essentially, this gives the DCO an entirely analog output.  
> (the fact that it's locked to a clock, however, gives it part of its 
> characteristic "cold" DCO sound -- the frequency never drifts)
That makes sense
> I'm not sure I understand your question.  Just to clarify, these chips 
> are used as programmable frequency dividers, not as "counters", per 
> se.  The official function name for the 8253 chip is "triple 
> programmable timer" (there are three independent timers in each chip, 
> each one can be used as a separate DCO voice).  The count value is a 
> binary word that's latched into the count value input of the counter 
> chip.  The actual counter output that's used by the synth is simply a 
> square wave at the frequency of the DCO.  More often than not, the 
> synth actually uses the direct counter output as the Square wave for 
> the voice, besides using it ti derive the reset pulse for the ramp 
> generator.  
>
Also clear.
> I don't know if it's true that the frequency resolution decreases at 
> lower frequencies.  If it does, I've never noticed it in the sound,  
> It is true, however, that these systems use fairly high frequency 
> master clocks for the counters.  Most of these synths seem to have 
> pretty decent fine pitch control (for pitch bend, fine tuning, DCO 
> detuning, etc.).  This type of system even allows for pretty 
> convincing keyboard portamento over fairly wide pitch intervals 
> (created by gradually changing the count value).
>
Actually, since the timer output is Fout = Fclk/N (where N is the value 
loaded by the CPU), the frequency resolution gets worse with smaller 
values of N. The best resolution would be at low output frequencies 
(large values of N). That agrees with the need for fairly high master 
clock rates in order to preserve resolution at the higher end of the 
musical scale.

DDS structures based on phase accumulators are much nicer than dividers 
since they maintain constant frequency resolution across the entire 
operating range. The downside is that the phase jitter you get when 
using waveforms that aren't band-limited and the need for reconstruction 
filters.

Thanks for the details.

Eric