Alternative to MIDI-CV revisited

[gstopp at fibermux.com]

     Yes I did throw together a discharging-cap frequency-voltage converter 
     based on Don's idea and tried it with a manual oscillator as input and 
     a scope as output. It looked like it was doing the basic f-v 
     conversion, with more travel in the higher frequency ranges compared 
     to the standard period measurement method, so I think the concept is 
     valid. However I did not try to control a VCO with it to compare 
     input-to-output frequencies. I realized that probably there would be 
     some trimming and adjustment with offsets to get the curve in the 
     right place, so to speak.
     
     Right at that time I had the exponential-VCO phase-locked loop idea in 
     the back of my mind, and that took over in importance so the 
     discharging-cap idea went by the wayside. I had this nagging thought 
     that the discharging-cap method (or linear cap plus log amplifier 
     method for that matter) are "blind" converters - that is, these 
     converters look at the input and put out what they think a VCO should 
     get to match the input frequency. They could be way wrong and never 
     know it. On the other hand a phase-locked loop would compare input to 
     output and *always* be within a hertz or two, and when the VCO in the 
     PLL is exponential, all you have to do is tap off the CV and it would 
     be perfect.
     
     The PLL method does produce a perfect CV. However the problem is that 
     the correction CV always has some ripple in it, so it's hard to 
     sample-and-hold it for those times when the input goes away and the 
     PLL VCO goes back down towards zero hertz. Each time the note stops 
     the S/H will be holding the CV at wherever it was as it was wavering 
     around the correct value, and a synthesizer driven by the held CV 
     would be left a little out of tune. The quick solution is to low-pass 
     this CV some more, but then you start adding excessive portamento. You 
     may think that you can filter the correction voltage *before* it gets 
     to the PLL VCO, but I tried this and the PLL looses its capture 
     efficiency. In fact the PLL works the best with *no* filtering on the 
     correction voltage, which makes it even more difficult to sample. BTW 
     you don't hear the ripple if you drive a Minimoog for example, because 
     the CV ripple is at the same frequency as the pitch so the Minimoog 
     sounds normal. It's just the S/H hold thing that gets screwed up.
     
     Perhaps the best fix to this is to have the S/H use a SPDT analog 
     switch, to track the rippling CV when the note is playing, and hold a 
     low-passed version of the CV when the note goes away - hmmmmm....
     
     Anyway if you turn down the release on the Mini's EG's the thing 
     sounds wonderful so I may just leave it as is. It's a back-burner toy 
     at the moment.
     
     Regarding your positive-infinity/negative infinity thoughts - the 
     trick here is to cover the *useable* range of musical frequencies, not 
     DC to white light. So if the cap discharge bottoms out, try to make 
     this below 20 hertz or so, and when the cap is fully charged, try to 
     make this a couple kilohertz. If this can work with some accuracy over 
     five or six octaves, it would be fine I think.
     
     I get the feeling that to do this right you will need some very fast 
     circuit components with low leakage to slam the cap to full charge, 
     then let it decay with a good buffer down to ground, and then sample 
     it with a S/H that has as the narrowest possible sample window. My 
     impression is that the component count would be comparable to the PLL 
     method after all is said and done and you will still have the same 
     input-just-went-away holding problem as the PLL method.
     
     But it is a very elegant idea and probably deserves more bench-time. 
     If anybody tries it, let us know how it goes!
     
     - Gene
     gstopp at fibermux.com


______________________________ Reply Separator _________________________________
Subject: Alternative to MIDI-CV revisited
Author:  Tom May <ftom at netcom.com> at ccrelayout
Date:    5/9/96 1:26 PM


I was thinking about the frequency-to-volts/octave conversion problem, 
wondering whether it would be possible to use the exponential decay of 
a cap (dis)charge somehow, when I happened to find the following in 
the archives while looking for something completely different.  On 
Thu, 8 Feb 1996 14:08:14 -0800 (a long time ago indeed), Don Tillman 
<don at till.com> was discussing frequency-to-voltage converters with 
Gene Stopp and wrote the following:
     
>There's a better way.  Start the beginning of each cycle of the input 
>signal with a cap charged to a reference voltage.  Discharge that cap 
>exponentially with a simple resistive load.  S&H that voltage at the 
>end of the input cycle.  You now have an volts-per-octave F-to-V. 
>Simple, all analog.  (No, I haven't built it yet.)
     
Wait, this won't work at all.  I could work out the math, but consider 
the two extreme cases: at infinite frequency the cap will not 
discharge at all so you will get the reference voltage.  What you 
really want is positive infinity.  At zero frequency the cap will 
discharge completely and you will get zero volts when what you really 
want is negative infinity.  This scheme compresses the entire 
frequency range into a voltage that ranges from 0 to the reference 
voltage with some kind of bizarre mapping.
     
Too bad.  It *is* simple.  Gene, did you say you actually built a 
converter like this?
     
Tom.