Pitch/Voltage converter board ready at EFM

[Harry Bissell]

The technique you describe is the Tachometer method. This has the disadvantage
of trading off step response for acceptable ripple. With a 6db/octave filter the
ripple will either be quite large, or the overall response unacceptably slow. It
IS accurate... but for lets say driving a VCO from a musical input... it won't
work wll at all.  The ripple will frequency modulate the VCO (which sounds like
shit unless that the effect you intend...) or it will glide slowly to the next
pitch... useless if you want a trill, hammer-on, etc....

BTW the PAiA theremax uses a "tach" P/V converter. Probably get away with it
here, since glide is all you can do on a Theremin (unless you are GOD). You can
count on the
input not changing fast (in mS...)

The "Ramp" method on the PV-1 is fast... it will acquire a step waveform in 2
cycles of the fundamental guaranteed. If it has to jump 4 octaves... it still
gets to the new pitch in
2 cycles. In fact most of those two cycles are spent holding the last pitch...
then it updates
in 1.5mS  (i figure this at 5 RC at the sample hold....). The "tach" method to
borrow a phrase... "can't touch this..."

If you can get a reliable square wave in the first place, the PV-1 will track it
flawlessly.
The trick of course is to get that square wave first.... (I'm working on that.)

Of all the things it is hard to recover a square wave from... I'd consider the
Human Voice the very hardest. It is almost nothing BUT harmonics, unstable in
pitch, amplitude, and there is absoultely NO WAY TO PREDICT where it is going
next....

H^)  Harry.

Maybe there should be a pitch to voltage design contest... who can track what
the longest ???  Remember that so far (axon guitar synth excepted - different
method) no one has beat the 2 cycle delay.

danial stocks wrote:

> >The problems I noted were these (I tried a simple plug in a mic test...)
> >
> >1) There is no input filtering, and the human voice usually has many
> >harmonics. If the fundamental frequency
> >isn't reasonably pure a garbled output will result.
> >
> >2) When a note decays it takes several cycles for it to stop. If these
> >drift in frequency, or if the last two detected cycles get chopped up, the
> >sample hold will latch the wrong value.
> >
> >3) Higher notes have a greater chance of mistracking for reason #2. If you
> >consider that the voice stops in (oh...) 5mS regardless of pitch... then a
> >200Hz input would be gone within 1 cycle, and the right value would still
> >be latched. But at 1KHz 5 cycles would occur, and the chances of latching
> >bad data would increase.
> >
> >There is a latch input that can be driven by a positive signal... you could
> >use a footswitch, or output from an envelope follower to latch the
> >sample/hold before the decay of the note. Then with careful singing it
> >would work...
> >
> >I want to do vocal tracking also. It will need additional circuitry. Its
> >not as easy as Guitar which will only need a little processing... because
> >the vocal note could get louder at any time (guitar always gets quieter on
> >the average...) I made this board with the intention of have a reliable
> >module to experiment with. The parts that are sure NOT to change are there,
> >the squaring circuits, the ramp, sample hold, and antilog converter. If you
> >experiment you MUST have these functions... and the "Ramp" type P/V
> >converter is always faster than the "tachometer" type.
> >
> >H^)
>
> I used an unusual technique for pitch vtg conv a while ago...
>
> 1. square the i/p - feed it thru diode clipper... nice sq wave.
> 2 feed it thru a 6db/oct lpf
> 3 use a precision op amp rectifier like the suggested ones in nat semi
> LM3914/5/6 display drivers data/appnotes
> this gives a nice [inverse] linear frq to vtg relationship - ie a 1 octave
> down will give double the vtg..
> If you want it the right way around ie vtg inc's w/freq then you need to
> take the reciprocal of this vtg.. I designed a linear reciprocator with a
> couple of op amps and a transcond. op amp, haven't built it yet.. If anyone
> wants to see the diagram I'll see if I can find someone with a scanner...
> the thing about this technique is .. it's fast .. it doesnt glitch on
> harmonics etc..
> you can't effectively do it the other way around ie with a hpf istead of lpf
> because of harmonics in the sq wav giving wrong readings [and you need the
> sq wave for input amplitude+ harmonic immunity.
> Cheers,
> Dan
>
> ______________________________________________________
> Get Your Private, Free Email at http://www.hotmail.com