[sdiy] Guitar Synth boards ready...

[harrybissell]

Ok we go down the lane...

The CEM3396 needs a sync pulse and a voltage proportional to both "frequency"
and "waveshape"

Normally the CPU will handle that. Without MIDI note information the internal
CPU has some big
problems.

I went down this path

1)  Use the Muffy for direct waveform generation... Disable the internal DCO and

inject the audio waves direct into the CEM3396.  There are two waveshapers, so
you could
run two different external waves and mix and process them

con:  The internal PWM process would be useless (unless you hacked in some
hysteresis). The
external audio signal (at rest) could/would be right at the comparator
threshold. This makes a
sh!tpile of bad noise

2)  Generate an external sawtooth from the guitar string directly.  It would be
in tune (perfectly)
with the guitar.  Making the sawtooth would require a current source
proportional to pitch... hence
we just entered the P/V conversion arena.   I tried several methods... the
problem with direct conversion
is you already have the pitch pulses... you need LINEAR current for the ramp and
you have 1/X.  Needs
an analog divider chip (I tried the RC4200) or a divider / feedback system
(TL072 and CA3280).  There
is an issue with how to keep the system stable when the input goes to zero...and
for the first couple of
cycles.  In the CEM3396 case it introduces an audible glitch I didn't like. I
could probably kill that... but
I moved to the next road

3)  Make the sawtooth with true (linear) VCO technology.  I tried the GR-300
method... very clever in that it does
not require an analog divider.  Very UN clever in that it does not produce an
amplitude stable sawtooth
(lower notes have bigger amplitudes).  They use a crude waveshaping to clip the
sawtooth so it goes from
full saw to a saw-like pulse (call it spaced saw).  Again... not my cup of tea.

The one NICE thing is that the true VCO allows tuning variations... LFO mod...
and decouples the synth
sound from the jitter of the guitar string sensing.  It starts to sound like a
true synth.

4) Make the sawtooth with a true V/oct VCO.  This is a cool idea in that the P/V
conversion needs only
a pair of matched transistors (and tempco) rather than a full analog divider
(and no tempco).  I 'think'
that if you did both the log-ratio and the expo current source in the same
pair... the tempcos would be
unnecessary (but now I'm over my own head, perhaps mssrs. Schmitz, Danielson,
Haible will rescue
me :^)

So I guess my take at this time would be to go with plan 4... inject those saws
into the CEM3396.
I would send a MIDI note on... of a KNOWN note (always the same).  Now the
voltage from the
CPU is proportional only to waveshape.  I could adjust the external VCO
amplitude from that CV and use
the internal waveshaping...

... or just do the waveshaping in external VCOs.  Probably I'd do an external
VCO bank... an easier solution than the all the hacking.

I think I will re-visit the other waveshaping ideas once I get my MONO_PV boards
in.   I'm just too tired
of that huge vectorboard I've been hacking for six months.  Now that I have
suceeded in making an acceptable Guitar P/V I could probably do more with the
simpler waveforming ideas.

Gate and trigger turn out to be not such a big deal... detecting the fact that a
note has stopped is
essential for the P/V so it comes for free after that.

I want access to all the cool Matrix features... and the patch storage.  OTOH I
have been working
(offlist) with a gentleman who has a Poly SEM system he wants to drive... so
that helped to get me
off my de at d @ss and do the R&D.   I was really on that road anyway, so he
inspired me to pass that last
exit :^P

H^) harry

john mahoney wrote:

> ----- Original Message -----
> From: "harrybissell" <harrybissell at prodigy.net>
> > Ahhh but...
> >
> > ... It is indeed 'trebley' (is that a word?)
>
> Well, almost! I should have removed the "e":
>     http://dictionary.reference.com/search?q=trebly
>
> Harry, we've discussed it a little before, but I'd like to review your
> concept of "hacking a Matrix 1000." The M1000 (and Matrix6) use the Curtis
> CEM3396, a chip that can do complex waveshaping and filtering and so on. The
> 3396 is normally driven with square waves from DCOs.
>
> Your intention, as I recall, is to use square waves from Muffy to drive the
> 3396s, right? But there was an issue that I can't recall...
>
> Do you also need a CV that's proportional to the pitch, and thus you must
> have pitch-to-voltage converters on each string to derive the CVs?
>
> Or, was it that you need a trigger and/or gate for each string? Now that
> I've written it, I seem to recall that the challenge was to produce a valid
> gate. The tricky part is that a string's output trails off, but this isn't
> really any different from using a noise gate on a guitar. (Oh, right --
> noise gates often chop off the sustain, which lead to special
> made-for-guitar gates.)
>
> A guitar-driven Matrix would be very nice.
> --
> john