[sdiy] all analog FM synthesis?

[Johannes Öberg]

(I'm replying to many authors at once; should I stop this? It's very
convinient if one is using Gmail (I have a few invites if anybody
wants an account))

>No. Actually it's easier than FM. (If you want to do it correctly and with
>a deep modulation.)

>OTOH, this particular part of the circuit may be quite a bit easier to
>implement via ADC / lookup table / DAC - that means a mixed signal uC.

Argh, I realized precisely these things right after I disconnected my
modem :-) The bad thing being that your analog FM synth will still
sound like an 8-bit 22kHz Yamaha soundcard! :-)

The trapezoid VCO seems really nice for something like an analog
PM-synt. After all, there's nothing saying you must have sinewaves do
to PM, in fact, triangles even sound pretty much the same.

>...and a vector multiplier. The vector multiplier reduces to a dual
VCA plus two mixers plus
>some signal bending at the two CV inputs if you also use the negated
>quadrature signals from the VCO (cos, sin, -cos, -sin) or if the
>quadrature VCO already has differential outputs. The CV signal bending
>is the hard part and involves folding the phase signal into wrapped
>quadrature phase signals and sin shaping of the result.

For a programmer (like me), this means something roughly equivalent to
like "take the (current sample * modulator depth) modulus (the size of
the look-up table)" right?

The sineshaper wouldn't really be necessary unless you're going for
authentic DX sounds is it?

>They really would sound the same as long as you do not chain
>"operators".

This is because the derivate of a sine wave is another (co-) sine
wave, right? Then it won't work if you're using complex modulating
waveforms, right? If I ever build an analog FM synth, I sure will want
to try out all sorts of non-standard oscillators, with PWM and stuff.
FM is so much more fun on the TX-81z (actually, I'm using the OPL-3)
than the DX-7 IMHO.

>But they are essentially almost the same:
>So phase modulation should sound like FM with a constant-slope HPF:ed
>modulator (i.e., derivative of the modulator), which means there is much
>more treble content in the modulator.
>To some extent, I think it should be possible to use a constant-slope LPF
>on the phase modulator in order to make it sound like FM (and vice versa).

Is this based on theory or experience? I can see why it *should* work,
but hardly why it would :-)

I know next to nothing about the math behind FM and PM, but I've been
playing around with both of them on my synthesizers, and in my
experience they sound very different, to the point where synced-FM
sounds more like PM. Where PM wants to sound like an electric piano,
FM always tend to sound like your using a circle saw on an electric
piano... (which might not be a bad thing depending on taste)

FWIW, the wierd (and cheap second-hand, because it's mostly crap)
Evolution EVS-1 has both FM and PM, check it out.

>I can't make any sense of this.  "Mst input"?  "Grey signal"?  "Clavia
>units"?  "Constant module"?  "Slv output"?  Can anybody translate this
>into regular terminology?

Look at the pretty pictures :-) It basicly boils down to hardsync
where you modulate the synced oscillators frequency with another
oscillator. That way you won't get the dissonance associated with real
FM.

I don't have a Nord Modular, but from what I figure, Mst input is
Master frequency input,  Clavia units are explained at the top of the
article, a constant module is a 'DC output' and "Slv" means "Slave
output" (hey, where's your LED display skillz? ^_^). I'm guessing that
a Grey cable is used to indicate that the module connected at the
other end isn't recieving data every sample but is really slaved in
some outer loop in the DSP code (for processing speed).

>Maybe I should make the lab-exercise for you to listen to?

Väldigt gärna!