Some phasemodulation thoughts

[Magnus Danielson]

From: Rene Schmitz <uzs159 at uni-bonn.de>
Subject: Some phasemodulation thoughts
Date: Sun, 03 Oct 1999 23:46:25 +0200

> Hi all!
> 
> I've been thinking about Maciej's proposal makeing a phase modulator from a
> wavetable oscillator. 
> 
> Now my idea would be to use a analog sinewave generator (like the AD639).
> (For all who don't know that chip: It can generate a voltage proportional
> to sin(x) for +/-500degrees!)
> However I would try to implement it using discretes, I've read the
> discussion of that chip on Martin Czechs website, where the idea of the
> internal circuits are shown. (http://www.geocities.com/SoHo/Museum/4459/)
> I simulated a circuit which can do +/-360 degrees. (But I'm not sure about
> temperature compensation. As of yet the "phase gain" depends still on
> temperature.) 
> 
> The task would be to add the output of a sawtooth VCO together with the
> modulation signal. 
> 
> To allow for deeper phasemodulation one could "wrap" the phase signal with
> a circuit similar to the serge waveshaper. The folding should occur at
> +/-(3/2)pi. Let pi=1 (volts :-) and set up a chain of full wave rectifiers
> to fold the modulation voltage at +/-1.5V +/-3V +/-4.5V ...
> The amplitude of the saw would have to be 1V (2Vpp).
>
> However, the sinewave shaper is not necessary at all, for makeing a
> triangle one can just use the waveshaper chain, and add a second signal to
> a sawtooth input. The foldover points would have to be at multiples of the
> sawtooth amplitude then. But doing a triangle like this and feeding it to a
> sineshaper afterwards will not yield the same results!

First, thanks for the reference into MC's pages, I didn't know he had
additional Serge Wave Multiplier material there.

Martin Czechs: You have old links to my material - it is now at
http://home.swipnet.se/cfmd/ but the tree looks the same from there.
You others may take a look at your links to me. I will soon go and actively
search for wrong links!

Then, I have also been thinking about a similar way of doing things.
For instance, if you look at http://home.swipnet.se/cfmd/synths/schematics/
you will find a very crude schematic on a sawtooth frequency doubler idea
that I got once. It basically builds on a comparator which compares against
some level (0V in this case) and uses a diode to only pull in one direction.
Then, the original signal is added to this squarewave and the R3 will put
the signal back on balanced levels (+/- 2.5 V) while the U1B will get it up
to level again. This scheme will produce a double frequency sawtooth in a
very rude and really glitchy way. People should note that the schematic shows
a curcuit that is intended to be used alongside the ASM-1 VCO. It's easy
enougth to rescale it. A version of this has been realized by Jörgen Bergfors
for his ASM-1 VCO based VC-LFO.

In a similar manor may you build up the necessary wrapper curcuit that you
where discussing. You would need to insert more comparators and set their
up their comparator values in the same fashion as for a flash A/D converter.
The end result of those outputs (mixed 1:1 just the same way as I did it)
would look like the result if you A/D and D/A converted the signal.

The mixing would have to be adjusted, so would the output gain. These
properties will depend on the number of comparators being used.

This form of phase modulation is limited in the number of cycles you may
modulate, those the depth. However, for small cycle counts it will work
pretty well.

Another method of doing it would be to do some brain surgery into the VCO core.
If we take a VCO core like the ASM-1 VCO (which is quite similar to many
others) you can do it like this:

1) Create a diffrential curcuit through which you send your modulation signal.

2) Add the output modulation signal to the load current signal of the
   capacitor.

3) Add a negative reset curcuit to the VCO, core. This is necessary for taking
   the oscillator multiple of cycles backwards. This basically means another
   LM311 comparator with associated curcuitry (including another reset JFET).

The benefit of this method of doing things is that you are not fundamentally
sensitive to the number of cycles you move backwards. However, the diffrential
curcuit is prone to leakage and imperfections just as a integrator is, so care
has to be taken into avoid it, this makes very low frequency modulations a
headache where as the other modulation technique excell in low frequency
modulation.

For those who is really eager to solve these things, you can do a hybrid
solution, so that you can actually compensate for leakage as well as take care
of low freqency modulation in the backend where as deep highfrequency goes
into the VCO core. This takes some additional enginneering to do, but i think
you can do it.

The diffrentiation of the modulation signal will basically take out the
integration done in the VCO core, so by feeding a diffrentiated modulation
signal will the sum signal after the VCO integration be the same, just that
the voltage limitation is not there.

I have been considering doing this to a ASM-1 VCO, but never come around to
actually do it. Note that you can also add deep linear FM modulation signal the
same way, but add the FM signal after the diffrentiation. Those one would end
up with a deluxe VCO having exp FM as well as deep lin FM and lin PM.

Should we try this?

Any killers anyone?

I could enginneer some untested schematics if really needed.

> Note the discussion of the Serge waveshaper at Martins site! He compared
> the waveforms from Yamaha PM, with what you get when variing the input
> amplitude on the waveshaper.

This is interesting, you should get further insigth by the above comparisions.

> This circuit will make high demands on the sawtooth input, to avoid
> glitches. But they could be compensated like in Jorgen Bergfors' sineshaper.

Which then built on my idea ;)

Cheers,
Magnus