Yamaha Synths, Phase Modulation

[Don Tillman]

   From: Martin Czech <martin.czech at intermetall.de>
   Date: Thu, 30 Sep 1999 13:51:57 +0200 (MET DST)

   click, rrrrrrrr (automatic endless tape starting),
   hisssssss, scratch, an automatic voice says: 

   Yamaha synths do phase modulation, not FM.

I've heard that before.  I didn't believe it though.

   This can be proved very simply:
   Modulate one OP with a second, so that a very fast vibrato is
   audible.  Now tune down the modulator frequency, viola, the
   frequency modulation  becomes inaudible. 

Okay, I tried it myself.  Yeah, that's a somewhat convincing
demonstration.  

I tried some more examples, looking at the waveforms as best I could
on a scope.  Hmmm, okay, it *could* be phase modulation.

Next I looked up the patents.  John Chowning's original patent is
clearly FM.  Check out US patent 4,018,121:

http://www.patents.ibm.com/details?pn=US04018121__

But it's also clear that John Chowning's original patent works nothing
like the actual DX-7.  His patent is all 7400-series TTL and uses 80
microinstructions for a single two-operator voice (!!!).  Some aspects
are really interesting though -- see how he interpolates between
successive sine lookups using a cosine lookup (cool!).

Searching some more I found the patent for the actual DX-7, US patent
4,554,857:

http://www.patents.ibm.com/details?pn=US04554857__

One important page apparently got jammed in the scanner.  Luckily you
can get a text version of it at the US Patent Office site:

http://www.uspto.gov/patft/index.html

Specifically:
http://164.195.100.11/netacgi/nph-Parser?Sect1=PTO1&Sect2=HITOFF&d=PALL&p=1&u=/netahtml/srchnum.htm&r=1&f=G&l=50&s1='4,554,857'.WKU.&OS=PN/4,554,857&RS=PN/4,554,857

Yes Martin, you are correct, the DX-7 really is phase modulation.
Excellent call.

Figure 1 shows the operator frequency set by the key being played and
the local pitch ratio parameter, and Figure 8 shows the phase
modulation from other operators.  It's all right there.

Something else...

A number of years ago I posted on Analogue Heaven a personal guess
about how the DX-7 worked.  I said that the cool thing about FM
synthesis was that it could be implemented without any real-time audio
multiplies and that was a big win because fast 16x16 multiplier chips
cost around $200.00 back in those days.  FM allowed you to build a
digital polyphonic synth for a reasonable price.

I proposed that the VCA operation could be performed by adding two
different sine lookup phases and you wouldn't need an actual
multiplier for your VCA.  Well, that's not *exactly* how they did
it...  The sine lookup table supplies the log of the sine, the VCA
control "voltage" is added to that, and then it goes to an exponential
lookup ROM to complete the multiply.  

Still, no real-time multiplies.

  -- Don