[net.music.synth] Stupid Question About FM

csdf@mit-vax.UUCP (Charles Forsythe) (09/12/85)

I can't take it anymore! I can't wait a year to take Bose's accoustics
course to learn this. What, exactly *is* FM sound synthesis and why is
it so great? Somebody out there must know! I expect that I am simply
missing something obvious because I do know:

1) What a Fourrier transform is.
2) What a Z-transform is
3) What FM is (as a process)
4) Just about any other "signal processing" jargon.

Unfortunately, MIT loves theory but not application, so they never
bothered to explain how FM could be used to create amazing synth effects
and such.

I appreciete any answers anyone cares to send, and if anybody else has
been silently ignorant, I will be happy to summarize to the net.
-- 
Charles Forsythe
CSDF@MIT-VAX

"What? With her?"

-Adam from _The_Book_of_Genesis_

mohan@topaz.RUTGERS.EDU (Sunil Mohan) (09/13/85)

> I can't take it anymore! I can't wait a year to take Bose's accoustics
> course to learn this. What, exactly *is* FM sound synthesis and why is
> it so great? Somebody out there must know! I expect that I am simply
> missing something obvious because I do know:

 An excellent reference would  be John Chowning's  article on  FM  for
music. This was  republished in the Computer Music  Journal (I  forget
the issue, but pre 1980 I think), and also in the book "Foundations Of
Computer Music", both published by MIT press. And you do not need Bose's
Accoustics course.

I think there are two main reasons why FM is "so great" (I  personally
am not too impressed, but then the kinds of sounds I need are not that
readily, or at least easily reproducible in FM):

- Something to do with its being digital allows it to produce nice 
  "clean"   and strident sounds,  which   can easily  ride over  other
  instruments, making it good for leads in pop/rock.

- The  ability  to specify   time-varying modulation  ratios allows  the
  overtones  structure  (timbre) of  a sound   to  vary over its  Volume
  Envelope, and   with note dynamics. This  is  a prime  requirement for
  duplicating natural accoustic timbres.

All FM instruments allow additive synthesis, which  is inherently more
flexible than subtractive  by allowing arbitrary  overtone structures.
Note that the voluble  proponent of additive  synthesis, Wendy Carlos,
uses individual oscillators for the  lower overtones, resorting to  FM
only at the upper reaches.
This brings me to a burning question:
  
   Why do all Analog synths I have seen (inc the mighty MATRIX-12)
   have only one Voltage Controlled Resonance Frequency in their
   filter sections ?  Wouldn't having more than one add flexibility ?

Also FM allows  a greater variety of overtone   structures (as in  the
shape  of   the locus  of amplitudes of    overtones) than traditional
subtractive synthesis (see flame above).

Has anyone experimented with any of the following:
-	More than one VCF on a voice ?
-	Adding VCFs to a DX7 or sthg similar ?
-	Using a choruser/whatever on a DX7 to try to get "thick
	  analog" sounds ?
-- 

_
Sunil

UUCP:   ...{harvard, seismo, ut-sally, sri-iu, ihnp4!packard}!topaz!mohan
ARPA:   Mohan@RUTGERS

rlr@pyuxd.UUCP (Rich Rosen) (09/16/85)

> I can't take it anymore! I can't wait a year to take Bose's accoustics
> course to learn this. What, exactly *is* FM sound synthesis and why is
> it so great? Somebody out there must know! [CHARLES FORSYTHE]

For a good summary of FM synthesis and its beginnings, read the paper by
John Chowning (of Stanford's CCRMA) called "Synthesis of Complex Audio
Spectra by Means of Frequency Modulation".  It appeared originally in
the Journal of the Audio Engineering Society (Vol. 21, No. 7, Sept. 1973),
and it was reproduced in the second issue (Vol. 1, No. 2) of the Computer
Music Journal.  Those first few issues of CMJ are chock full of absolutely
excellent theoretical articles about digital synthesis, simulation of
acoustic instruments and motion of sound sources.  I stopped subscribing
during the course of volume four.  Does anyone else out there still
subscribe and have recommendations (or not) about the journal?

(Charles, CMJ is published by MIT Press, or at least it was until recently.
In the early days it was published out of Menlo Park CA, hence the early
Stanford contributions.)
-- 
"Wait a minute.  '*WE*' decided???   *MY* best interests????"
					Rich Rosen    ihnp4!pyuxd!rlr

aurenz@uiucuxc.Uiuc.ARPA (09/17/85)

>  What, exactly *is* FM sound synthesis and why is it so great?
>  Somebody out there must know! I expect that I am simply
>  missing something obvious because I do know:
>  
>  1) What a Fourrier transform is.
>  2) What a Z-transform is
>  3) What FM is (as a process)
>  4) Just about any other "signal processing" jargon.

	If you understand FM as a process (e.g. as applied to radio 
	communication), then you understand FM as a synthesis 
	technique. All math is the same, only the operating parameters 
	are different. For example:

	1)  In FM radio, the modulator signal is in the audio band
	    (20-20Khz) and the carrier is around 100Mhz. With FM
	    synthesis, both carrier AND modulator are in the 
	    audio band, and are generally related by small integer
	    ratios (e.g. 2:1, 3:1 etc.) Non-integer ratios are also 
	    used to produce non-harmonic spectra (i.e. metal sounds).

	2)  In FM radio, the modulation index tends to stay constant;
	    in FM synthesis this index varys with time. Doing
	    this varys the spectra over time, which is what "real"
	    and interesting sounds do.

	3)  So for simple simple case of sinusoidal modulation
	    (one carrier one modulator):

			+-------+
			|  Mod  |
			+-------+
			    |
			    V
			+-------+
			|  Car  |
			+-------+
			    |
			    V
			   Out
	    
	    you can use the standard bessel functions to compute 
	    the output spectra. Of course, when you stack your 
	    modulators the computation gets much trickier.

	The reason FM is "so great" is that it's a relatively cheap
	way to generate very complex spectra with relatively few 
	parameters (as opposed to additive synth, which needs quite a lot). 
	So in that respect, FM is a more "powerful" technique
	than additive synth.

	The drawback of FM is, as mentioned in (3), it's very difficult
	to transform between (parameters <-> spectra) for all but very 
	simple cases. Hence the need for much heuristic knowledge of the
	behaviour of FM spectra. 

	By contrast, in additive synthesis one can basically "lift"
	the necessary synth parameters from a 2d spectral plot.

	Well, that's enough hot air for now. Hope it helps!

	-----------------------------------------------------------
						Scot Aurenz

		  { ihnp4! pur-ee! } uiucdcs!uiucuxc!aurenz

aurenz@uiucuxc.Uiuc.ARPA (09/20/85)

>  [KEVIN KARPLUS]
>  ...The balance between technical and musical articles seems
>  to have been lost a few years back.  I don't know whether to attribute
>  this to the loss of John Strawn as an editor, or to the dearth of
>  people submitting technical articles.

	I don't know about the former, but it certainly isn't
	due to the QUANTITY of articles submitted to CMJ.
	When I talked with Mr. Curt Roads, the current editor, 
	at this year's ICMC* he cofirmed that the lag time
	between an article's submission and publication is now
	about one YEAR (!), mainly due to the sheer number
	of articles to be processed.

	* International Computer Music Conference (Vancouver, BC)

>  If you (or someone you know) has been doing research in computer music,
>  I encourage you to submit an article to CMJ.  I could use some more
>  good reading.
	
	Actually, if any of you live in the Boston/Cambridge 
	vicinity, I think CMJ would perhaps appreciate a few
	more minds and hands for the task. And to those out
	there who are referees, etc. for CMJ: hey, what about
	this turnaround problem? Any suggestions?


						Scot Aurenz
		{ ihnp4! pur-ee! }   uiucdcs!uiucuxc!aurenz