Professional Documents
Culture Documents
Intro
Intro
Intro
Keith McMillen
Zeta Music/Gibson Western Innovation Zone
2560 9th St. Suite 212
Berkeley, California 94710 USA
McMillen@CNMAT.Berkeley.edu
The success of alternate controllers has been less than overwhelming in the
history of electronic music. The predominant controller for electronic music
synthesizers has been the piano or organ keyboard. Beside the widespread
availability of pianos and organs and the people who play them, the very
nature of the keyboard makes it an ideal choice from an implementor's point
of view.
For musicians trained on other instruments, the option of synthesis has not
been attractive. Woodwinds, bowed strings, and brass instruments all place
the player in direct physical contact with the vibrating element---reeds,
strings, or columns of air. Instead of limited control of dozens of notes
these instruments offer subtle and intimate control over one or a few notes.
Whether to "trade in" this control for a wider tonal palette is a difficult
decision.
MIDI has been serving our interface needs for over a decade. Although many
have criticized MIDI (Loy 1985; Moore 1988; Scholz 1991), no one has done
much about its obvious problems. Alternate controllers have not been a major
factor in the business of electronic music, and therefore have not been well
accommodated by the industry. They represent a challenging problem both
technically and economically. The persistence of an interface standard that
makes the necessary extensions for nuance and control difficult if not
impossible has not helped.
The connection of keyboards to sample playback sound modules is well served
by MIDI. Even the speed of MIDI (31.25 kBaud) is adequate for transmitting
data using the event-based nature of a keyboard. A ten-note chord can be
sent in 6.7 msec, a delay which is on the borderline of being imperceptible.
The continuous controller information generated from a keyboard usually has
no more than three parameters (pitch bend, modulation, and after-touch),
keeping the bit count low.
6 strings * (3 pitch bytes + 3 volume bytes) * 10 bits / 0.01 sec = 36.0 kBaud
MIDI does provide some facility for continuous volume change (controller
#7), and for pitch change without articulation. Some synthesizers respond to
legato-style commands. Pitch bend can vary pitch up and down up to one
octave but with a resolution of only 5.1 divisions per semitone (19.6 cents).
Do you remember the days before MIDI? Most available synthesizers were
analog and used analog voltages to represent musical values. Articulation
was separate from pitch and all controllable parameters were on equal
footing. Bandwidth and resolution were not concerns but good intonation was
a perpetual effort---a lot like playing a violin.
This did not preclude processing of the audio signal out of the synthesizer.
Several synthesizers have individual outputs for each voice. These could be
mapped to a specific string from the controller and modulated in the analog
domain based on information extracted from the string.
One of the most satisfying examples of this was the connection of a Zeta
Mirror 6 model fret-scanning guitar controller (Wait 1989) to a Yamaha
TX-802 FM synthesizer operating in legato mode. Each of six outputs from the
TX-802 was routed back to the Mirror 6 where it passed through a
voltage-controlled amplifier (VCA). This VCA was then controlled by an
envelope follower tracking the energy of one of the strings. The six VCA
outputs were summed and fed to an amplifier. Guitarists marveled at how
smoothly responsive the instrument felt and the great intimacy of the
control over synthesis. The simple use of continuous dynamic control
returned much of the musical nuance to the interface.
This did, however, limit the choice of synthesizers for the users of
alternate controllers. The emergence of sampling and its eventual
monopolization of the synthesizer market created new problems for
interfaces. The frequency of FM oscillators and VCOs are continuously
variable over the entire range of pitch. Samples, as the name implies, are
not, and require swapping of files that cover specific pitch ranges in order
to cover a wide range of pitches. Playing a trill across a sample boundary
results in discontinuous spectral envelopes for many sampled sounds.
Articulation for FM and VCOs come from external envelopes that can be varied
based upon input parameters from controllers. With sampling, the attack
character of the sampled instrument is inherent in the wave table. Timbral
changes are restricted to, at best, simple filtering or cross-fading between
fixed sounds. Even something as personal as vibrato is often captured by the
sample and is not under the performer's control.
The "skin-deep" beauty of sampling has left many musicians longing for a
more meaningful conversation with their instruments. Nostalgia has even
created a demand for older voltage-controlled synthesizers. Our group
recently completed the design of the Oberheim OBMx, an analog twelve-voice
subtractive synthesizer. One reviewer said of it, "it has what synthesizers
have been missing---a personality" (Aiken 1994). This willingness to give up
the accurate reproduction of acoustic instruments for control is
understandable but the situation demanding this choice is regrettable.
Breaking the Chain
==================
The first concepts for what was to become the ZIPI musical data language
started in the fall of 1989, coinciding with the start of intense
collaboration of Zeta Music and the Center for New Music and Audio
Technology (CNMAT) at the University of California, Berkeley. In order to
successfully improve the keyboard-MIDI-sampler path, replacements were
needed for all three of the elements in the chain. Since that date, research
has focused on the Infinity Box (a gesture-guided pitch and timbre to ZIPI
converter), the ZIPI network and its specification, and Frisco (an additive
resynthesis engine with a control structure that will respond to ZIPI MPDL
commands).
As of this writing (June 1994) the physical layer of ZIPI has been
implemented using a ring of Intel 80386-based personal computers with Zilog
8530 cards with ZIPI PALs and current loop hardware installed. Software for
the monitor and nodes has been written for interrupt, polled, and DMA access
methods. Data link and basic network services are functioning. The polled
approach is the only viable method for an Intel 80386-based machine since
the MS-DOS operating system's interrupt latencies are too great to allow
even 250 kBaud operation. DMA requires the capture of an entire packet by
the monitor before it can be parsed, thus slowing the ring. All code
development for ZIPI is written in the C programming language with an
emphasis on portability to other processors.
By the time this article is published (December 1994), we will post some
ZIPI-related software to the Computer Music Journal's ftp site,
mitpress.mit.edu, in the directory /pub/Computer-Music-Journal/Code/ZIPI.
Just as the RS-232 serial connection standard continues to exist along with
Ethernet, we have no illusions that ZIPI will replace MIDI. Likewise, just
as MIDI has been pressed into service in areas never intended by its
designers (machine control, mixer automation, lighting), we can't fully
anticipate other manufacturers' networking needs.
ZIPI was presented to the industry at the winter NAMM shows in January 1993
and 1994. Participation and suggestions of many companies have added much to
the scope and practicality of what is presented here. We encourage readers
to suggest details for additional application layers, such as machine
control, studio automation, and sample dump and audio standards.
ZIPI Group
G-WIZ
2560 9th St., Suite 212
Berkeley, California 94710, USA
electronic mail: zipi@CNMAT.Berkeley.edu
Acknowledgments
===============
This work was supported in part by Grant C92-048 from the California State
Department of Commerce Competitive Technologies Program to CNMAT and Zeta
Music Systems, Inc. We would also like to thank the following individuals
for their valued comments, sometimes very critical, regarding the ZIPI and
MPDL specification: Jim Aiken, David Anderson, Marie-Dominique Baudot,
Richard Bugg, Tim Canning, Chuck Carlson, Lynx Crowe, Rob Currie, Steve
Curtin, Peter Desain, Kim Flint, Adrian Freed, Guy Garnett, Mark Goldstein,
Henkjan Honing, Dean Jacobs, Henry Juszkiewicz, Michael Land, Carl Malone,
Dana Massie, Bill Mauchly, Peter McConnell, F. Richard Moore, Chris Muir,
David Oppenheim, Stephen Travis Pope, Rob Poor, Miller Puckette, John
Senior, Warren Sirota, John Snell, Michael Stewart, Tovar, and David
Zicarelli.
References
==========
Muir, C., and K. McMillen. 1986. "What's Missing in MIDI?" *Guitar Player*
June 1986.