Originally from
here.
Ask Uncle
Willy
#3: July 7, 1995
Uncle Willy has received several questions about the sounds and music in
games by Williams Electronics Games, Inc. Since this a rather extensive
subject, Uncle Willy has decided to devote this week's issue of Ask Uncle
Willy to covering this subject in depth. Uncle Willy apologizes for the
length of this article, but hopes this topic proves interesting.
Uncle Willy can be reached at
uncle_willy@wms.com
All comments and questions are welcome. Keep in mind that some questions
just do not have an answer. Others, of a proprietary nature, do not permit
an answer.
Uncle Willy enjoys hearing from you!
Question:
Can you tell me about the sounds and music from the games by
Williams Electronics Games, Inc.? How are the sounds made? How is
the music composed? How does the sound system recreate sounds?
Who are the people who compose music for Williams games?
Answer:
Introduction
Williams Electronics Games, Inc. designs and manufactures pinball
games under the Williams and Bally names and coin-op video games
under the Midway name. Anyone who has played these games knows that
the computing power and graphics behind them have become very
sophisticated and lifelike. Pinball games are equally sophisticated,
often containing several computer-controlled playfield toys that are
integrated into the pinball action.
To keep the audio on par with these state-of-the-art games, Williams
has developed a new sound system, called DCS (which stands for
Digital Compression System). The DCS sound board provides four
channels of 16-bit digital audio, with independent control over the
volume, looping and playback of each channel. Each channel can
dynamically play back anything from an entire piece of music to a
short sound effect, with a typical game using one channel for music
and the remaining three for sound effects, speech and foreground
music such as fanfares or breaks.
The first pinball game to use DCS was Indiana Jones. The first
video game to use DCS was Mortal Kombat II.
Background
DCS is not the first digital game sound system, but it is the first
truly high-fidelity, digital audio system designed specifically for
coin-op arcade pinball and video games. A comparison to past and
current game sound systems puts its sophistication and sound quality
in perspective.
The "beeps" and "blips" of early arcade games were made with analog
circuits and simple digital tone generators. These sounds dis-
appeared from arcades in the early 1980's when chip sets became
available to do FM synthesis. For several years, most sound systems
for arcade games combined a microprocessor, and FM synthesis chip
set, and a low sample rate, low quality digitizing system. These
systems were very similar to many of the sound cards currently
available for personal computers, and provided fairly complex music,
somewhat understandable speech and sonically interesting (if not
altogether realistic) sound effects.
FM synthesis was in turn replaced by sample playback systems,
implemented with custom hardware or with software running on a
Digital Signal Processor (DSP). Samples help to make music more
realistic, but the systems often suffer from a lack of polyphony and
and upper limit of a few seconds recording time for each sample.
One of the most recent advances in game sound is CD-ROM. Although
CD-ROM technology is popular in home and personal computer games,
there are several drawbacks that limit its usefulness in fast-paced,
multi-player arcade games. One problem is that CD-ROM audio schemes
are limited to a single channel of mono or stereo sound. Arcade
games require several independent sound channels for layering music,
sound effects and speech. Access time is another problem. Delays
as short as 20 milliseconds between action and sound can make an
arcade seem sluggish. Most of the CD-ROM game systems rely on a
separate sample playback or FM synthesis system for interactive sound
effects. Also, few CD-ROM systems can withstand the bumping and
shaking excited players can inflict on arcade game cabinets.
DCS was designed to overcome the limitations of other game sound
systems. There were three main design parameters: improving sound
quality, maintaining interactivity and streamlining sound develop-
ment. 16-bit audio played back at 32 kHz delivers near-CD sound
quality. The ability to instantly start, stop, loop and control the
volume of four independent channels in response to commands from the
game provides interactivity. Most importantly, the process of
developing sounds changes from programming synthesized sounds to
producing music and sound effects in profession recording studios.
Audio Data Compression
The biggest limitation with most game and multimedia sound systems is
storage. 16-bit digital audio at a rate of 31,250 samples per second
requires about 60 kilobytes of storage per second of sampled sound.
Overall cost of a game limits the ROM available for sounds to about
3 megabytes. The problem is that 3 megabytes of ROM at 60 kilobytes
a second is only enough for about one minute of total sound. Most
Williams games have a total of 10-15 minutes of non-repeating sound.
The solution to this storage problem for DCS was the development of
a proprietary transform coder algorithm that reduces a 500 kilobit/
second data rate by a factor of ten or more, and runs on a low cost
DSP chip. This algorithm is similar to the algorithms used in the
Sony MiniDisc format, the Philips Digital Compact Cassette format
and the emerging digital sound formats for movies such as Dolby
Stereo Digital, Sony Dynamic Digital Sound, and Digital Theater
Sound.
The encoding phase takes place on a PC, starting with digital audio
files. The fields are broken down into frames of 240 samples each.
A frame is 7.68 milliseconds of audio, and files can range from
one to several thousand frames in length. Each frame is transformed
by a 256-point Fast Fourier Transform, with simple cosine windowing
and eight samples of overlap on each end. The resulting spectrum is
further broken down into 16 bands and quantized according to masking
curves and user-controlled parameters. The quantizing levels and
the resulting data for each frame are entropy encoded into variable
length packets. The packets for each file are combined with header
blocks and stored as files that are used later to generate ROM
images.
The decompression phase takes place on the DSP chip. Starting with
the header, a packet of compressed data is read from ROM and
decompressed into a frame. This involves entropy decoding the data
stream, and then de-quantizing the data into a frame of frequency
domain data. The four channels are decompressed independently and
summed in the frequency domain. Volume operations, such as level
settings and cross-fades, are also performed in the frequency domain.
The final frame is inverse transformed and clocked out a serial to a
Digital to Analog Converter (DAC). Music and sound effects compress
to an average of 50 to 70 kilobits/second, and speech reduces to 20
to 40 kilobits/second. The resulting audio quality exceeds that of
other algorithms operating at similar bitrates.
Several aspects of the DCS algorithm capitalize on the nature of
sounds for games. The encoding, which is more complicated than
decoding, can be performed on a PC without any restrictions on
processing time. The algorithm has a variable bitrate, which means
that the amount of artifacts and the overall quality can be fine
tuned for each individual sound. Since the frame size is small,
it is possible to carefully edit sound effects and seamlessly loop
music.
Hardware
The DCS hardware is relatively simple. The major components, a
surface-mounted DSP chip, sockets for ROM, a 16-bit mono DAC,
and an audio power amplifier, are contained on a two-layer printed
circuit board measuring about eight inches square. There are
additional discrete analog and digital components making up filter,
interface and power supply circuits.
The DSP chip used is the ADSP-2105 by Analog Devices. The 2105 runs
at 40 MHz and executes over 10 million instructions per second.
The 2105's on-chip RAM is supplemented with additional static RAM
on the DCS board. This memory is used primarily as temporary
storage for the realtime decompression and inverse FFT operations
required for each of the four playback channels.
DCS contains a bi-directional, 8-bit interface between the sound
board and the game host. Playback commands from the game host
can trigger anything from a single sound effect to a one-minute
segment of music that loops indefinitely. Commands from the host
are asynchronously received and buffered and can arrive at a
rate greater than 25 thousand per second, although a typical rate
in game play is 10-100 per second.
The sound board can also send timed data back to the game host.
This data is useful for synchronizing animation, display effects and
light shows with music and sound effects.
Staff
Several full time composers and one freelance composer produce sound
effects, voice overs and music for all of the pinball and video game
projects.
Jon Hey has been at Williams for several years. The Jon Hey Band
recently headlined at the Chicago Film Makers benefit and performed
at the Bucktown Arts Festival in Chicago. Vince Pontarelli joined
Williams two years ago, having previously worked at Libman Music, a
well-known Chicago jingle and post-score house, and as a writer for
Todd Scales at Windsound. Dave Zabriskie also joined Williams fairly
recently, having come from Premier pinball. In addition to a long
list of game sound credits, Dave has written extensively for
orchestral and choral groups, including a recent string symphony for
the Chicago String Ensemble. Kevin Quinn is the newest addition to
the Williams sound department. Before coming to Williams, Kevin
owned and operated Concord Music, a successful commercial jingle house
in Chicago. Freelancer Dan Forden works out of his own project studio.
Dan has put together and worked out of several project studios over
the last several years in connection with bands that he has both
played in and recorded.
Facilities
Each staff composer has an office and a MIDI workstation. Supporting
the MIDI workstations is a fully equipped recording studio built
around a Tascam M-2516 16-channel mixer. Four patch bays connect
the mixer to a DS-30 DAT recorder, a CD player, an Alesis ADAT and
the outboard gear, which includes a Lexicon LXP-15, an Ensoniq DP/4,
and various equalizers and compressors. The studio is divided into
two rooms, one of which is built around an isolation booth.
The studio is used mostly for recording vocals and voice-overs in the
isolation booth and for creating and editing sound effects on a
Macintosh IIci running Digidesign Sound Tools. Sound files are
stored on removable 45 megabyte cartridges, making it easy for each
person to manage the files for the game projects he is working on.
Production
Original music is composed for each game. The MIDI workstations for
the staff composers consists of a Mac Quadra 630 running Mark the
Unicorn Performer software, a Kurzweil K2000 keyboard, one or more
Emu Proteus modules, Roland JV 1080 sound modules, and Ensoniq DP/4
effects processor, and Alesis Quadraverb II, and Alesis Quadraverb
GT and an Alesis D4 drum module. Each workstation also has its own
MIDI interface, mixer, compressor and equalizer. For variety, there
is also a Roland JD-990, a Korg 01W/FD, and Ensoniq SQR and two Akai
S1100 samplers and an Alesis ADAT digital multi-track tape machine.
In addition to the Mac Quadra, each composer also has a '486 or
Pentium PC. The PCs are used for hard disk recording and editing,
using custom hardware and software that generates files for the
data compression system.
All music must be uniquely arranged and recorded for a game, even
when the music is based on a film or television score. For example,
in Star Trek the Next Generation pinball, which was arranged and
scored by Dan Forden, the music package included an arrangement
of the show's well known theme (based on the movie theme by Jerry
Goldsmith) as well as several original Forden compositions.
Sound effects for games fall into three basic categories: sound
effects pulled from CD, field recordings and sounds taken from film
or video soundtracks. Williams owns several sound effects libraries.
A portable DAT recorder is used to make field recordings. Many
sounds, such as those used in the Indiana Jones pinball, are taken
from sound effects submixes of film soundtracks. Regardless of the
source, almost every sound effect is custom edited and processed to
work in the context of the game.
Speech and voice effects are also an important part of the sound
package for any game. Anywhere from a fourth to a third of the
available memory space is used for speech and voice effects which
inform and entertain the player. This is especially important in
games based on movies or other high profile themes. For example,
each major cast member of the television series Star Trek the Next
Generation recorded material expressly for the pinball game. Vocal
talent is also recorded in the isolation booth located in the
Williams sound department studio. This was the case with all of the
screams, grunts and moans heard in the Mortal Kombat II and 3 video
games.
Because WIlliams sound designers now work in a recording studio using
professional tools, as opposed to programming music and sounds in
assembly language, the time to complete a sound package for arcade
pinball and video games has been cut roughly in half. More import-
antly, the sound quality has been drastically improved with the
advent of DCS.
All text and images © 1995 Williams Electronics Games, Inc.