Music in Retro Gaming

For most of the classic era, a game console or home computer could not store or stream recorded audio — there was nowhere to put it and no spare processing power to decode it. Instead, machines shipped with a dedicated sound chip that synthesized audio on the fly. Music was data: a sequence of instructions telling the chip which notes to play, on which voices, with what envelope, updated dozens of times per second by the game's code.

Two broad approaches dominated. PSGs (Programmable Sound Generators) produced simple waveforms — square/pulse tones, triangle waves, and pseudo-random noise — giving the bright, buzzy 'chiptune' sound. FM synthesis, commercialized by Yamaha in chips derived from John Chowning's frequency-modulation research at Stanford, layered operators to build richer, more organic timbres and defined the sound of the 16-bit arcade and console era. Later machines added PCM sample playback, where short recordings of real instruments were pitched and looped.

Because the music existed as chip programs and register data rather than as audio waveforms, it can be preserved by storing that data and re-running it on an emulated chip — the basis of every format below.

The earliest console sound was austere. The Atari 2600's TIA (Television Interface Adaptor), the chip at the heart of the 1977 Atari Video Computer System, offered just two channels of one-bit sound, each with 32 pitch values and a 4-bit volume control. Its pitch divisions didn't map cleanly to a musical scale, so 2600 composers famously had to choose between melodies that were in tune and melodies that fit the available notes — a constraint that gives early Atari music its distinctive, slightly detuned character.

Arcade and home-computer audio quickly grew more capable through dedicated PSGs. Texas Instruments' SN76489 provided three square-wave tone generators plus one noise channel, and spread across the BBC Micro, ColecoVision, IBM PCjr, and Sega's consoles. General Instrument's AY-3-8910 (designed in 1978) likewise offered three tone channels plus noise and an envelope generator, powering the ZX Spectrum 128, Amstrad CPC, MSX, Intellivision, the Atari ST (via Yamaha's YM2149F variant), and dozens of arcade boards including Frogger and Gyruss.

These chips defined the vocabulary of 'chiptune': pulse-wave leads, noise-channel percussion, and clever tricks like rapid arpeggios to fake chords on a monophonic voice.

Nintendo's Famicom (1983 in Japan; released in the West as the NES in 1985) ran on Ricoh's 2A03, a single chip combining a modified MOS 6502 CPU with an integrated APU (Audio Processing Unit). The 6502 core had its binary-coded-decimal mode disabled, and the audio side delivered five channels: two pulse/square-wave channels for melody and harmony, one triangle-wave channel typically used for bass, one noise channel for percussion, and a DPCM sample channel for short drum hits and voice clips.

That five-voice palette shaped a generation of game scores. Some cartridges went further with extra sound chips on the cartridge itself — Konami's VRC6, Nintendo's FDS, Sunsoft's 5B and others — expanding the voice count beyond the console's built-in capabilities.

NES music is preserved in the NSF (NES Sound Format), which packages the 6502 audio driver code and music data so an emulator can re-run the original playback routine on a virtual 2A03 — including support for the cartridge expansion chips. SoniqBoom plays NSF by emulating that chip and its driver, reproducing the authentic console output rather than a re-recording.

No chip is more beloved by enthusiasts than the MOS Technology 6581 SID (Sound Interface Device), introduced with the Commodore 64 in 1982. Designed by Bob Yannes, the SID was practically a synthesizer on a chip: three independent voices, each with selectable waveforms (sawtooth, triangle, pulse with variable width, and noise), ADSR envelopes, ring modulation, oscillator sync, and — crucially — a programmable analog multi-mode filter (low-pass, high-pass, band-pass). That filter, plus the SID's willingness to be pushed in undocumented ways, gave C64 music a warmth and expressiveness far beyond a plain PSG. A later revision, the 8580, refined the design.

The C64's musical legacy runs in two parallel streams. Game composers like Rob Hubbard and Martin Galway turned the three-voice chip into something that sounded orchestral; and the demoscene — communities competing to push the hardware to its limits — kept SID music as a living art form long after the machine's commercial life, through ongoing 'SID tune' composition and competitions.

SID music is preserved in the PSID/RSID format (the .sid file), curated en masse by the High Voltage SID Collection (HVSC), which holds tens of thousands of tunes. SoniqBoom emulates the SID chip to play these files, and supports HVSC song-length metadata for accurate track durations.

Sega's 8-bit Master System (1985 in Japan as the Mark III) used a clone of the TI SN76489 PSG — three square tones plus noise — integrated into its video display processor. Bright and punchy, it was firmly in the PSG tradition.

The 16-bit Mega Drive / Genesis (1988) made the leap to FM. It paired a Z80-driven SN76489 PSG with Yamaha's YM2612, a six-channel, four-operator FM synthesis chip (a cost-reduced derivative of the YM2608). One of those six FM channels could be switched into a PCM mode to play digital samples. The combination — gritty FM brass and bass, noisy PSG accents, and sampled drums — became one of the most recognizable sounds of the era, heard across Sega's catalog.

Sega's chips are preserved in the VGM (Video Game Music) format, which is fundamentally different from program-based formats like NSF or SID: rather than storing driver code, a VGM file is a timestamped log of the exact register writes sent to the sound chips. On playback, an emulator replays those writes to emulated YM2612/SN76489 chips with sample-accurate timing. The format covers a wide range of FM and PSG hardware beyond Sega. SoniqBoom plays VGM via chip emulation.

Nintendo's Game Boy (1989) carried a custom Sharp LR35902 CPU — the SM83 core, a hybrid drawing from the Intel 8080 and Zilog Z80 — with an integrated sound unit. Its APU offered four channels: two pulse-wave channels (one with a frequency sweep), one programmable wave channel that played back a 32-sample waveform stored in RAM, and one noise channel. The wave channel in particular let composers craft custom timbres beyond plain square waves.

Constrained but characterful, the Game Boy's sound became iconic — and, decades later, a foundation of the LSDj/'chip music' live-performance scene, where the handheld is used as a hardware synthesizer.

Game Boy music is preserved in the GBS (Game Boy Sound) format, which packages the LR35902 audio driver and song data for re-execution on an emulated APU. SoniqBoom plays GBS through chip emulation, reproducing the handheld's four-channel output.

The Super Nintendo (Super Famicom, 1990 in Japan; SNES, 1991 in North America) took a different path from its FM-driven rivals: a fully sample-based audio subsystem co-developed with Sony. At its core sat an 8-bit SPC700 CPU (part of the S-SMP module) driving a 16-bit DSP, with 64 KB of dedicated audio RAM. The DSP mixed eight voices built from 16-bit samples compressed in the BRR format, and could apply hardware effects including an echo/reverb unit and ADSR envelopes.

Because every instrument was a (tiny, compressed) recording, SNES music could evoke real strings, choirs, guitars, and percussion — within the strict 64 KB budget that forced composers into a constant art of compression and clever looping. The results defined the lush sound of early-1990s console RPGs and platformers.

SNES music is preserved in the SPC format, which is essentially a complete snapshot of the 64 KB audio RAM plus the SPC700/DSP register state at the moment a track starts. An emulator loads that snapshot and runs the audio CPU to play the song — no main-CPU code required. SoniqBoom plays SPC by emulating the SPC700 and DSP.

The MSX standard (launched 1983, championed by Microsoft Japan and ASCII) put a common architecture in machines from many manufacturers, especially in Japan and parts of Europe. Every MSX shipped with a General Instrument AY-3-8910 PSG (three tone channels plus noise) as its baseline sound hardware.

What set the platform apart musically were the cartridge-based sound chips Konami built into its later games. The SCC (Sound Creative Chip), introduced around 1986, was a wavetable chip: it added five channels, each playing a 32-byte user-defined waveform looped to produce a pitch. Riding on top of the MSX's built-in three PSG voices, the SCC let games like Gradius 2, Salamander, and Snatcher sound dramatically richer than the bare standard allowed. Other cartridges added Yamaha FM sound (the MSX-MUSIC/MSX-AUDIO OPLL and OPL chips).

MSX music is preserved in the KSS format, which packages the Z80 driver code and music data so an emulator can re-run it against the appropriate emulated chips — PSG, SCC, and FM. SoniqBoom plays KSS through chip emulation.

Early IBM PCs had only the tinny internal PC speaker for sound, until add-on cards changed everything. The AdLib Music Synthesizer Card (1987) was the first sound card for IBM compatibles to win wide acceptance, becoming an early de facto standard. It was built around Yamaha's YM3812 (OPL2), a two-operator FM synthesizer offering nine channels of melodic voices (or six melodic plus a percussion mode). FM gave DOS games a far richer sound than the PC speaker's single square wave.

id Software's games leaned on this hardware directly. Commander Keen (1990) and Wolfenstein 3D (1992), with music by Bobby Prince, drove the OPL2 through id's own IMF (id Music Format) — a stream of timestamped OPL register writes, conceptually similar to VGM but specific to the OPL2/AdLib hardware and id's audio engine. Playing an IMF means replaying those FM register writes to an emulated OPL2 chip, reproducing the exact synthesized output PC gamers heard in the early 1990s. SoniqBoom plays IMF via OPL2 emulation.

Creative Technology's Sound Blaster (1989) extended the AdLib formula and seized the PC market. The original Sound Blaster 1.0 paired an AdLib-compatible YM3812 (OPL2) FM synthesizer with something AdLib lacked: an 8-bit digital sample channel (DAC/ADC) for playing and recording PCM audio at up to roughly 23 kHz, plus a game/MIDI port. Being AdLib-compatible meant it ran the entire existing FM music library, while the digital channel let games add sampled sound effects and speech.

Later models raised the bar — the Sound Blaster Pro added stereo and dual OPL2 (then OPL3), and the Sound Blaster 16 brought 16-bit CD-quality digital audio and the OPL3 (YM262) FM chip with up to 18 channels. 'Sound Blaster compatible' became the baseline every DOS game targeted through the 1990s.

This split — FM-synthesized music alongside digitally sampled effects — bridged the chiptune era and the later world of fully recorded, streamed game soundtracks. SoniqBoom's support for OPL-based formats reproduces the FM side of that history through chip emulation.

While the 2600 used the TIA, Atari's 8-bit home computers — the Atari 400 and 800 (1979) and their successors — used the far more capable POKEY (Pot Keyboard Integrated Circuit). POKEY provided four semi-independent audio channels, each 8-bit, with options to pair channels for 16-bit precision, plus distortion/polynomial-counter settings that produced a range of buzzy and noisy timbres. It also handled keyboard scanning and serial I/O. POKEY turned up in many arcade machines too — Centipede, Missile Command, Asteroids Deluxe, and Gauntlet among them — and even appeared on a few Atari 7800 cartridges for enhanced sound.

Atari 8-bit music is preserved in the SAP (Slight Atari Player) format, which packages the 6502 driver code and music data for re-execution on an emulated POKEY. The later 16-bit Atari ST (1985) used the Yamaha YM2149F — a variant of the AY-3-8910 PSG — and that machine became a hub of tracker-based music and the European demoscene. SoniqBoom plays SAP through POKEY emulation.

The chips set the limits; composers found the music inside them. On the Commodore 64, British composer Rob Hubbard (b. 1955) wrote scores for over 75 games between roughly 1985 and 1989 — including Monty on the Run, International Karate, Sanxion, and Commando — coaxing pseudo-polyphony and arpeggiated 'chords' out of the SID's three voices. His countryman Martin Galway (b. 1966) brought a melodic, sample-augmented style to titles like Wizball, Rambo: First Blood Part II, Arkanoid, and Times of Lore, and was an early pioneer of sampled sound on the C64.

In Japan, Koji Kondo (b. 1961) joined Nintendo in 1984 as its first dedicated composer and wrote the instantly recognizable themes for Super Mario Bros. (1985) and The Legend of Zelda — the Mario theme later became the first video-game music inducted into the U.S. National Recording Registry. Nobuo Uematsu (b. 1959) joined Square in 1986 and scored the Final Fantasy series across the NES, SNES, and beyond, earning a reputation as one of the form's most celebrated melodists.

Yuzo Koshiro (b. 1967) pushed game audio toward club and dance idioms, composing for The Revenge of Shinobi (1989), ActRaiser (1990, with a clever SNES sample-loading scheme), and the Streets of Rage / Bare Knuckle series (1991–1994), whose house- and techno-infused Genesis soundtracks were ahead of their time. These composers — and many peers — are why the formats in this addendum are worth preserving: the constraints were severe, and the artistry within them endures.