Cross-PlatformAudioBeginner

Four Ways to Make a Sound

The tune driver is the same on every machine — but what it talks to isn't. Paula streams a sample, the APU synthesises fixed voices, the SID shapes a rich one, and the Spectrum toggles a single bit. Four answers to 'how does a computer make a sound?', and each one shaped the music.

audiosound-chipssamplingsynthesiscomparison

Overview

The tune sequencer is nearly identical on every machine: a table of notes, a tick that serves them. This pattern is its mirror image — the part that is never the same. Ask “how does this computer turn a number into a sound?” and the four platforms give four fundamentally different answers, and those answers are why an Amiga tune and a Spectrum tune don’t just sound different, they sound like different kinds of thing.

Four models, cheapest hardware to richest:

  • Spectrum — a single bit. There is no sound chip. The CPU flips one speaker line on and off, and the rate of flipping is the pitch. You generate the waveform yourself, edge by edge.
  • NES — fixed synthesis. The APU has five voices with shapes baked in silicon: two pulse waves, a triangle, noise, a sample channel. You don’t make a waveform; you pick one and set its pitch and volume.
  • C64 — subtractive synthesis. The SID gives each of three voices a choice of rich oscillators, an ADSR envelope, and a real analogue filter. You start with a bright source and carve it into a sound.
  • Amiga — sample playback. Paula reads 8-bit samples straight from memory by DMA. You supply the actual recorded waveform; the chip just streams four of them. This is the only one where the sound can be anything.

The same note, four ways

Each snippet below is the minimal “make one sound” for its machine — the first pattern in each platform’s audio arc, boiled to its essence.

Spectrum — the CPU is the oscillator (Sound Beep):

loop:
    ld   a,$10
    out  ($fe),a       ; speaker high — bit 4 of port $FE
    call delay         ; the delay length IS the pitch
    xor  a
    out  ($fe),a       ; speaker low
    call delay
    djnz loop          ; every edge hand-made by the CPU

NES — pick a voice the chip already generates (A Square Wave):

    lda #$01
    sta $4015          ; enable pulse 1
    lda #%10111111
    sta $4000          ; the chip makes the square; you set duty + volume
    lda #$fd
    sta $4002          ; 11-bit period = pitch
    lda #$00
    sta $4003

C64 — shape a rich source (SID Note Trigger):

    lda #$67
    sta $d400
    lda #$11
    sta $d401          ; frequency
    lda #$f9
    sta $d406          ; ADSR — attack/decay/sustain/release
    lda #$41
    sta $d404          ; sawtooth waveform + gate on

Amiga — hand the chip a waveform (Playing a Sample):

    move.l  #sample,AUD0LC(a5)   ; YOU provide the waveform...
    move.w  #len,AUD0LEN(a5)
    move.w  #428,AUD0PER(a5)     ; ...Paula streams it by DMA
    move.w  #$8201,DMACON(a5)

The comparison

Machine Sound model You provide The hardware provides
Spectrum (beeper) 1-bit, CPU-timed every single edge nothing — it’s a wire
Spectrum 128K (AY) PSG period, volume, envelope shape 3 tone voices + noise
NES (APU) fixed-function synthesis which voice, pitch, volume 2 pulse, triangle, noise, DMC
C64 (SID) subtractive synthesis waveform, ADSR, filter cutoff rich oscillators + analogue filter
Amiga (Paula) 8-bit PCM playback the recorded waveform itself DMA streaming + mixing, 4 channels

The ladder runs from “the CPU does everything and the hardware is a single wire” to “the CPU does almost nothing and the hardware plays whatever you recorded”. Every step up hands more of the work to silicon and gives the programmer a richer starting point.

Why the model shaped the music

The hardware didn’t just constrain the music — it characterised it, and each scene grew around what its chip did best:

  • The Spectrum’s clicky ingenuity. With no chip, every Spectrum sound is a feat of CPU timing, which is exactly why its multi-channel beeper engines are so admired — musicians coaxed chords out of a component that could only be on or off.
  • The NES’s bright, instantly-recognisable timbre. Fixed pulse-and-triangle voices give a clean, hard-edged sound with a narrow palette — the reason a few bars of NES music are unmistakable.
  • The SID as an instrument in its own right. Oscillators plus envelopes plus a filter make the SID a genuine synthesiser, not just a tone generator; sweep the filter and it sings. It’s why people still compose SID music on hardware decades later.
  • The Amiga and the sample. Because Paula plays anything, Amiga music used real recorded instruments — and the four-channel MOD it enabled became the bridge from chiptune to the sampled, tracker-driven sound of the demoscene and early PC games.

Same job — play a tune — four philosophies of how a machine makes a noise, and four musical cultures that grew from them.

Patterns: Tune Sequencer — the driver that’s the same everywhere · foundational audio: Sound Beep (Spectrum) · A Square Wave (NES) · SID Note Trigger (C64) · Playing a Sample (Amiga)

Vault: Paula | APU | The SID Chip | ULA | AY-3-8912