Chiptune is the sound of constraint. Every bleep, every impossibly fast arpeggio, every buzzy bassline came out of a chip with a handful of voices and almost nothing to tweak — and that shortage is exactly why the music sounds the way it does. Recreate chiptune well and you are not loading a “retro” preset; you are rebuilding a specific, knowable instrument and playing by its rules. This guide starts at the chip that made the sound and works outward: the pulse wave and its single timbral control, the trick that fakes chords out of one voice, the bass and drums hiding in two more channels, and how to assemble all of it in a modern DAW. Do it right and a single square wave can carry a hook that outlives the console it came from.
Sound is generated live in your browser — nothing is recorded or uploaded. Press Play once to allow audio.
The chip that made the sound
The Nintendo Entertainment System generated its audio on the Ricoh 2A03 (the RP2A07 in PAL regions), a chip that also housed the console’s CPU. Its sound unit gives you exactly five voices and no more: two pulse channels, one triangle, one noise, and a DMC channel for short digital samples. There is no filter, no reverb, no effects rack — just five oscillators and a volume level for most of them. (The DMC is the outlier: it reconstructs 1-bit delta-encoded samples rather than generating a waveform, and because it shares the chip with the CPU, heavy sample playback could even nudge a game’s timing.)
Everything that reads as “8-bit” follows from that inventory. Because there is no fifth or sixth melodic voice to reach for, a composer has to make each of the five carry a defined job, and has to invent ways to imply more than the hardware plays. That single fact — a fixed, tiny palette — is the whole reason the techniques in this guide exist, and it is the last time this article will belabour it. From here on we take the voices one at a time and learn what each is for. If a huge, thick modern lead is the opposite pole of this sound, it is worth hearing the contrast — the supersaw stacks dozens of detuned oscillators to get its width, where a chip lead has to earn its size from a single square wave.
One detail matters before we go further: the chip ran at different clock speeds in NTSC and PAL territories, so the same music plays a touch lower and slower on a European console. You do not need to reproduce that quirk, but it explains why recordings of the same game can sit in slightly different keys.
These constraints are easiest to hear in the scores that defined the machine. The Super Mario Bros. theme leans on exactly the division of labour that follows below — a pulse lead over an arpeggiated pulse harmony, a walking triangle bass and noise-channel drums — while Mega Man 2 shows how far a single pulse lead and fast arpeggios can be pushed melodically. Some composers treated the ceiling as a dare: Tim Follin’s NES work is known for wringing rapid arpeggios and near-polyphony out of the same three-voice budget everyone else had. You need not copy any of them, but keeping a reference track open while you build trains your ear for how much these four voices can really carry.
The pulse wave and its one knob: duty cycle
The two pulse channels are the melodic workhorses, and each has a single control over its tone: the duty cycle. A pulse (or square) wave spends part of each cycle “high” and the rest “low,” and the duty cycle is simply the ratio between the two. The 2A03 offers four settings — 12.5%, 25%, 50% and 75% — but 25% and 75% produce an identical tone, so in practice you have three distinct timbres per channel.
Each pulse channel also carries a small piece of dedicated hardware called the sweep unit, which slides the pitch up or down automatically once a note is sounding. It is a narrow feature — it only bends the frequency at a set rate — but it is behind a surprising amount of the console’s vocabulary: the rising whoosh of a power-up, the quick blip of a coin, the falling pitch of something dropping. Because the slide runs in hardware it costs the composer nothing to trigger, so it turns up constantly in sound effects and the odd melodic flourish. Recreating those moments means reaching for a fast pitch envelope or a portamento glide rather than drawing the bend in by hand. One detail worth matching: the sweep recomputes the pitch on that same frame clock, so a fast hardware slide has a faint stair-stepped, zippering quality rather than a perfectly smooth glide — a stepped or quantised portamento gets closer than a clean one.
Those three shapes are the entire timbral palette of a pulse voice, and each has a character worth knowing. The 12.5% setting is thin and nasal, almost sawtooth-like — bright and cutting, good for a lead that has to sit over everything. 25% is fuller and reedier, the classic “chip lead” voice. The 50% square is the hollow, pure tone most people picture when they think of an old game — rounder and woodier than the narrower pulses. On top of duty, each pulse channel has a 4-bit volume, so sixteen loudness steps for envelopes and accents. Those steps are more useful than they sound: a fast volume ramp gives a plucked, staccato attack, and a repeating volume shimmer fakes a tremolo the chip has no dedicated effect for.
To rebuild this in a modern setup, reach for any oscillator that offers a pulse wave with adjustable width — pulse-width is standard on nearly every soft synth, and a quick pass through Serum or Vital gets you all three timbres in seconds. If you are shopping for something purpose-built, the round-up of the best synth plugins flags several with dedicated chip or pulse oscillators. The one rule: pick a duty and commit to it per part — switching duty mid-note is a movement trick, covered later, not a set-and-forget tone.
Faking a chord: the fast arpeggio
Here is the problem a chip composer faces the moment they want harmony: a single pulse channel plays one note at a time, and there are only two of them. Spend both on a three-note chord and you have nothing left for melody, bass, or movement. The solution — the single most recognisable move in the genre — is the fast arpeggio: instead of holding a chord, one voice cycles through its notes so quickly that the ear stops hearing separate pitches and fuses them into a single harmony.
How fast is fast enough is fixed by the console itself. The music engine updates once per video frame — sixty times a second on an NTSC machine — so swapping the note every frame cycles a three-tone chord at roughly twenty times a second, comfortably past the point where the ear resolves the pitches. The frame count becomes the expressive control: a one-frame arpeggio is so quick it reads as a buzzy, hollow timbre rather than a chord, while stretching each note to two or three frames slows the cycle until you begin to hear it strum. That is the source of the restless shimmer a smoothly held pad never has. In a modern DAW you recreate it not with a chord track but with a rapid one-note-per-step sequence — a note every 1/32 or faster — locked to that frame-sized grid.
The speed is the trick. A chiptune arpeggio runs at roughly a 1/16 to 1/64 rate, often changing note every single frame — about sixty times a second — which is far faster than a musical arpeggio and closer to audio-rate. That gives it a distinctive, slightly buzzy shimmer, because no two notes ever ring together the way they would in a real strummed chord; the harmony is implied, never actually sounded at once. The technique is old enough to have a lineage — Martin Galway is widely credited with bringing it into game music — and it is worth understanding as its own instrument, not just a fast version of a normal arpeggio.
Two habits make it musical. First, keep the note set tight: a root–third–fifth triad is the default, but dropping in a seventh or a suspension buys a lot of colour, and a little music theory goes a long way when every note counts. Second, let the arp drop out once the harmony is established — the listener’s ear holds the chord for a bar or two, so you can free the voice for a melodic run and bring the arp back when the harmony shifts. In a DAW the quickest way in is an arpeggiator locked to the chord tones at a 1/32 rate or faster; purists program the note-per-frame cycle by hand in the piano roll for full control over which tone lands on the beat. The same fast-arp shimmer drives a lot of modern dance music too; it is the backbone of plenty of trance leads, where the technique long outlived the hardware that forced it.
The triangle: bass with no volume knob
The third voice is the triangle channel, and it is the odd one out. Unlike the pulse channels, it has no duty control and — crucially — no volume control: it is either on at full level or off. It also plays an octave lower than the pulses. Those two traits make it almost useless for melody and perfect for one job: bass.
Historically this was a deliberate design choice. Where competing chips of the era spent all their oscillators on square waves, the NES traded one away for a triangle, which cost it a third melodic voice and any volume shaping on that channel, but bought a far richer low end than a square bass could give. The triangle’s waveform is also faintly stepped rather than perfectly smooth, which adds a little edge that helps it cut through on small speakers.
To recreate it, use a triangle oscillator an octave below your lead, and resist the urge to automate its volume — the flatness is part of the sound. A touch of bit reduction gets you the stepped edge. The triangle often pulls double duty, too: a very short, low triangle note — almost a click — stands in for a kick drum, which is why on many games the bass and the kick share this one voice and never quite sound at the same instant. Because the bass is doing so much structural work with so little tonal variation, how you sit it in the mix matters more than usual; the fundamentals of mixing bass apply directly. It is a useful contrast to study against a modern tuned-drum bass, too: where the chip triangle is a fixed-level sustain, the 808 is a pitched, enveloped boom — opposite answers to the same question of “what carries the low end.”
The noise channel: a drum kit from static
The fourth voice makes no pitched tone at all. The noise channel runs a pseudo-random bit generator — a shift register flipping between high and low — to produce static, and it is how nearly every NES drum was made. It offers sixteen preset frequencies and its own volume envelope, which is all you need to carve percussion out of noise. Under the hood it is a fifteen-bit shift register stepping through a 32,767-value pseudo-random sequence; a second mode shortens that to a buzzy, almost tonal loop — the metallic laser-zap timbre you have heard a thousand times.
The move is entirely in the envelope. A fast attack and an instant decay give you a hi-hat tick; a slightly longer decay with a lower noise pitch becomes a snare; a very short burst can stand in for a kick, though many games leaned on the triangle or the DMC for the low thud instead. There is also a second mode that swaps the random pattern for a shorter, more tonal, buzzy noise — useful for metallic or robotic effects. The later trick was the DMC sample channel, which played tiny, low-bitrate recordings; by the end of the console’s life, composers were streaming real drum hits through it, which is why later games have punchier kits than early ones.
Recreating this is simple: feed a white-noise source through a tight amplitude envelope, and shape the decay per drum. Filtering the noise a little sets the pitch of each hit. Program the pattern with the same instincts you would bring to any beat — the placement and dynamics matter more than the source, and a well-timed noise snare will always beat a flat one.
Four voices, four jobs
Put the pieces together and the discipline becomes clear: you are arranging an entire track inside four melodic voices, and every one has to earn its place. The standard division of labour is two pulse channels for lead and harmony, the triangle for bass, and the noise for percussion — the same map any classic game score follows.
The map leaves out a fifth claimant on those voices: the sound effects. A game’s jumps, coins and lasers came out of the same channels as the music, usually by commandeering a pulse voice for a fraction of a second — so the harmony or a fill would briefly drop out the instant the player acted. The better composers arranged for it, keeping the load-bearing line on the triangle or scoring a channel lightly enough that a stolen note went unmissed. Recreating a specific track then forces a choice: honour that ducking, or write the “complete” version the console could never sound all at once. Most modern covers quietly restore the missing notes, which is one reason a remake can sound fuller than the cartridge ever did.
The consequence is that every note is justified. There is no pad to hide behind, no stack of layers to thicken a weak idea; if a pulse channel is playing a countermelody, it is not playing an arpeggio, so you choose. That forces a kind of writing that is unusually deliberate — you are constantly deciding what each of four hands should do at every moment, and the arrangement gets its energy from those trade-offs. A classic move is call-and-response between the two pulse channels — the lead states a phrase on one, the other answers in the gaps — so the pair reads as a conversation rather than two parts crowding the same space. Want a big chorus? Split the two pulses into a lead and a fast arp for implied chords, drive the triangle harder, and let the noise carry momentum. Need a breakdown? Drop to a single pulse and the triangle and let the space do the work.
This voice-budget thinking is exactly why chip sounds keep resurfacing in modern genres built on bold, exposed parts — the chiptune lead is a staple of hyperpop, where a single bright pulse cutting over a track is a feature, not a limitation. Learn to voice a full song in four parts and every other arrangement you write gets tighter.
Making it move
A static chip patch sounds dead; the life is in the modulation, and the classic tricks are all ways of squeezing motion out of controls that were never meant for it. Four are worth mastering.
Duty sweeps. Change a pulse channel’s duty cycle while a note is sounding and the timbre morphs mid-pitch — a thin 12.5% snapping to a full 50% is a whole expressive gesture from one parameter. Vibrato comes from a small, fast pitch wobble; a delayed vibrato that fades in over a held note is the signature of a “singing” chip lead. Pitch bends and slides — sliding into a note rather than jumping to it — add swagger and are cheap on a chip, since pitch was one of the few things you could ramp smoothly.
The fourth is the cleverest: a pseudo-echo. With no delay effect available, composers faked one by re-triggering a note a beat later at lower volume, sometimes on the second pulse channel, so it read as an echo trailing the original. It is the same instinct behind a lot of retro lead design — the movement and glide that make a Reese or a synthwave lead feel alive come from automating exactly these kinds of small, continuous changes. There is one width trick worth adding to the list: play the same line on both pulse channels a few cents apart, and the slight detune thickens a lone square into something almost lush — the chip’s closest thing to a chorus. On a chip you just have fewer knobs to do it with, which makes each move count more.
One trick sits outside that list because it fakes an effect the hardware does not physically have: echo. There is no reverb or delay anywhere in the 2A03, so composers wrote the repeat by hand — a second, quieter voice trailing the melody a beat behind, or the lead itself dropping faint delayed notes into the gaps between phrases. The repeat is pushed down in volume, sometimes an octave, so it registers as depth rather than a doubled part. It costs a whole voice, which is why you mostly hear it in sparse sections where a channel is briefly free, and why it always feels deliberate. Rebuild it the same way rather than with a plug-in: duplicate the lead, shift the copy back a sixteenth or an eighth, and drop its level.
Beyond the 2A03: the expansion chips
Not every “NES” track you have heard came from five voices. On the Japanese Famicom, a cartridge could carry its own extra sound hardware and add channels to the console’s own — which is why some scores sound impossibly rich for the system. Recreating those tracks means knowing which chip was in the cart.
The most famous is Konami’s VRC6, used in the Japanese Castlevania III. It adds two more pulse channels — with eight duty settings each, far more timbral range than the 2A03’s three — plus a sawtooth channel, the one basic waveform the stock chip cannot make. Nintendo’s own MMC5 adds two further pulse channels; the Famicom Disk System contributes a wavetable voice with a user-drawable 64-step waveform; and chips like Namco’s 163 add still more wavetable channels.
Two things follow for a modern recreation. First, if a reference track has a sawtooth lead or more than four simultaneous melodic parts, it almost certainly used an expansion chip — do not fight to force it into the base five voices. Second, Western NES releases mostly could not use these add-ons, so if you are chasing a specifically American-console sound, hold to the plain 2A03 palette. Match the chip to the track and the arrangement limits fall into place on their own.
Rebuilding it in a modern DAW
You do not need a console or a tracker to make this. The whole instrument reduces to four parts you can build in any DAW, and the only real pitfall is aliasing.
Start with a synth that offers a pulse wave with variable width for the lead and harmony — stock synths in Ableton, Logic, FL and Bitwig all qualify, and any electronic-oriented soft synth will do. Set one to a 25% pulse for the lead, a second to 12.5% or 50% for contrast, and use your DAW’s arpeggiator (or draw the notes by hand) to get the fast triad cycling. Add a triangle oscillator an octave down for bass, and a white-noise source with a tight envelope for drums. That is the entire chain. If you want dedicated tools, purpose-built chip plugins exist — but for a first build, stock instruments are plenty, and any of the DAWs covered in the beginner DAW guide ships everything required. Concretely: in Ableton reach for Operator or Analog set to a pulse; in Logic, Retro Synth or the ES2; in FL Studio, 3x Osc or Sytrus; in Bitwig, Polysynth — every one does pulse-width straight out of the box.
The one technical trap is aliasing. A naive digital square wave generates harmonics above the audible range that fold back down into it as a harsh, metallic ring — not the clean buzz of the real chip. Use a band-limited pulse oscillator (most modern synths are band-limited by default) or gently roll off the extreme highs, and the harshness disappears. Once the parts are down, treat it like any other track: balance the four voices, and give it a light master to glue it — chip music benefits from loudness and clarity as much as anything else.
Keeping it authentic
Most failed chiptune fails the same handful of ways, and every one comes back to breaking a rule the hardware enforced for free.
The commonest mistake is too much polyphony. The instant you stack five or six notes into a lush chord, the illusion collapses — the real chip could never do it, and the ear knows. Hold yourself to the four-voice budget even in a plug-in that would happily give you sixty. Close behind is using the wrong waveform: a sawtooth or a soft analog wave reads as generic synth, not chip — the pulse wave’s specific hollow buzz is non-negotiable. Then there is flat duty (one setting for the whole track, so nothing has its own voice), forgetting the triangle bass (a chip tune with a modern sub bass sounds instantly wrong), and reverb — the 2A03 had none, and drenching a lead in hall reverb is the fastest way to make it sound like a synth pretending to be retro rather than the real article. If you want a sense of space, use the authentic substitute instead — a single quiet re-triggered note as a slap echo, exactly the pseudo-echo trick, which reads as period-correct where a reverb tail never will.
The through-line is restraint: the sound is defined as much by what the hardware could not do as by what it could. That is the same discipline behind any faithful sound recreation — getting the Daft Punk robot voice right, for instance, is likewise a matter of respecting the exact process rather than approximating the vibe. Honour the limits and the character takes care of itself.
Build the skill: three drills
Reading about the voices is one thing; feeling the budget under your hands is another. These three drills move you from “I made a chip tune” to “I know exactly which voice is carrying the track.”
1 · Two voices, eight bars (beginner). Using only a 25% pulse lead and a triangle bass an octave below, write eight bars of melody and bassline — no drums, no harmony. The goal is to make two voices feel complete. If it sounds empty, your melody and bass are not answering each other; fix that before adding anything.
2 · Add harmony and drums (intermediate). Bring in the second pulse channel as a fast arpeggio for chords, and a noise channel for hats and a snare. Now change the lead’s duty cycle at least once mid-phrase and hear the timbre shift. You are now running the full four-voice map — make sure no voice is idle for no reason.
3 · Movement and a budget (advanced). Take the intermediate piece and add a delayed vibrato on a held lead note, a pitch slide into one phrase, and a pseudo-echo on the second pulse. Then impose the rule that no more than four notes ever sound at once — if you break it, something has to give. That constraint is the whole craft in miniature.