The Screen Nobody Explained
Open a tracker — ProTracker on an Amiga 500, FastTracker 2 on a 486 PC — and the interface gives you almost nothing by way of welcome. A grid fills the screen: rows running top to bottom represent time steps, columns represent simultaneous voices. Early ProTracker gave you four channels; FastTracker 2 extended that to 32. Each cell in the grid holds a note value, a sample slot number, an effect command, and an effect parameter. That is the entire compositional surface. There is no piano roll, no waveform lane, no automation curve. If you want the hi-hat to open on the upbeat, you type it.
The sample slots are loaded first. A tracker file — the MOD format on Amiga, XM on PC — embeds the samples directly inside the module, which is why these files were self-contained and why they spread across bulletin board systems so efficiently. A producer in 1992 might load fifteen samples: a kick sampled from a Roland TR-909 via line-in, a snare, a few pitched percussion hits, a bass waveform held as a single cycle so it could be looped and transposed across the keyboard range. Memory on an unexpanded Amiga 500 was 512 kilobytes of chip RAM available for audio. Every second of sample time had to earn its place.
Building the Pattern, Column by Column
A pattern in ProTracker is sixty-four rows by default, playing at a tempo set in beats per minute via a global tempo command — the F effect in hexadecimal notation. Working at 125 BPM with a default speed of six ticks per row, sixty-four rows resolves to approximately four bars in 4/4. The producer sequences a kick on rows 0, 16, 32 and 48; a snare on 8 and 40. The hi-hat goes in on every eighth row in channel three, with an E9 effect command for a retrigger if the producer wants a rolled figure. Channel four carries the bassline: each cell specifies the sample slot, the note (C-3, D#3, F-3), and optionally a volume command or portamento.
Early ProTracker gave you four channels; FastTracker 2 extended that to 32.
Effect columns are where a tracker gains its expressiveness. The A command does a volume slide; 3 executes portamento-to-note, bending pitch between successive cells. A bass note with a portamento command and a volume fade produces something close to the glide behaviour of a monosynth, without requiring any hardware beyond the computer. The Roland TB-303's characteristic slide was approximated this way across thousands of early-1990s tracker compositions.
Once a pattern is programmed, you build a song by arranging patterns in sequence through the order list — a numbered playlist sitting above the pattern grid. Pattern 0 might be the intro: kick and bass only. Pattern 1 adds the hi-hat and a mid-range sample. Pattern 2 introduces the lead. The order list strings them: 0, 0, 1, 1, 2, 2, 1, 3. This is the composition. There is no timeline in the modern sense, no measure of absolute position in seconds; you navigate by pattern number and row.
What the Format Demanded
Tracking forced economy. With four to eight channels, every voice competed for space. A producer couldn't layer six variations of a snare; they chose one. Sample loops had to be set precisely — a missed loop point on a sustained bass note produced an audible click on every cycle. The hexadecimal notation for effect commands (digits 0 through F in each position) meant that a producer learned a small vocabulary of codes and applied them by muscle memory. There was no undo history in early trackers. Overwrite a cell and it was gone.
The resulting files were tiny — a complete two-minute track in under 200 kilobytes was routine — and entirely portable. This is what let tracker music circulate on floppy disks through the demoscene and into the early online networks that would eventually connect bedroom producers to each other, well before any commercial platform thought to serve them.