Converting DSM Tracker Modules to WAV: RIFF DSMF Multi-Channel Audio Synthesis

In the golden age of PC MS-DOS demoscene music, composers sought formats that went beyond the 4-channel hardware restrictions of the Commodore Amiga without sacrificing tracker precision. Finnish developer group DSIK (Digital Sound Interface Kit) created the Dynamic Studio Module (DSM) format, using a structured Resource Interchange File Format (RIFF) container.

Files saved in this format carry the .dsm extension.

Unlike older tracker formats that used fixed binary offset tables, DSM adopted a clean chunk-based architecture, paving the way for multi-channel tracker standards on PC sound hardware like Gravis UltraSound and Sound Blaster AWE32.

This technical guide dissects the RIFF DSMF chunk hierarchy, examines 16-channel voice mixing, and explains how convrtr's DSM to WAV converter synthesizes vintage tracker compositions into studio-quality 16-bit stereo WAV audio directly in your web browser.

The RIFF DSMF Container

Every DSM file begins with a standard 12-byte RIFF container header:

| Offset | Size (Bytes) | Field Name | Expected Value | | :--- | :--- | :--- | :--- | | 0x00 | 4 bytes | Chunk ID | "RIFF" | | 0x04 | 4 bytes (LE) | File Size | File length minus 8 bytes | | 0x08 | 4 bytes | Form Type | "DSMF" |

Following the root container, the file contains sequential data chunks padded to 2-byte word boundaries:

1. The SONG Chunk

The SONG chunk defines global playback metadata and sequence progression:

• Title: 32 ASCII characters describing the song name.
• Header Version & Flags: Version indicator (typically 1) and playback flags.
• Dimensions: numOrders, numSamples, numPatterns, and numChannels (supporting up to 16 simultaneous digital channels).
• Master Speed & Tempo: Default tick rate and BPM.
• Order Table: Array of pattern numbers to play sequentially.

2. The INST Chunk

Each instrument is declared in its own INST chunk:

• Sample Name: 32-character ASCII instrument label.
• Playback Tuning: 32-bit C2Spd playback frequency (typically 8363 Hz for middle C).
• Loop Points: 32-bit start and end sample offsets for sustained instruments.
• Volume: Native volume range (0 to 64).
• PCM Data: 8-bit PCM audio samples, encoded either as unsigned or two's-complement signed bytes.

3. The PATT Chunk

Pattern data is stored across sequential PATT chunks:

• Pattern Number: 16-bit integer index.
• Row Count: Number of rows per pattern (typically 64).
• Event Streams: Compact event byte streams indicating channel number, note (semitones 1 to 84), instrument index, and effect parameters.

Software Mixing & 16-Bit Stereo WAV Synthesis

convrtr's in-browser audio engine simulates the DSIK playback routine at 44.1 kHz:

1. Channel Polyphony: Up to 16 independent channel voices with configurable stereo separation (default 75%).
2. Linear Interpolation: Smooth fractional sample stepping to eliminate harsh aliasing during pitch shifting.
3. Loop Point Wrapping: Seamless repeat looping for pads and sustained instruments.
4. Soft Limiting: Automatic clamping to prevent digital clipping when all 16 voices trigger simultaneously.

All decoding and audio synthesis run in pure client-side TypeScript inside your browser, producing lossless 16-bit stereo WAV audio with zero cloud reliance.