Converting Impulse Tracker IT Modules to WAV: Architecture and Synthesis

In 1996, Australian programmer Jeffrey Lim released Impulse Tracker (.it) for MS-DOS. It quickly established itself as the zenith of tracker music production, marking the final evolutionary leap of the classic demoscene tracker quartet alongside Amiga ProTracker (.mod), Scream Tracker 3 (.s3m), and FastTracker II (.xm).

While earlier formats struggled with channel constraints, rigid Amiga clock periods, or primitive envelope controls, Impulse Tracker introduced up to 64 polyphonic audio channels, resonant low-pass digital filters, New Note Actions (NNA) for intelligent voice management, 16-bit compressed samples, and precision pitch calculation based on arbitrary C-5 sampling frequencies. Video game composers embraced IT to create iconic soundtracks for titles like Jazz Jackrabbit 2, Deus Ex, and Unreal Tournament.

This technical guide breaks down the forensic byte layout of Impulse Tracker files, explains how 64-channel bit-packed patterns and IMPS sample headers are decoded, and demonstrates how convrtr's Impulse Tracker to WAV engine synthesizes multi-channel tracker arrangements into standard 16-bit linear stereo PCM audio directly inside your browser.

The IMPM Header Specification

Every valid .it file begins with a 192-byte primary module header starting with the 4-byte ASCII signature IMPM (0x49 0x4D 0x50 0x4D):

| Offset (Bytes) | Field Name | Data Type | Forensic Description | | :--- | :--- | :--- | :--- | | 0x00 - 0x03 | Magic Signature | 4 ASCII chars | Constant "IMPM" (0x49 0x4D 0x50 0x4D) | | 0x04 - 0x1D | Song Name | 26 ASCII chars | Song title, null-padded (0x00) | | 0x1E - 0x1F | Row Highlights | 2 bytes | Pattern row highlight markers (minor, major) | | 0x20 - 0x21 | Order Count | 16-bit uint (LE) | Number of playback order entries (OrdNum) | | 0x22 - 0x23 | Instrument Count | 16-bit uint (LE) | Number of instrument definitions (InsNum) | | 0x24 - 0x25 | Sample Count | 16-bit uint (LE) | Total sample headers in module (SmpNum) | | 0x26 - 0x27 | Pattern Count | 16-bit uint (LE) | Total unique 64-row pattern chunks (PatNum) | | 0x28 - 0x29 | Created Version | 16-bit uint (LE) | Tracker version (0x0214 = IT 2.14) | | 0x2A - 0x2B | Compatible Version | 16-bit uint (LE) | Minimum compatible tracker version | | 0x2C - 0x2D | Flags | 16-bit bitfield | Bit 0: Stereo, Bit 2: Instruments, Bit 3: Linear Slides | | 0x2E - 0x2F | Special Flags | 16-bit bitfield | Bit 0: Song message attached, Bit 3: MIDI pitch bend | | 0x30 | Global Volume | 1 byte | Master song volume (0 to 128) | | 0x31 | Mix Volume | 1 byte | Mixing attenuation volume (0 to 128) | | 0x32 | Initial Speed | 1 byte | Ticks per pattern row (typically 6) | | 0x33 | Initial Tempo | 1 byte | Beats per minute clock rate (typically 125) | | 0x34 | Panning Separation | 1 byte | Stereo separation factor (0 to 128) | | 0x40 - 0x7F | Channel Pan Table | 64 bytes | Initial panning for channels 0–63 (0=left, 32=center, 64=right, 100=surround, 128=disabled) | | 0x80 - 0xBF | Channel Vol Table | 64 bytes | Initial volume for channels 0–63 (0 to 64) | | 0xC0 onwards | Order List | Byte array | Sequence array of order indices (255 = End, 254 = Skip) |

32-Bit Absolute Parapointers

Unlike Scream Tracker 3, which relied on 16-bit paragraph pointers multiplied by 16 (offset << 4), Impulse Tracker adopted clean 32-bit little-endian absolute file offsets for addressing data structures:

Following the OrdNum order table:

• InsNum * 4 bytes: 32-bit absolute offsets pointing to instrument blocks.
• SmpNum * 4 bytes: 32-bit absolute offsets pointing to IMPS sample headers.
• PatNum * 4 bytes: 32-bit absolute offsets pointing to compressed pattern blocks.

The IMPS Sample Header and C-5 Frequency

Each sample header begins with the 4-byte ASCII signature IMPS (0x49 0x4D 0x50 0x53) at its declared 32-bit file offset:

• DOS Filename: 12 ASCII characters storing the legacy 8.3 filename.
• Flags:
  • Bit 0: Sample data present.
  • Bit 1: 16-bit sample format (vs 8-bit).
  • Bit 2: Stereo sample data.
  • Bit 3: Compressed sample payload.
  • Bit 4: Loop enabled.
  • Bit 5: Sustain loop enabled.
  • Bit 6: Bidirectional (ping-pong) loop.
• Sample Name: 26-character ASCII sample label.
• Conversion Flags (Cvt): Bit 0 indicates signed PCM samples; Bit 2 indicates differential delta encoding.
• Length, LoopStart, LoopEnd: 32-bit integers specifying sample bounds.
• C5Speed: A critical 32-bit integer representing the exact sample playback frequency (in Hertz) when playing note C-5 (e.g. 8363 Hz, 22050 Hz, or 44100 Hz). Unlike Amiga modules where pitch was dictated by hardware clock dividers, C5Speed allows arbitrary sample rates to be tuned accurately.
• Sample Pointer: 32-bit absolute file offset indicating where the raw audio waveform data begins.

Pattern Packing and 64-Channel Matrix

Impulse Tracker patterns can contain up to 64 polyphonic channels. To eliminate redundant storage on empty rows, IT uses a dynamic channel packing protocol:

Each row consists of channel tokens ending with a null byte (0x00):

1. Read byte chvar:
   • If chvar == 0x00: End of row. Increment the row counter.
2. Decode channel number and mask update flag:
   • channelIndex = (chvar - 1) & 63 (Channels 0 to 63).
   • If chvar & 128: A new maskvar byte follows immediately. Otherwise, reuse the previous maskvar recorded for this channel.
3. Unpack channel data according to the active maskvar bits:
   • Bit 0 (0x01): Read Note byte (0 to 119 where 60 is C-5; 254 = Note Cut, 255 = Note Fade).
   • Bit 1 (0x02): Read Instrument byte (1-based sample index).
   • Bit 2 (0x04): Read Volume byte (0 to 64 = volume, 128 to 192 = panning).
   • Bit 3 (0x08): Read Command effect letter and CommandValue parameter byte.

In-Browser Multi-Channel Audio Synthesis

Rendering an .it module into universal 16-bit linear stereo PCM audio requires a multi-voice digital mixing pipeline:

1. Tempo and Ticks:
   • tickDuration = 2.5 / tempo (seconds).
   • samplesPerTick = sampleRate * tickDuration.
   • Each row advances after speed ticks (default 6).
2. Frequency Calculation:
   • For note n (where C-5 is 60) and sample rate sampleRate: frequency = C5Speed * (2 ** ((n - 60) / 12))
   • The playback step per output audio frame is: step = frequency / sampleRate
3. Voice Mixing and Loop Points:
   • For each active channel voice, sample position advances by step.
   • If a voice passes LoopEnd with looping enabled, it wraps cleanly back to LoopStart.
   • Voice amplitudes are panned across stereo output buffers according to channel pan settings.
4. Encoding to Standard WAV:
   • Left and right float channels are normalized and converted to signed 16-bit little-endian words, enveloped in a canonical RIFF WAV container.

100% Client-Side Retro Audio Preservation

Vintage demoscene modules and game audio archives should be playable anywhere without installing obsolete MS-DOS emulators or third-party executable trackers. Because convrtr executes the entire Impulse Tracker decoding and synthesis pipeline client-side in pure TypeScript, your chiptune files convert instantly in your browser tab with complete privacy.