Reference · Engine

Music and sound effects

partialfrom#26

The game's music is libsnd sequences (SEP) played through VAB banks on the SPU. Sound effects are key-ons on the same voices. Speech and the cinematics' sound are XA (cinematics).

Song sets (0x8006e96c)

A song set is a small struct the opener loads and keeps at 0x800c4338:

• The menu's set is at 0x800b8030: music 109/110 (MENU.VB/VH), effects 151/152, SEP 230 (RUGRATS.SEP), 5 songs.
• 0x800b8180: music 133/134, effects 151/152, SEP 223 (INTRO.SEP), 1 song.
• A level's set is picked by database (0x80018598) and opened by the level setup (0x80017c0c).

Playing a song (MidiPlaySong, 0x8006eac4)

MidiPlaySong(song, loops) stops the song playing (0x8006ebb4), sets the SEP song's volume (0x800a91a0) and plays it (0x800a84e0, mode 1; loops 0 is forever).

• The volume (0x8006e74c) is the settings' music volume (s32 0x800c5684, 90 at boot). Some levels raise it for one song: game mode 2 song 2 by 30, mode 5 song 2 by 20, mode 18 by 10. It is clamped to 127.
• The title's loop (0x8003d884) plays song 0.

A room's song (0x8006efcc)

RoomChange (0x80077a54) calls 0x8006efcc(_, song, 30) for room flag 128 with +0x64 of 1-4. The set's record keeps the song playing (+0xa) and one fading out (+0xb), with a fade timer +0x10:

• The song is +0xa or +0xb: nothing.
• No song playing (+0xa -1, or it has ended, 0x800a5e88): MidiPlaySong at once.
• A fade under way (+0x10 not 0): the fading song stops (0x800a8d00), +0xb = +0xa, +0xa = the song, played and set to volume 0 (0x800a91a0), then SsSeqSetCrescendo (0x800a59f4) to the music volume over 30.
• Else: +0xb = +0xa, +0x10 = 30 * 50, SsSeqSetDecrescendo (0x800a5bbc) on +0xb from its volume over 30, and the new song at 0 with the crescendo as above. So the two overlap: a crossfade.

0x8006ef0c (from the level loop 0x80018ad0) and 0x8006ef64 (0x8007424c) fade the playing song out and back in over 60.

Effects (0x8006f26c)

Effect n (under 180) reads 3 bytes at 0x800b84fc + 3n: program, tone, note. It keys that tone on in the effects bank (0x800a99a8, SsUtKeyOn) at the effects volume (s16 0x800c5688, 100 at boot) on both sides. The voice is taken in turn from 17 to 23 (0x800c434c).

A sound at a place (0x8006f644) is heard from the camera: dx^2 / 16384

• dy^2 / 4096 + dz^2 / 16384 through the falloff 0x800b8008 (full within 128, none past 16384, 127 at most). It is panned by the heading from the eye to it (0x80021590: asin of x / length where |x| < |z|, else acos of z / length) less the heading the camera faces. That facing is the heading toward the eye plus the camera matrix's third column: 0x80010b54 copies the matrix transposed. With the third row the pan comes out mirrored, as the port's did until now.

Sound sources (0x8006fe04, 0x8006f88c)

A looping source is a 40-byte record at 0x800e3218, 0x800c434b of them: +0x00 its sound, +0x04 its voice, +0x08 on, +0x0c its place, +0x1c and +0x20 the distances it is full within and gone past, +0x24 its volume. They are started by:

• the loader for an LP kind 3 record (0x800454dc at 0x80045b24: the sound the template word's high half, near and far the rotation's first two words, the volume the third's low half), in LP order;
• the object set-up for some templates (91 and 614 in level 3, 227 in 16, 657 in 23, 1191, 920 and 535 in 28, 789 in 31, and 1178 with its sound rand % 3);
• level 4's start (one), and level 10's start, which empties the list first and adds three.

A source's sound is the level's table 0x800b847c[level] (s16 program, tone, note) at its sound. Each frame (0x8006f88c, after the draw) each source is heard as a placed sound is, but through its own falloff (0x8006fd68 with its record; mono uses 0x800b8008):

• heard and on, it is keyed on a free voice (0x800c4be8: 7 at a level's start, so 15, 16 and 17; 0x800a99a8 with the voice) or turned (SsUtSetVVol, 0x800b0788);
• not heard, it is keyed off (0x800a9cf0) and its voice freed.

Leaving a level (0x80094bfc) silences them (0x8006f828). The port's world::game::sources does this, through retro_rt's Spu::key_on_voice, set_voice_volume and key_off. npc_check SRCDBG=1 compares the turns each frame: the same where the cameras are.

SEP

See rrt::kit::libsnd::sep:

• "pQES", then u16 version.
• Then the songs back to back, each with: u16 id, u16 resolution, u24 tempo, u16 rhythm, u32 length (all big-endian), then MIDI events with running status, ending FF 2F 00.
• libsnd marks loops with NRPN 99 = 20 (start, data entry the count) and 99 = 30 (end).

Port

• The banks, songs, voices, envelopes, sequencer, slot volumes and fades are retro_rt's rrt::kit::libsnd (rrt-kit.md, moved there from this port): 24 voices, the SPU's ADSR, Gaussian interpolation, and libsnd's sequencer reading the SEP events.
• rugrats::audio::sound keeps the game's side: the song sets and their files, the effect table, effects on voices 17-23, and the room switch (0x8006efcc) over two of libsnd's song slots with the songs' ids.
• Voice volumes multiply velocity, tone, program, channel (CC 7), expression (CC 11) and bank master volumes, each out of 127; a slot's volume is applied as the voices mix, so SsSeqSetVol and the fades change notes already sounding.
• libsnd's ticks are 60 a second: the sound init (0x8006e7e8) sets SsSetTickMode(0x1000), SS_NOTICK, and the VSync callback (0x800192b4) calls SsSeqCalledTbyT (0x800a48b8) every vblank. The switch's 30 is half a second.
• The sequencer has 14 voices (SsSetReservedVoice(14), 0x800ad588); effects key on 17-23 directly.
• No tone in the game's 1779 has vibrato or portamento, and one has the reverb bit, so libsnd's unmodelled vibrato, portamento and reverb do not change what Rugrats plays.
• Reverb is silent: the SPU init (0x800add98, from the sound init) clears SPUCNT and the reverb volumes. The only paths to SpuSetReverb (0x800af2f8) and the reverb depth and type (0x800af4d8) are libsnd's SEQ handlers that SsSepOpen (0x800a5ec8) installs: CC 91, the NRPN data entries. The game's 33 SEPs send no CC 91; besides the loop marks (NRPN 20, 30) their only NRPN is 40, always with data 0.
• Pans add up around 64.
• Mix puts the SPU's voices over the XA soundtrack for the audio device.
• song_wav renders a song offline. Song 0 at 90: peak 21990, RMS 2859.

Unknown

• libsnd's exact volume and pan law and its pitch formula (tone shift taken as 1/128 semitone).
• Reverb (the SPU's reverb is not modelled).
• Vibrato and portamento from the tone attributes.

Around cinematics

A cinematic's table byte +7 (0x800b8bd0 + 16 n, read by 0x80074b80) decides the song:

• PostCine in a level (0x80018ad0): with 32, the song fades down over 60 ticks (0x8006ef0c, SsSeqSetDecrescendo). The screen fades over 127 dt unless 16 is set. The level is drawn as it stopped (0x8001886c, no update) under a full-screen POLY_F4 with draw mode 0x640 (subtract) whose grey 0x800c3f2c steps by 256 / (127 / dt) to 255 (0x8001cf2c, from the flip 0x8001d00c).
• At the start (0x8007424c), the case depends on the byte:
  • with neither 8 nor 32, the song stops (0x8006ee14: SsSeqStop on both slots, the effect voices keyed off);
  • with 32, it pauses (0x8006ec40);
  • with 8 alone, it plays on.
• At the end, with 32: SsSeqReplay (0x8006ed30), then SsSeqSetVol 0 and SsSeqSetCrescendo to the music setting over 60 ticks (0x8006ef64).

A stopped song stays stopped. The song saved at +9 has no reader. The next room with a song (flag 128) starts one again.