Entry 327

"The jukebox clears the reverb; ccSnd +98 and +0x110 settled"

2026-10-01audiodecomp

Filescrates/piney-audio/src/driver.rscrates/piney-audio/src/lib.rscrates/piney-audio/tests/driver.rscrates/piney-audio/tests/driver_fixture.txtcrates/piney-audio/tests/stream.rstools/sound_ee.pydocs/engine/sound.md

Four sound questions were left open: 56's two, 184's +0x110, and 42's "who asks for sdCommand cases 4 and 5". All addresses are INF SLUS_202.67.

sdCommand 4 and 5: ccSndChangeData

sdCommand (0x001816f0) walks ccSound.sdRemote[8] (+0x1c, DWARF). A scan of the 459 relocations to ccSnd in main and the four overlays finds four writers:

• ccSetMainVol: [0];
• ccSetOutputMode: [2];
• ccSndChangeData: [3] = 4 at 0x00183628, after the old music is stopped and before the load, which every path reaches, the desktop's first load (old -1) included;
• ccSndChangeData: [4] = 5 at 0x00183888, once the data is in.

Case 4 sets the effect mode to 0x100 (off, area cleared) on both cores (sceSdRemote 0x8130), the effect return volume to 0 (0x8010, params 0x0b80/0x0c80) and disables effects (0x8070, 2 and 3). Case 5 sets mode 0x105 (hall, area cleared), depth 0x7fff and delay 127, return volume 0x3fff, and enables effects.

The game's own run shows when they happen. tools/sound_ee.py now logs sceSdRemote(0x8130) as reverb CORE MODE. Its desktop step now runs the sound task in ccSndChangeData's breaths, as the game's thread would. Both 4 and 5 then run in the one sdCommand after the load: load X; reverb 0 100; reverb 1 100; reverb 0 105; reverb 1 105; seq ..., in all 51 desktop rows and the 10 jukebox changes. The volume scenarios no longer carry a stray pair. The effect is that the reverb starts empty with the new music: the old music's tail does not ring on.

The port's driver keeps sd_reverb (sdRemote[3] and [4]). load sets it, and task, at sdCommand's place, sends Command::Reverb(false) then Reverb(true). The engine turns the reverb off, then gives both cores a fresh hall. The driver test (driver_fixture.txt, regenerated) holds the order. It fails without the change, since every desktop and jukebox line now has the four reverb entries.

ccSnd +98

+98 is +0x62, which the sound page already describes. ccSndMoviePlayer mode 0 sets it (0x00180c44) and mode 1 clears it (0x00180e1c). ccSndStreamCtrl (0x0017d1a8) and ccSndStreamSE (0x0017caac) do nothing while it is set. piney-audio's stream code has the same rule: the caller makes no stream sound calls while the Audio screen's movie plays.

The port-0 volume branch

Every SQTBL row on the four volumes plays on HD ports 1-3, so ccSqPlay's port-0 branch (the SE volume) is never taken by a sequence:

Wave uses these same tables. ccPortVolSet with port 0, which the event sound command 8 can ask for, is modelled: Driver::port_vol_set gives port 0 se_vol. The SE volume is 0-256, so the 24-bit sign extension never shows.

ccSnd +0x110

The DWARF names +0x108-+0x117 int free[4], scratch for the stage's scene-sound routine (stageParam +0x105). Over the 459 ccSnd references, only three places touch the array:

• bgmChurch uses free[0] and free[1];
• bgmBreed uses free[0] and free[3];
• ccSndSQLoad's loop clears all four (0x00182448).

free[2] (+0x110) is never read: the game leaves it unused. The scan follows registers loaded from ccSnd. A pointer to the array handed to another function would escape it, and none was seen.