FF7/PSX/Sound/AKAO sequence

Contents

Introduction

AKAO sequence is the most complicated part of the FF7 sound system. ("AKAO" is the signature string, which implies that the sound format has developed by Square Enix sound programmer Minoru Akao.)

AKAO sequence is similar to MIDI sequence - it's custom tracker format for playing sequence sound, well tuned specially for PSX.

The sequence data can be found in all FF7 game modules: Field, Battle, Worldmap and in minigames.

All files with exension *.SND are AKAO.

• MINI/ASERI2.SND - Battle Arena theme
• MINI/SENSUI.SND - used in Submarine minigame
• ENEMY6/OVER2.SND - game over sequence
• ENEMY6/FAN2.SND - battle win "fanfare" sequence
• MOVIE/OVER2.SND - same game over sequence, don't know, why to duplicate data

Other AKAO sequences are hard-wired in other files.

File Structure

Header (size: 16 bytes)

struct AkaoHeader
{
  static const uint8_t magic[4];  // "AKAO" C-string
  uint16_t id;                    // song ID, used for playing sequence
  uint16_t length;                // data length - sizeof(header)
  uint16_t reverb_type;           // reverb type (range from 0 to 9)
  struct AkaoTimeStamp
  {
    uint8_t year_bcd;             // year (in binary coded decimal)
    uint8_t month_bcd;            // month (in binary coded decimal, between 0x01 - 0x12)
    uint8_t day_bcd;              // day (in binary coded decimal, between 0x01 - 0x31)
    uint8_t hours_bcd;            // hours (in binary coded decimal, between 0x00 - 0x23)
    uint8_t minutes_bcd;          // minutes (in binary coded decimal, between 0x00 - 0x59)
    uint8_t seconds_bcd;          // seconds (in binary coded decimal, between 0x00 - 0x59)
  } timestamp;
};

Channel info (size: 4 bytes + 2 bytes * <channels count>)

struct AkaoChannelInfo
{
  uint32_t mask;                         // represents bitmask of used channels in this song
  uint32_t start_offsets[num_channels];  // offsets to channel opcode data
};

int num_channels = 0;
while (int bit = 0; ((info.mask & 0xFFFFFF) & (1 << bit)) != 0; bit++)
  num_channels++;

First there is 32-bit number (offset 0x10), which represents bitmask of used channels in this song.

After this bitmask, there is <channels count> offsets to channel opcode data counting from current offset. Each offsets is a relative offset based on the address *next to* the offset itself. (This is a general rule for early versions of AKAO to interpret relative offsets.)

Channel Commands [AKAO Opcodes]

For every channels in an AKAO sequence, there is a set of commands to perform. This is similar to Field opcodes. Here I will call this sound commands "opcodes". Every opcode has its own number of arguments (from no-arguments, to 3 arguments).

Example (home-created AKAO sequence):

Header

41 4b 41 4f - AKAO string

34 12 - song ID: 0x1234

16 00 - data length 0x16 in hex or 22 in decimal

04 00 - reverb type: 4 (large studio)

96 12 18 22 46 28 - created at 1996-12-18 22:46:28

Channel info

01 00 00 00 - this indicates, that used only one channel

00 00 - offset to first channel opcodes: in our example 0x00 means that next to this offset is opcodes for first channel

Channel commands

e8 a8 66 - sets tempo, parameter 0x66a8

ea 00 50 - sets reverb depth

a8 55 - load sample 0x55 from INSTR.ALL to channel

aa 40 - sets channel volume

c2 - turns on reverb effect

a1 0c - sets volume pan

c8 - sets loop point

66 - 0x66 % 11 = 3 (3 means to take 3rd number from play length table), 0x66 / 11 = 9 (9 means to take pitch[9] from loaded instrument record index)

ca - returns to saved loop point with opcode c8

This example plays Chocobo "Whoo-Hoo" (instrument number 0x55) repeatedly.

Sound Opcode list

Opcode	Summary	Length	Operands	Note
0x00-0x99	Note, Tie, Rest	1		The opcode indicates the note key and length. The note will be keyed off 2 ticks before the next note.
0x9A-0x9F	Unimplemented	1		Should not be used.
0xA0	Finish Channel	1
0xA1	Load Instrument	2	instrument: byte
0xA2	Overwrite Next Note Length	2	length: byte	Ignores the regular length (delta-time) of the next note and overwrites it with the specified length.
0xA3	Channel Master Volume	2	volume: byte (0-127)
0xA4	Pitch Bend Slide	3	length: byte, semitones: signed byte	When length is 0, it will be translated to 256 ticks.
0xA5	Set Octave	2	octave: byte (0-15)
0xA6	Increase Octave	1
0xA7	Decrease Octave	1
0xA8	Channel Volume	2	volume: byte (0-127)
0xA9	Channel Volume Slide	3	length: byte, volume: byte (0-127)	When length is 0, it will be translated to 256 ticks.
0xAA	Channel Pan	2	pan: byte (0-127)	64 is the center.
0xAB	Channel Pan Slide	3	length: byte, pan: byte (0-127)	When length is 0, it will be translated to 256 ticks.
0xAC	Noise Clock Frequency	2	clock: byte (0x00-0x3f)
0xAD	ADSR: Attack Rate	2	attack_rate: byte (0x00-0x7f)
0xAE	ADSR: Decay Rate	2	decay_rate: byte (0x00-0x0f)
0xAF	ADSR: Sustain Level	2	sustain_level: byte (0x00-0x0f)
0xB0	ADSR: Decay Rate & Sustain Level	3	decay_rate: byte (0x00-0x0f), sustain_level: byte (0x00-0x0f)
0xB1	ADSR: Sustain Rate	2	sustain_rate: byte (0x00-0x7f)
0xB2	ADSR: Release Rate	2	release_rate: byte (0x00-0x1f)
0xB3	ADSR: Reset ADSR	1
0xB4	Channel Pitch LFO (Vibrato)	4	delay: byte, rate: byte, type: byte (0-15)	When rate is 0, it will be translated to 256 ticks. The interpretation of the first parameter may change to depth depending on 0xF3.
0xB5	Channel Pitch LFO Depth	2	depth: byte	The most significant bit of the depth determines the amplitude range. When it is 0, the range is up to about ± 50 cents (amplitude is 15/256 compared with range type 1). When it is 1, the range is up to about ± 700 cents.
0xB6	Turn Off Channel Pitch LFO	1
0xB7	ADSR: Attack Mode	2	attack_mode: byte (1 or 5)
0xB8	Channel Volume LFO (Tremolo)	4	delay: byte, rate: byte, type: byte (0-15)	When rate is 0, it will be translated to 256 ticks. The interpretation of the first parameter may change to depth depending on 0xF3.
0xB9	Channel Volume LFO Depth	2	depth: byte
0xBA	Turn Off Channel Volume LFO	1
0xBB	ADSR: Sustain Mode	2	sustain_mode: byte (1, 3, 5 or 7)
0xBC	Channel Pan LFO	3	rate: byte, type: byte (0-15)	When rate is 0, it will be translated to 256 ticks.
0xBD	Channel Pan LFO Depth	2	depth: byte
0xBE	Turn Off Channel Pan LFO	1
0xBF	ADSR: Release Mode	2	release_mode: byte (3 or 7)
0xC0	Channel Transpose (Absolute)	2	semitones: signed byte
0xC1	Channel Transpose (Relative)	2	semitones: signed byte
0xC2	Turn On Reverb	1
0xC3	Turn Off Reverb	1
0xC4	Turn On Noise	1
0xC5	Turn Off Noise	1
0xC6	Turn On Frequency Modulation	1
0xC7	Turn Off Frequency Modulation	1
0xC8	Loop Point	1		Remember the current offset as a loop point and increase the nesting level of the loop.
0xC9	Return to Loop Point Up to N Times	2	times: byte	On the Nth repeat, this instruction will end the current loop by decrementing the nesting level of the loop. Otherwise, it will increment the loop counter and return to the loop point. When times is 0, it will be translated to 256 times.
0xCA	Return to Loop Point	1		This instruction will increment the loop counter.
0xCB	Reset Sound Effects	1		Reset sound effects such as noise, frequency modulation, reverb, overlay voice and alternate voice.
0xCC	Turn On Legato	1		This instruction will stop the regular key on and key off performance of the subsequent notes and update the pitch only.
0xCD	Turn Off Legato	1
0xCE	Turn On Noise and Toggle Noise On/Off after a Period of Time	2	delay: byte	When delay is 0, it will be translated to 257 ticks.
0xCF	Toggle Noise On/Off after a Period of Time	2	delay: byte	When delay is 0, it will be translated to 257 ticks.
0xD0	Turn On Full-Length Note Mode	1		This instruction will stop the regular key off performance of the subsequent notes.
0xD1	Turn Off Full-Length Note Mode	1
0xD2	Turn On Frequency Modulation and Toggle Frequency Modulation On/Off after a Period of Time	2	delay: byte	When delay is 0, it will be translated to 257 ticks.
0xD3	Toggle Frequency Modulation On/Off after a Period of Time	2	delay: byte	When delay is 0, it will be translated to 257 ticks.
0xD4	Turn On Playback Rate Side Chain	1		Duplicate and use the playback frequency of the previous voice channel.
0xD5	Turn Off Playback Rate Side Chain	1
0xD6	Turn On Pitch-Volume Side Chain	1		Multiply the playback frequency of the previous voice channel to the output volume. Lower pitch will make the volume smaller.
0xD7	Turn Off Pitch-Volume Side Chain	1
0xD8	Channel Fine Tuning (Absolute)	2	amount: signed byte	The pitch can be changed in the range of one octave above and below. The amount is the multiplicand for the playback frequency, not a log-based parameter like cents. For example, when the amount is 127, the pitch will be multiplied by 127/128 (about +1 octave). Negative value ​​will lower the pitch.
0xD9	Channel Fine Tuning (Relative)	2	amount: signed byte	The amount will be added to the current tuning amount.
0xDA	Turn On Portamento	2	speed: byte	When speed is 0, it will be translated to 256 ticks.
0xDB	Turn Off Portamento	1
0xDC	Fix Note Length	2	length_to_add: signed byte	Ignore the regular length (delta-time) of subsequent notes and set to the fixed length. The length_to_add parameter will be added to the current length value. (the initial value is 0)
0xDD	Channel Pitch LFO Depth Slide	3	length: byte, depth: byte	When length is 0, it will be translated to 256 ticks.
0xDE	Channel Volume LFO Depth Slide	3	length: byte, depth: byte	When length is 0, it will be translated to 256 ticks.
0xDF	Channel Pan LFO Depth Slide	3	length: byte, depth: byte	When length is 0, it will be translated to 256 ticks.
0xE0-0xE7	Unimplemented	1		Code-referenced to 0xA0. Should not be used.
0xE8	Tempo	3	tempo: uint16	bpm = tempo / 214.998204 (approximate) More strictly, bpm = 60.0 / (48 * (65536.0 / tempo) * (0x44E8 / (33868800.0 / 8))) Note that this coefficient is different from other games with PlayStation AKAO.
0xE9	Tempo Slide	4	length: byte, tempo: uint16	When length is 0, it will be translated to 256 ticks.
0xEA	Reverb Depth	3	depth: uint16
0xEB	Reverb Depth Slide	4	length: byte, depth: uint16	When length is 0, it will be translated to 256 ticks.
0xEC	Turn On Drum Mode	3	drum_map_offset: signed int16	The drum_map_offset is a relative offset pointing to the drum instrument map table, which determines the instrument for each keys.
0xED	Turn Off Drum Mode	1
0xEE	Unconditional Jump	3	destination_offset: signed int16	This instruction can be used to make an infinite loop.
0xEF	CPU-Conditional Jump	4	condition: byte, destination_offset: signed int16	Jump if the condition variable matches condition. The value of the condition variable can be set from the game program.
0xF0	Jump on the Nth Repeat	4	times: byte, destination_offset: signed int16	When times is 0, it will be translated to 256 times.
0xF1	Break the Loop on the Nth Repeat	4	times: byte, destination_offset: signed int16	Unlike 0xF0, this instruction will end the current loop by decrementing the nesting level of the loop. When times is 0, it will be translated to 256 times.
0xF2	Load Instrument (No Attack Sample)	2	instrument: byte	Unlike 0xA1, the sample before the loop point is replaced by a short, silence sample.
0xF3	Use No Delay for Channel Pitch/Volume LFO	1		Specific to FF7. This instruction changes the interpretation of the 0xB4 and 0xB8 parameters. There is no actual use in music.
0xF4	Turn On Overlay Voice	3	instrument: byte, instrument2: byte	Play the same melody by different instruments on two voice channels. A free voice channel is required to work. Note that the two channels share the playback rate, and the pitch is not calculated for each instruments. Used in the song "Anxious Heart".
0xF5	Turn Off Overlay Voice	1
0xF6	Volume Balance of Overlay Voice	2	balance: byte (0-127)	When the balance is 0, the volume of the primary voice will be 100% (127/128) of original and that of the secondary voice will be 0%. 127 is the opposite.
0xF7	Volume Balance Slide of Overlay Voice	3	length: byte, balance: byte (0-127)	When length is 0, it will be translated to 256 ticks.
0xF8	Turn On Alternate Voice	2	release_rate: byte (0x00-0x1f)	This instruction allows subsequent notes to be played on two alternating channels. At the same time, the ADSR release rate will be set to the specified value. A free voice channel is required to work. Check "Opening - Bombing Mission", "Tifa's Theme" and "Fortress of the Condor" for actual usage.
0xF9	Turn Off Alternate Voice	1
0xFA-0xFC	Unimplemented	1		Code-referenced to 0xA0. Should not be used.
0xFD	Time Signature	3	ticks_per_beat: byte, beats_per_measure: byte	Note that two parameters can be 0. This pattern is used for initialization.
0xFE	Measure Number	2	measure: byte
0xFF	Unimplemented	1		Code-referenced to 0xA0. Should not be used.

Drum Instrument Map Table

When a song uses a drum kit with opcode 0xEC, a drum instrument map table will be placed at the end of the sequence. The table determines the instrument, channel volume and pan for each keys.

The table consists of a repetition of 5-byte items.

struct AkaoDrumNoteAttr
{
  uint8_t instrument;   // corresponding to opcode 0xA1
  uint8_t note_number;  // note number when playing the drum note
  uint16_t volume;      // corresponding to opcode 0xA8 (lower 8bit is a fractional part of the volume)
  uint8_t pan;          // corresponding to opcode 0xAA
}