Sound
The WonderSwan features the following sound hardware:
- Four audio channels:
- channel 1 - wavetable (32 x 4-bit samples),
- channel 2 - wavetable or 8-bit unsigned PCM sample,
- channel 3 - wavetable with optional hardware sweep,
- channel 4 - wavetable or LFSR noise,
- Hyper Voice(color) - headphone output exclusive 16-bit stereo PCM output,
- 24000 Hz output:
- internal speaker - 8-bit, mono,
- headphone output - 16-bit, stereo.
Hyper Voice functionality is documented on its own sub-page.
Mixing diagram
The sound is mixed as follows:
Ch1 Ch2 Ch3 Ch4 Glossary: L = left, R = right, M = mono
| | | | Mnn = nn-bit mono
8 8 8 8 port $98 port $nn = value port
| | | | /
_V___V___V___V_ / _____ _____
| |--L10------+----->| | | |
| + | | | + |--M11-->| >>r |--M8--> Speaker output
|_______________|--R10------|-+--->|_____| / |_____|
/ _|_|_ / (r=0...3)
/ | | / \
port $96 | <<5 | port $9A \
|_____| \
| | port $91
| |
port $68 port $64 | |
/ _______________ \ __V_V__ _____
L8-->| (Color only) |--L16-->| |----------L16-->| |
| Hyper Voice | | + + | | I2S |------> Headphone output
R8-->|_______________|--R16-->|_______|----------R16-->|_____|
\ /
port $69 port $66
I/O Ports
Sound Channel Frequency ($80, $81; $82, $83; $84, $85; $86, $87)
15 bit 8 7 bit 0
---- ---- ---- ----
.... .ddd dddd dddd
||| |||| ||||
+++--++++-++++- Frequency divisor
This frequency is used for wavetable output. It is calculated as follows: sample rate = 3072000 Hz / (2048 - divisor).
Note that this refers to the sample rate of the wavetable, and needs to be scaled accordingly for a given waveform when performing music playback.
Sound Channel Volume ($88; $89; $8A; $8B)
7 bit 0 --------- llll rrrr |||| |||| |||| ++++- Right channel volume (0-15) ++++------ Left channel volume (0-15)
This volume is used for wavetable output.
The calculation to get an unsigned 8-bit sample out of a 4-bit wavetable sample / 4-bit volume pair is simple: out_sample = sample * volume. This means that the maximum sample in wavetable output mode is 15 * 15 = 225.
Sound Channel 2 Voice Sample ($89)
7 bit 0 ---- ---- ssss ssss |||| |||| ++++-++++- Unsigned 8-bit PCM sample (0-255)
This sample is used for voice output.
Sound Channel 3 Sweep Amount ($8C)
7 bit 0 ---- ---- vvvv vvvv |||| |||| ++++-++++- Value; 8-bit, signed.
This port is used for wavetable output with sweep enabled.
The signed value in this port will be added to the frequency divider for channel 3 every sweep step (as determined by the Sweep Ticks port). Wraparound is present - adding 1 to a frequency divider value of 2047 will cause it to roll over back to 0.
Sound Channel 3 Sweep Ticks ($8D)
7 bit 0 ---- ---- ...t tttt | |||| +-++++- Ticks per step - 1
This port is used for wavetable output with sweep enabled.
Every ttttt + 1 ticks, clocked at 375 Hz, the value in the Sweep Amount port will be added to the frequency divider for channel 3.
Sound Channel 4 Noise Control ($8E)
7 bit 0 ---- ---- ...e rttt | |||| | |+++- LFSR tap mode | +---- LFSR reset: 1 = reset shift register +------ LFSR enabled: 0 = off, 1 = on
This port is used for noise output.
| Tap mode | Tap bit | Sequence length |
|---|---|---|
| 0 | 14 | 32767 |
| 1 | 10 | 1953 |
| 2 | 13 | 254 |
| 3 | 4 | 217 |
| 4 | 8 | 73 |
| 5 | 6 | 63 |
| 6 | 9 | 42 |
| 7 | 11 | 28 |
Noise output works using the following algorithm, performed once per sample generated:
- Create a new bit by inverting the XOR of bit 7 with the tap bit.
- Shift the LFSR register one bit to the left.
- Write the new bit as bit 0.
- Use the new bit as if it were a wavetable sample: 0 = 0, 1 = 15.
Sound Wavetable Address ($8F)
7 bit 0 ---- ---- wwww wwww |||| |||| ++++-++++- Wavetable address (bits 6-13)
The wavetable is 4 x 16 bytes long; each 16-byte block contains 32 4-bit samples for the relevant channel:
Address $00 $01 $02 $03 $04 $05 $06 $07 $08 $09 $0A $0B $0C $0D $0E $0F
98 CB ED FF FF EF BD 8A 57 24 01 00 00 21 43 76
Sample
15 .. .. .. ## ## #. .. .. .. .. .. .. .. .. .. ..
14 .. .. .# || || |# .. .. .. .. .. .. .. .. .. ..
13 .. .. #| || || || #. .. .. .. .. .. .. .. .. ..
12 .. .# || || || || |. .. .. .. .. .. .. .. .. ..
11 .. #| || || || || |# .. .. .. .. .. .. .. .. ..
10 .. || || || || || || #. .. .. .. .. .. .. .. ..
9 .# || || || || || || |. .. .. .. .. .. .. .. ..
8 #| || || || || || || |# .. .. .. .. .. .. .. ..
7 || || || || || || || || #. .. .. .. .. .. .. .#
6 || || || || || || || || |. .. .. .. .. .. .. #|
5 || || || || || || || || |# .. .. .. .. .. .. ||
4 || || || || || || || || || #. .. .. .. .. .# ||
3 || || || || || || || || || |. .. .. .. .. #| ||
2 || || || || || || || || || |# .. .. .. .# || ||
1 || || || || || || || || || || #. .. .. #| || ||
0 || || || || || || || || || || |# ## ## || || ||
Note that the higher bits - the first hexadecimal number - specify the later samples; that is to say, the nybbles of each sample are swapped relative to the wave drawing.
Sound Channel Control ($90)
7 bit 0 ---- ---- nsv. 4321 ||| |||| ||| |||+- Channel 1 enable ||| ||+-- Channel 2 enable ||| |+--- Channel 3 enable ||| +---- Channel 4 enable ||+------- Channel 2 mode: 0 = wavetable, 1 = voice |+-------- Channel 3 sweep: 0 = disable, 1 = enable +--------- Channel 4 mode: 0 = wavetable, 1 = noise
Sound Output Control ($91)
7 bit 0 ---- ---- H... hrrs | |||| | |||+- Internal speaker output enable | |++-- Internal speaker shift/volume: | | 0 = 100% 1 = 50% | | 2 = 25% 3 = 12.5% | +---- Headphone output enable +--------- Headphones connected, 1 if true
For correct playback on the internal speaker, it is important to set the shift/volume correctly, to match the peak summed output of the first four channels. If the value is too low, music using multiple loud channels will wrap, causing audible distortion. If the value is too high, single-channel PCM sample playback will be very quiet.
Sound Channel 4 LFSR Register ($92, $93 read)
15 bit 8 7 bit 0 ---- ---- ---- ---- .rrr rrrr rrrr rrrr ||| |||| |||| |||| +++-++++--++++-++++- Shift register value
Sound Channel 2 Voice Volume ($94)
7 bit 0
---- ----
.... lLrR
||||
|||+- Right channel 100% volume
||+-- Right channel 50% volume
|+--- Left channel 100% volume
+---- Left channel 50% volume
This port is used for voice output.
The 50% volume have no effect if their respective 100% volume bits are set.
Sound Test ($95)
7 bit 0 ---- ---- ??5? llsh | |||| | |||+- Hold Ch1-4 output updates | ||+-- Use 3072000 Hz CPU clock for sweep | ++--- Hold noise LFSR output | (Differences between bit 2/3 | behavior are unknown) +------- Force L10/R10 outputs to (channel 2 voice output * 5) & 0x3FF
Sound Channel Output Right ($96, $97 read)
15 bit 8 7 bit 0
---- ---- ---- ----
.... ..ss ssss ssss
|| |||| ||||
++--++++-++++- Unsigned 10-bit sample
Sound Channel Output Left ($98, $99 read)
15 bit 8 7 bit 0
---- ---- ---- ----
.... ..ss ssss ssss
|| |||| ||||
++--++++-++++- Unsigned 10-bit sample
Sound Channel Output Sum ($9A, $9B read)
15 bit 8 7 bit 0
---- ---- ---- ----
.... .sss ssss ssss
||| |||| ||||
+++--++++-++++- Unsigned 11-bit sample
Sound Speaker Main Volume ($9E)
15 bit 8 7 bit 0
---- ---- ---- ----
.... .... .... ..vv
||
++- Built-in speaker output volume
Writing to this port changes the output volume of the built-in speaker.
While it functions the same way as the SOUND button (and affects the same internal register), pressing the SOUND button does not update this port's value. However, manual writes do, and those values can be read out.
This port does not affect the headphone output in any way, just like the SOUND button.
TODO: The exact volume levels have not been verified; it's probably either 0, 25, 50, 100% or 0, ~33, ~67, 100%.
