// license:BSD-3-Clause // copyright-holders:Devin Acker /*************************************************************************** NEC/Casio uPD931 synthesis chip Many details of this implementation are based on research and notes by Robin Whittle: https://www.firstpr.com.au/rwi/casio/Casio-931-2006-06-17.txt Any references to MT-65 behavior are based on this document. TODO: - implement vibrato register (CT-8000 doesn't use it) - a few other unknown/unclear bits in the flags shift register ***************************************************************************/ #include "emu.h" #include "upd931.h" #include #include DEFINE_DEVICE_TYPE(UPD931, upd931_device, "upd931", "NEC uPD931") upd931_device::upd931_device(const machine_config &mconfig, const char *tag, device_t *owner, u32 clock) : device_t(mconfig, UPD931, tag, owner, clock) , device_sound_interface(mconfig, *this) , device_memory_interface(mconfig, *this) , m_io_config("io", ENDIANNESS_LITTLE, 8, 8, 0, address_map_constructor(FUNC(upd931_device::io_map), this)) , m_retrig_timer(nullptr) , m_filter_cb(*this) , m_sync_cb(*this) , m_master(true) { } /**************************************************************************/ device_memory_interface::space_config_vector upd931_device::memory_space_config() const { return space_config_vector{std::make_pair(AS_IO, &m_io_config)}; } /**************************************************************************/ void upd931_device::io_map(address_map &map) { map(0x20, 0x27).w(FUNC(upd931_device::note_w)); map(0x30, 0x37).w(FUNC(upd931_device::octave_w)); // waveform write position: ct8000 uses 40, mt65 uses 60 map(0x40, 0x40).mirror(0x20).w(FUNC(upd931_device::wave_pos_w)); map(0xa7, 0xa7).w(FUNC(upd931_device::wave_data_w)); map(0xb0, 0xb7).w(FUNC(upd931_device::flags_w)); // lower address bits are ignored map(0xb8, 0xbf).w(FUNC(upd931_device::status_latch_w)); // vibrato/sustain: ct8000 uses c0-c7, mt65 uses d0-d7 (lower two bits are also ignored) map(0xc0, 0xc3).mirror(0x10).w(FUNC(upd931_device::vibrato_w)); map(0xc4, 0xc7).mirror(0x10).w(FUNC(upd931_device::sustain_w)); map(0xe0, 0xe7).w(FUNC(upd931_device::note_on_w)); map(0xf4, 0xf4).lw8(NAME([this] (offs_t offset, u8 data) { m_filter_cb(data); })); } /**************************************************************************/ void upd931_device::device_start() { space(AS_IO).specific(m_io); m_stream = stream_alloc(0, 1, clock() / CLOCKS_PER_SAMPLE); if (m_master) { m_retrig_timer = timer_alloc(FUNC(upd931_device::timer_tick), this); reset_timer(); } m_last_clock = clock(); m_db = 0xf; m_i1 = m_i2 = m_i3 = 1; m_status = 0xf; // generate note frequencies based on CT-8000 crystal frequency (probably not entirely accurate) const unsigned sample_rate = 4'946'864 / CLOCKS_PER_SAMPLE; for (int i = 0; i < 73; i++) { // A4 is note 33, 442 Hz const double freq = 442.0 * pow(2, (i - 33) / 12.0); m_pitch[i] = (1 << PITCH_SHIFT) * (freq * 16 / sample_rate); } save_item(NAME(m_db)); save_item(NAME(m_i1)); save_item(NAME(m_i2)); save_item(NAME(m_i3)); save_item(NAME(m_addr)); save_item(NAME(m_data)); save_item(NAME(m_status)); save_item(NAME(m_wave)); save_item(NAME(m_wave_pos)); save_item(NAME(m_vibrato)); save_item(NAME(m_sustain)); save_item(NAME(m_reverb)); save_item(NAME(m_flags)); save_item(NAME(m_last_clock)); save_item(STRUCT_MEMBER(m_voice, m_note)); save_item(STRUCT_MEMBER(m_voice, m_octave)); save_item(STRUCT_MEMBER(m_voice, m_pitch)); save_item(STRUCT_MEMBER(m_voice, m_pitch_counter)); save_item(STRUCT_MEMBER(m_voice, m_timbre_shift)); save_item(STRUCT_MEMBER(m_voice, m_wave_pos)); save_item(STRUCT_MEMBER(m_voice, m_wave_out)); save_item(STRUCT_MEMBER(m_voice, m_env_state)); save_item(STRUCT_MEMBER(m_voice, m_env_counter)); save_item(STRUCT_MEMBER(m_voice, m_env_level)); save_item(STRUCT_MEMBER(m_voice, m_force_release)); } /**************************************************************************/ TIMER_CALLBACK_MEMBER(upd931_device::timer_tick) { m_sync_cb(1); sync_w(1); } /**************************************************************************/ void upd931_device::device_reset() { std::fill(std::begin(m_wave[0]), std::end(m_wave[0]), 0); std::fill(std::begin(m_wave[1]), std::end(m_wave[1]), 0); m_wave_pos = 0; m_vibrato = 0; m_sustain = 0; m_reverb = 0; m_flags = 0; std::fill(m_voice.begin(), m_voice.end(), voice_t()); m_filter_cb(0); m_sync_cb(0); } /**************************************************************************/ void upd931_device::device_clock_changed() { m_stream->set_sample_rate(clock() / CLOCKS_PER_SAMPLE); if (m_retrig_timer) { const u64 ticks_left = m_retrig_timer->remaining().as_ticks(m_last_clock); const attotime remaining = attotime::from_ticks(ticks_left, clock()); const attotime period = attotime::from_ticks(RETRIG_RATE, clock()); m_retrig_timer->adjust(remaining, 0, period); } m_last_clock = clock(); } /**************************************************************************/ void upd931_device::sound_stream_update(sound_stream &stream, std::vector const &inputs, std::vector &outputs) { for (int i = 0; i < outputs[0].samples(); i++) { s32 sample = 0; for (voice_t &voice : m_voice) { update_env(voice); update_wave(voice); sample += voice.m_wave_out[0] * (voice.m_env_level[0] >> VOLUME_SHIFT); sample += voice.m_wave_out[1] * (voice.m_env_level[1] >> VOLUME_SHIFT); } outputs[0].put_int_clamp(i, sample, 1 << 16); } } /**************************************************************************/ void upd931_device::i1_w(int state) { if (!m_i3 && m_i1 && !state) { // address high nibble on I1 falling edge m_addr &= 0x0f; m_addr |= (m_db << 4); } m_i1 = state; } /**************************************************************************/ void upd931_device::i2_w(int state) { if (!m_i3) { if (!m_i2 && state) { // address low nibble on I2 rising edge m_addr &= 0xf0; m_addr |= (m_db & 0xf); // apply register write m_stream->update(); m_io.write_byte(m_addr, m_data & 0xf); } else if (m_i2 && !state) { // data on I2 falling edge m_data = m_db; } } m_i2 = state; } /**************************************************************************/ void upd931_device::i3_w(int state) { m_i3 = state; } /**************************************************************************/ void upd931_device::db_w(u8 data) { m_db = data; } /**************************************************************************/ u8 upd931_device::db_r() { if (m_i1 && m_i2 && !m_i3) return m_status; return 0xf; } /**************************************************************************/ void upd931_device::sync_w(int state) { if (BIT(m_flags, FLAG_RETRIGGER)) { m_stream->update(); for (voice_t &voice : m_voice) { if (voice.m_env_state == ENV_DECAY1 || voice.m_env_state == ENV_DECAY2) { voice.m_env_state = ENV_ATTACK1; voice.m_env_counter = 0; } } } } /**************************************************************************/ void upd931_device::note_w(offs_t offset, u8 data) { m_voice[offset].m_note = data; } /**************************************************************************/ void upd931_device::octave_w(offs_t offset, u8 data) { m_voice[offset].m_octave = m_data; } /**************************************************************************/ void upd931_device::wave_pos_w(u8 data) { m_wave_pos = data; } /**************************************************************************/ void upd931_device::wave_data_w(u8 data) { const u8 sel = BIT(m_flags, FLAG_WAVE_SEL); m_wave[sel][m_wave_pos & 0xf] = data; } /**************************************************************************/ void upd931_device::flags_w(u8 data) { m_flags >>= 1; m_flags |= ((data & 1) << FLAG_WAVE_SEL); } /**************************************************************************/ void upd931_device::status_latch_w(offs_t offset, u8 data) { // TODO: more details. ct8000 only checks if bits 1-3 are all zero or not if (m_voice[offset].m_env_state == ENV_IDLE) m_status = 0xf; else m_status = 0; } /**************************************************************************/ void upd931_device::vibrato_w(u8 data) { // TODO: implement this. ct8000 always writes 0 since it uses the external VCO vibrato instead m_vibrato = data & 3; } /**************************************************************************/ void upd931_device::sustain_w(u8 data) { m_sustain = data & 3; m_reverb = BIT(data, 1, 2) ? 1 : 0; } /**************************************************************************/ void upd931_device::note_on_w(offs_t offset, u8 data) { voice_t &voice = m_voice[offset]; if (BIT(data, 0)) note_on(voice); else voice.m_env_state = ENV_RELEASE; // mt65 turns off sustain and reverb when changing tones, ct8000 does this instead // (and also when playing a new note that is already playing) voice.m_force_release = BIT(data, 3); } /**************************************************************************/ void upd931_device::note_on(voice_t &voice) { if (voice.m_note >= 0x2 && voice.m_note <= 0xe) { const u8 note = voice.m_note - 2; u8 octave = voice.m_octave & 7; if (octave >= 2) octave -= 2; // octave values 0-1 are the same as 2-3 /* setting bit 3 of the octave reduces the duty cycle of individual notes, which is implemented here by changing which part of the phase counter to use as the sample address. ct8000 uses this for a few of its presets to produce a simple key-scaling effect. */ if (BIT(voice.m_octave, 3)) voice.m_timbre_shift = 3 - octave; else voice.m_timbre_shift = 0; voice.m_pitch = m_pitch[octave * 12 + note]; } else { voice.m_pitch = 0; } voice.m_pitch_counter = 0; voice.m_wave_pos = 0xff; voice.m_wave_out[0] = voice.m_wave_out[1] = 0; voice.m_env_state = ENV_ATTACK1; voice.m_env_counter = 0; if (m_master) reset_timer(); } /**************************************************************************/ void upd931_device::reset_timer() { const attotime period = attotime::from_ticks(RETRIG_RATE, clock()); m_retrig_timer->adjust(period, 0, period); } /**************************************************************************/ void upd931_device::update_env(voice_t &voice) { const unsigned shift = BIT(m_flags, FLAG_ENV_SHIFT, 2); switch (voice.m_env_state) { case ENV_IDLE: return; case ENV_ATTACK1: { static const u16 rates[] = { 0, 2048, 512, 256, 160, 80, 32, 8 }; const u8 val = BIT(m_flags, FLAG_ATTACK1, 3); u32 rate; if (val == 0) rate = VOLUME_MAX; // value 0 = instant else if (val < 4 && voice.m_env_counter >= (0xe0 << VOLUME_SHIFT)) rate = 160 << shift; // values 1-3 slow down at 7/8 of max volume else rate = rates[val] << shift; voice.m_env_counter = std::min(voice.m_env_counter + rate, VOLUME_MAX); // only increase wave A level if it isn't set to rise during attack2 instead if (BIT(m_flags, FLAG_ATTACK2_A)) voice.m_env_level[0] = 0; else voice.m_env_level[0] = voice.m_env_counter; voice.m_env_level[1] = voice.m_env_counter; if (voice.m_env_counter >= VOLUME_MAX) { voice.m_env_counter = 0; voice.m_env_state = ENV_ATTACK2; } break; } case ENV_ATTACK2: { static const u32 rates[] = { 0, 2048, 256, 128, 64, 32, 16, 8 }; const u8 val = BIT(m_flags, FLAG_ATTACK2, 3); u32 rate; if (val == 0) rate = VOLUME_MAX; else rate = rates[val] << shift; voice.m_env_counter = std::min(voice.m_env_counter + rate, VOLUME_MAX); // fade wave A in, if specified if (BIT(m_flags, FLAG_ATTACK2_A)) voice.m_env_level[0] = voice.m_env_counter; // fade wave B out, if specified if (BIT(m_flags, FLAG_ATTACK2_B)) voice.m_env_level[1] = VOLUME_MAX - voice.m_env_counter; if (voice.m_env_counter >= VOLUME_MAX) voice.m_env_state = ENV_DECAY1; break; } case ENV_DECAY1: { static const u32 rates[] = { 2048, 640, 160, 32, 16, 8, 2, 0 }; const u8 val = BIT(m_flags, FLAG_DECAY1, 3); const u32 rate = rates[val] << shift; if (voice.m_env_counter < rate) { voice.m_env_counter = 0; voice.m_env_state = ENV_IDLE; } else { voice.m_env_counter -= rate; } voice.m_env_level[0] = voice.m_env_counter; // only fade wave B if it didn't already fade out during attack2 if (voice.m_env_level[1]) voice.m_env_level[1] = voice.m_env_counter; if (!BIT(m_flags, FLAG_DECAY2_DISABLE)) { // transition to decay2 at 1/2 or 1/4 of max volume, if enabled const u8 decay2_level = BIT(m_flags, FLAG_DECAY2_LEVEL) ? 0x40 : 0x80; if (voice.m_env_counter < (decay2_level << VOLUME_SHIFT)) voice.m_env_state = ENV_DECAY2; } break; } case ENV_DECAY2: { // apply reverb rate (below 1/8 max volume) or decay2 rate u16 rate; if (m_reverb && voice.m_env_counter < (0x20 << VOLUME_SHIFT)) rate = 1 << shift; else if (BIT(m_flags, FLAG_DECAY2)) rate = 3 << shift; else rate = 6 << shift; if (voice.m_env_counter < rate) { voice.m_env_counter = 0; voice.m_env_state = ENV_IDLE; } else { voice.m_env_counter -= rate; } voice.m_env_level[0] = voice.m_env_counter; if (voice.m_env_level[1]) voice.m_env_level[1] = voice.m_env_counter; break; } case ENV_RELEASE: { // apply reverb rate (below 1/8 max volume), sustain rate (if enabled) or default release rate u16 rate = 512 << shift; if (!voice.m_force_release) { if (m_reverb && voice.m_env_counter < (0x20 << VOLUME_SHIFT)) rate = 1 << shift; else if (m_sustain == 1) rate = 16 << shift; else if (m_sustain == 2) rate = 12 << shift; } if (voice.m_env_counter < rate) { voice.m_env_counter = 0; voice.m_env_state = ENV_IDLE; } else { voice.m_env_counter -= rate; } // fade each wave individually because if keyed off during attack, they may be different from each other voice.m_env_level[0] = std::min(voice.m_env_level[0], voice.m_env_counter); voice.m_env_level[1] = std::min(voice.m_env_level[1], voice.m_env_counter); break; } } } /**************************************************************************/ void upd931_device::update_wave(voice_t &voice) { voice.m_pitch_counter += voice.m_pitch; const u8 cycle = BIT(voice.m_pitch_counter, PITCH_SHIFT + 4, 2); /* the part of the counter which is treated as the sample address depends on if key scaling is active. if it is active, then the voice alternates between a narrowed waveform and silence, which alters the timbre without affecting pitch */ u8 pos = BIT(voice.m_pitch_counter, PITCH_SHIFT - voice.m_timbre_shift, 4 + voice.m_timbre_shift); if (pos == voice.m_wave_pos || pos >= 0x10) return; voice.m_wave_pos = pos; // play every other cycle backwards, if enabled if (BIT(m_flags, FLAG_MIRROR) && BIT(cycle, 0)) pos ^= 0xf; // flag bits determine which of 4 consecutive cycles to apply waveforms A and B const unsigned cycle_mode[] = { BIT(m_flags, FLAG_MODE_A, 2), BIT(m_flags, FLAG_MODE_B, 2) }; static const u8 cycle_mask[] = { 0xf, // always on 0x5, // on, off, on, off 0x1, // on 1x, off 3x 0x3 // on 2x, off 2x }; static const s8 steps[] = { 0, 1, 2, 2, 4, 4, 8, 8, 0, -1, -2, -2, -4, -4, -8, -8 }; for (int i = 0; i < 2; i++) { // check if this waveform is enabled for this cycle if (!BIT(cycle_mask[cycle_mode[i]], cycle)) continue; s8 step = steps[m_wave[i][pos] & 0xf]; // invert waveform on every other cycle, if enabled if (BIT(m_flags, FLAG_INVERT) && BIT(cycle, 0)) step = -step; voice.m_wave_out[i] += step; voice.m_wave_out[i] = util::sext(voice.m_wave_out[i] & 0x3f, 6); } }