diff options
Diffstat (limited to 'src/mame/audio/upd1771.c')
-rw-r--r-- | src/mame/audio/upd1771.c | 554 |
1 files changed, 554 insertions, 0 deletions
diff --git a/src/mame/audio/upd1771.c b/src/mame/audio/upd1771.c new file mode 100644 index 00000000000..96b3b2bb062 --- /dev/null +++ b/src/mame/audio/upd1771.c @@ -0,0 +1,554 @@ +// license:BSD-3-Clause +// copyright-holders:David Viens +/********************************************************************** + + NEC uPD1771-017 as used in the Epoch Super Cassette Vision (SCV) + + Made using recording/analysis on a Yeno (PAL Super Cassete Vision) + by plgDavid + + Full markings on my 2 specimens are + "NEC JAPAN 8431K9 D1771C 017" (31st week of 1984, mask rom #017) + + I've since (October 2012) got a Grandstand Firefox F-7 Handheld game + (AKA Epoch GalagaX6/Epoch Astro Thunder 7/Tandy Astro Thunder), + (http://www.handheldmuseum.com/Grandstand/Firefox.htm) + which includes a + "NEC JAPAN 8319K9 D1771C 011" (19th week of 1983, mask rom #011) + Thanks to user 'Blanka' from Dragonslairfans for the nice catch! + (http://www.dragonslairfans.com/smfor/index.php?topic=3061.0) + + Since the chip generates tones using ROM wavetables, + it is perfectly possible to generate other sounds with different rom code and data. + + Most upd17XXX devices are typically 4bit NEC MCUs, however based on information + in "Electronic Speech Synthesis" by Geoff Bristow (ISBN 0-07-007912-9, pages 148-152) + the upd1770/1771 is not one of these 4-bit ones. + + The uPD1770/uPD1771 SSM is a 16-bit-wide rom/ram mcu with 8kb (4kw) of rom code, + 64 bytes of ram (16x16bit words addressable as 16 or 2x8 bits each, the + remaining 32 bytes acting as an 8-level stack), 182 instructions, a complex + noise and tone internal interrupt system, external interrupts, + and two 8-bit ports with multiple modes allowing for chips to operate as master + or slave devices. + SSM stands for "Sound Synthesis Microcomputer". + + People who I *THINK* worked on the uPD1771 and what part I think they worked on: + Toshio Oura - Project Lead(?), VSRSSS/TSRSSS speech synthesis engine (on upd1776C), master/slave i/o controls, author of bristow article and primary author of the IEEE article + Hatsuhide Igarashi - Clock oscillator and pad layout, coauthor on the IEEE article, other IEEE stuff + Tomoaki Isozaki - ? (senior NEC engineer?), coauthor on the IEEE article + Sachiyuki Toufuku - ?, coauthor on the IEEE article + Tojiro Mukawa - IGFETs and the DAC + M. Sakai ? - digital filtering for VSRSSS? (IEEE 4131979, 1169295) + M. Endo ? - digital design system or speech synthesis? (IEEE 4069656, another? person: IEEE 150330, 225838) + H. Aoyama ? - logic design system used to assemble/lay out the chip? (IEEE 1585393) + I. Fujitaka ? (no IEEE) + Eiji Sugimoto - cpu design? 1156033 1155824 + F. Tsukuda ? (no IEEE) + N. Miyake ? switched capacitor stuff? (IEEE nnnnnn) + + + The uPD1771 internal workings are described to some extent by the Bristow book + and the IEEE article "A Single-Chip Sound Synthesis Microcomputer" which complements the book + and are covered by at least four US patents: + 4184152 - on IGFET-based DAC stuff + 4488061 - on the IGFET-based drive circuit part of the DAC. + 4408094 - covers the 3 pin DAC with the volume control/vref pin. Not all that interesting, + except it might describe to some extent how the (9->5bit?) PWM works in the text. + 4470113 - covers the multiplexed PB0/1/2/3 pins and their use as /CS /WR /RD and ALE + note as I have marked the pins below I assume the final pins connected + to /CS /WR /RD and /ALE are PB7,6,5,4 but this is just a guess of mine: + The actual order may well match the patent. + 4577343 - covers the VSRSSS implementation as discussed in the Bristow book. + This patent has an internal diagram of the workings of the chips and + a limited description of how many registers etc it has. + 4805508 - on the operation of the tone divider register and correction for accurate period when + the tone interrupt frequency is not perfectly divisible from the clock. + These next two may not be specific to the 1771 or even related at all! + 4321562 - on a self-adjusting circuit for internal coupling to the clock crystal inputs. + This may be a generic NEC invention and probably isn't limited to the upd1771. + 4656491 - on a new method of distributing resistors and transistors on anti-ESD pin buffers + This may be a generic NEC invention and probably isn't limited to the upd1771. + + + Based on the 4577343 patent mostly, plus the bristow and IEEE article: + * these are the registers: + 8bits: + AH, AL (forming the 16-bit A' accumulator), + B, C (a pair of general purpose registers), + 4bits (may be technically part of ALU): + H -> points to one of the 16 words of ram + 1bit: + L -> selector of left or right half of the ram word + ?bits: + D (having to do with the DAC) + N (3 bits? having to do with the pseudorandom noise interrupt, namely setting the clock divider ratio for the PRNG clock vs cpu clock) + MODE (5 or more bits? enabling/disabling/acking the noise interrupt, and the tone interrupts (there are four!)) + SP (the stack pointer, probably 5 bits, points to the stack ram; may encompass H and L as above!) + FLO: unsure. quite possibly 'flag overflow' used for branching. there likely exists other flags as well... + ODF: 'output data flag?', selects which half of a selected ram word is output to the dac not really sure of this? + + + Mask roms known: + uPD1776C: mentioned in the bristow book, implements VSRSSS speech concatenation + (see US Patent 4577343 which is a patent on this VSRSSS implementation) + uPD1771C-006: used in NEC APC for sound as the "MPU" + -011: used on Firefox F-4/Astro Thunder handheld + -015: unknown, known to exist from part scalper sites only. + -017: used on Epoch Super Cassete Vision for sound; This audio driver HLEs that part only. + + Used pinout in the SCV: + + NC 1 28 NC + NC 2 27 NC + NC 3 26 ACK + !WR 4 25 D7 + !CS 5 24 D6 + RESET 6 23 D5 + NC 7 22 D4 + VCC 8 21 D3 + 6Mhz XIN 9 20 D2 + 6Mhz XOUT 10 19 D1 + AUDOUT 11 18 D0 + NC(recheck!)12 17 GND + AUDOUT(inv) 13 16 VCC + GND 14 15 ? tied to pin 16 (VCC) through a resistor (pullup?) + + Pinout based on guesses and information in "Electronic Speech Synthesis" by Geoff Bristow + (ISBN 0-07-007912-9, pages 148-152), and the data on page 233 of the Nec APC technical manual at + http://bitsavers.trailing-edge.com/pdf/nec/APC/819-000100-1003_APC_System_Reference_Guide_Apr83.pdf + I/O pin purpose based on testing by kevtris. + [x] is unsure: + PB3 <> 1 28 <> PB2 + PB4(/ALE) <> 2 27 <> PB1 + PB5(/RD) <> 3 26 <> PB0 + PB6(/WR) <> 4 25 <> D7(PA7) + PB7(/CS) <> 5 24 <> D6(PA6) + /RESET -> 6 23 <> D5(PA5) + [/TSTOUT?] <- 7 22 <> D4(PA4) + VCC -- 8 21 <> D3(PA3) + XI(CLK) -> 9 20 <> D2(PA2) + XO <- 10 19 <> D1(PA1) + D/A OUT + <- 11 18 <> D0(PA0) + D/A POWER -- 12 17 <- CH2 + D/A OUT - <- 13 16 <> /EXTINT and [/TSTOUT2?] (test out is related to pin 15 state) + GND -- 14 15 <- CH1 tied to pin 16 (VCC) through a resistor, on APC to VCC thru a 12k resistor and thru a 10uf cap to gnd + + CH1 and CH2 are mode selects, purpose based on testing by kevtris: + CH1 CH2 + H L - 'master' mode, pb4-pb7 are i/o? /EXTINT is an input + L L - 'slave' mode where pb4-pb7 are /ale /rd /wr /cs? /EXTINT may be an output in this mode? + X H - test mode: the 512 byte 16-bit wide ROM is output sequentially on pins pa7-pa0 and pb7-pb0 for the high and low bytes of each word respectively + D/A POWER is the FET source for the D/A OUT+ and D/A OUT- push-pull dac drains; it should be tied to VCC or thereabouts + + In the SCV (info from plgDavid): + pin 5 is tied to the !SCPU pin on the Epoch TV chip pin 29 (0x3600 writes) + pin 6 is tied to the PC3 pin of the upD7801 CPU + pin 26 is tied to the INT1 pin of the upD7801 (CPU pin 12), + + 1,2,3,28,27 don't generate any digital signals + 6 seems to be lowered 2.5 ms before an audio write + 7 is always low. + 12 is always high + + (NOTE: the photomicrograph in the bristow book makes it fairly clear due to + pad thicknessess that the real VCC is pin 8 and the real GND is pin 14. + The function of pin 7 is unknown. + + Pins 11 and 13 go to a special circuit, which according to kevtris's analysis + of my schematics, consist of a balanced output (not unlike XLR cables), + which are then combined together then sent to the RF box. + (The bristow book explains that there are two DAC pins and one DAC + VREF/volume pin. The dac+ and dac- are pins 11 and 13, and based on the + photomicrograph it looks like dac vref is probably pin 12) + + HLE: + All writes are made through address 0x3600 on the upD7801 + Instead of using register=value, this chip require sending multiple + bytes for each command, one after the other. + +**********************************************************************/ + +#include "emu.h" +#include "audio/upd1771.h" + + +#define LOG 0 + +/* + Each of the 8 waveforms have been extracted from the uPD1771c-017 internal + ROM, from offset 0x1fd (start of first waveform) to offset 0x2fc (end of + last waveform). + (note: given test mode dumping offset non-clarity it may be 0x200-0x2ff) + The waveforms are stored in an 8-bit sign-magnitude format, so if in ROM the + upper bit is 0x80, invert the lower 7 bits to get the 2's complement result + seen here. + Note that only the last 4 waveforms appear to have been intended for use as + waveforms; the first four look as if they're playing back a piece of code as + wave data. +*/ +const char WAVEFORMS[8][32]={ +{ 0, 0,-123,-123, -61, -23, 125, 107, 94, 83,-128,-128,-128, 0, 0, 0, 1, 1, 1, 1, 0, 0, 0,-128,-128,-128, 0, 0, 0, 0, 0, 0}, +{ 37, 16, 32, -21, 32, 52, 4, 4, 33, 18, 60, 56, 0, 8, 5, 16, 65, 19, 69, 16, -2, 19, 37, 16, 97, 19, 0, 87, 127, -3, 1, 2}, +{ 0, 8, 1, 52, 4, 0, 0, 77, 81,-109, 47, 97, -83,-109, 38, 97, 0, 52, 4, 0, 1, 4, 1, 22, 2, -46, 33, 97, 0, 8, -85, -99}, +{ 47, 97, 40, 97, -3, 25, 64, 17, 0, 52, 12, 5, 12, 5, 12, 5, 12, 5, 12, 5, 8, 4,-114, 19, 0, 52,-122, 21, 2, 5, 0, 8}, +{ -52, -96,-118,-128,-111, -74, -37, -5, 31, 62, 89, 112, 127, 125, 115, 93, 57, 23, 0, -16, -8, 15, 37, 54, 65, 70, 62, 54, 43, 31, 19, 0}, +{ -81,-128, -61, 13, 65, 93, 127, 47, 41, 44, 52, 55, 56, 58, 58, 34, 0, 68, 76, 72, 61, 108, 55, 29, 32, 39, 43, 49, 50, 51, 51, 0}, +{ -21, -45, -67, -88,-105,-114,-122,-128,-123,-116,-103, -87, -70, -53, -28, -9, 22, 46, 67, 86, 102, 114, 123, 125, 127, 117, 104, 91, 72, 51, 28, 0}, +{ -78,-118,-128,-102, -54, -3, 40, 65, 84, 88, 84, 80, 82, 88, 94, 103, 110, 119, 122, 125, 122, 122, 121, 123, 125, 126, 127, 127, 125, 118, 82, 0} +}; + + +#define NOISE_SIZE 255 + + +static unsigned char noise_tbl[]= +{ + 0x1c,0x86,0x8a,0x8f,0x98,0xa1,0xad,0xbe,0xd9,0x8a,0x66,0x4d,0x40,0x33,0x2b,0x23, + 0x1e,0x8a,0x90,0x97,0xa4,0xae,0xb8,0xd6,0xec,0xe9,0x69,0x4a,0x3e,0x34,0x2d,0x27, + 0x24,0x24,0x89,0x8e,0x93,0x9c,0xa5,0xb0,0xc1,0xdd,0x40,0x36,0x30,0x29,0x27,0x24, + 0x8b,0x90,0x96,0x9e,0xa7,0xb3,0xc4,0xe1,0x25,0x21,0x8a,0x8f,0x93,0x9d,0xa5,0xb2, + 0xc2,0xdd,0xdd,0x98,0xa2,0xaf,0xbf,0xd8,0xfd,0x65,0x4a,0x3c,0x31,0x2b,0x24,0x22, + 0x1e,0x87,0x8c,0x91,0x9a,0xa3,0xaf,0xc0,0xdb,0xbe,0xd9,0x8c,0x66,0x4d,0x40,0x34, + 0x2c,0x24,0x1f,0x88,0x90,0x9a,0xa4,0xb2,0xc2,0xda,0xff,0x67,0x4d,0x3d,0x34,0x2d, + 0x26,0x24,0x20,0x89,0x8e,0x93,0x9c,0xa5,0xb1,0xc2,0xde,0xc1,0xda,0xff,0x67,0x4d, + 0x3d,0x33,0x2d,0x26,0x24,0x20,0x89,0x8e,0x93,0x9c,0xa5,0xb1,0xc2,0xdd,0xa3,0xb0, + 0xc0,0xd9,0xfe,0x66,0x4b,0x3c,0x32,0x2b,0x24,0x23,0x1e,0x88,0x8d,0x92,0x9b,0xa4, + 0xb0,0xc1,0xdc,0xad,0xbe,0xda,0x22,0x20,0x1c,0x85,0x8a,0x8f,0x98,0xa1,0xad,0xbe, + 0xda,0x20,0x1b,0x85,0x8d,0x97,0xa1,0xaf,0xbf,0xd8,0xfd,0x64,0x49,0x3a,0x30,0x2a, + 0x23,0x21,0x1d,0x86,0x8b,0x91,0x9a,0xa2,0xae,0xc0,0xdb,0x33,0x2b,0x24,0x1f,0x88, + 0x90,0x9a,0xa4,0xb2,0xc2,0xda,0xff,0x67,0x4c,0x3e,0x33,0x2d,0x25,0x24,0x1f,0x89, + 0x8e,0x93,0x9c,0xa5,0xb1,0xc2,0xde,0x85,0x8e,0x98,0xa2,0xb0,0xc0,0xd9,0xfe,0x64, + 0x4b,0x3b,0x31,0x2a,0x23,0x22,0x1e,0x88,0x8c,0x91,0x9b,0xa3,0xaf,0xc1,0xdc,0xdc +}; + + + +#define STATE_SILENCE 0 +#define STATE_NOISE 1 +#define STATE_TONE 2 +#define STATE_ADPCM 3 + + +const device_type UPD1771C = &device_creator<upd1771c_device>; + + +upd1771c_device::upd1771c_device(const machine_config &mconfig, const char *tag, device_t *owner, UINT32 clock) + : device_t(mconfig, UPD1771C, "NEC uPD1771C 017", tag, owner, clock, "upd1771c", __FILE__), + device_sound_interface(mconfig, *this), + m_ack_handler(*this) +{ +} + +//------------------------------------------------- +// device_start - device-specific startup +//------------------------------------------------- + +void upd1771c_device::device_start() +{ + /* resolve callbacks */ + m_ack_handler.resolve_safe(); + + m_timer = machine().scheduler().timer_alloc(timer_expired_delegate(FUNC(upd1771c_device::ack_callback),this)); + + m_channel = machine().sound().stream_alloc(*this, 0, 1, clock() / 4); + + save_item(NAME(m_packet)); + save_item(NAME(m_index)); + save_item(NAME(m_expected_bytes)); + save_item(NAME(m_state)); + save_item(NAME(m_pc3)); + save_item(NAME(m_t_timbre)); + save_item(NAME(m_t_offset)); + save_item(NAME(m_t_period)); + save_item(NAME(m_t_volume)); + save_item(NAME(m_t_tpos)); + save_item(NAME(m_t_ppos)); + save_item(NAME(m_nw_timbre)); + save_item(NAME(m_nw_volume)); + save_item(NAME(m_nw_period)); + save_item(NAME(m_nw_tpos)); + save_item(NAME(m_nw_ppos)); + save_item(NAME(m_n_value)); + save_item(NAME(m_n_volume)); + save_item(NAME(m_n_period)); + save_item(NAME(m_n_ppos)); +} + +//------------------------------------------------- +// device_reset - device-specific reset +//------------------------------------------------- + +void upd1771c_device::device_reset() +{ + m_index = 0; + m_expected_bytes = 0; + m_pc3 = 0; + m_t_tpos = 0; + m_t_ppos = 0; + m_state = 0; + m_nw_tpos = 0; + memset(m_n_value, 0x00, sizeof(m_n_value)); + memset(m_n_ppos, 0x00, sizeof(m_n_ppos)); +} + + + +READ8_MEMBER( upd1771c_device::read ) +{ + return 0x80; // TODO +} + +/* +*************TONE***************** +Tone consists of a wavetable playback mechanism. +Each wavetable is a looping period of 32 samples but can be played with an offset from any point in the table +effectively shrinking the sample loop, thus allowing different pitch "macros ranges" to be played. +This method is rather crude because the spectrum of the sound get heavily altered... +unless that was the intent. + +Tone Write (4 bytes): + +Byte0: 0x02 + +Byte1: 0bTTTOOOOO + MSB 3 bits of Timbre (8 wavetables) + LSB 5 bits offset in the table. + +Byte2: 0bPPPPPPPP + 8bits of clock divider/period + Anything under <= 0x20 give the same value + +Byte3: 0b???VVVVV + MSB 3 bits unknown + LSB 5 bits of "Volume" + + Note: volume is not a volume in a normal sense but some kind + of bit cropping/rounding. +*/ + + +/* +*************NOISE***************** +Noise consists on 4 different components +A weird Wavetable LFSR (for lack of a better term), +and three independent (but equal) low frequency +50/50 pulse wavs. + +The 4 components are mixed in a mysterious way, +a weird ORing with volume having a huge effect. + +Byte0: 0x01 + +Byte1: 0bTTTOOOOO + MSB 3 bits of LFSR Timbre (8 wavetables) + LSB 5 bits ????????? + +Byte2: 0bPPPPPPPP + 8bits of clock divider/period + +Byte3: 0b???VVVVV + MSB 3 bits unknown + LSB 5 bits of "Volume" + + +Byte4: 0bPPPPPPPP Low Freq0 period(if not 0 this periodically resets the Wavetable LFSR) +Byte5: 0bPPPPPPPP Low Freq1 period(if not 0 this periodically resets the Wavetable LFSR) +Byte6: 0bPPPPPPPP Low Freq2 period(if not 0 this periodically resets the Wavetable LFSR) + +Byte7: 0b???VVVVV Low Freq0 volume +Byte8: 0b???VVVVV Low Freq1 volume +Byte9: 0b???VVVVV Low Freq2 volume +*/ + +WRITE8_MEMBER( upd1771c_device::write ) +{ + //if (LOG) + // logerror( "upd1771_w: received byte 0x%02x\n", data ); + + m_ack_handler(0); + + if (m_index < MAX_PACKET_SIZE) + m_packet[m_index++] = data; + else + { + logerror("upd1771_w: received byte 0x%02x overload!\n", data); + return; + } + + switch (m_packet[0]) + { + case 0: + m_state = STATE_SILENCE; + m_index = 0; + //logerror( "upd1771_w: ----------------silence state reset\n"); + break; + + case 1: + if (m_index == 10) + { + m_state = STATE_NOISE; + m_index = 0; + + m_nw_timbre = (m_packet[1] & 0xe0) >> 5; + m_nw_period = ((UINT32)m_packet[2] + 1) << 7; + m_nw_volume = m_packet[3] & 0x1f; + + //very long clocked periods.. used for engine drones + m_n_period[0] = (((UINT32)m_packet[4]) + 1) << 7; + m_n_period[1] = (((UINT32)m_packet[5]) + 1) << 7; + m_n_period[2] = (((UINT32)m_packet[6]) + 1) << 7; + + m_n_volume[0] = m_packet[7] & 0x1f; + m_n_volume[1] = m_packet[8] & 0x1f; + m_n_volume[2] = m_packet[9] & 0x1f; + + //logerror( "upd1771_w: ----------------noise state reset\n"); + } + else + m_timer->adjust(attotime::from_ticks(512, clock())); + break; + + case 2: + if (m_index == 4) + { + //logerror( "upd1771_w: ----------------tone state reset\n"); + m_t_timbre = (m_packet[1] & 0xe0) >> 5; + m_t_offset = (m_packet[1] & 0x1f); + m_t_period = m_packet[2]; + //smaller periods don't all equal to 0x20 + if (m_t_period < 0x20) + m_t_period = 0x20; + + m_t_volume = m_packet[3] & 0x1f; + m_state = STATE_TONE; + m_index = 0; + } + else + m_timer->adjust(attotime::from_ticks(512, clock())); + break; + + case 0x1f: + //6Khz(ish) DIGI playback + + //end capture + if (m_index >= 2 && m_packet[m_index - 2] == 0xfe && m_packet[m_index - 1] == 0x00) + { + //TODO play capture! + m_index = 0; + m_packet[0] = 0; + m_state = STATE_ADPCM; + } + else + m_timer->adjust(attotime::from_ticks(512, clock())); + break; + + //garbage: wipe stack + default: + m_state = STATE_SILENCE; + m_index = 0; + break; + } +} + + +WRITE_LINE_MEMBER( upd1771c_device::pcm_write ) +{ + //RESET upon HIGH + if (state != m_pc3) + { + logerror("upd1771_pc3 change!: state = %d\n", state); + m_index = 0; + m_packet[0] = 0; + } + + m_pc3 = state; +} + + +TIMER_CALLBACK_MEMBER( upd1771c_device::ack_callback ) +{ + m_ack_handler(1); +} + + +//------------------------------------------------- +// sound_stream_update - handle a stream update +//------------------------------------------------- + +void upd1771c_device::sound_stream_update(sound_stream &stream, stream_sample_t **inputs, stream_sample_t **outputs, int samples) +{ + stream_sample_t *buffer = outputs[0]; + + switch (m_state) + { + case STATE_TONE: + //logerror( "upd1771_STATE_TONE samps:%d %d %d %d %d %d\n",(int)samples, + // (int)m_t_timbre,(int)m_t_offset,(int)m_t_volume,(int)m_t_period,(int)m_t_tpos); + + while (--samples >= 0) + { + *buffer++ = (WAVEFORMS[m_t_timbre][m_t_tpos]) * m_t_volume * 2; + + m_t_ppos++; + if (m_t_ppos >= m_t_period) + { + m_t_tpos++; + if (m_t_tpos == 32) + m_t_tpos = m_t_offset; + + m_t_ppos = 0; + } + } + break; + + case STATE_NOISE: + while (--samples >= 0) + { + *buffer = 0; + + //"wavetable-LFSR" component + int wlfsr_val = ((int)noise_tbl[m_nw_tpos]) - 127;//data too wide + + m_nw_ppos++; + if (m_nw_ppos >= m_nw_period) + { + m_nw_tpos++; + if (m_nw_tpos == NOISE_SIZE) + m_nw_tpos = 0; + m_nw_ppos = 0; + } + + //mix in each of the noise's 3 pulse components + char res[3]; + for (int i = 0; i < 3; ++i) + { + res[i] = m_n_value[i] * 127; + m_n_ppos[i]++; + if (m_n_ppos[i] >= m_n_period[i]) + { + m_n_ppos[i] = 0; + m_n_value[i] = !m_n_value[i]; + } + } + //not quite, but close. + *buffer+= ( + (wlfsr_val * m_nw_volume) | + (res[0] * m_n_volume[0]) | + (res[1] * m_n_volume[1]) | + (res[2] * m_n_volume[2]) + ) ; + + buffer++; + } + break; + + default: + //fill buffer with silence + while (--samples >= 0) + { + *buffer++ = 0; + } + break; + } +} |