 |
|
| If you can't view the Datasheet, Please click here to try to view without PDF Reader . |
|
|
|
| If you can't view the Datasheet, Please click here to try to view without PDF Reader . |
|
|
| Datasheet File OCR Text: |
| RC SYSTEMS RC8650 VOICE SYNTHESIZER DoubleTalk RC8650 CMOS, 3.3 Volt / 5 Volt Voice Synthesizer Chipset The RC8650 is a highly versatile voice and sound synthesizer, integrating a text-to-speech (TTS) processor, real time and prerecorded audio playback, musical and sinusoidal tone generators, telephone dialer and A/D converter, into an easy to use chipset. Using a standard serial or eight bit bus interface, virtually any ASCII text can be streamed to the RC8650 for automatic conversion into speech by the TTS processor. The audio playback modes augment the TTS processor for applications requiring very high voice quality and a relatively small, fixed vocabulary, or applications requiring special sounds or sound effects. The audio output is delivered in both analog and digital PCM audio formats, which can be used to drive a speaker or digital audio stream. The RC8650's integrated TTS processor incorporates RC Systems' DoubleTalkTM TTS technology, which is based on a patented voice concatenation technique using real human voice samples. The DoubleTalk TTS processor also gives the user unprecedented real-time control of the speech signal, including pitch, volume, tone, speed, expression, articulation, and so on. Up to 3.5 MB of nonvolatile memory is included in the RC8650 for the storage and on-demand playback of up to 15 minutes of prerecorded messages and sound effects. A programmable "greeting" message can be stored that is automatically played whenever the RC8650 is powered up, allowing a custom message to be played or the RC8650's default settings to be reconfigured. A user-programmable dictionary allows the pronunciation of virtually any character string to be rede- GENERAL DESCRIPTION fined, or even trigger the playback of tones, prerecorded messages and sounds based on specific input patterns. All of these features can be programmed and updated via a standard serial port, even in the field after the RC8650 has been integrated into the end-product. The RC8650 is comprised of two surface-mounted devices. Both operate from a +3.3 V or +5 V supply and consume very little power. Most applications require only the addition of a lowpass filter/audio power amplifier to implement a fully functional system. RC8650 FUNCTIONAL BLOCK DIAGRAM DoubleTalk RC8650 User's Manual Rev 2G Revised 06/16/03 1 (c) 1999-2003 RC Systems, Incorporated RC SYSTEMS * Integrated text-to-speech processor: - - - - - High voice quality, unlimited vocabulary Converts any ASCII text into speech automatically Capable of very high reading rates Add/modify messages by simply editing a text file On-the-fly control of speed, pitch, volume, etc. RC8650 VOICE SYNTHESIZER * * * * * * * * * * * * * FEATURES APPLICATIONS Robotics Talking OCR systems ATM machines Talking pagers and PDAs GPS navigation systems Vending and ticketing machines Remote diagnostic reporting Dial-up information systems Handheld barcode readers Electronic test and measurement Security systems Aids for the orally or visually disabled Meeting federal ADA requirements * Playback of sound files: - Real-time PCM and ADPCM - Prerecorded on chip, up to 15 minutes * Tone generation: - Three voice musical - Dual sinusoidal - DTMF (Touch-Tone) dialer * On-chip A/D converter: - Four channels, 8-bit resolution - One-shot, continuous, single sweep, and continuous sweep modes of operation - Software and hardware triggering - Support for external op amp * * * * * Analog and digital audio outputs Stop, pause, and resume controls Serial and bus interfaces User programmable greeting and default settings Flexible user exception dictionary: - Change the pronunciation of any input string based on spelling and context - Convert encrypted data into meaningful messages - Trigger tone generation, recorded message playback, voice parameter changes In-circuit, field programmable 2 KB input buffer for virtually no-overhead operation Available in 3.3 V and 5 V versions Low power (typ @ 3.3 V): - 8.8 mA active - 700 A idle - 2 A standby RC8650 Product Summary * * * * * Based on 8 kHz sampling rate with ADPCM encoding 2 RC SYSTEMS TYPICAL APPLICATION CIRCUIT RC8650 VOICE SYNTHESIZER 3 RC SYSTEMS RC8650 VOICE SYNTHESIZER SECTION 1: SPECIFICATIONS PINOUTS Figure 1.1. Pin Assignments 4 RC SYSTEMS PIN DESCRIPTIONS RC8650 VOICE SYNTHESIZER Table 1.1. Pin Descriptions 5 RC SYSTEMS RC8650 VOICE SYNTHESIZER Table 1.1. Pin Descriptions (Continued) 6 RC SYSTEMS RC8650 VOICE SYNTHESIZER Table 1.1. Pin Descriptions (Continued) 7 RC SYSTEMS The RC8650 chipset includes a number of features that make it ideally suited for any design requiring voice output. The RC8650's major features are described below. RC8650 VOICE SYNTHESIZER All data is sent to the RC8650 through its built in serial and/or parallel ports. For maximum flexibility, including infield product update capability, use of the serial port is recommended whenever possible. The RC8650's audio output is available in both analog and digital formats. The analog output should be used in applications where no further processing of the audio signal is required, such as driving a speaker or headphones (the output still needs to be filtered and amplified, however). The digital output is for applications that require further processing of the audio signal, such as digital mixing or creating sound files for later playback. FUNCTIONAL DESCRIPTION Versatile I/O Text-to-Speech Synthesizer The RC8650 provides text-to-speech conversion with its integrated DoubleTalkTM text-to-speech synthesizer. Any English text written to the RC8650 is automatically converted into speech. Commands can be embedded in the input stream to dynamically control the voice, even at the phoneme level (phonemes are the basic sound units of speech). A greeting message can be stored in the RC8650 that is automatically spoken immediately after the RC8650 is reset. Most any of the commands recognized by the RC8650 may be included as part of the greeting message, which can be used to set up custom default settings and/or play a prerecorded message or tone sequence. An integrated nonvolatile memory area is also provided for storing a custom pronunciation dictionary, allowing the pronunciation of any character string to be redefined. RECOMMENDED CONNECTIONS Power and ground connections are made to multiple pins of the RC8650 and RC46xx chips. Every VCC pin must be connected to power, and every VSS pin must be connected to ground. To minimize noise, the analog and digital circuits in the RC8650 use separate power busses. These busses are brought out to separate pins and should be tied to the supply as close as possible. Make sure adequate decoupling is placed on the AVREF pin, as noise present on this pin will also appear on the AO output pins and affect A/D converter accuracy. In systems where the power supply is very quiet, AVREF can be connected directly to VCC. Designs incorporating a switching power supply, or supplies carrying heavy loads, may require filtering at the AVREF pin; a 150 series VCC resistor in combination with a 100 F capacitor to ground should suffice. Connect any unused input pins to an appropriate signal level (see Table 1.1). Leave any unused output pins and all NC pins unconnected. Power/Ground An integrated, three-voice musical tone generator is capable of generating up to three tones simultaneously over a four-octave range. Simple tones to attention-getting sounds can be easily created. Musical Tone Generator Touch-Tone Generator The RC8650 includes an integrated DTMF (Touch-Tone) generator. This is useful in telephony applications where standard DTMF tones are used to signal a remote receiver, modem, or access the public switched telephone network. Sinusoidal Tone Generator A precision, dual sinusoidal tone generator can synthesize the tones often used in signaling applications. The tone frequencies can be independently set, allowing signals such as call-progress tones to be generated. Pins IC0 through IC32 and PIO0 through PIO7 must be connected between the RC8650 and RC46xx chips. IC30, IC31, and IC32 must have 47 k - 100 k pullup resistors to VCC. Chip Interconnects Recorded Audio Playback Up to 15 minutes of prerecorded messages and sound effects can be stored in the RC8650 for on-demand playback. Recordings are stored in on-chip nonvolatile memory, providing zero-power message storage. Additionally, the RC8650 can play eight-bit PCM and ADPCM audio in real time, such as speech and/or sound effects stored in an external memory or file system. The RC8650 has an internal oscillator and clock generator that can be controlled by an external 7.3728 MHz crystal, ceramic resonator, or external 7.3728 MHz clock source. If an external clock is used, connect it to the XIN pin and leave XOUT unconnected. See Figure 1.2 for recommended clock connections. Clock Generator Analog-to-Digital Converter The four channel, 8-bit A/D converter can be used to monitor battery cell voltages, temperature, and other analog quantities. The ADC can be programmed on the fly to convert any single channel, or scan up to four channels repetitively. Figure 1.2. Clock Connections 8 RC SYSTEMS INTERFACING THE RC8650 The RC8650 contains both asynchronous serial and 8 bit bus interfaces. All text, commands, tone generator data, real time audio data, etc., are transmitted to the RC8650 via one of these ports. For maximum flexibility, use of the serial port is recommended whenever possible. Not all RC8650 functions are supported through the bus interface. In particular, index markers, operating system updates, chipset identification, current operating settings, and A/D conversion are only supported through the serial interface. RC8650 VOICE SYNTHESIZER Table 1.2. Baud Rate Options Serial Interface The serial port operates with 8 data bits (LSB first), 1 or more stop bits, no parity, and any standard baud rate between 300 and 115200 bps. A typical RS-232C interface is shown in Figure 1.3. Note that the MAX232A transceiver is not required if the host system's serial port operates at logic levels compatible with the RC8650 (0/+5 or 0/+3.3 V). The RC8650's serial port may be connected directly to the host system in this case. The CTS# pin should be used to control the flow of serial data to the RC8650. It is not necessary to check CTS# before transmitting every byte, however. All data is routed through a high speed 16-byte buffer within the RC8650 before being stored in the primary buffer. CTS# may be checked every eight bytes with no risk of data loss. The serial port's baud rate can be programmed using any of three methods: pin strapping, auto-detect, and by command. Pin strapping sets the baud rate according to the logic levels present on the BRS0- BRS2 pins, as shown in Table 1.2. Auto-detect enables the serial port to automatically detect the baud rate of the incoming data. The baud rate command (described in Section 2) allows the baud rate to be changed at any time, effectively overriding the first two methods. Pin strapping cannot be used to program baud rates higher than 19200; to do this, auto-detection or the baud rate command must be used. The automatic baud rate detection mechanism is enabled when the BRS0-BRS2 pins are all at a High logic level and the BRD pin is connected to the RXD pin. The baud rate is determined by the shortest High or Low period detected in the input stream. This period is assumed to be the bit rate of the incoming data. In addition to the baud rates listed in Table 1.2, auto-detect mode also supports 38400, 57600, and 115200 baud rates. In order for the RC8650 to determine the incoming baud rate, there must be at least one isolated "1" or "0" in the input character. The CR character, 0Dh, is recommended for locking the baud rate. The character is not otherwise processed by the RC8650; it is discarded. If the measured bit period is determined to be a valid baud rate, the RC8650 acknowledges lock acquisition by transmitting the ASCII character "l" (6Ch) on the TXD pin. The baud rate will remain locked unless changed with the baud rate command, or the RC8650 is reset. Baud rate selection Figure 1.4. Baud Rate Detection Timing Figure 1.3. RS-232C Interface 9 RC SYSTEMS Note The measurement cycle ends when there have been no Highto-Low nor Low-to-High transitions on the BRD pin for 75 ms or longer. Consequently, the RC8650 will ignore any data sent to it for a period of 75 ms after the "lock-on" character has been received. The CTS# pin is driven High during this time, and the acknowledgment character is not transmitted until the RC8650 is actually ready to accept data. See Figure 1.4. Real-time status information is provided via the TXD pin. Status are transmitted as one-byte messages, shown in Table 1.3. Each message correlates to a status flag in the Status Register, shown in Table 1.4. The specific character used, and whether it will be transmitted, are functions of the VC and STM bits of the Protocol Options Register. (The Protocol Options Register is described in Section 2.) For information about how to obtain reading-progress status, see the Index Marker command description. RC8650 VOICE SYNTHESIZER Because the RC8650 can take up to 15 s to accept data written to it (AC Characteristics, tYHWH parameter), software drivers should wait for RDY to drop to 0 after a byte is written in order to avoid overwriting it with the next data byte. Not doing so could result in the loss of data. Waiting for RDY to drop to 0 ensures that RDY will not falsely show that the RC8650 is ready the next time the driver is called. If a system interrupt can occur while waiting for RDY to become 0, or if RDY cannot otherwise be checked at least once every 8 s, a software timeout should be enforced to avoid hanging up in the wait loop. The time RDY stays 0 is relatively short (8 s min.) and can be missed if interrupted. The timeout should be at least 15 s, which is the maximum time for RDY to drop to 0 after writing a byte of data. In non time-critical applications, the output routine could simply delay 15 s or longer before exiting, without checking for RDY = 0 at all. Figure 1.5 illustrates the recommended method of writing data to the RC8650's bus interface. This method should be used for writing all types of data, including text, commands, tone generator and real time audio data. Status messages Table 1.3. Status Messages Bus/Printer Interface The RC8650's bus interface allows the RC8650 to be connected to a microprocessor or microcontroller in the same manner as a static RAM or I/O device, as shown in Figure 1.6. The microprocessor controls all transactions with the RC8650 over the system data bus using the RD and WR# signals. RD controls the reading of the RC8650's Status Register; WR# controls the transfer of data into the RC8650. The Status Register bits and their definitions are shown in Table 1.4. A registered bus transceiver is required for communication between the RC8650 and microprocessor; two 74HCT374s placed back to back may be substituted for the 74HCT652 shown in the figure. Prior to each write operation to the RC8650, the host processor should verify that the RC8650 is ready by testing the RDY status flag. The RC8650 can also be interfaced to a PC's printer port as shown in Figure 1.6. A 74HCT374 can be used in place of the 74HCT652, since bidirectional communication is not necessary. Handshaking is performed automatically via the BUSY pin. Figure 1.5. Recommended Method of Writing Data Via the Bus Interface 10 RC SYSTEMS RC8650 VOICE SYNTHESIZER Table 1.4. Bus Interface Status Register Bit Definitions Figure 1.6. Bus/Printer Interface 11 RC SYSTEMS RC8650 VOICE SYNTHESIZER Figure 1.7. Method of Capturing Status Information for Driving External Circuitry The analog output pins AO0 and AO1 are high impedance (10 k typical) outputs from the RC8650's internal D/A converters. When using these outputs, the addition of an external low-pass filter is highly recommended. When laying out the printed circuit board, avoid running digital lines near the AO lines in order to minimize induced noise in the audio path. If space permits, run a guard ground next to the AO traces. The circuit shown in Figure 1.8 is a low-pass filter/power amplifier capable of delivering 1.1 W to an 8 load, when operating from a +5 V power supply (power output will be less when operating from +3.3 V). Analog Audio Output The amplifier's shutdown pin can be controlled by the TS0 pin to minimize current drain when the RC8650 is inactive. The digital audio pin DAOUT outputs the RC8650's audio signal as a digital audio stream consisting of 8 data bits per sample. The normalized sampling rate for all text to speech modes and the DTMF generator is 84 kbs (10,500 bytes/sec). The sinusoidal generator, prerecorded and real time audio playback mode rates are user programmable, so their normalized rates will vary. See the Pin Descriptions and Audio Control Register command description for further details. Digital Audio Output Figure 1.8. 3 kHz Low-Pass Filter/Power Amplifier 12 RC SYSTEMS ELECTRICAL SPECIFICATIONS RC8650 VOICE SYNTHESIZER Figure 1.9. Test Circuit Supply voltage, VCC and AVCC . . . . . . . . . . . . . . . . . . -0.3 V to +6.5 V DC input voltage, VI . . . . . . . . . . . . . . . . . . . . . . . -0.3 V to VCC +0.3 V Operating temperature, TA . . . . . . . . . . . . . . . . . . . . . . . 0 C to +70 C Storage temperature, TS . . . . . . . . . . . . . . . . . . . . . -55 C to +125 C ABSOLUTE MAXIMUM RATINGS* * WARNING: Stresses greater than those listed under "Absolute Maximum Ratings" may cause permanent damage to the device. This is a stress rating only; operation of the device at any condition above those indicated in the operational sections of these specifications is not implied. Exposure to absolute maximum rating conditions for extended periods may affect device reliability. 13 RC SYSTEMS DC CHARACTERISTICS RC8650 VOICE SYNTHESIZER TA = 0 C to +70 C, VCC = AVCC = AVREF = 3.3 V / 5 V, VSS = AVSS = 0 V, XIN = 7.3728 MHz TA = 0 C to +70 C, VCC = AVCC = AVREF = 3.3 V / 5 V, VSS = AVSS = 0 V AC CHARACTERISTICS External Clock Input Timing Figure 1.10. External Clock Waveform 14 RC SYSTEMS Bus Interface Timing RC8650 VOICE SYNTHESIZER Figure 1.11. Bus Interface Waveforms 15 RC SYSTEMS Analog Audio Timing RC8650 VOICE SYNTHESIZER Figure 1.12. Analog Audio Waveforms Digital Audio Timing Figure 1.13. Digital Audio Waveforms 16 RC SYSTEMS Standby Timing RC8650 VOICE SYNTHESIZER Figure 1.14. Standby Waveform Reset Timing Figure 1.15. Reset Waveform 17 RC SYSTEMS PACKAGE INFORMATION RC8650 VOICE SYNTHESIZER 100 Pin Plastic 14 x 20 mm QFP (measured in millimeters) 18 RC SYSTEMS RC8650 VOICE SYNTHESIZER 48 Pin Plastic 12 x 20 mm TSOP (measured in millimeters) Recommended PCB Layouts (measured in millimeters) 19 RC SYSTEMS ORDERING INFORMATION RC8650 VOICE SYNTHESIZER The RC8650 is available in several audio capacity and voltage ranges. The ordering part number is formed by combining several fields, as indicated below. Refer to the "Valid Combinations" table, which lists the configurations that are planned to be supported in volume. All configurations include the RC8650AFP chip; the companion chip is shown in parentheses. For example, the RC8650-1, a 5 V part with 130 seconds of recordable audio memory, is composed of the RC8650AFP and RC4651FP. 20 RC SYSTEMS RC8650 VOICE SYNTHESIZER SECTION 2: PRINCIPLES OF OPERATION This section describes the operating characteristics of the DoubleTalk RC8650 chipset. Because the RC8650 must handle the highly irregular spelling system of English, as well as proper names, acronyms, technical terms, and borrowed foreign words, there inevitably will be words that it will mispronounce. If a word is mispronounced, there are three techniques for correcting it: 1. Spell the word phonetically for the desired pronunciation. 2. Redefine the way the word should be pronounced by creating an exception for it in the RC8650's exception dictionary. This method allows words to be corrected without having to modify the original text, and it automatically corrects all instances of the word. Exception dictionaries are covered in detail in Section 4. 3. Use the RC8650's Phoneme mode. The first technique is the easiest way to fine tune word pronunciations--by tricking the RC8650 into the desired pronunciation. Among the more commonly mispronounced words are compound words (baseball), proper names (Sean), and foreign loan words (chauffeur). Compound words can usually be corrected by separating the two words with a space, so that "baseball" becomes "base ball." Proper names and foreign words may require a bit more creativity, so that "Sean" becomes "Shon," and "chauffeur" becomes "show fur." Heteronyms (words with identical spelling but different meanings and pronunciations) can also be modified using this technique. For example, if the word read is to be pronounced "reed" instead of "red," it can simply be respelled as "reed." TRANSLATION ACCURACY The RC8650 has four primary operating modes and two inactive modes designed to achieve maximum functionality and flexibility. The operating mode can be changed anytime, even on the fly, by issuing the appropriate command to the RC8650. Note The RC8650 will not begin speaking until it receives a CR (ASCII 13) or Null (ASCII 00) character--this ensures that a complete contextual analysis can be performed on the input text. If it is not possible for the application to send a CR or Null at the end of each text message, use the Timeout Delay command. The RC8650 does not make any distinction between uppercase and lowercase characters--text and commands may be sent in any combination of uppercase and lowercase. All data sent to the RC8650 is buffered in an internal 2 KB input buffer, allowing additional text and commands to be queued even while the RC8650 is producing output. Text-to-speech mode. All text sent to the RC8650 is automatically translated into speech by the integrated DoubleTalk TTS engine. TTS mode can be further subdivided into three translation modes: Text, which reads text normally; Character, which reads (spells) one character at a time; and Phoneme, which allows the TTS engine's phonemes to be directly accessed. TTS mode is the default operating mode. Real Time Audio Playback mode. Data sent to the RC8650 is written directly to the RC8650's audio buffer. This results in a high data rate, but provides the capability of producing the highest quality speech, as well as sound effects. PCM and ADPCM data types are supported. Prerecorded Audio Playback mode. This mode allows recorded messages and sound effects that have been stored in the RC8650 to be played back. PCM and ADPCM data types are supported. Tone Generator modes. These modes activate the RC8650's musical tone generator, sinusoidal generator, or DTMF generator. They can be used to generate audible prompts, music, signaling tones, dial a telephone, etc. Idle mode. To help conserve power in battery-powered systems, the RC8650 automatically enters a reduced-power state whenever it is inactive. Data can still be read and written to the RC8650 while in this mode. Current draw is typically 700 A @ 3.3 V. Standby mode. This mode powers down the RC8650, where current draw is typically only 2 A. Standby mode can be invoked from either the STBY# pin or with the Sleep command. Data cannot be read from or written to the RC8650 in this mode. OPERATING MODES The commands described in the following pages provide a simple yet flexible means of controlling the RC8650 under software control. They can be used to vary voice attributes, such as the volume or pitch, to suit the requirements of a particular application or listener's preferences. Commands are also used to change operating modes. Commands can be freely intermixed with the text that is to be spoken, allowing the voice to be dynamically controlled. Commands affect only the data that follows them in the data stream. COMMANDS All RC8650 commands are composed of the command character, a parameter n comprised of a one to four-digit number string, and a single string literal that uniquely identifies the command. Some commands simply enable or disable a feature of the RC8650 and do not require a parameter. The general command format is: Command Syntax 21 RC SYSTEMS If two or more commands are to be used together, each must be prefaced with the command character. This is the only way the RC8650 knows to treat the remaining characters as a command, rather than text that should be spoken. For example, the following commands program pitch level 40 and volume level 7 (CTRL+A is the default command character): CTRL+A "40P" CTRL+A "7V" The default RC8650 command character is CTRL+A (ASCII code 01). The command character itself can be spoken by the RC8650 by sending it twice in a row: CTRL+A CTRL+A. This special command allows the command character to be spoken without affecting the operation of the RC8650, and without having to change to another command character and then back again. RC8650 VOICE SYNTHESIZER This section describes the software commands that affect the text-tospeech synthesizer. TTS COMMANDS The command character This command places the RC8650 in the Text operating mode. The optional delay parameter n is used to create a variable pause between words. The shortest, and default delay of 0, is used for normal speech. For users not accustomed to synthetic speech, the synthesizer's intelligibility may be improved by introducing a delay. The longest delay that can be specified is 15. If the delay parameter is omitted, the current (last set) value will be used and the exception dictionary will be disabled. This feature is useful for returning from another operating mode or disabling the exception dictionary (see Enable Exception Dictionary command). Text Mode/Delay (T/nT) Changing the command character The command character can be changed to another control character (ASCII 01-26) by sending the current command character, followed by the new character. To change the command character to CTRL+D, for example, issue the command CTRL+A CTRL+D. To change it back, issue the command CTRL+D CTRL+A. It's generally a good idea to change the command character if the text to be read contains characters which may otherwise be interpreted as command characters (and hence commands). The command character can be unconditionally reset to CTRL+A by sending CTRL+^ (ASCII 30) to the RC8650. Command parameters are composed of one to four digit number strings. The RC8650 supports two types of parameters: absolute and relative. Absolute parameters explicitly specify the parameter's new value, such as 9S or 3B. Relative parameters specify a displacement from a parameter's current value, not the actual new value itself. Relative parameters can specify either a positive or negative displacement from a parameter's current value. For example, the Volume command +2V increases the volume level by two (V+2V). If the current volume is 4, the volume will increase to 6 after the command has executed. The command -2V will have a similar effect, except the volume will be decreased by two. If the value of a parameter falls outside the command's range, the value will either wrap around or saturate, depending on the setting of the SAT bit of the Protocol Options Register. For example, if parameters are programmed to wrap, the current volume is 7 and the command +4V is issued, the resultant volume will be (7+4)-10 = 1, since the volume range is 0-9. If parameters are programmed to saturate, the resultant volume would be 9 instead. When writing application programs for the RC8650, it is recommended that relative parameters be used for temporarily changing voice attributes (such as raising the pitch of a word), using absolute-parameter commands only once in the program's initialization routine. This way, if the base value of an attribute needs to be changed, it only needs to be changed in the initialization routine. This command puts the RC8650 in the Character operating mode. The optional delay parameter n is used to create a variable pause between characters. Values between 0 (the default) and 15 provide pauses from shortest to longest, respectively. Values between 16 and 31 provide the same range of pauses, but control characters will not be spoken. If the delay parameter is omitted, the current value will be used and the exception dictionary will be disabled. Character Mode/Delay (C/nC) Command parameters This command disables the text-to-phonetics translator, allowing the RC8650's phonemes to be accessed directly. Table 2.1 lists the phonemes that can be produced by the RC8650. When concatenating two or more phonemes, each phoneme must be delimited by a space. For example, the word "computer" would be represented phonetically as K AX M P YY UW DX ER Phoneme Mode (D) Phoneme attribute tokens The RC8650 supports a number of phoneme attribute tokens that can be used in addition to the standard commands. These tokens do not require the command character or any parameters, but can only be used in Phoneme mode and exception dictionaries. As indicated in Table 2.2, the / and \ tokens temporarily increase and decrease the pitch by m steps. Besides being temporary, the difference between using the pitch tokens and the Pitch command is that the effective pitch range is extended beyond the normal 0-99 range by approximately 20 steps, and if the pitch should fall out of range, it will always saturate, regardless of the Protocol Options Register SAT setting. All other phoneme attribute token commands remain in effect until explicitly changed. 22 RC SYSTEMS Table 2.1. DoubleTalk Phoneme Symbols Phoneme Symbol A AA AE AH AW AX AY B CH D DH DX E EH EI ER EW EY F G H I IH IX IY J K KX L Example Word das (Spanish) cot cat cut cow bottom bite bib church did either city ser (Spanish) bet mesa (Spanish) bird acteur (French) bake fee gag he libro (Spanish) bit rabbit beet age cute ski long Phoneme Symbol M N NG NY O OW OY P PX R RR S SH T TH TX U UH UW V W WH Y YY Z ZH space , . Example Word me new rung nino (Spanish) no (Spanish) boat boy pop spot ring tres (Spanish) sell shell tin thin stick uno (Spanish) book boot valve we when mayo (Spanish) you zoo vision variable pause * medium pause long pause RC8650 VOICE SYNTHESIZER to change the stress or emphasis of specific words in a phrase. This is because Phoneme mode allows voice attributes to be modified on phoneme boundaries within each word, whereas Text mode allows changes only at word boundaries. This is illustrated in the following examples. CTRL+A "d" CTRL+A "m" "//h aw +<\\yy uw s p \iy k t uw t -\w ey .+/" -/d>/eh r \m iy dh ae Note that expression is disabled in this example, since the pitch variations due to the internal intonation algorithms would otherwise interfere with the pitch tokens. Compare this with the same phrase produced in Text mode with expression enabled: CTRL+A "t" CTRL+A "e" "How dare you speak to me that way!" Phoneme mode is also useful in applications that provide their own text-to-phoneme translation, such as the front end of a custom textto-speech system. Speed (nS) The synthesizer's speech rate can be adjusted with this command, from 0S (slowest) through 9S (fastest). The default rate is 1S (5S if the VC bit of the Protocol Options Register is set to 0). Voice (nO) The text-to-speech synthesizer has eight standard voices and a number of individual voice parameter controls that can be used to independently vary the voice characteristics. Voices are selected with the commands 0O through 7O, shown in Table 2.3. Because the Voice command alters numerous internal voice parameters (articulation, pitch, expression, tone, etc.), it should precede any individual voice parameter control commands. * Normally used between words; duration determined by nT command Table 2.3. Voice Presets Table 2.2. Phoneme Attribute Tokens Articulation (nA) Applications of Phoneme mode Phoneme mode is useful for creating customized speech, when the normal text-to-speech modes are inappropriate for producing the desired voice effect. For example, Phoneme mode should be used This command adjusts the articulation level, from 0A through 9A. Excessively low articulation values tend to make the voice sound slurred; very high values, on the other hand, can make the voice sound choppy. The default articulation is 5A. 23 RC SYSTEMS Expression, or intonation, is the variation of pitch within a sentence or phrase. When expression is enabled (n > 0), the RC8650 attempts to mimic the pitch patterns of human speech. For example, when a sentence ends with a period, the pitch drops at the end of the sentence; a question mark will cause the pitch to rise. The optional parameter n determines the degree of intonation. 0E provides no intonation (monotone), whereas 9E is very animated sounding. 5E is the default setting. If the parameter is omitted, the current (last set) value will be used. This is useful for re-enabling intonation after a Monotone command. RC8650 VOICE SYNTHESIZER Table 2.4. Punctuation Filter Expression (E/nE) Monotone (M) This command disables all intonation (expression), causing the RC8650 to speak in a monotonic voice. Intonation should be disabled whenever manual intonation is applied using the Pitch command or phoneme attribute tokens. Note that this command is equivalent to the 0E command. The values of n listed in Table 2.4 cause number strings to be read one digit at a time (e.g., 0123 = "zero one two three"). ORing 04h to the values listed in the table (n = 4-7) forces number strings to be read as numbers (0123 = "one hundred twenty three"). N = 6 and n = 7 also force currency strings to be read as they are normally spoken--for example, $11.95 will be read as "eleven dollars and ninety five cents." Finally, ORing 08h to these values (n = 8-15) disables leading zero suppression; number strings beginning with zero will always be read one digit at a time. The default filter setting is 6B (Some punctuation, Numbers mode, leading zero suppression enabled). Effect on number strings Formant Frequency (nF) This command adjusts the synthesizer's overall frequency response (vocal tract formant frequencies), over the range 0F through 9F. By varying the frequency, voice quality can be fine-tuned or voice type changed. The default frequency is 5F. CONTROL COMMANDS This is a global command that controls the RC8650's output volume level, from 0V through 9V. 0V yields the lowest possible volume; maximum volume is attained at 9V. The default volume is 5V. The Volume command can be used to set a new listening level, create emphasis in speech, or change the output level of the tone generators. This command varies the synthesizer's pitch over a wide range, which can be used to change the average pitch during speech production, produce manual intonation, or create sound effects (including singing). Pitch values can range from 0P through 99P; the default is 50P. Pitch (nP) Volume (nV) Tone (nX) The synthesizer supports three tone settings, bass (0X), normal (1X) and treble (2X), which work much like the bass and treble controls on a stereo. The best setting to use depends on the speaker being used and personal preference. Normal (1X) is the default setting. Timeout Delay (nY) This command is used to add reverberation to the voice. 0R (the default) introduces no reverb; increasing values of n correspondingly increase the reverb delay and effect. 9R is the maximum setting. Reverb (nR) The RC8650 defers translating the contents of its input buffer until a CR or Null is received. This ensures that text is spoken smoothly from word to word and that the proper intonation is given to the beginnings and endings of sentences. If text is sent to the RC8650 without a CR or Null, it will remain untranslated in the input buffer indefinitely. The RC8650 contains a programmable timer that is able to force the RC8650 to translate its buffer contents after a preset time interval. The timer is enabled only if the Timeout Delay parameter n is non-zero, the RC8650 is not active (not talking), and the input buffer contains no CR or Null characters. Any characters sent to the RC8650 before timeout will automatically restart the timer. The Timeout parameter n specifies the number of 200 millisecond periods in the delay time, which can range from 200 milliseconds to 3 seconds. The default value is 0Y, which disables the timer. Punctuation Filter (nB) Depending on the application, it may be desirable to limit the reading of certain punctuation characters. For example, if the RC8650 is used to proofread documents, the application may call for only unusual punctuation to be read. On the other hand, an application that orally echoes keyboard entries for a blind user may require that all punctuation be spoken. The RC8650 supports four primary levels of punctuation filtering as shown in Table 2.4. These levels determine which punctuation characters will be spoken and which will not. In addition to the four base levels, the command can be expanded to control how number strings will be read. This is done by ORing the values 04h and/or 08h to the base parameter range, as described below. Sleep Timer (nQ) The sleep timer is used to force the RC8650 into Standby mode after a programmed time interval. For example, the RC8650 can power down automatically if the user forgets to turn off the power at the end of the day. An audible "reminder" tone can even be programmed to sound every ten minutes to remind the user that the power was left on, before shutdown occurs. 24 RC SYSTEMS Table 2.5. Timeout Delays RC8650 VOICE SYNTHESIZER Index markers are nonspeaking "bookmarks" that can be used to keep track of where the RC8650 is reading within a passage of text. The parameter n is any number between 0 and 99; thus, up to 100 unique markers may be active at any given time. When the RC8650 has spoken the text up to a marker, it transmits the marker number to the host via the TXD pin. Note that this value is a binary number between 0 and 99, not a literal ASCII number string as was used in the command to place the marker. This allows the marker to be transmitted as a one-byte value. There is no limitation to how many index markers can be used in a text string. The frequency depends on the resolution required by the application. In Text mode, for example, one marker per sentence or one marker per word would normally be used. In Phoneme mode, markers can be placed before each phoneme to monitor phoneme production, which is useful for synchronizing an animated mouth with the voice. Index Marker (nI) The sleep timer is stopped and reset whenever the RC8650 is active, and begins running when the RC8650 enters Idle mode. In this way, the RC8650 will not shut itself down during normal use, as long as the programmed timer interval is longer than the maximum time the RC8650 is inactive. The command parameter n determines when Standby mode will be entered. You can place the RC8650 in Standby mode immediately, program the sleep timer to any of 15 ten-minute intervals (10 to 150 minutes), or disable the sleep timer altogether (Table 2.6). Note that the delay interval is simply n x 10 minutes for 0 < n < 16. ORing 10h to these values (16 < n < 32) also enables the reminder tone, which sounds at the end of each ten minute interval. Programming n = 0 disables the sleep timer, which is the default setting. Setting n = 16 forces the RC8650 to enter Standby mode as soon as all output has ceased. If the sleep timer is allowed to expire, the RC8650 will emit the ASCII character "p" from the TXD pin and the STBY status flag will be set to 1, just before entering Standby mode. This enables the host to detect that the RC8650 has entered Standby mode. Once the RC8650 has entered Standby mode, it can be re-awakened only by a hardware reset or by driving the STBY# pin low for 250 ns or longer, then High again. All of the RC8650 handshake signals (BUSY, CTS#, and RDY#) are forced to their "not ready" states while the RC8650 is in Standby. Baud Rate (nH) The serial port's baud rate can be programmed to the rates listed in Table 2.7. If included as part of the greeting message, the command will effectively override the baud rate set by the BRS pins. Table 2.7. Programmable Baud Rates Table 2.6. Sleep Timer The TS pins provide talk status information for each audio channel, which can be used to activate a transmitter, take a telephone off hook, enable an audio power amplifier, etc., at the desired time. Each pin's state and polarity can be configured as shown in Table 2.8. The programming of the TS pins do not affect the Status Register TS flag in any way. The default setting is 1K. If a TS pin is programmed High or Low, it will remain so until changed otherwise. This feature can be used to activate a transmitter, for example, before speech output has begun. In the automatic mode, the TS pin is asserted as soon as output begins; it will return to its false state when all output has ceased. Note that because RC8650 commands work synchronously, the TS pin will not change state until all text and commands, up to the TS Pin Control command, have been spoken and/or executed. TS Pin Control (nK) 25 RC SYSTEMS Table 2.8. TS Pin Control RC8650 VOICE SYNTHESIZER Zap Commands (Z) This command prevents the RC8650 from honoring subsequent commands, causing it to read commands as they are encountered (useful in debugging). Any pending commands in the input buffer will still be honored. The only way to restore command recognition after the Zap command has been issued is to write CTRL+^ (ASCII 30) to the RC8650 or perform a hardware reset. Reinitialize (@) This command reinitializes the RC8650 by clearing the input buffer and restoring the voice parameters and control registers to their factory default settings. The exception dictionary, prerecorded audio, greeting message, baud rate, nor TS pin control setting are affected. This command controls various internal RC8650 operating parameters. The command parameter n is calculated by ORing together the individual control bits shown in Table 2.9. For example, 193G (193 = 128 + 64 + 1) disables V8600 emulation, enables all status messages and specifies that parameters should saturate. 128G is the default setting. Bit POR.7 (VC) programs the RC8650 to emulate RC Systems' original V8600 voice synthesizer module. When this bit is set to 0 (which V8600 application programs do, as this bit was undefined in the V8600), the overall voice speed range is reduced and the default speed is changed from 1S to 5S, matching the characteristics of the V8600. The serial port status messages (see Table 1.3) are also affected by the setting of this bit. Note Relative parameters work differently than usual with this command. Instead of specifying a displacement from the register's current value, relative parameters allow you to set ("+") and clear ("-") individual register bits. For example, +65G sets bits POR.0 and POR.6; -65G clears POR.0 and POR.6. Protocol Options Register (nG) Stop (CTRL+X), Skip (CTRL+Y) The Stop command stops the RC8650 and flushes its input buffer of all text and commands. The Skip command skips to the next sentence in the buffer. Neither command affects any of the RC8650's settings. Note The format of these commands is unique in that the command character (CTRL+A) is not used with them. The CTRL+X (ASCII 24) and CTRL+Y (ASCII 25) characters are written directly to the RC8650, which enables the RC8650 to react immediately, even if its input buffer is full. To be most effective, the states of the RC8650 handshaking signals should be ignored. Table 2.9. Protocol Options Register Bit Definitions 26 RC SYSTEMS Audio Control Register (nN) The Audio Control Register determines whether the RC8650's audio stream will be output as an analog signal on the AO pins or as serial digital data on the DAOUT pin. See Table 2.10 for the definition of each register bit. The default register setting is 0N. In the digital audio modes, data is transferred from the DAOUT pin in 8 bit linear, offset binary format (midscale = 80h). The DARTS# pin can be used to regulate the flow of data--it must be Low for transfers to begin. In the synchronous mode, do not attempt to read the data at an average rate faster than 10 kbytes/sec. At clock rates above 80 kHz RC8650 VOICE SYNTHESIZER the host must pause between reading each byte in order to keep the average transfer rate from exceeding 10 kbytes/sec. Figure 2.1 illustrates the synchronous data transfer mode. Note how either DARTS# or DACLK can be used to regulate the flow of data from the RC8650. Note Relative parameters work differently than usual with this command. Instead of specifying a displacement from the register's current value, relative parameters allow you to set ("+") and clear ("-") individual register bits. For example, +40N sets bits ACR.3 and ACR.5; -5N clears ACR.0 and ACR.2. Table 2.10. Audio Control Register Definitions 27 RC SYSTEMS RC8650 VOICE SYNTHESIZER Figure 2.1. Synchronous Digital Audio Transfer Timing Enable Exception Dictionary (U) The exception dictionary is enabled with this command. If the RC8650 is in Phoneme mode, or if an exception dictionary has not been loaded, the command will have no effect. The exception dictionary can be disabled by issuing one of the mode commands D, T, or C. The "pause" tone can be used to generate longer inter-digit delays in phone number strings, or to create precise silent periods in the RC8650's output. The generator's output level can be adjusted with the Volume command (nV). DTMF commands may be intermixed with text and other commands without restriction. TONE GENERATION COMMANDS The musical and sinusoidal tone generators are activated with these commands. Refer to Section 3 for more information. Table 2.11. DTMF Dialer Button Map Musical/Sinusoidal Tone Generators (J/nJ) The DTMF (Touch-Tone) generator generates the 16 standard tone pairs commonly used in telephone systems. Each tone is 100 ms in duration, followed by a 100 ms inter-digit pause--more than satisfying telephone signaling requirements (both durations can be extended to 500 ms by setting the DDUR bit of the Protocol Options Register). The mapping of the command parameter n to the buttons on a standard telephone is shown in Table 2.11. DTMF Generator (n*) 28 RC SYSTEMS AUDIO PLAYBACK COMMANDS A virtually unlimited number of prerecorded sound effects and messages can be stored in the RC8650, limited only by the amount of available on-chip audio memory. RCStudio, a Windows-based application available from RC Systems, makes it easy to create, manage, and download sound libraries composed of standard Windows wave files to the RC8650. Sound libraries created with RCStudio can also be downloaded to the RC8650 by simply transmitting the library file in its entirety. Each sound file (message or sound effect) in a sound library is automatically assigned a record number, beginning with zero. The first file is record 0, the second is record 1, and so on. The playback command plays records in any random order, using n to specify the desired record. The playback level can be adjusted with the Volume (nV) command. A volume setting of 5 will cause the files to be played back at their original volume level. Text and/or commands may be freely intermixed with the playback command. For example, ^A "11*" "Hello" ^A "-3V" ^A "3&" ^A "+3V" ^A "9&" plays the Touch-Tone "#" key and says "hello" at the current volume setting, followed by the fourth sound file at a reduced volume level, and finally the tenth sound file at the original volume level. RC8650 VOICE SYNTHESIZER The output sampling rate can be programmed to any rate between 4 and 11 kHz (32,000-88,000 bps) by choosing the appropriate parameter value. The relationship between the command parameter n and the sampling rate fs is n = 155 - 617/fs fs = 617/(155 - n) where fs is measured in kHz. For example, to program an 8 kHz sampling rate, choose n=78. The range of n is 0-99, hence fs can range from 4 to 11 kHz. The following procedure should be used for sending PCM or ADPCM audio data to the RC8650 in real time: 1) Program the desired volume level with the Volume (nV) command. A volume setting of 5 will cause the data to be played back at its original volume level. This step is optional. 2) Issue the Real Time Audio Playback Mode command n# if PCM data is being sent, or n% for ADPCM data. The RC8650 expects the audio data to immediately follow the command; therefore, be sure not to terminate the command with a CR or NUL. The TS pin and TS flag will be asserted at this time. 3) If the RC8650's serial port is being used for transferring the audio data, change the host system's baud rate to 115,200 baud at this time. 4) Begin transferring the audio data to the RC8650. The same methods employed for sending any other type of data to the RC8650 should be used. Note that the DAC will not begin taking samples from the audio buffer until at least 100 bytes have been sent or the value 80h is sent, whichever occurs first. 5) After the last byte of audio data has been sent to the RC8650, send the value 80h (-128). This signals the RC8650 to terminate Real Time Audio Playback mode and return to the text-to-speech mode of operation. Note that up to 2048 bytes of data may still be in the audio buffer, so the RC8650 may continue producing sound for as long as 0.5 second (at 4 kHz sampling rate) after the last byte of data has been sent. The TS pin/TS flag will not be cleared until all of the audio data has been output to the DAC, at which time the RC8650 will again be able to accept data from the host. If the host's serial port baud rate was changed in step 3, it should now be changed back to its original rate. Prerecorded Audio Playback Mode (n&) This mode allows audio samples to be written directly to the RC8650's digital-to-analog converter (DAC) via the RC8650's serial and parallel ports. All data sent to the RC8650 is routed directly to the RC8650's internal audio buffer; the RC8650 then outputs samples from the buffer to the DAC at the rate programmed by n. Because the audio data is buffered within the RC8650, the output sampling rate is independent of the data rate into the RC8650, as long as the input rate is equal to or greater than the programmed sampling rate. The RC8650 supports PCM and ADPCM audio data formats. RC Systems' RCStudio software can convert standard Windows wave files to PCM and ADPCM formats for use with the RC8650. ADPCM compression yields data files that are half the size of PCM files, thereby reducing the required data bandwidth and storage requirements. Real Time Audio Playback Mode (n#/n%) 29 RC SYSTEMS A/D CONVERTER COMMANDS The ADC Control Register controls the operation of the integrated analog-to-digital converter. All ADC results are transferred via the TXD pin. The following is an overview of the ADC: - - - - Four channels, 8-bit resolution (2 LSB precision) One-shot, continuous, single sweep, and continuous sweep modes of operation Selectable software or hardware triggering Support for external amplification/signal conditioning of all four ADC channels RC8650 VOICE SYNTHESIZER Figure 2.2 is a functional block diagram of the ADC input stage; Figure 2.3 illustrates the ADC in operation. Table 2.12 lists the definitions of each bit of the ADC Control Register. The default register setting is 0$. Operation of the ADC is not mutually exclusive of other RC8650 functions. The ADC can operate concurrently with text-to-speech, tone generation, audio playback, etc. The effective sampling rate in continuous mode is one-tenth the serial port baud rate (e.g., 115200 baud = 11.52 ksps). Note Relative parameters work differently than usual with this command. Instead of specifying a displacement from the register's current value, relative parameters allow you to set ("+") and clear ("-") individual register bits. For example, +34$ sets bits ADR.1 and ADR.5; -16$ clears ADR.4. ADC Control Register (n$) Table 2.12. ADC Control Register Definitions 30 RC SYSTEMS RC8650 VOICE SYNTHESIZER Figure 2.2. ADC Input Block Diagram Figure 2.3. ADC Transfer Timing 31 RC SYSTEMS MISCELLANEOUS COMMANDS Anytime the RC8650 is reset, an optional user-defined greeting message is automatically played. The message may consist of any text/command sequence up to 234 characters in length. Modal commands can be included, such as tone generator and audio playback commands. Caution The exception dictionary is erased whenever a new greeting message is written to the RC8650. To create a new greeting message, perform the following steps: 1) Write the command CTRL+A "255W". 2) Write the exact text/command sequence you want to store, up to 234 characters. For example, the string CTRL+A "3S" CTRL+A "2O" "ready" will program the RC8650 to use voice speed 3, Big Bob's voice, and say "ready" whenever it is reset. 3) Write a Null (ASCII 00) to terminate the command and store the greeting in the RC8650's nonvolatile memory. The RCStudio software, available from RC Systems, can automatically create and download greeting messages for you. Greeting messages created with RCStudio include the commands necessary to allow the file to be downloaded to the RC8650 by simply transmitting the file in its entirety. RC8650 VOICE SYNTHESIZER Interrogate (12?) This command retrieves the current operating settings of the RC8650. Table 2.13 lists the parameters in the order they are transmitted from the TXD pin, the command(s) that control each parameter, and each parameter's range. The parameters are organized as a byte array of one byte per parameter. Write Greeting Message (255W) Table 2.13. Parameters Returned by Interrogate Command Load Exception Dictionary (L) This command purges the RC8650's exception dictionary and stores subsequent output from the host in the RC8650's nonvolatile dictionary memory. The maximum dictionary size is 16 KB. Exception dictionaries must be compiled into the format required by the RC8650 before they can be used. The RCStudio software, available from RC Systems, includes a dictionary editor and compiler for performing this task. Dictionaries that have been compiled with RCStudio include the Load command in the file header, allowing the file to be downloaded to the RC8650 by simply transmitting the file in its entirety. Exception dictionaries are covered in detail in Section 4. Chipset Identification (6?) This command returns RC8650 system information that is used during factory testing. Eight bytes are transmitted via the TXD pin. The only information that may be of relevance to an application is the internal microcode revision number, which is conveyed in the last two bytes in packed-BCD format. For example, 13h 01h would be returned if the version number was 1.13. 32 RC SYSTEMS COMMAND SUMMARY RC8650 VOICE SYNTHESIZER Table 2.14. RC8650 Command Summary 33 RC SYSTEMS RC8650 VOICE SYNTHESIZER SECTION 3: MUSICAL & SINUSOIDAL TONE GENERATORS The RC8650 contains a three-voice tone generator that can be used for creating music and sound effects. This section explains how to program the generator. Note The musical tone generator output is available only from the AO pins. Digital audio output from the DAOUT pin is not possible. The musical tone generator is activated with the J command (no parameter). Once activated, all data output to the RC8650 is directed to the musical tone generator. MUSICAL TONE GENERATOR Note The RC8650 expects the tone generator data to immediately follow the J command; therefore, be sure not to terminate the command with a CR or Null. The tone generator is controlled with four, four-byte data and command frames, called Initialize, Voice, Play, and Quit. With these, the programmer can control the volume, duration, and frequencies of the three voices. Figure 3.1. Musical Tone Generator Command Formats Initialize Command The Initialize command sets up the tone generator's relative amplitude and tempo (speed). The host must issue this command to initialize the tone generator before sending any Voice frames. The Initialize command may, however, be issued anytime afterward to change the volume or tempo on the fly. The Initialize command consists of a byte of zero and three parameters. The parameters are defined as follows: KA KTL KTH Voice amplitude (1-255) Tempo, low byte (0-255) Tempo, high byte (0-255) Voice frames contain the duration and frequency (pitch) information for each voice. All Voice frames are stored in a 2 KB buffer within the RC8650, but are not played until the Play command is issued. If the number of Voice frames exceeds 2 KB in length, the RC8650 will automatically begin playing the data. Voice frames are composed of three frequency time constants (K1-K3) and a duration byte (KD), which specifies how long the three voices are to be played. The relationship between the time constant Ki and the output frequency fi is: fi = 16,768/Ki where fi is in Hertz and Ki = 4-255. Setting Ki to zero will silence voice i during the frame. KD may be programmed to any value between 1 and 255; the larger it is made, the longer the voices will play during the frame. Voice Frame Initialize command format Voice frame format The range of the tempo KT (KTL and KTH) is 1-65,535 (1-FFFFh); the larger the value, the slower the overall speed of play. The amplitude and tempo affect all three voices, and stay in effect until another Initialize command is issued. If the command is issued between Voice frames to change the volume or tempo on the fly, only the Voice frames following the command will be affected. 34 RC SYSTEMS Table 3.1. Musical Note Pitch/Ki Values RC8650 VOICE SYNTHESIZER mediate note values to be played, while maintaining the same degree of accuracy. This is important when, for example, a thirty-second note is to be played staccato, or a note is dotted (multiplying its length by 1.5). Table 3.2. Musical Note Duration/KD Values Using the suggested values, it turns out that most musical scores sound best when played at a tempo of 255 or faster (i.e., KTH = 0). Of course, the "right" tempo is the one that sounds the best. Play Command The Play command causes the voice data in the input buffer to begin playing. Additional Initialize commands and Voice frames may be sent to the RC8650 while the tone generator is operating. The TS pin and TS flag are asserted at this time, enabling the host to synchronize to the playing of the tone data. TS becomes inactive after all of the data has been played. The task of finding Ki for a particular musical note is greatly simplified by using Table 3.1. The tone generator can cover a four-octave range, from C two octaves below Middle C (Ki = 255), to D two octaves above Middle C (Ki = 14). Ki values less than 14 are not recommended. For example, the Voice frame DATA 24,64,0,0 will play Middle C using voice 1 (K1 = 64). Since K2 and K3 are zero, voices 2 and 3 will be silent during the frame. The duration of the note is a function of both the tempo KT and duration KD, which in this case is 24. As another example, DATA 48,64,51,43 plays a C-E-G chord, for a duration twice as long as the previous example. Table 3.2 lists suggested KD values for each of the standard musical note durations. This convention permits shorter (1/64th note) and inter- The Quit command marks the end of the tone data in the input buffer. The RC8650 will play the contents of the buffer up to the Quit command, then return to the text-to-speech mode that was in effect when the tone generator was activated. Once the Quit command has been issued, the RC8650 will not accept any more data until the entire buffer has been played. Quit Command Example Tune The Basic program shown in Figure 3.2 reads tone generator data from a list of DATA statements and LPRINTs each value to the RC8650. The program assumes that the RC8650 is connected to a PC's printer port, although output could be redirected to a COM port with the DOS MODE command. The astute reader may have noticed some "non-standard" note durations in the DATA statements, such as the first two Voice frames in line 240. According to the original music, some voices were not to be played as long as the others during the beat. The F-C-F notes in the first frame are held for 46 counts, while the low F and C in the second frame are held for two additional counts. Adding the duration (first and fifth) bytes together, the low F and C do indeed add up to 48 counts (46 + 2), which is the standard duration of a quarter note. Choosing note durations and tempo 35 RC SYSTEMS RC8650 VOICE SYNTHESIZER 100 110 120 130 140 150 160 170 180 190 200 210 220 230 240 250 260 270 280 290 300 310 LPRINT ' ensure serial port baud rate is locked LPRINT CHR$(1);"J"; ' activate tone generator READ B0,B1,B2,B3 ' read a frame (4 bytes) LPRINT CHR$(B0); CHR$(B1); CHR$(B2); CHR$(B3); IF B0 + B1 + B2 + B3 > 0 THEN 120 ' loop until Quit END ' ' ' Data Tables: ' ' Init (volume = 255, tempo = 86) DATA 0,255,86,0 ' ' Voice data DATA 46,48,64,192, 2,0,64,192, 48,48,0,0, 48,40,0,0, 48,36,0,0 DATA 94,24,34,0, 2,24,0,0, 24,0,36,0, 24,0,40,0, 48,0,48,0 DATA 48,40,0,192, 46,36,0,0, 2,0,0,0, 48,36,0,0, 48,24,34,0 DATA 46,24,34,0, 2,0,34,0, 46,24,34,0, 2,24,0,0, 24,0,36,0 DATA 24,0,40,0, 48,0,48,0 ' ' Play, Quit DATA 0,0,1,1, 0,0,0,0 Figure 3.2. Example Musical Tone Generator Program 36 RC SYSTEMS The musical tone generator is capable of producing three tones simultaneously, and works well in applications which require neither precise frequencies nor a "pure" (clean) output. The output is a pulse train rich in harmonic energy, which tends to sound more interesting than pure sinusoids in music applications. The sinusoidal tone generator enables the simultaneous generation of two sinusoidal waveforms. Applications for this generator range from generating simple tones to telephone call-progress tones (such as a dial tone or busy signal). The frequency range is 0 to 2746 Hz, with a resolution of 4 to 11 Hz. The sinusoidal tone generator is activated with the command nJ, where n is an ASCII number between 0 and 99. Note the similarity to the musical tone generator command, J, which uses no parameter. The parameter n programs the internal sampling rate, much like the Real Time Audio Playback command does; in fact, the sampling rate fs has the same relationship to n as the Real Time Audio Playback command: fs = 617 / (155 - n) Immediately following the nJ command are three binary parameter bytes: nJ Kd K1 K2 where Kd determines the tone duration, and K1 and K2 set the output frequencies of generators 1 and 2, respectively. The tone duration and frequencies are not only functions of these parameters, but of n as well. The output amplitude is a function of the Volume command (nV). The command and parameter values are buffered within the RC8650, and can be intermixed with text and other commands without restriction. The tone duration Td is calculated as follows: Td = Kd x 256 / fs (sec) where 0 Kd 255. Substituting the relationship fs = 617 / (155 - n) into the above equation, Td = Kd x (155 - n) / 2410 (sec) Setting Kd = 1 yields the shortest duration; Kd = 0 (treated as 256) the longest. Depending on the value of n, Td can range from 23 ms to 16.5 sec. RC8650 VOICE SYNTHESIZER The tone frequencies F1 and F2 are computed as follows: Fi = Ki x fs / 1024 (Hz) where 0 Ki 255. Substituting the relationship fs = 617 / (155 - n) into this equation, Fi = Ki x 603 / (155 - n) (Hz) Depending on the value of n, Fi can range from 0 Hz to 2746 Hz. If only one tone is to be generated, the other tone frequency may be set to 0 (Ki = 0), or equal in frequency. Note, however, that due to the additive nature of the tone generators, the output amplitude from both generators running at the same frequency will be twice that of just one generator running. Both K1 and K2 may be set to 0 to generate silence. Note that the frequency step size and frequency range are strictly functions of n. In general, the larger n is, the larger the step size and range will be. The parameter Ki can be thought of as a multiplier, which when multiplied by the step size, yields the output frequency. For example, setting n = 95 (corresponding to an internal sampling rate of 10.28 kHz) results in a frequency step size of 603 / (155 - 95) Hz, or 10 Hz. Thus, the output frequency range spans 0 Hz to 255 x 10 Hz, or 2550 Hz, in 10 Hz steps. As an example, suppose your application needed to generate the tone pair 440/350 Hz (a dial tone) for say, 2.5 seconds. We will choose n = 95, because it yields a convenient step size of 10 Hz. The tone duration parameter Kd is calculated as follows: Kd = 2410 x Td / (155 - n) substituting Td = 2.5 (sec) and n = 95, Kd = 2410 x 2.5 / (155 - 95) = 100 K1 (440 Hz) is computed as follows: K1 = F1 x (155 - n) / 603 = 440 x (155 - 95) / 603 = 44 In like manner, K2 (350 Hz) is computed to be 35. In order to embed the command in a text file, the computed values must be converted into their ASCII equivalents: 100 = "d", 44 = "," and 35 = "#". The complete command becomes ^A95Jd,# which can be embedded within normal text for the synthesizer. SINUSOIDAL TONE GENERATOR 37 RC SYSTEMS RC8650 VOICE SYNTHESIZER SECTION 4: EXCEPTION DICTIONARIES Exception dictionaries make it possible to alter the way the RC8650 interprets character strings it receives. This is useful for correcting mispronounced words, triggering the generation of tones and/or the playback of prerecorded sounds, or even speaking in a foreign language. In some cases, an exception dictionary may even negate the need of a text pre-processor in applications that cannot provide standard text strings. This section describes how to create exception dictionaries for the RC8650. The text-to-speech modes of the RC8650 utilize an English lexicon and letter-to-sound rules to convert text the RC8650 receives into speech. The pronunciation rules determine which sounds, or phonemes, each character will receive based on its relative position within each word. The integrated DoubleTalk text-to-speech engine analyzes text by applying these rules to each word or character, depending on the operating mode in use. Exception dictionaries augment this process by defining exceptions for (or even replacing) these built in rules. Exception dictionaries can be created and edited with a word processor or text editor that stores documents as standard text (ASCII) files. However, the dictionary must be compiled into the internal format used by the RC8650 before it can be used. The RCStudio software, available from RC Systems, includes a dictionary editor and compiler. Table 4.1. Context Tokens Exceptions have the general form L(F)R=P which means "the text fragment F, occurring with left context L and right context R, gets the pronunciation P." All three parts of the exception to the left of the equality sign must be satisfied before the text fragment will receive the pronunciation given by the right side of the exception. The text fragment defines the input characters that are to be translated by the exception, and may consist of any combination of letters, numbers, and symbols. Empty (null) text fragments may be used to generate sound based on a particular input pattern, without actually translating any of the input text. The text fragment (if any) must always be contained within parentheses. Characters to the left of the text fragment specify the left context (what must come before the text fragment in the input string), and characters to the right define the right context. Both contexts are optional, so an exception may contain neither, either, or both contexts. There are also 15 special symbols, or context tokens, that can be used in an exception's context definitions (Table 4.1). Note that although context tokens are, by definition, valid only within the left and right context definitions, the wildcard token may also be used within text fragments. Any other context token appearing within a text fragment will be treated as a literal character. The right side of an exception (P) specifies the pronunciation that the EXCEPTION SYNTAX text fragment is to receive, which may consist of any combination of phonemes (Table 2.1), phoneme attribute tokens (Table 2.2), and commands (Table 2.14). Using the tone generator and prerecorded audio playback commands, virtually limitless combinations of speech, tones, and sound effects can be triggered from any input text pattern. If no pronunciation is given, no sound will be given to the text fragment; the text fragment will be silent. A dictionary file may also contain comments, but they must be on lines by themselves (i.e., they cannot be on the same line as an exception). Comment lines must begin with a semicolon character (;), so the compiler will know to skip over them. An example of an exception is C(O)N=AA which states that o after c and before n gets the pronunciation AA, the o-sound in cot. For example, the o in conference, economy, and icon would be pronounced according to this exception. Another example is 38 RC SYSTEMS $R(H)= which states that h after initial r is silent, as in the word rhyme (the $ context token represents any non-alphabetic character, such as a space between words; see Table 4.1). Punctuation, numbers, and most other characters can be redefined with exceptions as well: (5)=S I NG K O (CHR$)=K EH R IX K T ER (Spanish five) (Basic function) RC8650 VOICE SYNTHESIZER The first exception states that o followed by e, i, or y is to be pronounced OW, the o-sound in boat. The second exception does not place any restriction on what must come before or after o, so o in any context will receive the UW pronunciation. If the exceptions were reversed, the (O)+ exception would never be reached because the (O) exception will always match o in any context. In general, tightlydefined exceptions (those containing many context restrictions) should precede loosely-defined exceptions (those with little or no context definitions). (RAT)=R AE T (RATING)=R EY T IH NG (R)=R This is an example of how not to organize exceptions. The exception (RATING) will never be used because (RAT) will always match first. According to these exceptions, the word rating would be pronounced "rat-ing." It can be beneficial to group exceptions by the first character of the text fragments, that is, all of the A exceptions in one group, all the B exceptions in a second group, and so on. This gives an overall cleaner appearance, and can prove to be helpful if the need arises to troubleshoot any problems in your dictionary. In order to better understand how an exception dictionary works, it is helpful to understand how the DoubleTalk text-to-speech engine processes text. Algorithms within the DoubleTalk engine analyze input text a character at a time, from left to right. A list of pronunciation rules is searched sequentially for each character until a rule is found that matches the character in the correct position and context. The algorithm then passes over the input character(s) bracketed in the rule (the text fragment), and assigns the pronunciation given by the right side of the rule to them. This process continues until all of the input text has been converted to phonetic sounds. The following example illustrates how the algorithm works by translating the word receive. The algorithm begins with the letter r and searches the R pronunciation rules for a match. The first rule that matches is $(RE)^#=R IX, because the r in receive is an initial r and is followed by an e, a consonant (c), and a vowel (e). Consequently, the text fragment re receives the pronunciation R IH, and the scan moves past re to the next character: receive. (E is not the next scan character because it occurred inside the parentheses with the r; the text fragment re as a whole receives the pronunciation R IX) The first match among the C rules is (C)+=S, because c is followed by an e, i, or y. C thus receives the pronunciation S, and processing continues with the second e: receive. (EI)=IY is the first rule to match the second e, so ei receives the sound IY. Processing resumes at the character receive, which matches the default V rule, (V)=V. The final e matches the rule #:(E)$=, which applies when e is final and follows zero or more consonants and a vowel. Consequently, e receives no sound and processing continues with the following word or punctuation, if any. Thus, the entire phoneme string for the word receive is R IX S IY V. THE TRANSLATION ALGORITHM It is possible that some input text may not match anything in a dictionary, depending on the nature of the dictionary. For example, if a dictionary was written to handle unusual words, only those words would be included in the dictionary. On the other hand, if a dictionary defined the pronunciation for another language, it would be comprehensive enough to handle all types of input. In any case, if an exception is not found for a particular character, the English pronunciation will be given to that character according to the built in pronunciation rules. Generally, the automatic switchover to the built in rules is desirable if the dictionary is used to correct mispronounced words, since by definition the dictionary is defining exceptions to the built in rules. If the automatic switchover is not desired, however, there are two ways to prevent it from occurring. One way is to end each group of exceptions with an unconditional exception that matches any context. For example, to ensure that the letter "a" will always be matched, end the A exception group with the exception (A)=pronunciation. This technique works well to ensure matches for specific characters, such as certain letters or numbers. If the exception dictionary is to replace the built in rules entirely, end the dictionary with the following exception: ()= This special exception causes unmatched characters to be ignored (receive no sound), rather than receive the pronunciation defined by the built in rules. TEXT NOT MATCHED BY THE DICTIONARY Since DoubleTalk uses its translation rules in a sequential manner, the position of each exception relative to the others must be carefully considered. For example, consider the following pair of exceptions: (O)+=OW (O)=UW RULE PRECEDENCE EFFECT ON PUNCTUATION 39 RC SYSTEMS Punctuation defined in the exception dictionary has priority over the Punctuation Filter command. Any punctuation defined in the dictionary will be used, regardless of the Punctuation Filter setting. Note If the dollar sign character ($) is defined within the text fragment of any exception, currency strings will not be read as dollars and cents. RC8650 VOICE SYNTHESIZER No Cussing, Please The reading of specific characters or words can be suppressed by writing exceptions in which no pronunciation is given. (????)= (YOU fill in the blanks!) When Zero Isn't Really Zero Exceptions are defined independently for the Character and Text modes of operation. The beginning of the Character mode exceptions is defined by inserting the letter C just before the first Character mode exception. No exceptions prior to this marker will be used when the RC8650 is in Character mode, nor will any exceptions past the marker be used in Text mode. For example: . . ()= C . . . ()= (Text mode exceptions) (optional; used if built in rules are not to be used in no-match situations) (Character mode exceptions marker) (Character mode exceptions) (optional; used if built in rules are not to be used in no-match situations) CHARACTER MODE EXCEPTIONS When reading addresses or lists of numbers, the word "oh" is often substituted for the digit 0. For example, we might say 1020 North Eastlake as "one oh two oh North Eastlake." The digit 0 can be redefined in this manner with the following exception: (0)=OW Acronyms and Abbreviations Acronyms and abbreviations can be defined so the words they represent will be spoken. $(KW)$=K IH L AH W AA T $(DR)$=D AA K T ER $(TV)$=T EH L AX V IH ZH IX N String Parsing & Decryption Sometimes the data that we would like to have read is not available in a "ready-to-read" format. For example, the output of a GPS receiver may look something like this: $GPGGA,123456,2015.2607,N,... The following examples illustrate some ways in which the exception dictionary can be used. APPLICATIONS The first 14 characters of the string consists of a fixed header and variable time data, which we would like to discard. The following exception ensures that the header will not be read: ($GPGGA,,)= Note how wildcard tokens are used for handling the time data (8th-13th characters), since the content of this field is variable. The 15th-16th and 17th-18th characters represent the latitudinal coordinate in degrees and minutes, respectively. The three exceptions shown below handle the latitudinal component of the GPS string. Note in the first exception how a null text fragment is used in the appropriate position to generate the word "degrees," without actually translating any of the input characters. ,\\()\\.=D IX G R IY Z , , (.)=M IH N IH T S , , (,N,)=N OW R TH L AE T IH T UW D The four exceptions together will translate the example string as "20 degrees, 15 minutes, north latitude." (Additional exceptions for handling the seconds component, and digits themselves, are not shown for clarity). Correcting Mispronounced Words Correcting mispronounced words is the most common application for exception dictionaries. S(EAR)CH=ER $(OK)$=OW K EY The first exception corrects the pronunciation of all words containing search (search, searched, research, etc.). As this exception illustrates, it is only necessary to define the problem word in its root form, and only the part of the word that is mispronounced (ear, in this case). The second exception corrects the word ok, but because of the left and right contexts, will not cause other words (joke, look, etc.) to be incorrectly translated. 40 RC SYSTEMS Heteronyms Heteronyms are words that have similar spellings but are pronounced differently, depending on the context, such as read ("reed" and "red") and wind ("the wind blew" and "wind the clock"). Exceptions can be used to fix up these ambiguities, by including non-printing (Control) characters in the text fragment of the exception. Suppose a line of text required the word "close" to be pronounced as it is in "a close call," instead of as in "close the window." The following exception changes the way the s will sound: (^DCLOSE)=K L OW S Note the CTRL+D character (^D) in the text fragment. Although a nonprinting character, the translation algorithms treat it as they would any printing character. Thus, the string "^D close" will be pronounced with the s receiving the "s" sound, wherever it appears in the text stream. Plain "close" (without the CTRL+D) will be unaffected--the s will still receive the "z" sound. It does not matter where you place the Control character in the word, as long as you use it the same way in your application's text. You may use any non-printing character (except LF and CR) in this manner. RC8650 VOICE SYNTHESIZER lines of the bus interface (see Figure 1.6) and hardwiring the remaining four to the appropriate logic levels, virtually any set of 16 ASCII characters can be generated, which in turn can be interpreted by the exception dictionary. For example, by connecting the four control bits to DB0 through DB3, DB4 and DB5 to VCC, DB6 and DB7 to ground and the strobe to PWR#, ASCII codes 30h through 3Fh (corresponding to the digits "0" through "9" and the six ASCII characters following them) can be generated by the four control bits. Message strings would then be assigned to each of these ASCII characters. For example, you could make the character "0" (corresponding to all four control bits = 0) say, "please insert quarter," with the following dictionary entry: (0)=P L IY Z IH N S ER T K W OW R T ER The Timeout timer should also be activated (1Y, for example) in order for the "message" to be executed. Otherwise, the RC8650 will wait indefinitely for a CR/Null character that will never come. The timer command could be included in the greeting message. Dictionaries can be created that enable the RC8650 to speak in foreign languages. It's not as difficult as it may seem--all that is required in most cases is a pronunciation guide and a bit of patience. If you don't have a pronunciation guide for the language you're interested in, check your local library. Most libraries have foreign language dictionaries that include pronunciation guides, which make it easy to transcribe the pronunciation rules into exception form. Foreign Languages Make sure that your exceptions aren't so broad in nature that they do more harm than good. Exceptions intended to fix broad classes of words, such as word endings, are particularly notorious for ruining otherwise correctly pronounced words. Take care in how your exceptions are organized. Remember, an exception's position relative to others is just as important as the content of the exception itself. TIPS Exception dictionaries even allow the RC8650 to read foreign language text in English! The following exceptions demonstrate how this can be done with three example Spanish/English words. (GRANDE)=L AA R J (BIEN)=F AY N (USTED)=YY UW The sense of translation can also be reversed: (LARGE)=G RR A N D EI (FINE)=B I EI N (YOU)=U S T EI DH Language Translation On rare occasions, an exception may not work as expected. This occurs when the built in pronunciation rules get control before the exception does. The following example illustrates how this can happen. Suppose an exception redefined the o in the word "process" to have the long "oh" sound, the way it is pronounced in many parts of Canada. Since the word is otherwise pronounced correctly, the exception redefines only the "o:" PR(O)CESS=OW But much to our horror, the RC8650 simply refuses to take on the new Canadian accent. It so happens that the RC8650 has a built in rule which looks something like this: $(PRO)=P R AA This rule translates a group of three characters, instead of only one as most of the built in rules do. Because the text fragment PRO is translated as a group, the o is processed along with the initial "pr," and consequently the exception never gets a shot at the o. If you suspect this may be happening with one of your exceptions, include more of the left-hand side of the word in the text fragment (in the example above, (PRO)CESS=P R OW would work). When Things Don't Work as Expected Message Macros Certain applications may not be able to send text strings to the RC8650. An example of such an application is one that is only able to output a four bit control word and strobe. Sixteen unique output combinations are possible, but this is scarcely enough to represent the entire ASCII character set. You can, however, assign an entire spoken phrase to a single ASCII character with the exception dictionary. By driving four of the data bus 41 RC SYSTEMS RC8650 VOICE SYNTHESIZER SECTION 5: RC8650 EVALUATION KIT The RC8650 Evaluation Kit comes with everything required to evaluate and develop applications for the RC8650 chipset using a Windowsbased PC. The included RCStudioTM software provides an integrated development environment with the following features: The following components are included in the DoubleTalk RC8650 Evaluation Kit: EVALUATION KIT CONTENTS * * * * Read any text, either typed or from a file Easy access to the various RC8650 voice controls Manage collections of sound files and store them in the RC8650 Exception dictionary editor/compiler, and much more... * * * * * Printed circuit board containing the RC8650-1 chipset AC power supply Speaker Serial cable RCStudioTM development software CD The evaluation board can also be used in stand-alone environments by simply printing the desired text and commands to it via the onboard RS-232 serial or parallel ports. EVAL BOARD OUTLINE 42 RC SYSTEMS CONNECTOR PIN ASSIGNMENTS & SCHEMATICS RC8650 VOICE SYNTHESIZER Table 5.1. P1 Pin Assignments (Audio Output & Control) Table 5.4. P101 Pin Assignments (RS-232 Serial Interface) Table 5.5. P102 Pin Assignments (TTL Serial Interface) Table 5.2. P2 Pin Assignments (A/D Converter) JP4-JP6 must be open in order to use the TTL interface Table 5.6. P103 Pin Assignments (Printer/Bus Interface) Table 5.3. JP1-JP3 Pin Assignments (Baud Rate) 43 RC SYSTEMS 44 RC8650 VOICE SYNTHESIZER RC SYSTEMS 45 RC8650 VOICE SYNTHESIZER RC SYSTEMS 46 RC8650 VOICE SYNTHESIZER RC SYSTEMS 47 RC8650 VOICE SYNTHESIZER Specifications written in this publication are believed to be accurate, but are not guaranteed to be entirely free of error. RC Systems reserves the right to make changes in the devices or the device specifications described in this publication without notice. RC Systems advises its customers to obtain the latest version of device specifications to verify, before placing orders, that the information being relied upon by the customer is current. In the absence of written agreement to the contrary, RC Systems assumes no liability relating to the sale and/or use of RC Systems products including fitness for a particular purpose, merchantability, for RC Systems applications assistance, customer's product design, or infringement of patents or copyrights of third parties by or arising from use of devices described herein. Nor does RC Systems warrant or represent that any license, either express or implied, is granted under any patent right, copyright, or other intellectual property right of RC Systems covering or relating to any combination, machine, or process in which such devices might be or are used. RC Systems products are not intended for use in medical, life saving, or life sustaining applications. Applications described in this publication are for illustrative purposes only, and RC Systems makes no warranties or representations that the devices described herein will be suitable for such applications. 1609 England Avenue, Everett, WA 98203 Phone: (425) 355-3800 Fax: (425) 355-1098 http://www.rcsys.com |
Price & Availability of RC8650L
 |
|
|
|
|
All Rights Reserved © IC-ON-LINE 2003 - 2022 |
| [Add Bookmark] [Contact Us] [Link exchange] [Privacy policy] |
|
Mirror Sites : [www.datasheet.hk]
[www.maxim4u.com] [www.ic-on-line.cn]
[www.ic-on-line.com] [www.ic-on-line.net]
[www.alldatasheet.com.cn]
[www.gdcy.com]
[www.gdcy.net] |