![background image](http://html.mh-extra.com/html/texas-instruments/msp53c691/msp53c691_user-manual_1097147058.webp)
General Constraints
4-2
4.1
General Constraints
The sound quality of the speech produced by the MSP53C691 is sensitive to
the timing of speech data transfers from the master microprocessor. The
speech data is stored in a circular buffer. The buffer sizes are 26 bytes for FM
and 68 bytes for CELP, MELP, and ADPCM. If the data is written to the slave
too infrequently, the buffer becomes empty and the synthesis process stops
or becomes corrupted. It is very important that the master supplies the speech
data to the slave as soon as it is requested.
Each of the synthesis algorithms use the data in the buffer at various rates. In
general, the slave requests speech data (sets INRDY low) as soon as there
are at least two bytes empty in the buffer. The rate at which the slave requests
speech data depends upon two things: 1) The rate at which the data is
consumed by the synthesizer algorithm (that is, the bit rate of the algorithm)
and 2) The rate at which the slave queries the buffer to check for empty bytes.
The slave generally requests new data in bursts; that is, while processing a
frame the slave requests new data as soon as there are two empty bytes in
the CELP/MELP buffer. Once the processing of a frame is done, the slave
ceases to request data until it is time to process the next frame. Once the
MSP53C691 has pulled INRDY low to request new data, the data must be
loaded quickly until the buffer is again full and the slave stops pulling INRDY
low. It is highly recommended that the slave read and write status pins (INRDY
and OUTRDY) be tied to an interrupt of the master for most efficient data
transfer. A polling method can be used to find the status of the INRDY and
OUTRDY pins; however, it is not a suggested method for mix mode operation
or for playing high bit rate CELP or ADPCM.
The master can send commands to the slave in four-bit mode while speaking.
However, this feature makes the precise characterization of timing more diffi-
cult because of the widely varied command set. The number of commands that
can be sent to the slave while speaking varies and depends on the command
being sent and on the the bit rate of the synthesizer.
The following sections show timing waveforms which have been found to work
with tested data sets. It is often difficult for the system designer to exactly repli-
cate the timing shown in these sections. The timing must be adjusted to opti-
mize the sound quality for the specific system being designed. In each case,
only the speech data was sent to the slave without any interleaved commands.
The following general considerations must be observed:
-
Use interrupt service routines to send and receive data. This provides a
fast response and allows speech data to be sent even while the master is
performing other tasks.
-
Avoid disabling the interrupts tied to INRDY and OUTRDY. If it is neces-
sary to turn off the interrupts tied to INRDY and OUTRDY, make sure that
they are not kept off so long that the slave’s buffer underflows. The
MSP53C691 has a 68-byte internal buffer to store incoming speech data.
If CELP data for 11.2 kbps at 8-kHz sampling rate is sent, the buffer can
hold approximately 48 ms of speech. If the speech file is all voiced or has
a higher sampling rate, the buffer can hold an even shorter duration of
speech. It is important to consider the state of the buffer before turning off
the interrupts.
Summary of Contents for MSP53C691
Page 1: ...MSP53C691 Speech Synthesizer December 2000 MSDS Speech User s Guide SPSU020 ...
Page 22: ...2 10 ...
Page 96: ...C 6 ...