Version: 1.18
PRELIMINARY/CONFIDENTIAL
TrueSpeech® Co-Processor
6 Host Data Transfer Modes
Data transfer between the CT8022 and Host can be performed in two ways:
•
•
via the transmit and receive Data Buffers (recommended for new designs)
via the Software Control and Software Status Registers (CT8015-compatible mode)
6.1 Data Transfer using the Data Buffers
This transfer mode is recommended for all new designs.
In this mode, the CT8022 supports DMA transfers as well as low-overhead Host processor-based transfers. Once the
CT8022 playback/record options have been selected and activated, speech frames can be transferred with minimal or
zero control overhead required for device polling or supervision.
6.1.1
DMA Transfers
The CT8022 provides a full-duplex DMA interface with independent Transmit and Receive request and
acknowledge signals. The Host issues the commands to the CT8022 to start playback and/or record and then enables
DMA transfers by programming the Hardware Control Register. Once playback/record operation has been set up,
the CT8022 can continue to transfer speech frames without further involvement of the Host. The Host processor can
optionally use the buffer management commands to observe and control the status of the speech data streams.
6.1.2
Host Processor Transfers
In systems where no DMA controller is available, the Host processor can be used to perform the data transfers. The
playback and/or record operation is set up as before. Then, the Host processor can configure the Hardware Control
Register to generate an interrupt to the Host on the TX/RX ready condition, or the Host may simply periodically poll
the TX/RX ready status. Once the Host is ready to perform the data transfer, it can perform read/write bursts of up
to 16 words of data without needing to recheck the TX/RX ready status. The data buffers can contain a maximum of
16 words of data. This size is sufficient to accommodate a complete TrueSpeech data frame. This allows
microprocessors with repeat instruction prefixes (e.g. Intel X86: REP INS, REP OUTS) to perform block moves of
compressed speech frames. When performing record/playback of non-compressed data, where the speech data frame
size is greater than 16 words, the Host may divide the data transfer into 16-word sub-blocks and recheck the TX/RX
ready status between the sub-blocks.
CT8022A11AQC FW Revision 0118 DSP GROUP, INC., 3120 SCOTT BOULEVARD
SANTA CLARA, CA 95054 PH: 408 986 – 4300 FAX: 408 986 – 4490
43
All specifications are subject to change without prior notice.
ETC [ ETC ]
