Version: 1.18
PRELIMINARY/CONFIDENTIAL
TrueSpeech® Co-Processor
9.8.4
Playback Frame Alignment Using Transmit Data Buffer
CT8022 provides an additional TFR_MODE = 01 [binary] which maintains strict 30ms data frame alignment for all
playback rates. This mode was not available in older versions of the CT8020. The following section describes non-
frame-aligned playback using TFR_MODE = 11 [binary].
If performing Playback via the Transmit Data Buffer with TFR_MODE = 11 [binary], the following should be noted:
1. When the CT8022 completes transfer of a speech data frame from the Transmit Data Buffer to its playback
queue (in the external data SRAM), the CT8022 immediately returns the (empty) Transmit Data Buffer to the
Host and grants the Host access to this buffer. This causes the TX READY signal (and interrupt) to be re-
asserted. This happens even when the CT8022’s internal memory playback buffer is full.
2. In the case of compressed speech playback (TrueSpeech 8.5 - 4.1), the effect is simply to allow the Host to
transfer one extra compressed speech frame to the CT8022. When the Host writes the next compressed speech
frame to the CT8022, the CT8022 is unable to empty the Transmit Data Buffer to its internal memory, since it is
full. In this case, the speech data remains in the Transmit Data Buffer until the next 30ms frame period when
space becomes available in the internal memory. In this memory full + transmit data buffer full state, the TX
READY will not be set. This provides an indication to the Host that the CT8022 is not ready to accept more
data. The effect of all this is to allow the CT8022 to effectively buffer one extra frame. This extra frame is not
included in the frame count returned by the Buffer Monitoring Command.
3. In the case of non-compressed speech, the effect of this behavior is not as simple. With non-compressed
playback (16-bit linear, 8-bit linear and 8-bit µ-law/A-law), the empty Transmit Data Buffer is also returned to
the Host upon conclusion of transfer of each complete speech frame (120 or 240 words). However in this case,
the Transmit Data Buffer is not large enough to take the entire next frame. Instead it is able to hold only the first
16 words of the next frame. The result is that the frame-by-frame transfer of data to the CT8022 will occur with
a 16-word offset relative to the 120 or 240-word frame boundaries. For example:
In 16-bit linear playback, the 240-word speech frame is broken into 15 sub-frames of 16 words each for transfer
via the Transmit Data Buffer. Therefore, the data transfer pattern, broken down into 30ms frame periods, will
appear as follows:
Playback Frame Number
Data Transferred
Number Of Sub-Frames
Transferred
240 words of frame 1
+
16 words of frame 2
224 words of frame 2
+
16 words of frame 3
224 words of frame 3
+
16 words of frame 4
224 words of final frame
16 (256 words)
15 (240 words)
15 (240 words)
14 (224 words)
1
2
3
Final Frame
CT8022A11AQC FW Revision 0118 DSP GROUP, INC., 3120 SCOTT BOULEVARD
SANTA CLARA, CA 95054 PH: 408 986 – 4300 FAX: 408 986 – 4490
79
All specifications are subject to change without prior notice.
ETC [ ETC ]
