W681307 Datasheet, PDF (80/160 Pages) Winbond – USB1.1 CODEC Microprocessor Control Unit with 32KB Mask ROM and 4KB RAM

W681307
12. SPEECH PROCESSOR
： The Speech Processor provides a complete implementation, including
• 4 duplex channels
• 1 channel Echo Canceller
• A Mixer Block
• Programmable Tone Generators
The Speech Processor supports A-law, µ-law and 16-bits linear PCM formats. Echo canceller channels support 16 bits linear PCM only.
The Speech Processor is implemented by an optimized micro-coded DSP, an external FIR Filter Engine and one digital gain multiplexer.
The architecture of the Speech Processor can be shown in Figure 12-1.
The Microcode DSP Core:
• Performs Tone Generation
• Performs PCM Mixing
The FIR Engine:
• Performs real-time Echo Estimation
• Implements Network Echo Suppression
• Calculates the Echo Cancellation filter coefficients
• Performs intermediate calculations on the Echo Estimation error
12.1
Transcoder DSP
The Transcoder DSP is a low power implementation and has the following features:
• Low power consumption and low gate count
• Group delay under 14 µs / channel
• DTMF and call progress tones
• Sidetone generation and Volume control
• Requires only 4.1 MIPS per channel
The Transcoder DSP supports A-law, µ-law and 16-bits linear PCM formats. Format selection is programmable on a by-channel basis.
： The signal flow (per channel) is shown in Figure 12-1 below
PCM_IN
INPUT
GAIN
+
MIXER
MATRIX
PCM
FORMAT
TONE
GEN
SIDE
TONE
PCM_OUT
+
MIXER
MATRIX
+
OUTPUT
GAIN
Figure 12-1 Transcoder signal flow
There are dual-tone generators for each PCM channel. These can generate DTMF and common signaling tones, as well as user
notification tones. The tones may be added in either direction. A mixer function is enabled by setting 1401[0]. Four full-duplex PCM
- 80 -
Publication Release Date: May, 2007
Revision 1.3