AMBE-2000™ Vocoder Chip Users Manual
Version 4.92, June, 08
DVSI Confidential Proprietary, Subject to Change
Page 44
Visit us at
www.dvsinc.com
Table 7-B Rate Selection Using Rate Info 0-4, AMBE-2020™ only (AMBE+™)*
RATE_SEL4
Pin
RATE_SEL3
Pin
RATE_SEL2
Pin
RATE_SEL1
Pin
RATE_SEL0
Pin
Speech Rate
(bps)
FEC Rate (bps)
Total Rate
(bps)
1
1
1
1
1
2000
0
2000
1
0
0
0
0
3600
0
3600
1
0
0
0
1
4000
0
1
0
1
1
0
2400
1600
4000
1
0
0
1
0
4800
0
1
1
0
0
0
4000
800
1
0
1
1
1
3600
1200
1
1
0
0
1
2400
2400
4800
1
0
0
1
1
6400
0
1
1
0
1
0
4000
2400
6400
1
1
0
1
1
4400
2800
7200
1
0
1
0
0
8000
0
1
1
1
0
0
4000
4000
8000
1
0
1
0
1
9600
0
1
1
1
0
1
3600
6000
1
1
1
1
0
2400
7200
9600
*Note: AMBE+ is only used for speech rates at 3600 bps and higher. Any rates below 3600 bps
use a modified algorithm similar to the AMBE algorithm.
7.3
Voice Activity Detection (VAD), Comfort Noise Insertion (CNI)
The Voice Activity Detection (VAD) algorithm along with the Comfort Noise Insertion (CNI) feature of the AMBE2020™
chip performs useful functions in systems trying to convert periods of silence, that exist in normal conversation, to savings
in system bandwidth or power.
With the VAD functions enabled, periods of silence will be denoted by the encoder in two ways. First, the encoder will
output a silence frame (in-band). This silence frame contains information regarding the level of background noise, which
allows the corresponding decoder to synthesize a “Comfort Noise” signal at the other end. The comfort noise is intended to
give the listener the feeling that the call is still connected, as opposed to producing absolute silence, which can give the
impression that, the call has been “dropped”. Second, the Encoder Silence Detected flag is set in Control Word 1 of the
Framed Output format described in section 5.3.3.
VAD is enabled when a logic high is on the hardware pin VAD_EN (pin 86), subject to the restrictions of section 7.1. Once
the VAD is enabled the VAD function may be controlled ON and OFF through the use of control word 2 as described in
section 5.2.9.
If the VAD features are being used to reduce transmit power during times of conversational silence, DVSI recommends that
a silence frame be transmitted at the start of the period and approximately each 500-1000 milliseconds thereafter. This is to
ensure that the parameters regarding the levels of background noise are transmitted to the decoder for the smoothest audible
transitions between synthesized speech and synthesized silence.
The silence threshold value is -25 dBm0 in the VAD algorithm. Each frame that exceeds this level will be classified as
voice. If the frame level is less than –25 dBm0 the voice/silence decision will be determined based upon various adaptive
thresholds.
The synthesis of a Comfort Noise frame by the decoder is not dependant on VAD being enabled. The decoder will produce
a comfort noise frame if it receives an in-band silence frame (produced only by an encoder with VAD enabled).
