DSP subsystem for VoIP

Ittiam's Voice Over Packet (VoIP) software is a complete software sub-system targeted at client (residential gateways/IP Phones) and server class (carrier-class and enterprise gateway) applications. The software solution includes all required DSP algorithms such as speech compression/decompression, echo cancellation and associated telephony/signaling functions. Key benefits are

• Software based on open architecture that allows feature enhancements and customer-specific proprietary features to be integrated
• C callable and reentrant implementations
• Highly optimized to provide high channel densities and low cost-per-channel

Target Applications

• Voice over DSL (VoDSL)
• Voice over ATM (VoATM)
• Enterprise/residential gateways
• Carrier Class gateways
• IP-PBX
• IP (LAN) Phones
• Video Phones
• IADs

DSP subsystem Components

Ittiam's implementation of the VoIP Software components support a uniform C callable interface suitable for use in any framework or operating environment.

• Reentrant, interruptible, multi-channel implementation
• Highly optimized implementation with minimum memory and MCPS utilization
• Efficient memory management that minimizes load on system-stack
• Fully compliant to standards/specifications

Given below is the list of optimized software supported by the DSP subsystem

• Full Range of speech coders
• Narrow Band
• G.711 PCM @ 64kbps
• G.726 ADPCM @ 16/24/32/40 kbps
• G.728 LD-CELP @ 16kbps
• G.723.1A MP-MLPQ/CELP @ 5.3/6.3 kbps with VAD/CNG/DTX support
• G.729AB CS-ACELP @ 8kbps with VAD/CNG/DTX support
• Wideband
• G.722 (Subband ADPCM @ 48/56/64 kbps)
• G.722.1 (MLT @ 24/32 kbps)
• G.722.2 (GSM-AMR Wide Band)
• Line Echo cancellation (LEC)
• Conformance to ITU-T G.168-2000 specification
• Echo Tail length Configurable from 16 to 128 ms
• Rapid Convergence
• Robust performance against Narrowband signals and background noise
• Acoustic Echo Canceller for speaker phones (AEC)
• Configurable echo tail lengths
• Rapid Convergence
• Robust double talk detection
• Silence Suppression
• Voice activity detection (VAD)
• Silence insertion descriptor (SID) support
• Comfort noise generation (CNG)
• Telephony Functions
• DTMF Tone Generator and Detector
• Call Progress Tone Generator and Detector
• Fax/Voice/Data mode discriminator
• Bellcore/ETSI standard Caller ID Detector and Generator
• Adaptive Jitter Buffer
• Packet Loss Concealment
• RTP

DSP subsystem Framework

The DSP subsystem framework integrates the voice and call processing components and provides for format conversion from circuit switched networks (PCM interface) to packetized networks. Each voice channel from the PCM side is echo canceled, compressed, and packetized for transmission over the packet network. In the reverse direction, each packetized voice channel from the packet network is buffered for de-jittering, decompressed, and terminated at the PCM interface. The figure below shows a basic VoIP DSP subsystem with different software components.

Key features of the DSP subsystem include

• Highly optimized DSP software
• Voice feature selection on a per-channel basis
• Multi Channel Reentrant implementation
• C callable API
• System built on the DSP/BIOS™
• Open architecture allows feature enhancements
• High Channel Density / Low Cost-per-port
• Easy to integrate with other Signaling and Protocol Subsystem

Platforms

• Currently supported on TI C64x™ (TMS320C64xx™)
• For other platforms contact mkt@ittiam.com