|
In the past 10 years, Voice over IP (VoIP) has emerged as an alternative to traditional telephony (PSTN), allowing a reduction in costs and increased capacity on long-distance telephone calls. While providing for lower costs and greater flexibility on converged voice and data networks, VoIP in general failed to deliver better-than-PSTN quality, mainly due to the 3.4 KHz bandwidth limitation of legacy narrowband speech codecs. Capitalizing on expanding IP broadband networks, wideband speech codecs—which encode 7.1 KHz of the voice spectrum—can now be effectively deployed in VoIP. AudioCodes has now ported its VoIPerfectHD™ into the MIPS32® family of RISC processors, allowing the elevation of daily voice communication to be similar to that of FM radio and/or conference room quality.
AudioCodes’ HD VoIP technology offers a broad range of features that enable a more natural and comprehensible voice quality such as popular wideband coders G.722 and G.722.2 and the new ITU standard G.711.1. Enhanced algorithms that enrich the HD VoIP experience included in the AudioCodes HD VoIP offering are:
-
Innovative Packet Loss Concealment
-
Acoustic Echo Cancelation (for IP Phone Applications)
-
Adaptive Jitter Buffer with Reorder
-
Wideband Voice Conferencing
-
Wideband Media Transcoding
VoIP Quality of Service
One of the main challenges for Integrated Access Device (IAD) developers when including VoIP software within their product is the Quality of Service (QoS) for VoIP. A single processor running both data and voice applications may “starve” the VoIP application due to other, more computationally heavy tasks, such as data routing and security. Such starvation will result in poor voice quality when compared to PSTN, and finally a rejection of the product.
AudioCodes’ VoIPerfect Solution running on the MIPS32® 34K® processor core offers the distribution of software tasks between hardware Thread Contexts (TCs) and allows the programmer to control the QoS of each task. Thus the VoIP task can be allocated to single or multiple TCs which is intended to allow the VoIP task to get the required QoS. The development environment can be based on MIPS Technologies’ Symmetric Multi-Thread Context (SMTC
) Linux Kernel, by performing dynamic allocation of the TCs under the Linux scheduler—expanding Symmetric multi-Processing (
SMP
) to multi-threading capabilities.
Multi-threading
In the 34K core, multi-threading is used for better CPU processing power utilization. This approach brings the parallelism era to VoIP and data applications, resulting in better overall performance. AudioCodes’ VoIPerfect engine is designed and implemented with a high degree of concurrency, allowing two or more independent tasks to be executed simultaneously. This parallel execution results in a speed-up of between 1.2x and 1.5x compared to a single thread execution, depending on the exact algorithms being used.
AudioCodes’ VoIPerfect™ Software Structure
Solution Demonstration
AudioCodes’ software is optimized for MIPS cores, including the high-performance MIPS32® 24K® core and the multi-threaded 34K core; and is scalable to the next level of performance with the superscalar MIPS32® 74K™ core and the multi-threaded, multi-core MIPS32® 1004K™ Coherent Processing System. The solution can be currently demonstrated on the MIPS
Malta™ Development System. The demo, picture below, presents AudioCodes’ unique proprietary handling and management of QoS that preserves voice quality.
About the Author: Shaul Weissman is the Vice President of the Chips Business Unit at AudioCodes.
For more information about AudioCodes’ VoIPerfect availability and features, contact Walter Juras at Walter.Juras@audiocodes.com.
|
