Audio Transcoding

Delivering Voice and Other Audio Services Using Efficient DSPs


Audio transcoding, especially for voice, is a critical function in telecom applications such as session border controllers, media gateways and media servers, or media resource appliances.

It is the essential technology behind conference call applications such as Zoom, Microsoft Teams and countless other voice-over-IP applications, including the provision of voice services on 5G and some LTE wireless networks (VoLTE/Vo5G).

There are well-established standards for audio codecs such as EVRC, OPUS, SILK and the newer and more complex EVS codec, ensuring interoperability and simplifying system design.

Transcoding multiple channels in real time at high densities is a computationally intensive task, which is performed most efficiently using specialty processing, such as digital signal processors (DSPs) or ASICs, rather than relying on the server processors alone.

The conventional solution is to add more processing power through more servers, which not only takes up space and adds provisioning issues, but also increases both capital and operating costs for the operator.

An alternative approach is to use a PCI Express add-in card with DSPs and software-definable transcode capability to offload from the main processor and greatly increase the total transcode availability within the server. This not only avoids the expense of a server ‘rip-and-replace’, but also conserves space as the increased processing capacity still takes up the same amount of space in the rack.

Introducing PCIE-8130

The PCIE-8130 from SMART Embedded Computing is a PCI Express media processing card that offers high performance voice transcoding based on DSP technology.

Each board features an array of low power DSP devices running optimized voice processing software. The transcoding performance scales linearly according to the number of DSPs that are fitted: a half-length PCIe card variant offers options for four or six DSPs; while the full-length PCIe card variant features up 12 DSPs.

Even with four DSPs consuming less than 60W of power, the PCIE-8130 delivers a voice transcoding performance comparable to a typical server consuming 300W or more, estimated at over 5,000 channels of G.711 20ms to AMR transcoding.

SMART EC estimates that using a PCIE-8130 add-in acceleration card offers up to 4X improvement in cost-per-channel transcode, depending on specific configurations and applications, and up to a 10X reduction in watts-per-channel transcode.

Application developers interact with the board via a simple object-oriented application programming interface (API). All common standards-based audio codecs can be accessed through a simple API to a sophisticated network-proven software framework, providing transcode, signaling and call management functionality. In addition, the software-definable nature allows future codec developments to be realized on the same hardware.

This greatly reduces development and maintenance costs, allowing for a dramatic acceleration in time-to-market.

The PCIE-8130 uses the latest OCT3032 DSPs from Octasic.

The PCIE-8120 uses OCT2224 DSPs from Octasic.

Let us Help You With Product Selection