Design challenges ahead for media gateway

by By James Awad, Octasic Inc. , TechOnline India - December 15, 2008

Media gateway developers, manufacturers and service providers are searching for cost-effective ways to meet market demand within reasonable time periods.

Media gateway developers, manufacturers and service providers face a daunting task: looking for ways to lower cost and meet time-to-market demands as consumers' insatiable desire for more advanced applications accelerates at an unprecedented pace. Nowhere is this challenge more evident than in the digital signal processor (DSP) development community where cost, time-to-market and shelf life are the mantra to usher media gateways into the future.

Media gateway vendors are continuously challenged by new network requirements that demand advanced voice signal processing, video support and more stringent quality metrics, while DSP vendorsmust demonstrate a broad experience in both voice and video processing in order to deliver on the promise of better quality.

Initially, media gateways, based on industry standards, were designed using generic DSPs for a specific purpose--Voice over Internet Protocol (VoIP). As the market grew, DSP vendors began to provide system-on-a-chip (SoC) solutions that included all the necessary hardware to perform TDM-to-IP (time-division multiplexing to IP) or IP-to-IP conversion. This often involves having the right balance of performance-to-power consumption and the appropriate physical input/output (I/O) interfaces.

Click on image to enlarge.

Standards are important to ensure interoperability between systems and to help drive down cost by offering a common language between vendors and developers. The large number of standards for voice and video codecs, for example, is the main "raison d'être" of media gateways. Media gateways serve as the universal translators between many of these different standards, and therefore, as new standards are developed and deployed, the gateways must also evolve and provide the platform for new and emerging technologies and applications. This accelerated technology development has caused a shift in the way media gateway and DSP engine developers prepare for the future.

VoIP is not just VoIP anymore

About eight years ago, the industry began to see the emergence of complete media gateway products, offering both hardware and firmware components, which enabled original equipment manufacturers (OEMs) to focus on upper-layer application software. These products have matured over the past decade to a point where the underlying DSP technology has stabilized and the basic feature set is well understood.

While the basic TDM-to-IP media gateway market has matured, a slew of new requirements--to support not only pristine quality in VoIP but also emerging applications--have appeared. The emergence of session border controllers (SBCs), fixed mobile convergence (FMC) and video communications, as examples, have created specific design challenges.


After reaching a critical mass of VoIP subscribers, many network providers, thanks to SBCs, can connect their networks to that of their competitors directly, without bridging back to TDM anymore. SBCs are also used to interface with wireless networks, serving as converters between these IP-only networks. With TDM out of the way, the focus is now on obtaining the best voice quality in the packet-only domain.

One of the most important factors in packet-to-packet conversion is latency. Traditional media gateway processors incur processing latencies of 10–15 milliseconds (ms), due to their fixed TDM packetization scheduling. Newer DSP solutions can cut this latency to less than 5 ms.

SBCs must also address voice-quality issues. The traffic entering SBCs can come from a variety of sources, ranging from mobile phones (often in noisy areas) to residential gateways (ATAs) that often have less than stellar echo-cancellation performance. These types of environments pose a challenge to the media gateway, which must "clean up" the voice signal intelligently--that is without making things worse! Tandem echo cancellation can be challenging for both acoustic and line echo.

FMC, which has been a long time coming, allows mobile subscribers to use their mobile phones near a wireline base station, whether it is a Wi-Fi router or a cellular femtocell. The first wave of dual-mode mobile phones were introduced a few years ago, but only now are carriers getting some traction in the market, either due to more aggressive pricing schemes or improved battery life for the mobile phones.

FMC places specific requirements on the gateway as well, for example, the need to support longer Real-time Transport Protocol (RTP) packets. Due to the data-oriented model of Wi-Fi transmission, voice packets must contain more bytes in order to accommodate the long periods of time when nothing can be transmitted. Certain DSP solutions are not able to support these large packets.

While both SBCs and FMC are interesting technologies that are coming into their own, video support will likely be the next driver for huge revenue growth. It is only a matter of time before the cost is reduced sufficiently to enable widespread adoption.


Currently, video support is a premium feature that comes with a higher price tag. Although most customers are not interested in video, it becomes very expensive to add later into the DSP design. As a result, it is important that the video and audio processing be tightly coupled in order to ensure things such as audio/video synchronization (lip-sync) and proper handling of packet loss.

From a hardware perspective, video presents the same challenges that voice did a decade ago. The amount of processing required is so great that power consumption becomes a limiting factor when establishing the number of video streams that a blade can support. Most DSPs have gone multi-core in an attempt to squeeze more million instructions per second (MIPS) into the same power budget, although this alone is not sufficient.

New DSP architectures are emerging that have improved power efficiency by orders of magnitude. For example, the use of asynchronous logic in the DSP core allows for two to three times better power efficiency than traditional DSPs.

With improved power efficiency, voice gateways can now achieve increased density, which allows for cost reductions. For video gateways this also means that higher video resolutions can be supported within one device, which is tantamount in the race to support high-definition video. Therefore, the ideal media gateway processor should support both voice and video processing, as well as enable video through software upgrades.

There is no doubt that DSP and gateway vendors must adopt a new business model that meets these new network requirements. One way to achieve this is through a "mass customization model" where the hardware can accept any software-driven application with the flip of a switch. These hardware and software solutions must be designed with the licensing mechanism as an integral part of the media gateway software developer's kit.

Mass customization will drive the next wave of cost savings for equipment manufacturers, delivering optimal price points through software licensing and adding new features such as video services. This will help OEMs maintain higher margins in the face of fierce competition.

James Awad is a product manager at Octasic Inc. He received his BE from Concordia University in Montreal.


blog comments powered by Disqus