The rapid growth of Voice over IP (VoIP) is pushing IP phone suppliers to add new capabilities and features while continually improving the quality of voice service. To meet these demands, IP phone designers need processors that provide a combination of real-time performance, low power consumption, and the flexibility to add features.
In this article, we will look at the requirements of today's VoIP phones. We will then look at emerging technologies such as high-definition (HD) voice, fixed/mobile convergence and PC/telephony integration. Along the way we will explain how current and emerging technology dictates the requirements for a DSP platform, and explain what phone designers should look for in a DSP.
[If you're new to VoIP, check out How VoIP works: protocols, codecs, and more for a great intro.]
Real Time Voice Processing
Looking back a few years, the first hurdle for VoIP technology was measuring up to the long established quality-of-service (QoS) of analog plain old telephone service (POTS). Once this hurdle was cleared, designers began to implement IP phones that not only met this "carrier grade" audio quality QoS, but exceeded it.
SD and HD Voice
The latest shift in voice quality is the transition from standard definition (SD) to high definition (HD). Other digital media such as television, DVDs, radio and music are migrating from standard definition to high definition, spurring growth in a wide range of markets. Although the telecommunications transport network was the first media to embrace digital technology, it has yet to transition to HD.
When digital technology was first deployed for the telecommunications market, the decision was made to standardize on SD voice with a frequency range of 200 Hz to about 3,300 Hz, in order to conserve network bandwidth. This range clips speech well below the normal hearing range, which can range up to 20,000 Hz. This helps explain why a phone conversation over the POTS network is never quite as good as speaking with someone in person.
Over time there have been many upgrades to the telephone backbone network, including conversion to digital technologies. However, the "last mile" connecting homes and small businesses to the network remained dominated by low-bandwidth twisted-pair copper wiring.
Now, subscribers are using broadband technologies without the 3,300 Hz limit for their last mile connections to the infrastructure. In some areas, fiber optic and other high-capacity media are replacing copper wiring to the home. The data throughput rates of broadband connections have also continued to accelerate. Typical residential broadband speeds have reached five megabits per second (Mbps) for downloads and two Mbps for uploads. At the same time, the DSP computing power needed for voice has reached a very low price point. These factors contributed to the rise in VoIP services, which have arrived in homes via IP phones, routers and other equipment. These same factors are also enabling the use of HD codecs.
Since the human ear has a range of hearing from 20 Hz to 20,000 Hz, HD voice is a much better option than SD voice. HD voice can provide sounds over a much wider spectrum (Figure 1). As a result, the ear will be able to perceive the slightest and most subtle nuance during a typical HD telephone conversation. If the same codec is used for both HD and SD voice, the data bandwidth needed for the wider HD spectrum (256 Kbps channel) is significantly greater than SD voice (64 Kbps). However, advanced codecs such as the AAC-LD codec can deliver near-CD quality audio at data-rates of 48-64 Kbps.
Figure 1: SD and HD codec comparison.
Aside from offering the benefit of a more compelling audio experience, HD voice will become the basis for many new VoIP applications, including enhanced natural language voice control and real-time translation services. In addition, HD voice could provide the impetus needed to make voice the primary user interface for all internet-connected devices.
[For a more detailed discussion of HD voice, see HD Voice is Coming.]