Given the increasing availability of broadband communications and WiFi access points and the growing popularity of Voice-over-IP (VoIP), it was only a matter of time before the two technologies converged. VoIP-over-WiFi (VoWiFi) promises to be the ideal application to drive both VoIP and broadband adoption forward. Already, according to a recent research report from Infonetics, some 113,000 VoWiFi handsets were sold in 2004, with strong adoption rates expected through 2009.
However, for VoWiFi to become accessible and adopted by the masses, a number of technical hurdles must be addressed by handset OEMs and ODMs. For example, much of the current WiFi installed base is running the older 802.11b standard of 11 Mbits/s, even as new standards such as 802.11g deliver speeds up to 54 Mbits/s.
Coupled with the shared nature of a wireless network, varied network data rates means consideration must be given to the amount of bandwidth consumed for each voice call. The likelihood of bandwidth constraints is best addressed in the selection of voice codecs that are supported on a VoWiFi device. In addition, bandwidth issues within a WiFi network drive the need for an inherent level of quality of service (QoS), both on the WiFi network and on the overall end-to-end network.
VoWiFi handsets are viewed by many as the VoIP replacement for today's cordless or Digital European Cordless Telecommunications (DECT) phone. So users will expect similar standby and talk times, along with increased functionality. Power management and the ability to extend overall battery life are therefore crucial elements of any VoWiFi device.
VoWiFi Bandwidth Considerations/Codec Selection
When comparing the typical bandwidth per channel of VoIP against the bandwidth available from most WiFi access points, it would seem that a discussion of bandwidth availability for a VoIP application should not even be necessary. After all, even the "old" 802.11b provides speeds of 11 Mbit/s. Since even G.711 runs at a relatively paltry 64 kbit/s, this seems to be vastly more than the minimum required for VoIP codecs.
However, consider the myriad of applications contending for the WiFi airwaves: streaming MP3 audio from home media servers, multi-player gaming, Web surfing, and other device-to-device data traffic. Then, consider the various environmental factors that can rapidly degrade an 11 Mbits/s connection down to 5.5, 2 or even 1 Mbits/s. The "vast difference" in the bandwidth needed for a single VoIP call and the actual remaining bandwidth available for that call is no longer quite so vast.
VoIP is designed to operate in a variety of network conditions, including networks with limited bandwidth, so techniques have been developed to compress a standard telephony-grade signal into a more efficient bitstream. Traditional circuit-switched plain old telephone service (POTS) captures a voice call at a 64 kbits/s data rate. This coding scheme, defined by G.711, is still widely used for VoIP to provide "toll-quality voice" due to its simplicity of implementation and its inherent compatibility with circuit-switched networks.
To provide a more efficient use of network bandwidth for voice transmission, a number of alternative voice codecs have been developed. These codecs, such as G.729, G.723.1 and iLBC, use a variety of complex processing techniques to compress the captured voice stream, meaning that substantially less network bandwidth is required to transmit the same channel of voice. However, the consequence of compressing the voice channel into a smaller bitstream is varying levels of impact on the voice quality, as well as an increase in the processing power required to execute the codecs.
Taking advantage of the processing power available in today's communications devices, a number of techniques have been developed to enable VoIP calls to sound even better than calls on the POTS network. Among these techniques is the use of wideband codecs that have been developed with the goal of providing higher fidelity speech replication than G.711. With higher fidelity, more bandwidth is used to packetize the speech. In most cases, wideband codecs still consume less bandwidth than G.711; however, they typically consume much more of the processing cycles of the DSP or CPU on which they execute. As such, the trade-off expands to one that includes balancing perceived voice quality, network bandwidth consumption, and processor performance.
