Voice-over-Wi-Fi Implementation with Single Stream 802.11n
Sep 21, 2009
By Narasimhan Venkatesh, Chief Wireless Architect; Peddi Indukuri, Product Manager; and Subba Reddy, Engineering Manager, Redpine SignalsThe growth of wireless networks based on the IEEE 802.11 Wireless LAN family of standards has been one of the most outstanding success stories of the technology industry in recent years. Apart from the standards themselves, the universal pervasion of WLANs has been assisted and accelerated by the availability of interoperability testing and certification by the Wi-Fi Alliance – so much so that the term “Wi-Fi” is widely used interchangeably with “WLAN.”
The initial growth of WLAN was in its intended role of providing a wireless data networking capability as a replacement to a wired LAN connection. However, as its capabilities grew—with the standards being enhanced to offer higher data rates, better quality of service and special modes such as power-save—it quickly became an integral part of a large variety of electronic devices including phones, gaming devices, music players, sensors and other consumer devices. Among these, one of the fastest growing applications has been the transport of voice over the wireless network—thanks in part to the popularity of several commercial Voice-over-IP (VoIP) services.
The Wi-Fi Alliance, taking cognizance of the significant potential of Voice-over-Wi-Fi (VoWiFi), has released a certification program called Voice-Personal that helps ensure that the underlying requirements of VoWiFi in Wi-Fi devices are met, in a home or small office environment. Expanding their focus of certification beyond protocol adherence and interoperability, the Wi-Fi CERTIFIED Voice-Personal program focuses on a specific application and is based on performance testing. The program was released in July 2008 and Redpine Signals’ Lite-Fi is among the first products to support the certification.
In this article, we provide a background to VoWiFi performance by examining the factors that enable it to provide a satisfying user experience. We elaborate on some of these requirements and describe how they are implemented in VoWiFi devices.
Requirements of VoWiFi
Voice has traditionally been carried over fixed latency, connection-oriented, low error rate transport medium. These attributes have to be specially provided for in networks based on WLAN. From a user’s point of view, the experience of a voice call over WLAN would be similar to that over an alternative designed-for-voice network if certain requirements are met. These include the following:
Latency: A two-way, interactive communication like voice requires the medium to introduce limited packet latency. The connection should permit a tempo of speech to that in a face-to-face conversation. The generally accepted limit on latency in a VoWiFi network is 50 ms.
Jitter: This is the variation of the time of arrival of packets. Although handled by a jitter buffer at the receiving end, the network is nevertheless required to curb this.
Packet drops: The protocols employed to carry voice packets do not provide for a mechanism to re-transmit packets lost in transmission. Indeed, even if they did, it would not be effective since a retransmitted packet would almost certainly have latency requirements violated. The WLAN protocol, of course, does provide for packet retries at the MAC level and this mechanism helps take care of the occasional packet errors that occur even in good channel conditions. VoWiFi devices are expected to limit packet loss to a few percent in good to average channel conditions.
Power savings: A wireless device would be actually communicating only part of the time—wireless phone, for instance, would be “in use” typically only an hour or two a day. However, the devices would be expected to be “on” all the time—ready to receive a call if one was to come in, and ready to respond to the user to make a call instantly. The original 802.11 specifications do provide for a client device to be in a “sleep” mode and wake up occasionally to check for pending packets, but this power-save specification is not helpful for voice calls where packets arrive every 20 to 30 ms. The 802.11e standard and the Wi-Fi certification of WMM-PS provide for viable power-save states while maintaining other requirements specific to voice calls. This article describes this in detail.
Range of operation: Voice requires a bandwidth that is only a small fraction of the data rate possible in a Wi-Fi network, but it requires it reliably in a location-independent manner. The wireless client implementation at the physical layer is therefore required to handle the multipath and interference scenarios that occur in corners and other remote locations of offices and homes.
Roaming: A VoWiFi user in a large office could easily wander beyond the range of his Access Point while attending a call. Although enterprises are equipped with sufficient access points so as to provide Wi-Fi connectivity throughout the premises, the switch-over from one AP to the other—roaming—must be done quickly enough to keep the voice connection unbroken to the user. The mechanisms involved in roaming are described in greater detail in the sections that follow.
Read the full article.