(This article is sponsored by The Boston Group)

Introduction

VoIP stands for Voice Over IP. IP is the familiar Internet Protocol, the procedure for information (in packet form) transfer used on the Internet. VoIP is about the use of IP networks to provide telephone services.

Traditional phone services refer to telephone and fax. The players and technologies behind traditional phone service include the local service provider, the local loop (line from the central office (CO) to your home), the voice switch at the CO, the long-distance carrier and its signaling network. Voice is converted from audio to electrical signal by the microphone in the handset, digitized (converted into discrete values, specifically, 1s and 0s) by a voice coder in the CO voice switch and then carried, in digital form, across the telephone circuit network. At the far end, the digital signal is converted back to analog electrical form in the destination local loop and finally, transduced into audible voice by the speaker in the handset of the called party. In addition to transporting voice, the telephone service features the dial-tone whenever the the phone is lifted off-hook, the telephone number and a host of other features, such as, call forwarding and 3-way calling.

The VoIP framework is made up of the local service provider, with its voice switch , the Internet Telephone Service Provider (ITSP) with a device called gateway that transfers the voice between the telephone network and the IP network, the Internet Service Provider (ISP) and of course, an IP network. Another device called the gatekeeper is used for translating between telephone number and IP address. In VoIP, voice is converted from audio to electrical signal by the microphone in your handset and digitized by the voice switch. The digitized voice is then packaged by the gateway into an IP packet and carried across an IP network. At the far end, the digital voice is extracted from the IP packet and converted back to analog electrical form in the destination local loop and finally, transduced into audible voice by the speaker in the handset of the called party.

Although, this sounds straight forward, VoIP deals with a number of challenges arising from the characteristics of voice and the differences between the circuit-oriented telephone network and the packet-oriented IP network.

The Challenges

Voice is characterized by intolerance to delay (the time it takes for a voice sample to travel to the other end) and delay variation (changes in the delays from one voice sample to the next). Excessive delay results in echo. To be intelligible, voice has to be played out at a fixed rate. Thus, variation in the rate at which packets arrive at the receiver could render voice unintelligible.

Traditional telephone network is well suited for voice transmission as it provides a low, guaranteed delay and a dedicated path (and hence bandwidth) for each call, once the call is established .

IP network, on the other hand, is characterized by delays, delay variations, non-dedicated or shared path, packet loss and packets arriving out of order. IP network was designed for data traffic, which is bursty in nature.

End-to-end delay in the VoIP environment is composed of,
· Accumulation delay, the time taken to compress of block of voice samples,
· Processing or packetization delay, time taken to package voice samples in IP packet,
· Network delay, time taken by an IP packet to traverse the IP network and
· De-jitter delay, time to decompress, decode and play out the received voice samples.

Accumulation delay is controlled by using higher coding rate in the voice coder. Packetization delay could be caused by an over-utilized CPU on the gateway and can be improved by using faster processors. Network delay varies based on the traffic load on the network. It is minimized using special Qos (Quality Of service) techniques. Finally, de-jitter delay is managed by configuring the play out. In addition to designing to minimize the various delays, echo cancellation techniques are used to eliminate echo.

Delay variation is managed by using a jitter buffer at the called party end. Voice is played out from the jitter buffer. A side effect of jitter buffer however, is to increase the delay. Thus, trade off between total delay and jitter buffer size is an important design aspect. Another way to control the delay variation is to establish a "guaranteed bandwidth" pipe across the IP network, using resource reservation signaling protocols.

Order of packets is managed by using the Real Time Protocol (RTP), which carries timing or order information with each voice sample. This timing information is used at the destination to place the voice sample properly in the jitter buffer.

The voice decoder at the destination fills gaps caused by loss of packets, using interpolation techniques.

The Motivation

So, why migrate such reliable, feature-rich service to the VoIP framework? The initial drivers are economics, i.e. reduction in the cost of long distance and international telephone calls. This is because the IP network, unlike the traditional telephone network, is not subject to access fee and international settlement charges. Also, VoIP calls are not subject to duration-sensitive billing since IP network has a flat fee associated with it. For the service providers, savings result from having to maintain only one network (for carrying IP) rather than two (separate networks for IP and voice). However, as long distance becomes very cheap, the cost benefit may lose its attractiveness. The more significant benefits however, will come from the sophisticated applications such as multi-media conversations and unified messaging (for example, using a PDA to access your voice mail as well as email).

|