Voice Over IP

Introduction

Internet Protocol (IP) telephony can yield big cost savings to both corporations and consumers. It is more efficient than the plain old telephone service (POTS) and is poised to undergo huge growth. Before that growth can occur, however, designers who want to use the technology have to vault hurdles concerning latency, quality, and security. Quality of service (QoS) is the primary problem impeding this growth. A traditional problem with this technology, QoS must improve enough to enable Internet-based services to compete with traditional telephony providers.

IP telephony promises free, feature-rich telephone services, but quality, reliability and security issues keep some industry experts doubting. To successfully implement the technology, designers must consider chips, software, code longevity, and vendor support.

The term "IP telephony" covers a range of technologies, including voice-over-IP (VoIP) and fax-over-IP services, which are carried over both the Internet and private IP-based networks. IP telephony is part of packet voice, which includes voice-over-asynchronous-transmission-mode (ATM) and frame-relay networks, which run faster than IP but are less common. IP telephony connects across combinations of PCs, Web-based telephones, and phones connected via public telephone lines to remote voice gateways. Because information travels in discrete packets, it doesn't need to rely on a continuously available switched circuit. Consequently, it's very bandwidth- and cost-efficient

IP Telephony Yields Big Cost Savings

The most common reason to use Internet Protocol (IP) telephony is to save money. Corporations can save by running intracompany voice and fax in the spare bandwidth of leased lines. Consumers can save by connecting computers to computers and by using ordinary telephones to connect to Internet-telephony service providers (ITSPs), which use IP to provide low-cost voice/fax connections through combinations of the Internet, leased lines, and the public switched-telephone network (PSTN). Using IP telephony saves money because it is more efficient than ordinary plain old telephone service (POTS) and because it avoids most of the tariffs and tolls telephone companies are subject to, especially in the monopolistic international telephone-service market.

IP telephony is efficient; IP voice conversations require less than 10% of the bandwidth of POTS. This situation results from two factors: First, compression techniques, such as G.723.1, compress the 64 kbps POTS takes to 6.3 kbps. It is true that the 6.3 kbps grows when adding the IP overhead of about 40 bytes per packet, but an overall reduction of 6-to-1 is realistic. Second, POTS requires full duplex-equivalent to 64 kbps in both directions-to support a telephone conversation. But that feature is wasteful because in conversations, only one person is speaking most of the time. Voice-over-IP (VoIP) products sense the silence to cease transmission when one party is quiet. This technique almost halves the required bandwidth. In the end, IP telephony commonly takes as little as one-twelfth the bandwidth of POTS to transmit conversations.

Voice Quality

Although reasons exist to use IP telephony, Quality of Service (QoS) has been poor enough to single-handedly limit the adoption of this technology. The two most serious problems with QoS, latency and lost packets, relate closely to each other. Every VoIP application involves converting speech into a series of packets, each containing about 30 of voice. After the speaker-side gateway receives the voice transmission, it converts it to packets and shuttles the packets onto the network. Packets traverse the network and reunite at the receiver's end. The network may lose some packets, and others arrive too late to use in the reconstructed speech. In either case, the speech plays back without these lost packets.

Processing and transmission delay all packets, and this delay causes latency in conversations. The transmission leg across the network is the longest, especially on the Internet. Users expect latency of 250 msec or less-equivalent to the delay of a satellite link for international calls-for "toll-quality" service. Unfortunately, the Internet induces latencies that can far exceed 500 msec

Jitter buffers also contribute to latency. Jitter is the speed variation between quickly and slowly traveling packets. The jitter buffer stores packets, allowing most of the slower packets to catch up. The less control in routing, the more jitter that results, and more jitter means a longer jitter buffer. But a longer jitter buffer introduces more latency. Too short a jitter buffer loses too many packets, causing voice quality to tumble.

When the network loses a packet, VoIP products "reconstruct" it. The products cannot determine the information in the packet, but, like CD players smoothing over scratches, VoIP algorithms produce transitions that are less distracting than silence. Still, too many lost packets degrade voice quality to unacceptable levels. The maximum level of lost packets for toll-quality service is difficult to define, but 10% is a common level.

One cure for latency is to use a network that allows complete control of routing: Corporations use private intranets or Internet Telephony Service Providers (ITSPs), companies that use the IP networks to provide low-cost telephone services, usually relying on their own backbones. Private networks improve latency but at significant cost.

Compression Techniques

Some early voice-compression techniques were so severe that the reconstructed speech sounded robotic. The G.723.1 algorithm, probably the most popular compression/decompression (codec) algorithms in the H.323 standard, seems to solve this problem. G.723 can transmit voice on a 6.3-kbps stream, not counting IP overhead, and still score high in voice quality.

In addition to QoS, two other issues arise with IP telephony: IP telephony may offer neither the reliability nor the privacy most of us take for granted with conventional telephones. IP telephony may address security for VoIP, but the solution may be encryption rather than laws enforcing privacy, as with conventional telephones.

Choosing The DSP

Most IP-telephony products rely on DSPs for the heavy processing that voice codecs require. So, once you decide to build an IP-telephony product, how do you choose a DSP? First, you need to understand the algorithms the DSP can run. Most chip vendors provide the software for most of these algorithms either directly or through third parties. Many protocols exist for voice, fax, and data modems.

In most products, the voice chip does not handle the IP call stack. This fact is especially true in products for voice gateways that carry hundreds or thousands of channels. The intense computational requirements of voice mean that even a powerful DSP can handle just a few channels. The centralized nature of the call stack means that one process manages many channels, and this constraint implies that one chip processes the call stack and controls many voice-processing DSPs. Even in simpler products, such as Internet phones, processing the call stack differs so much from processing voice that separate chips are often required.

High Density

IP-telephony products fall into two broad categories: infrastructure products, such as voice gateways, which simultaneously handle many channels, and terminals, such as Web phones and videoconferencing systems, which simultaneously handle one or perhaps a few channels. High density is the goal of infrastructure products, and this density drives the main hardware features for VoIP chips: high speed, compactness, and low power. High-speed processing supports many connections in one chip, small chips mean that you can fit many onto a card, and low-power requirements reduce heat buildup. All three work together to maximize the number of simultaneous calls a system can handle in a small space.

Another factor driving density is that customers add voice cards to existing equipment, such as routers, and, often, just a few slots may be left. Low power is critical because the power supply is fixed. The amount of hardware a chip integrates also affects its size.

Density is not an issue for terminals such as Internet phones. Most of today's DSPs can handle a voice connection, so speed is less of a concern. Because only one channel exists, minimizing the power is less of a concern. However, cost is an issue for terminals.

Beyond power, size, and speed, you must also consider whether the chip includes software from the vendor or a third party. Because voice codecs are notoriously difficult to write and test, writing your own software may make little sense.

Software longevity is another factor in processor selection. Some companies are careful to keep future processor families backward compatible.

Ease of programming is a subjective feature that depends on the tools and on the processor structure. Also, consider upgrading in the field. Because IP-telephony standards are evolving, some vendors focus on helping customers build adaptable products. Be sure to evaluate the hardware and supporting software with this feature in mind . The key for chip designers is to think about the total solution for the application. The solution includes the chip, the preprogrammed software, the software tools, ease of programming, code longevity, and vendor support.

Software Architecture

The general modules of an IP Telephony system are : User Interface, Voice Processing, Telephony Signaling Gateway, Network Interface Protocols, Network Management Agent, and System Services. These subsystems are described below.

User Interface

The User Interface subsystem provides the software components that handle the interface to the user of the IP Telephone.

Voice Processing

The Voice Processing software is composed of the following software modules:

PCM Interface Unit

Receives PCM samples from the analog interface and forwards them to the appropriate DSP software module for processing. It also forwards processed PCM samples to the analog interface.

Tone Generator

Generates call progress tones to the user and generates in-band DTMF digits to the network based on key presses relayed from the User Interface. For certain voice codecs, the compression algorithm does not permit faithful transmission of DTMF tones. For those algorithms, e.g., G.723.1, the software generates an in-band message to the network that is used by the remote IP telephone (or gateway) to regenerate the DTMF tone.

Echo Canceller Unit

Performs ITU G.165 (line echo cancellaion) & G.168 (network echo cancellation) compliant echo cancellation on sampled, full-duplex voice port signals. Echo in a telephone network is caused by signal reflections generated by the hybrid circuit that converts between a 4-wire circuit (a separate transmit and receive pair) and a 2-wire circuit (a single transmit and receive pair). These reflections of the speaker's voice are heard in the speaker's ear. Echo is present even in a conventional circuit switched telephone network. However, it is acceptable because the round trip delays through the network are smaller than 50 msec. and the echo is masked by the normal side tone every telephone generates. Echo becomes a problem in Voice over Packet networks because the round trip delay through the network is almost always greater than 50 msec. Thus, echo cancellation techniques are required. ITU standard G.168 defines performance requirements that are currently required for echo cancellers. Echo is cancelled toward the packet network from the telephone network. The echo canceller compares the voice data received from the packet network with voice data being transmitted to the packet network. The echo from the telephone network hybrid is removed by a digital filter on the transmit path into the packet network.

Voice Activity Detector (VAD)

Detects voice activity and activates or deactivates the transmission of packets in order to optimize bandwidth. When activity is not detected, the encoder output will not be transported across the network. This software also measures Idle Noise characteristics of the interface and reports this information to the Packet Voice Protocol for periodic forwarding to the remote IP Telephone or gateway. Idle noise is reproduced by the remote end when there is no voice activity so that the remote user does not feel that the line went "dead."

Voice Codec Unit

Performs packetization of the 64 kbps data stream received from the user. Various compression algorithms exist which have different performance characteristics: G.711 PCM which operates at 64 kbps (no compression), G.726 ADPCM which operates at 16, 24, 32 and 40 kbps, G.723.1 which operates at 5.3 or 6.3 kbps and G.729 which operates at 8 kbps. Typically, voice algorithms that perform greater compression require much more processing power. It should be noted that high fidelity audio quality compression algorithms can also be used since an IP Telephone is not subject to the 4 kHz bandwidth restrictions found in the PSTN. This would provide better sounding audio than PCM and allow music to be faithfully reproduced.

Packet Playout Unit

Performs compensation for network delay, network jitter and dropped packets. Many proprietary techniques are used to address these problems since there are currently no standards in place for packet playout.

Packet Protocol Encapsulation Unit

Performs encapsulation of the packet voice data destined for the network interface. For VoIP this encapsulation is per the Real-time Transport Protocol (RTP) which runs directly on top of UDP.

Voice Encryption

Provides optional encryption of the voice packet data prior to transmission over the network to ensure privacy.

Control Unit

Coordinates the exchange of monitor and control information between the Voice Processing Module and Telephony Signaling and Network Management modules. The information exchanged includes software downline load, configuration data, signaling information and status reporting.

Telephony Signaling Gateway

The Telephony Signaling Gateway (TSG) subsystem performs the functions for establishing, maintaining and terminating a call.

Network Management

The Network Management subsystem supports remote administration of the IP Telephone by a Network Management System.

GAO Solution

GAO offers designers a complete VoIP package, which may include modules for fax and data over IP. This reduces the development time and time to market.

The following modules are part of the package:

Call discrimination	Detection of a voice, fax, or data call so that it can be switched to the appropriate functional module.
Data modem/fax interface	Interfaces to these modules
Echo cancellation	Line echo cancellation according to G.165 and Digital network echo cancellation according to G.168
Voice activity detection and Comfort noise generation	Voice connections are almost entirely one-way. GAO VoIP software halves the transmission bandwidth by recognizing the periods when a party is silent and ceasing transmission during those periods. GAO software also provides comfort noise generation.
Tone Detection/Generation	DTMF detection/generation
Speech vocoder (audio codec)	Any codec can be chosen from G.711, G.722, G.726, G.723.1, G.729A, G.729AB
Lost packet detection and reconstruction	GAO VoIP products detect when packets are lost and reconstructs them either by repeating information in subsequent frames or using an upper layer protocol i.e. TCP
Jitter buffer management and sequencing	Packets are stored in a jitter buffer and sequenced so that the slower packets can catch up to the conversation.
Packet protocol	The Protocol Processing module receives signaling information that has been interpreted by the Telephony Processing module and converts it from the telephony signaling protocols to the specific packet signaling protocol that is used to set up connections over the data network. It also adds appropriate protocol headers to both voice and signaling packets before transmission.