Electronic Design

VoIP: Packetized Voice Over Everything

Less maintenance costs, lower consumer monthly bills, and ease of convergence are pushing VoIP into the forefront past the tried-and-true POTS.

Though it has bounced around for over a decade, Voice over Internet Protocol (VoIP) has only recently seen a surge in activity, with clear movement toward widespread adoption. Slowly but surely, companies and individuals are replacing their standard plain-old telephone service (POTS) on the public switched telephone network (PSTN) with VoIP on the Internet.

Why on earth would we give up one of the best, most highly developed, and ultrareliable electronic systems ever built? Simply because we can. A more practical answer, though, is the array of benefits it offers— lower monthly telephone bills, the long-term reduced cost of maintaining both voice and data networks, and the ability to adopt a variety of converged services including video.

Gradually, companies are replacing their legacy phone systems and private branch exchanges (PBXs) with VoIP systems. Many consumers are buying phone service from cable TV companies and independents, such as AT&T and Vonage. Phone companies like Verizon (and SBC via its acquisition of AT&T) additionally are beginning to offer VoIP to their DSL customers. They won't be tearing down the PSTN any time soon, if ever. But as they lose revenue from their PSTN business, they're replacing it with VoIP services and eventually television.

Ann Powell, a consumer from San Antonio, Texas, switched her SBC standard phone service for Time Warner Cable's Internet phone service. She said the initial experience was rocky until the bugs were worked out, but now it works great. However, she didn't expect to lose phone service during a power outage.

The PSTN is totally self-powered, so customers have phone service even during the longest of power outages. But when the cable box and the analog terminal adapter (ATA) lose power, consumers can't use their VoIP phones. Also, there's no 911 service yet.

"I guess that is what cell phones are for," says Powell. Otherwise, she is happy with the service because it really has cut down her overall phone bill, including long distance.

The 911 issue is still being resolved. The FCC has required all VoIP carriers to provide a link to the local 911 service or cut off service to the subscribers. While most providers have complied, a few have yet to meet the requirement. Some companies have applied for an extension, which the FCC recently granted.

Jeanne Gibson, a telecom guru with Apache Oil Corp. in Houston, said her company just went through a major switch from standard phone service to a full IP-based system from Cisco. The switch at corporate headquarters went well, and the conversion at the various field offices is proceeding nicely. Long-distance service alone will save Apache costs in the mid-six figures annually, which is more than enough to pay for the switch.

Yet Gibson's main concern as a telecom person and not an IT person is to ensure that the IT people eventually get a voice service mindset, which currently isn't offered. The Holy Grail in any company is dial tone, not e-mail. While individuals will occasionally endure a short network outage, rarely will they sit still for a phone service failure. Quality of service (QoS) seems to be the first priority for IT departments that switch over.

All sorts of complicated operations take place during an Internet phone call, making POTS look as complex as a flashlight. Lots of different protocols are used. During the "dialing," the caller enters the phone number. This initiates a "connection," but it won't be a continuous connection like a POTS call. Instead, a packet protocol is used to establish an initial link.

Protocol H.323 from the International Telecommunications Union (ITU) is used to set up the call and maintain transmission. It doesn't guarantee QoS. However, H.323 is now giving way to a newer protocol called the Session Initiation Protocol (SIP). This Internet Engineering Task Force (IETF) protocol includes the individual protocols called Request for Comments (RFC), specifically RFC 3261, RFC 3262, and RFC 2543. The Media Gateway Control Protocol (MGCP) is another IETF standard, dubbed RFC 2805. The Media Gateway Control (Megaco) ITEF RFC 3015 also is defined by the ITU as H.248.

Next, an analog-to-digital converter samples the voice signal and converts it to serial digital format. The usual sample rate is 8 kHz, creating 8-bit voice words. This process creates a 64-kbit/s serial data rate signal, just as in other digital telephony. This is too fast for some connections, so a voice encoder is used to compress the digital data and reduce the serial data rate.

The ITU also defines the compression and decompression standards. Standard G.711 defines the 64-kbit/s signal, more commonly known as pulse code modulation (PCM). An 8-kbit/s voice signal known as G.729a probably is the most commonly used compression standard.

The G.723 compression standard produces a lower-quality 5.3-kbit/s signal, while G.723.1 produces a 6.4-kbit/s voice stream. Such high compression is possible because so much of a voice conversation is silence, including pauses between words and the conversation exchange. A DSP usually performs the compression as well as echo cancellation, a common problem in analog or digital telephony.

At the other end of the conversation, the compressed voice is decompressed and converted back into its PCM format. This format is then converted back to voice with a digitaltoanalog converter (DAC). The ADC and DAC plus the serialize/deserialize function usually are called the codec (coder-decoder). Then the compressed voice is packetized and transmitted over a local-area network, and subsequently the Internet. In many corporate systems, the packetized data is transmitted via the widely available asynchronous transfer mode (ATM).

The key problem in VoIP is the latency, or the time it takes for voice packets to traverse through all of the various LAN, metro-area-network (MAN), and wide-area-network (WAN) connections. Up to a 150-ms delay is permitted, as longer delays produce an annoying voice lag. For example, international reporters on television often take several seconds to respond to questions, which must pass through the 22,300-mile uplink and back. In two-way phone conversations, such delays are unacceptable.

Several different types of IP phone arrangements exist. Each phone in an enterprise system is an IP phone. This means it contains the codec, compression/decompression, SIP, and other circuits that generate Real Time Protocol (RTP) and User Datagram Protocol (UDP) packets, which eventually get translated into TCP/IP. These can go on the Internet. Or, the VoIP packets can be transmitted via ATM.

In most modern systems, VoIP packets are translated into Ethernet packets for connection to a LAN. The usual connection from a desk phone to the LAN is the ever-popular CAT5 UTP cable with RJ-45 connectors. The phone simply appears as a node on the enterprise LAN, and servers deal with it accordingly. In most cases, an employee has a PC as well as a phone. Thus, most current VoIP chips feature a built-in, twoport Ethernet switch so that both the phone and PC only need one LAN connection.

The typical home IP phone differs. While homeowners can buy an IP phone, in most cases any existing phone can be plugged into the RJ-11 modular jack on a piece of equipment called an analog terminal adapter (ATA) supplied by the IP phone provider. The internal home wiring can be disconnected from the phone company's incoming line, and all of it can connect to the RJ-11 input on the ATA. As a result, any phone in the house can make a call as usual.

The ATA contains the circuitry that terminates the standard phone and then takes the resulting analog voice and processes it (Fig. 1). It contains a subscriber line interface (SLIC) that the standard telephone expects to see. Afterward, the usual IP processing occurs, with the resulting Ethernet packets sent out over the broadband connection—either a DSL line or cable TV connection. With VoIP becoming more common, cable-TV companies and DSL providers are simply building the VoIP circuitry directly into set-top boxes and modems, readying them for VoIP service.

VoIP via wireless is possible, too. The most common arrangement is 802.11/Wi-Fi. NEC, Motorola, and Nokia already make dual-mode phones that include a standard cell phone plus Wi-Fi VoIP. Subscribers can make a call using the traditional cell-phone network or by Wi-Fi through an available hotspot or company wireless access point.

Such phones should be attractive for larger companies, because employees can be issued cell phones that also double as cordless internal telephones. Consumers can use such an arrangement in which the Wi-Fi phone talks to the wireless router connected to the broadband line, turning it into a cordless home phone.

This technology, Unlicensed Mobile Access (UMA), is also known as the Generic Access Network (GAN). It has been formally standardized by the 3rd Generation Partnership Project (3GPP), the standard organization associated with 3G UMTS WCDMA cell-phone technology. The UMA system uses the existing ITU standard, which includes H.323 for call setup.

A similar system, seamless converged communications across networks (SCCAN), was developed by Motorola, Avaya, and Proxim. It also uses GSM and Wi-Fi but incorporates the more popular SIP standard of the IETF. In addition, the IEEE is working on a standard (802.21) that will provide a common method of handoffs between 802 networks and non-802 networks.

Another IEEE effort, the 802.11r standard, defines a protocol that facilitates the use of IP telephone over Wi-Fienabled phones. It will speed handoffs between access points in a wireless LAN. However, CDMA manufacturers and providers like Qualcomm, Verizon, and Sprint haven't yet bought into this technology.

Don't forget WiMAX, either. These broadband wireless networks are expected to spring up in many locales beginning next year. With many local WiMAX cell sites installed throughout the area, a WiMAX cell phone can roam around the vicinity communicating with standard VoIP technology. Without a doubt, we can expect at least one cell phone to incorporate all three modes.

VoIP's heart lies in its software. This includes the voice compression/decompression, jitter buffering, and echo cancellation usually handled by digital signal processing. Some systems have standalone DSP chips, while others have built-in DSP on a chip. Still others run on an ARM or MIPS core with DSP instructions. The SIP or H.323 and other protocols run in stacks on an embedded controller like the ARM, MIPS, or Power PC.

For instance, Broadcom's BCM1101 and BCM1103 form the core of many enterprise IP phones. The BCM1103 features dual 16-bit ADCs and DACs (codecs). Its DSP core handles the voice compression and decompression with G.711, G.729a/b, G.723.1, G.726, and Broadcom's own BroadVoice16. It also has broadband voice coders such as G.722, G.722.1, and Broad-Voice32. A MIPS32 core runs the show and can handle the SIP, H.323, MGCP, and Megaco/H.248 protocol stacks. The chip features two 10/100/1000 Ethernet MACs, two 10/100 Ethernet transceivers, and a three-port 10/100/ 1000 Ethernet switch. Two RGMII ports are available with external Gigabit Ethernet transceivers.

Centillium makes chips for the customer-premise-equipment (CPE) and infrastructure ends of the VoIP path. The Entropia III CT-GCW4672 is a system-on-a-chip (SoC) processor for implementing voice and media gateways, wireless-infrastructure gateways, class 4 and 5 switchers, digital loop carriers, voice-enabled IP routers, and IP PBX systems. With multiple embedded DSPs and packet processors, its primary function is to bridge standard telephone TDM and VoIP wireless and wireline converged voice calls.

It has an amazing 1008 G.711 (PCM) VoIP or voice-over-ATM (VoATM) carrier-class voice channels with 128-ms echo cancellation. It can handle any of the ITU voice-compression standards as well as voice codecs used in CDMA, WCDMA, GSM, and UMTS cell phones. Interfaces include 16 standard TDM circuit pairs to a standard switched network and MII and GMII Ethernet physical-layer (PHY) ports, POS-PHY II, and Utopia for ATM network connection. A host processor interface handles PCI or Freescale buses. A 504-channel version called the 4002 is available, too.

LSI Logic offers a chip for implementing voice-over-Wi-Fi products. Its 7- by 7-mm LSI403US facilitates adding VoIP to cell phones, PDAs, and other handheld devices (Fig. 2). This extremely lowpower, 16-bit, fixed-point DSP is based on LSI Logic's well-known ZSP400 DSP core. The maximum 150-MHz clock rate produces 600 MIPS or millions of multiply and accumulates (MMACs). Its two high-speed serial/TDM ports are compatible with T1/E1 framers and H.100/H.110 bit-stream I/O. In addition, the DSP possesses an asynchronous, 8-bit, parallel host processor interface (HPI). It's ideal for low-cost and portable products where low power is essential.

Mindspeed Technologies focuses on the infrastructure side of the system. Its variety of chips and software takes care of almost all infrastructure options. The M82710 Comcerto 700 Series Carrier Convergence Processor offers multiple DSPs and ARM processors.

With the processor, telecom equipment manufacturers can create a product that transmits highly secure, carrier-class quality voice across wireless and wireline networks with integrated voice encryption, authentication, and denial-of-service protection. It contains 404 G.711 PCM channels and 168 channels of G.729a/b voice. Besides the TDM lines, interfaces include Ethernet, UTOPIA L2, and POS L2 for both Ethernet and ATM connections. A PCI bus connects to an external host processor where the signaling protocols are implemented.

Mindspeed's M82820 includes the host processor on chip so vendors can customize it using Linux. Also incorporated are 32 VoIP ports, encryption, and a PCI bus for connection to an external 802.11a/b/g chip for wireless operation.

Texas Instruments offers chips for virtually any application. The TNETV1050 combines TI's popular TMS320C55 x DSP with a MIPS32 RISC processor and Telogy software to take on all voice processing and signaling. On top of that, it handles H.323, SIP, and MGCP and all of the voice coding protocols. The popular VxWorks real-time operating system is embedded. An IPsec engine that features DES, 3DES, and AES encryption provides security. Then there's the TNETV1600 gateway chip, which includes connections for wireless for a Wi-Fi transceiver for IP wireless LAN phones.

Not to be left out are the SLIC manufacturers. The SLIC is the set of circuits usually installed at the telephone company's central office to support standard POTS telephones. This includes the BORSCHT functions of battery feed, overvoltage protection, ringing, line supervision, codec, hybrid, and test. SLIC chips are required inside the home ATAs and gateways to plug in any standard POTS phone. Chips used in VoIP ATAs and gateways include Legerity's VE880 series and Silicon Labs' Si3216.

Version 2.0 of Trinity Convergence's VeriCall Edge software implements full VoIP on a standard processor (e.g., ARM, MIPS, Freescale). No separate DSP is needed. It runs under embedded Linux and permits voice and video over IP (see "Video Over IP" at www.elecdesign.com, Drill Deeper 11114). Manufacturers will be able to make voice and full-feature video phones. VeriCall 4.0 targets infrastructure platforms (Fig. 3).



Broadcom Corp.


International Telecommunications Union

Internet Engineering Task Force (IETF)

Legerity Inc.

LSI Logic

Mindspeed Technologies



Silicon Labs

Texas Instruments

3rd Generation Partnership Project

Trinity Convergence

Hide comments


  • Allowed HTML tags: <em> <strong> <blockquote> <br> <p>

Plain text

  • No HTML tags allowed.
  • Web page addresses and e-mail addresses turn into links automatically.
  • Lines and paragraphs break automatically.