Today’s ILECs and CLECs are encountering unprecedented opportunities as well as some tough decisions. It’s a matter of timing, not unlike that faced by the New York Mets in the last game of the 2000 World Series. How long should you continue to use the starting pitcher—the conventional circuit-switched telephone technology?
The answer can be very complicated. Replacing the Mets’ starter would have made the bottom of the ninth inning awkward because he was to have been the leadoff batter. But not replacing him proved to be fatal because his pitching deteriorated more rapidly than anticipated.
Changing to a packet-switched technology is part of future planning for most carriers but will take many years to implement. Economically, the best short-term plan seems to be to continue using the well-proven PSTN and SS7 infrastructure for voice communications but to provide more flexible access to it. However, many carriers are going further by proactively partnering with technology developers to accelerate the deployment of services such as VoIP, which eventually may displace the traditional phone network.
For several years, early adopters of PCs and the Internet used relatively low-speed modems for data access via the PSTN. At the same time, businesses used separate packet-switched data networks operating with protocols such as IP and Ethernet and achieved much higher speeds. Many large businesses with several sites located in different time zones already are using IP-based intranets for total data, voice, and conferencing requirements. Of course, they have complete control of their internal network configuration and operation.
Proponents of public Internet telephony are confident that today’s questionable security, QoS, and network management performance, often cited as deficiencies of IP, will become less contentious as capabilities are added to the protocol. For example, IPsec provides payload encryption and, in some cases, source and destination identity encryption. IPv6 increases the address field to support a larger number of addresses and improves routing and network autoconfiguration.
The recent availability of fast, affordable computers has driven the general demand for higher-speed Internet data links and heightened interest in broadband technologies. Simultaneously, deregulation of telephone services has encouraged competition to provide the voice and data services carried by existing copper-pair local-loop wiring. These factors have opened up a substantial market for lower-cost multiline services for small companies.
The Last Mile
xDSL refers to several different flavors of high-speed Internet access technology. ADSL delivers downstream and upstream bandwidth at a ratio of about eight to one, respectively, and preserves POTS service. SDSL provides equal upstream and downstream bandwidths but does not allow simultaneous POTS service. G.Lite is a lower-cost version of ADSL with lower data rates and simplified installation.
Internet access is added to existing POTS service by running the IP protocol over the ADSL connection. This requires installation of an ADSL modem at each customer premises as well as a DSLAM at each local CO.
Separate or built-in filters split the upper and lower frequency bands, separating POTS from DSL bandwidth. For incoming downloads, the modem recovers IP packets from the modulated DMT signal. For uploads, it produces a suitably modulated DMT signal and adds it to POTS transmissions.
A splitter passes the 4-kHz voice band signals to a Class 5 switch for digitization and conversion to a TDM signal compatible with the standard T1/T3/OC-3…64-kb/s-based PSTN hierarchy. The DSL bandwidths go to the CO DSLAM, which demodulates and multiplexes ATM cells from each user’s copper-pair local loop onto the ATM network.
The Economic Case
For small companies requiring several phone and fax lines in addition to Internet access, the cost of a 24-channel T1 line can be a deterrent to business expansion. A requirement of 16 phone lines often is quoted as the minimum number for which leasing a T1 line can be economically justified.1 Sixteen phone lines include phones, faxes, and conventional modems. These companies also may use xDSL for Internet access.
Because of the big jump in cost between one or two lines and extensions and a T1 line, a large market exists among smaller companies for voice services that provide less than 16 lines. Recent xDSL schemes that use packet switching to handle a number of voice channels allow CLECs to address only the non-POTS part of the local-loop spectrum but deliver multiple lines of toll-quality voice service (Figure 1). Being left with just the 4-kHz POTS band, the ILEC can supply only the traditional single-line phone service.
When voice-signal packetizing is handled at the premises-end of the xDSL link, the modem becomes a more functionally advanced instrument termed an IAD. The IAD provides the sockets into which you plug your fax, phones, and PC. Internally, it performs a number of specialized tasks.
It must digitize voice signals, compress them, and apply echo cancellation using exactly the same standards as would a Class 5 switch. This is necessary to ensure that the customer receives voice service that is indistinguishable from POTS. It’s also important because modern CLASS, Custom Calling, and Centrex features such as call waiting and caller ID are provided via the Class 5 switch.
The Technical Case
The IAD must prioritize sending voice packets in preference to data packets. This is because of the strict end-to-end delay limitations placed on voice communications. ITU-T G.114 recommends 150 ms as the maximum one-way delay between the sender and the receiver before ordinary conversation becomes awkward. Beyond 150 ms, both parties must pause after each statement, or they can begin to talk over each other and soon become frustrated.
In operation, an IAD establishes queues for data bytes and digitized voice samples ahead of the packetizing circuit. If nothing is done to avoid the situation, it is quite possible for voice samples to be waiting in a queue while a 1,500-byte data packet is assembled. IP packets are of variable length and can easily be this long.
As a result, some mechanism must be used at the link level to deal with the different requirements of voice and data packets. ATM is the obvious protocol for this purpose, comprising both an adaptation layer and the actual ATM layer. For example, AAL5 is intended to streamline handling of large blocks of data. Up to 65,535 bytes of data are accommodated with little overhead in addition to the usual SAR sublayer segmentation into 48-byte ATM cells with 5-byte headers.
In contrast, AAL2 specifically addresses real-time media. AAL2 not only includes basic header and segmentation specifications, but also deals with transport of associated voice and call-control signaling. There is a means provided for carrying network management data packets over the same VCC as AAL2 voice packets, which significantly reduces the complexity faced by network operators.
However, if ATM is used with IP, packetization overheads increase and corresponding bandwidth efficiencies decrease. In an example provided by CopperCom, an IP-over-ATM scheme required 160 bytes to transmit an 80-byte voice payload and incurred a packetizing delay of 20 ms. Alternatively, using a 44-byte payload and only ATM AAL2 had a latency of 11 ms.2
CopperCom’s Stefan Knight, director of product marketing, stated, “Our packet voice technology is based on voice over ATM, which can use ADSL, SDSL, and T1 to reach the largest available market of small businesses. With toll-quality compression, we deliver up to 24 voice lines plus high-speed data (Figure 2).”
At the CO, the DSLAM still may recover the POTS signal to provide fail-safe emergency communications even if the customer’s requirements ordinarily are met entirely by the VoP service. In any case, the demodulated xDSL signal now contains data packets interspersed with voice packets. ATM allows these to be identified, separated, and kept in strict time order.
Because the voice signal has been digitized, the Class 5 switch will require the same interface as from any other digital-loop carrier system. This function is provided by a gateway. It accepts packetized voice from a number of DSLAMs and produces the 64-kb/s TDM signal expected by the Class 5 switch and defined by specifications such as GR-303, TR-008, or V5.
Although a CLEC may co-locate its DSLAM with the ILEC’s local CO equipment, the CLEC’s Class 5 switch and voice gateway may be remotely located. This means that data can be separated at the CO and multiplexed onto the IP network, but voice traffic must be back-hauled to the remote Class 5 switch to gain access to the PSTN. There must be an intervening protocol such as ATM or frame relay used in the access network for this to be accomplished. Fortunately, ATM cells and frame-relay frames are sized so interworking between the two protocols is a standard practice.
Commenting on his company’s voice-over-broadband technology, Rich Farana from TollBridge, said “Our solution is VoIP over AAL5. Since we use native IP, it is very natural for us to provide a seamless path to the next-generation voice network based on IP.
“Because we provide IP over AAL5, our gateway works with pure ATM-based DSLAM as well as IP routing-based DSLAM,” he continued. “The latter allows us to do VC aggregation, making the VC management easier.”
An Alternative Approach
Although the demand for basic PSTN service in the United States has remained at a low 3% growth rate for several years, the residential second-line market has increased by 15% per year. Consequently, it is reasonable to assume there is a market for low-cost residential service that provides up to four voice lines over xDSL.3
Aware has addressed this opportunity with VeDSL. This technology, shown in Figure 3 and referred to as Channelized VoDSL in the DSL Forum standards organization, is designed for residential ADSL services. It requires a VeDSL modem at the customer’s premises to digitize voice according to standard 64-kb/s PSTN requirements. The digitized voice is modulated directly within the ADSL tones, without packetization. A voice-enabled DSL line card in the local CO demodulates the ADSL signal and recovers the TDM voice signals from it, separating the data signal in the process.
Like many of the VoP systems, VeDSL dynamically allocates the amount of bandwidth available for data. If none of the phone circuits is being used, the full bandwidth is available for data. Analog dial-up data and fax modems are supported. Voice channel compression also is possible.
VeDSL is strictly a physical-layer process. Because packetizing is not done, latency is low and echo cancellation is not required. Comparatively simple circuitry sends and receives data and voice at each end of the link without higher-level protocols. This means that what might be a 20-ms latency in a VoP system reduces to 2 ms with VeDSL.
Bandwidth efficiency improves, which is particularly important to residential customers served by ADSL. Since VeDSL creates dedicated channels of physical-layer bandwidth for voice, there is no competition between voice and data traffic; QoS is inherent.
Costs also reduce throughout the system. For example, no gateway is needed because PCM-encoded voice is recovered immediately by the VeDSL line card in the DSLAM and transported directly to the PSTN. Alternatively, a next-generation DSLAM can perform packetizing/depacketizing at the CO to allow voice back-hauling over ATM or transport to next-generation packet networks if required.
Test Implications
The network areas that most likely will require attention are those that may have changed to accommodate VoP. In particular, QoS as it relates to voice signals will be a problem in systems that add significant latency or jitter to voice packets.
Improving latency or jitter is a trade-off. A CODEC may introduce greater latency but have improved sound quality. The receive buffer size also adds to latency, but if it’s not large enough, jitter may be excessive.
According to Dave Benini, marketing manager at Aware, “One of VeDSL’s advantages is the minimization of risk associated with the addition of new equipment to the network. Not only is the cost of a VoDSL gateway eliminated, but so is much of the associated integration complexity. There are no PVCs to configure and no customer equipment requiring ongoing management.”3
For packet systems, RADCOM stresses the importance of simultaneous two-port measurements that support network element ingress vs. egress data timing, even across mixed protocols. Correlating actual in/out data packets not only determines per-conversation latency, but also highlights lost packets.
Bandwidth requirements can be reduced up to 50% through silence suppression. By not transmitting silent parts of conversations, other active channels make better use of the available bandwidth. Jitter is improved by reducing CODEC latency because less data is being converted. But if silence suppression is being used, it must be taken into account when calculating jitter.
Probably the most important aspect of VoP systems, assuming all else is operating correctly, is loading. Packets will be dropped as loading increases, so voice quality eventually will suffer. RADCOM cites 5% packet loss as excessive and noticeable. On the other hand, most of the RADCOM comments regard IP networks and are less applicable to guaranteed QoS available over ATM.4
Troubleshooting VoP systems requires the same understanding and test techniques as data. In addition, the real-time signal constraints highlight network areas that may add delay because, for example, VoP packets haven’t been prioritized or a slow CODEC has been used. Troubleshooting should only proceed after you have understood the self-diagnostic information provided by the system elements.
Efficient Network’s Senior Product Marketing Manager Alvin Ono explained, “Our IADs are designed around industry standards and have the capability to manage and supply diagnostics for the services we provide. Industry standards have defined requirements that determine the way the IADs interface to the network from a signaling as well as a management standpoint.”
CopperCom’s Mr. Knight quantified the kinds of problems typically seen when installing his company’s equipment: “VoDSL does not introduce much of an additional installation burden. Installation is simplified because all VoDSL settings are dynamically downloaded from the gateway upon power-up. Once the DSL loop is up and the ATM VCC between the IAD and gateway is up, calls can be made. If not, three problems are possible:
- Is the ATM VCC properly configured end to end, and is the voice port activated on the gateway?
- Is there a problem in the ATM network such as a link that is down?
- Is there a problem with the gateway to the Class 5 switch interface?
“The ATMF-LES group’s EOC approach is used to monitor our IAD,” he explained. “Configuration, performance, and troubleshooting information can be passed between the gateway and IAD. In addition, an ATM F5 cell is used as a heartbeat so the IAD and gateway know they are perpetually virtually connected if the F5 cell issued by the gateway travels to the IAD and is reflected.”
References
- Shinal, J., “Voice Is the Killer DSL App,” Forbes.com, Feb. 29, 2000.
- Taylor, M., “Mastering Voice Over DSL: Network Architecture,” Copper-Com White Paper, 1999, p. 9.
- “Voice-Enabled DSL—New Voice Over DSL Technology for the Residence,” Aware White Paper, 2000.
- Boger, Y., “Fine-Tuning Voice Over Packet Services,” RADCOM White Paper.
Acknowledgements
The following companies provided information for this article:
| Aware | 781-276-4000 |
| CopperCom | 408-987-8500 |
| Efficient Networks | 972-852-1000 |
| RADCOM Equipment | 201-529-2020 |
| Spectrum Signal Processing | 800-663-8986 |
| TollBridge Technologies | 408-585-2100 |
Glossary
Published by EE-Evaluation Engineering
All contents © 2001 Nelson Publishing Inc.
No reprint, distribution, or reuse in any medium is permitted
without the express written consent of the publisher.
January 2001