|
||||||||||||||||||||||||||||||||||||
|
|
||||||||||||||||||||||||||||||||||||
|
||||||||||||||||||||||||||||||||||||
|
|
<
By Mike Mitchell, Senior Systems Engineer, Broadband Connectivity Group, Texas Instruments Incorporated If you're reading this column, you know the problem: Everyone who has ever accessed the World Wide Web has spent more time than they would like tapping a foot, savoring a long sip of coffee, or otherwise idly waiting while pages and files were downloaded. By and large, the Internet is too slow and it is no secret that if the Web is to become the information superhighway of the future, its speed limits must soon be raised dramatically. Fortunately, several technologies are emerging that can rectify this situation. One is ADSL (Asymmetrical Digital Subscriber Line) and as standards and implementation procedures evolve and become more stable in the years ahead, the development of ADSL products will move into high gear. As this occurs, developers will realize that the analog front-end of an ADSL modem presents the greatest opportunity for product differentiation, which, in turn, will be critical for the success of the product in the marketplace. Beyond Blockbuster The original application that gave impetus to the development of ADSL was video-on-demand which, unfortunately for the original developers, was never able to supplant the local video store as the distribution channel for movies. Fortunately for the developers the performance parameters of ADSL lend themselves nicely to the way the Internet operates; it was reasoned that a video-on-demand application would involve uploading a small command file from a user and downloading a large multimedia file at high speeds from the provider of the video movies. The asymmetrical nature of this data-communications traffic parallels the flow of traffic on the Internet. Since one of the objectives of ADSL was to provide broadband data and voice communications over the existing public telephone network, the technology must be able to deal with the limitations of the existing twisted-pair copper wiring which connects most homes to the public service telephone network (PSTN). Currently, the ADSL specification provides downloading data at communication speeds up to 8 Mbit/s from the telephone network's central office to end users. In the other direction, from users to the central office, a speed of 800 kbit/s is achievable. Additionally, a voice channel is maintained even when the data channels are not functioning. The Power Challenge The signals used in ADSL have a very high peak-to-peak average ratio meaning that the line driver must be capable of driving high-output peaks, even though they don't come along very often. The output of the line driver must be able to drive signals to a peak of 10 V, but only needs to drive about 1.6 Vrms, and this extra headroom in the line driver is dissipated as heat. In essence, ADSL calls for the building of a 'mega-lane' data-communications superhighway to accommodate the demands of peak rush-hour traffic even though rush-hour will occur very seldom: Most of the time, a modest four-lane highway would suffice for handling the communications traffic over ADSL links. Addressing the issue of heat dissipated from ADSL modems will be a major challenge for designers in the years ahead. Recent implementation trends on the part of the telephone network have called for a shortening of the final copper loop that connects an end user to the network. To achieve this goal phone companies have moved a portion of their central office (CO) switching equipment closer to the user, placing line cards and other such technology in what's known as digital loop carriers (DLCs.) DLCs serve a relatively small geographic area and are connected to the large CO switches by high-speed fiber-optic links. The boxes that house DLCs are often installed in utility easements such as highway medians or close to the curb in residential neighborhoods and, as a result, DLCs are subjected to the vagaries of the weather; ranging from the blistering heat of summer to the below-freezing cold of winter. Since ADSL consumes more power than plain old telephone service (POTS), a great deal of heat could be generated in the very confined area of a DLC. With no control over the environmental conditions in DLCs, excess heat will present very real problems and have a detrimental effect on the performance of ADSL modems, as well as the POTS line cards sharing a DLC box. The Distortion Issue ADSL employs a line-coding scheme called Discrete Multi Tone (DMT) in which the signals are composed of 256 discrete analog channels, or tones, each approximately 4.3 kHz wide. The entire bandwidth begins at approximately 20 kHz and ends at 1.1 MHz. The requirements of the specification make the design of a line interface a somewhat difficult task and one factor that is of critical importance to the performance of an ADSL modem, and specifically the ADSL codec, is distortion. For the designer of ADSL modems distortion must be addressed early in the development process, as it can drastically reduce the overall data rate of the connection from the quality of the individual tone fidelities that make up the DMT signal. Each circuit on the signal path can contribute to the link's total distortion: Line drivers, analog filters, ADCs, DACs, and gain amplifiers all contribute to the distortion. Careful consideration and thorough analysis of the distortion characteristics of ADSL components is required. The traditional measurement of distortion of Total Harmonic Distortion (THD) is usually performed with a single tone somewhere near the middle of the bandwidth of the device. However, THD test results don't offer a clear indication of how a codec will perform in an ADSL system with DMT signals because the link is comprised of many tones being generated simultaneously. A different measurement, known as the missing tone test, (MTT) gives the designer more relevant performance data. The test is performed by constructing a DMT signal with an individual tone turned off, and the distortion of the codec will cause a very small tone to appear in the missing tone's location. The relative amplitude of this distortion-created tone is known as missing tone rejection, and the higher the amplitude of the tone, the higher the distortion of the codec. The cumulative results of missing tone tests performed for every tone give the designer a much clearer picture of how the distortion of the codec will effect the performance of an ADSL modem. Low-Level Signals ADSL modems will be faced with the limitations of twisted-pair copper wiring which may attenuate data-communications signals by as much as 100 dB or more. As a consequence ADSL receivers must be very sensitive so that they can find these very low-level signals; to achieve the sensitivity the designer must include significant gain in the system, but care must be taken to ensure that noise or distortion is not introduced into the amplifiers. Unwanted signals must also be filtered out before the incoming signal is amplified, but the filtering process must not interfere with the data signal. Another issue related to handling of low-level signals is resolution and ADSL codecs are available with both 14-bit and 12-bit resolutions. Some experts say 12-bit resolution is all that is required while others assert that 14-bit resolution helps ensure higher modem performance: Using a codec with a higher resolution may not guarantee better modem performance, but it will minimize the risk that the codec will become a bottleneck or limiting factor to overall system performance. Integration One way for a designer to simplify the development process would be the use of a highly-integrated codec and some are available today that typically have integrated filtering and programmable gain amplifiers, but others still require external circuitry for signal gain and filtering. With a more integrated codec, the designer's task is much simpler. This has several significant advantages: Keeping An Eye On The Standards Aside from the technical issues, an added element of uncertainty comes from the state of the pertinent specifications. Issue 2 of the T1.413 ADSL specification was on-schedule for approval until an issue was raised in March concerning the specification's filtering requirements. Discussion over filtering may delay finalizing the standard for some time to come. Another debate over a slimmed down ADSL standard known as G.Lite has been brewing for some time. This specification would provide only half the bandwidth of a full-blown ADSL implementation, but by relaxing some of the requirements G.Lite would be decidedly less expensive. And since the Internet cannot provide the 8-Mbit/s bandwidth ADSL is capable of, some in the industry have questioned whether a better strategy might not be to provide programmable G.Lite modems that could be stepped up in performance with software upgrades as the bandwidth of the Internet improves. Mixed-Signal Expertise Full-scale ADSL implementation faces many challenges, but we can say with certainty that developing effective components will require an established base of mixed-signal expertise. For example, much of the high-order filtering that will take place in an ADSL modem's analog front-end will be done digitally, because this allows less silicon to be used which, as a consequence, drives down the cost of the components and the power consumed. In addition, mixed-signal expertise gives the semiconductor vendor the ability to integrate onto one chip many of the functions needed in an ADSL modem, such as amplifiers, analog and digital filters, equalizers, and converters. This will also drive down the cost of ADSL and hasten its market acceptance. Analog Main | Product of the Week | Columns | Editorial | Tech Notes
|
|||||||||||||||||||||||||||||||||||
|
Copyright © 2003 ChipCenter-QuestLink About ChipCenter-Questlink |
||||||||||||||||||||||||||||||||||||