ADSL Products Must Conform and Perform

ADSL modems may speed up home Internet access, but only after you test for conformance to specs and for performance that the user will accept.

Martin Rowe, Senior Technical Editor -- Test & Measurement World, 2/1/1997

Every so often, I find a Web page that I think can help my wife with her work. I download the site’s home page, call her into the room, and we begin reading the page. When I click on a link, a new page begins to load. The page usually takes some time to appear because it contains many graphic images. That’s when she walks out of the room saying, "Call me when there’s something to read."

If you’ve ever used the Web or downloaded a file over a plain old telephone system (POTS) modem, then you know that file transfers seem to take forever. Integrated Services Digital Network (ISDN) service offers data rates up to 128 kbps as opposed to 33.6 kbps for POTS modems, but even ISDN speeds can seem slow when loading large programs or graphics. And if you want real-time video, forget about either technology. The only way to get fast Internet access and real-time video is with data rates in the megabits per second range. Fiber-optic communications right to your PC or TV would solve the speed problem, but nobody wants to pay the cost of rewiring homes with fiber. Most people want high-speed Internet access using the twisted-pair wires currently in their homes (see "Megabits Over Twisted Pairs," below).

One of the technologies promising to deliver megabit data rates over twisted-pair copper wires is called Asymmetric Digital Subscriber Line (ADSL). ADSL, one of several DSL technologies, has proven effective in early field testing.1 If ADSL is successful, then ADSL modems will replace the millions of POTS modems currently in use. Before that happens, engineers must design and test ADSL modems for use in your home and at your local phone company.

As a test or design engineer, you first need to know which measurements are important for conformance to standards. Once you’ve proven that an ADSL modem’s design can meet standards and transmit bits reliably, you need to run performance tests using live data and protocols. Conformance and performance tests are important, but they can’t guarantee that all ADSL modems will talk to each other.

An ADSL Standard
The ANSI T1.413 standard—drafted by the ANSI T1E1.4 committee—defines ADSL modems that use the discrete multitone (DMT) technique for data transmission.2 DMT is one of two modulation techniques used by ADSL modem manufacturers. The other is called carrier amplitude phase (CAP). The ANSI T1.413 document also defines conformance tests for DMT ADSL modems. In those tests, you must simulate a line of twisted-pair wires (called a test loop), add electrical impairments to the lines, and measure bit error rate (BER).

ANSI T1.413 defines several tests for ADSL modems. These tests are:

• crosstalk tests;

• impulse-noise tests; and

• POTS interference tests.

Figure 1 shows the lab test setup for testing an ADSL modem’s conformance to ANSI T1.413. The ATU-C is the ADSL phone company’s central office (CO) ADSL modem, and the ATU-R is the modem in a home or a business. The ATU-C is sometimes called the transmitter because it transmits at a rate of megabits per second; the ATU-R is called the receiver.

You simulate the test loop using a line simulator, or you can use a spool of wire. The splitters separate POTS signals from ADSL signals. During the tests, you connect a POTS phone or modem to the ATU-R splitter and a CO simulator to the ATU-C splitter. You’ll also need a high-impedance network to introduce impairments into the test loop without altering the characteristics of the test loop. Be sure to perform conformance testing with any error correction enabled.

The first conformance test is the crosstalk test. Crosstalk testing is necessary because phone lines are typically bundled between the CO and their destinations. The tests are based on interference from high-speed DSL (HDSL), T1, ISDN (called DSL in ANSI T1.413), and ADSL transmissions on other lines in the same bundle (bundles use either 25 or 50 pairs of wires).

ANSI T1.413 Annex B defines the characteristics of each type of signal. Each interference signal (called a disturber) has its own power spectral density (PSD). Therefore, the noise you generate for each form of crosstalk will differ. You must also be sure that the PSD of the test system’s background noise is flat. One of the tests you must perform is the 24-disturber DSL near-end crosstalk (NEXT) test. In this test, you feed the ADSL signals over one pair of a bundle of 25 wire pairs. You vary the PSD level of the noise to simulate the number of disturbers.

The noise you use to simulate the disturbers must be Gaussian white noise that has a crest factor of 5. (The T1E1.4 committee is investigating this spec and may change it.) Connect your noise source to the test loop, which must have an impedance greater than 4 kW.

Simulate Wire Loops
In addition to simulating disturbers, you must also simulate several configurations of twisted-pair lines. These lines are revised-resistance design (RRD) lines (for 1.544-Mbps data rates) and carrier service area (CSA) lines (for 6-Mbps data rates). The Bellcore SR-TSV-002275 and ANSI T.601 specs define the lengths and gauges of the test loops.3,4 They also define the locations of bridge taps in the loops. Bridge taps are unterminated connections to a subscriber loop. Those unterminated wires cause reflections and can distort signals. Figure 2 shows two subscriber loops with bridge taps used for testing. The T1.601 Loop #13 has both 24 AWG and 26 AWG wires; the CSA Loop #4 is entirely 26 AWG wire.

While performing a crosstalk test, set your noise source to the appropriate level for the type of crosstalk you are simulating. That becomes your reference level, 0 dB. You have to establish communication between the ATU-C and ATU-R at the reference noise level (assume that the ATU-R is the unit under test). The process of establishing the connection between the two ADSL modems is called training.

Once training is complete, measure BER with a BER tester while using a pseudo-random test pattern of 223–1 bits. Increase the noise level in 1-dB increments until you find the highest noise level that produces a BER that is less than 10–7. The difference between your noise level and the reference level is called a margin. To pass crosstalk conformance tests, you need a margin of 6 dB.

Because the noise signals are random, you must perform the tests long enough to accurately measure BER. Table 1 shows the minimum test times for crosstalk conformance testing.

After the crosstalk test, you can perform the impulse-noise test. Impulse-noise testing subjects your ADSL product to wideband noise delivered in short bursts. The impulses specified in ANSI T1.413 Annex C are reconstructions of pulses recorded at field sites. There are two impulse test waveforms in ANSI T1.413 (Fig. 3 ). To generate those waveforms, you need an arbitrary waveform generator or a line simulator that generates impulse impairments.

The impulse tests begin after you train the ADSL modems using the 0-dB noise level that you used for the crosstalk tests. After you’ve established communications between the two modems, apply each impulse 15 times at intervals of at least 1 s. Repeat the tests by adjusting the amplitudes of the impulses until you measure errored seconds on the BER tester. What you’re measuring is the amplitude that produces errors in half of the seconds in a test. You can then follow a formula in ANSI T1.413 to calculate the probability that a given amplitude will produce an errored second. To pass the test, that probability must be less than 0.14%.

The above tests don’t account for errors created by POTS equipment. ADSL and POTS equipment must share the same wires, so you must test your ADSL equipment with interference from POTS equipment on the same line. Connect a CO simulator to the ATU-C splitter (Fig. 1), and connect a POTS phone to the ATU-R splitter. The POTS equipment should not interfere with ADSL operation. During the POTS interference test, you must simulate the conditions that POTS equipment produces—signaling, alerting, ringing, answering, and going off hook and on hook.

Performance Counts
While conformance testing is necessary, it’s not the end of your testing. You also must conduct performance testing to determine how well a line and an ADSL modem can provide a particular grade of service. Performance testing is similar to conformance testing, but this time you must turn off a protocol’s error correction before sending data over the test loop. You should measure errors with a protocol analyzer. Performance testing may also include testing the signals from an ADSL transmitter.

Because ADSL is most likely to be used for Internet access, you should run performance tests using TCP/IP. According to Michael Tzannes, senior vice president for telecommunications at Aware (Bedford, MA), you should test data throughput by transferring data with protocols rather than just by sending bit patterns. Because protocols have overhead caused by headers and error correction, the actual data throughput will be lower when you use protocols. Tzannes points out that TCP/IP works best with a 10:1 downstream-to-upstream data rate. Even if a modem’s downstream data rate is greater than 10:1, the upstream data rate will become a bottleneck for downstream communications with data transported over TCP/IP.

You should test ADSL modem performance under several RFI conditions. Transmission should be tested for susceptibility to RFI from AM radio transmitters, LORAN aircraft positioning equipment, and ham-radio transmitters. RFI can also come from ring trip caused by POTS equipment. In addition, you should test for RFI emissions to be sure that ADSL communications won’t interfere with local ham radio, AM radio, or over-the-air TV reception. Also, ADSL transmissions should not produce RFI that interferes with other CO equipment. In other words, you must perform RFI testing for compliance with regulations such as FCC Part 15.

While all the above tests are important, they still won’t guarantee that your ADSL equipment will operate with ADSL equipment from other manufacturers. Right now, no one can guarantee interoperability between ADSL modems from different manufacturers. According to Fadi Daou, communication marketing specialist at GenRad (Concord, MA), an ATU-R made by one manufacturer may not communicate with an ATU-C from another manufacturer. Therefore, the ATU-C and ATU-R in an ADSL test system must be from the same company. The ADSL Forum has established a test working group whose goal is to present recommendations and a working text for test-related issues. At a meeting in September 1996, the group decided to investigate functional interoperability issues.

FOOTNOTES
1. Taylor, Kieran, "DSL Modems: So Far, So Good," tele.com magazine, McGraw-Hill, New York, NY. Web: www.teledotcom.com October 1996.

2. ANSI T1.413-1995. Telecommunications—Network and Customer Installation Interfaces—Asymmetric Digital Subscriber Line (ADSL) Metallic Interface, American National Standards Institute (ANSI), New York, NY.

3. SR-TSV-002275, BOC Notes on the LEC Networks, Bellcore, Piscataway, NJ, April 1994.

4. ANSI T.601-1992, Telecommunications—Integrated Services Digital Network (ISDN)—Basic Access Interface for Use on Metal Loops for Applications on the Network Side of the NT (Layer 1) Specification, ANSI, New York, NY.

FOR FURTHER READING
Kampainen, Stephen, "ADSL: The End of the Wait for Home Internet?," EDN, October 10, 1996, p. 53.

The ADSL Forum’s Web site provides links to other articles, technical papers, and news. Visit the site at www.adsl.com. ADSL Forum, Foster City, CA, 415-378-6680; e-mail: adslforum@adsl.com