This article shows that ASDL (G.Lite or otherwise) is not going to be
EASY....and may even fail. And please note that only 89 people work
for the AWARE COMPANY....
HAL
.................................................
Managing ADSL Signals and Contending with Noise
By Jim Quarfoot
The next great leap forward for Internet data communications speeds will almost certainly come from the implementation of asymmetrical digital subscriber line (ADSL) technology over standard copper telephone lines. Although sophisticated digital signal processing (DSP) algorithms are at the heart of ADSL, other difficult issues must be dealt with, including signal management and handling noise and interference.
Managing a wide dynamic range
Handling the very wide dynamic range of ADSL signals at the receiving end of a link will be a very critical issue as ADSL moves into widespread deployment. Whenever an ADSL link is established, the central office (CO) will output a great deal of power with every data transmission to deal with the possibility that the receiving end or client end may be several miles of telephone lines away. However, the receiver may not be very far away at all. In fact, the receiving end of an ADSL link could be anywhere from several hundred feet to three miles distant. This means that if the CO outputs the same amount of power with every transmission, regardless of the distance to the receiving end, received signals can vary greatly in magnitude when they arrive at the end of the line. Dealing with signals that cover a large range, while contending with the background noise and interference that telephone lines are susceptible to, presents quite a challenge for ADSL receive circuitry.
Powering ADSL signals
The requirements for an ADSL signal are defined in ANSI T1E1.413. This specification has established -40 dBm/Hz as the maximum transmitting power density at the CO end of an ADSL link over a telephone line. With a frequency band of transmission between 25 kHz and 1.1 MHz (Figure 1), the bandwidth for the CO transmitter is approximately 1.1 MHz. This means that the maximum power in an ADSL link is 20 dBm. This number is commonly referred to in the industry as that which defines the power content of an ADSL link. It is critical for the line drivers at the CO end of the line to reach 20 dBm, because these drivers must deliver this much power onto the line. But at the far end of the line, this level of power will never be received. The signal will either be attenuated by extended distances and long telephone lines, or, in the case of shorter lines, the signal will be throttled back by the CO to a lower power level. Determining the proper power level is accomplished during the initial stages of an ADSL link, or during the link's “training” period.
Figure 1: Full rate ADSL spectrum utilizing echo cancellation.
What's an ADSL link?
In theory, ADSL can provide up to 8 Mbps of data bandwidth from the CO downstream to users, and up to 800 kbps upstream from the subscriber to the CO. While all of this data is being transferred, a voice conversation can take place on the same telephone line. The discrete multitone (DMT) implementation of ADSL involves a multicarrier protocol that uses multiple sine-wave carriers on the same telephone line. Each carrier tone differs in phase and frequency. DMT allows constant reallocation of data bins within each ADSL channel to continually re-optimize the quality of the data transmission.
With DMT, the CO effectively transmits up to 256 separate 4.3125-kHz tones downstream to the remote user. During the initial period, while the ADSL channel is being set up, the remote end measures the quality of the tones from the CO and decides whether a particular tone has sufficient quality to be used to transmit data bits. After a tone is judged to be sufficient, the amount of data to be transferred by the tone is determined. This set-up process is called the “training” period for an ADSL link. After processing the signals from the CO and going through this training period, the remote end sends the optimized bit distribution to the CO by using a secure low-speed reverse transmission over the same phone line. The CO then implements this distribution. Following the completion of the CO-to-remote end training period, the same training process is performed for the opposite upstream direction, or from the remote end to the CO. The upstream direction uses just 32 tones instead of the 256 used for downstream transfers.
While this training period is going on, the CO is able to throttle back the power on the line by measuring the strength of the received signal from the remote or client end of the line. If the received signal is of sufficient magnitude, the CO reduces the power of its transmitted signal because the training period has indicated that the connection is short and the maximum power level is not needed. The amount of the power reduction initiated by the CO is programmable in increments of 2 dB up to a maximum of 12 dB. As a result, a short ADSL link will have at least 8 dBm of power transmitted by the CO. For the purpose of this discussion, we have assumed that 8 dBm represents the maximum power received at the client end of the link.
Many factors that affect signal strength must be considered in order to determine the minimum power received at the client end of an ADSL link. Line length and signal frequency are primary factors, but wire gauge, line stubs, and peaking coils all have contributing effects on ADSL signal levels.
For example, a 1,000-ft line will attenuate a 30-kHz signal minimally, while a 1.1-MHz signal will lose about 8.2 dB. However, at 14,000 ft, a 30-kHz signal will be attenuated or lose 20 dB, while the 1.1-MHz signal will lose 115 dB. An average attenuation loss of 40 dB will be used here to simplify the discussion.
Crest factors/voltage levels
Since a DMT implementation of ADSL provides for 256 separate carrier tones in the 25-kHz to 1.1-MHz ADSL frequency band, each carrier tone occupies a 4-kHz slot, or bin, and can have its own rate of data modulation.
A variable modulation rate causes very high peak voltages as ADSL signals travel over telephone lines. Each carrier tone has an independent frequency and phase, creating a statistical distribution of composite signal amplitude. Thus, there will be times when the phases of multiple tones will align in a manner that causes the peak amplitudes of the individual tones to add to each other so that the composite waveform has a large peak amplitude. The ratio of this peak amplitude to the RMS value is called the “crest factor” and for DMT ADSL, it is typically chosen to be 5.6, or 15 dB (Figure 2).1 At this crest factor value, the probability of having even greater voltages is reduced to approximately 1 x 10–7.
Given this crest factor, the maximum voltage at the client end can be determined. Assuming a maximum power level of 8 dBm, the resulting RMS voltage level on a 100-ohm line will be 0.794V. Converting from RMS to peak by using a 5.6 crest factor yields a peak voltage of 4.45V, or 8.89V, peak-to-peak.
Figure 2: Crest Factor (peak to average ratio).
If the A/D converter is implemented using 3.3V logic, as is often the case, then this maximum incoming signal must be attenuated from 8.89V down to approximately 2V. At 2V, the converter is able to digitize the signal. This means that for maximum-level signals, the receiver circuitry must be capable of attenuating the signal by about 13 dB.
Conversely, the receiver circuitry must apply gain to minimum-level signals. To assure proper data recovery, the minimum signal-to-noise ratio (SNR) must be maintained. As a first estimate, the receiver circuitry must compensate for -40-dB attenuation, which is the number we are using for the average loss in the longest lines. Adding 40 dB of gain to our minimum setting of -13 dB results in a maximum receiver gain of 27 dB. However, at this level of gain, noise within the channel becomes an issue that must be examined.
Combating noise
Background noise levels in the telephone lines have been studied, and the results2 show that these noise levels vary widely and are frequency band-dependent. The number typically given as the average ADSL background noise power density is -140 dBm/Hz, although some lines are quieter, at -153 dBm/Hz.
For a 100-ohm line, these levels equate to a noise density range at the receiver of 31.6 nV/root Hz to as low as 7.1nV/root Hz. For a circuit designer, this noise level is critical because it determines the required noise performance of the amplifier that is used in the first stage of the receiver circuitry. Most designers endeavor to keep amplifier noise from contributing significantly to overall system noise. If the designer can assume that the system noise is 31.6 nV/root Hz, then selecting an amplifier for the receiver is not difficult, because amplifiers with significantly lower noise are readily available. However, if the goal is to keep the amplifier from significantly degrading system performance for the quieter lines as well, then an amplifier that does not significantly add to the 7.1 nV/root Hz noise density level of these lines is needed. Combating noise is more difficult in this case, because an amplifier with a noise level of 2 to 3 nV/root Hz must be used, and amplifiers at this low level of noise performance are difficult to implement.
Another circuit implementation that directly affects the noise level is the turns ratio of the transformer at the client end. To simplify the implementation of the client-end line driver, a transformer with a turns ratio greater than 1:1 is sometimes used, although doing so degrades the noise performance of the line receiver. For example, if a turns ratio of 1:3 is used to boost the line driver voltage by a factor of three, the received signal and noise is reduced by a factor of three. This means that the typical noise density of 31.7 nV/root Hz is reduced to 10.6 nV/root Hz, forcing a more stringent requirement on the noise performance of the receive amplifier.
Carefully consider the receive channel
With an ADSL link, careful attention must be paid to signal and noise levels in the receive channel. Large variations in signal amplitude require some form of gain control to keep the signal's dynamic range under control and within acceptable limits for digitization.
Signal echoing in a receive channel is another consideration that complicates the design of an ADSL system. When a signal is transmitted from the client or user end of an ADSL link to the CO end, some of the signal will echo back to the client-end receive channel. This is caused by imperfections in the impedance of the telephone line as well as imperfections in the components that were used to match the impedance of the line. Obviously, the echo signal is unwanted, but the designer must recognize and account for its presence when determining the dynamic range of the signals in the receive channel. Specific methods for dealing with this echo signal will vary, depending upon the use of echo canceling or frequency division multiplexing (FDM) in the ADSL link.
Noise considerations also dictate careful design trade-offs. A high turns ratio in the coupling transformer is helpful for boosting transmitted signal power levels, but this can be detrimental to the receive levels. To achieve an acceptable level of system noise, the client-side receive amplifier must have a noise performance level approaching 2 nV/root Hz. Although this can be reached in CMOS by using higher currents and larger die areas, it can be achieved more effectively in bipolar technology. Thus, implementing the gain change and noise performance needed in ADSL receive channels is achieved most efficiently by using a bipolar device.
Effectively implementing ADSL
ADSL, like any new technology that leapfrogs existing methods, requires special implementation considerations. Managing the wide dynamic range of ADSL signals at the remote or client end of an ADSL link is one such challenge. Another consideration that will require careful planning is how to contend with the level of noise commonly found on twisted-pair copper telephone lines.
Jim Quarfoot is a systems engineer and a senior member of the technical staff for Texas Instruments' development of high-speed amplifier product strategy. He received his BSEE from the University of Michigan, and his MSEE from Southern Methodist University.
References
Chen, W.Y., DSL Simulation Techniques and Standards Development for Digital Subscriber Line Systems, Macmillan Technical Publishing, IN, 1998, p. 126.
Ibid., p. 71.
|
