The main physical (PHY) layer structure of asymmetric digital subscriber line (ADSL) technology, as explored in part one of this article, involves a modulation scheme known as discrete multitone (DMT). In this second part, we examine other PHY layer components, move on to the handshake and initialization process of ADSL, and finally end with a look at the network and application layer protocols.
The phone line interface in ADSL is an analog circuit, but the signal processing for a DMT signal is done digitally, so an interface is needed between these two circuits. An analog-to-digital converter (ADC) is connected in the receive path of the hybrid to interface downstream signals at the home, while a digital-to-analog converter (DAC) is used to connect into the transmit path for upstream data.
Multicarrier systems contain many simultaneous signals. If each signal's peak amplitude is represented by x, and all signals simultaneously reach peak signal level, the resulting level would be x*20logN, where N is the number of signals. The signals in a DMT system have a statistical nature - they can be considered uncorrelated random processes (that is, their cross-correlation is equal to zero). The possible peak amplitude may be large, but the probability of this level occurring is low.
The peak-to-average ratio (PAR) is used to define the ratio between a signal's peak level and its average level over time. Most multicarrier systems use a modified definition for PAR that is based on the statistical likelihood of exceeding a certain peak level (such as the probability of clipping in the DAC output). For 256 subcarriers and a clipping probability of 10-7, the PAR value is around 5 (14 dB). The PAR value partially determines the operating parameters for the ADC and DAC.
The important parameters associated with the design of the converter for ADSL include the PAR factor mentioned above, the number of bits per subcarrier, and the required signal-to-noise ratio (SNR). Typical DACs use 10-b resolution, a level considered acceptable for up to 8 b per subcarrier.
The ADC must take into account all of these parameters, plus additional bits of resolution for input noise (receiver SNR is lower than transmitter SNR) and for echo energy. Typically one to two extra bits are employed in the ADC.
Other ADSL components
The POTS splitter is a passive three-terminal circuit which is key to the coexistence of the existing phone service and ADSL service. The splitter contains a common terminal (from the phone plant end of the service), a low-pass filter to the POTS side, and a high-pass filter to the ADSL side. The filters are designed to diplex the signals onto the outside phone line and have stopband impedance characteristics that minimize the effect of changing from "on hook" to "off hook" condition.
The residence modem also contains several other blocks. A hybrid simultaneously connects the transmitter and receiver to the same copper wire. The hybrid contains the line conditioning and gain control for proper operation over a range of signal levels.
The digital circuits connected to the DAC and ADC converters contain the signal processing and memory necessary to perform the demodulation and data conversion to get the information from the phone wire to the home device.
The equipment in the central office (CO) is similar to the equipment used in the home except it is structured as a bank of modems. These modems, one for each home in use, along with some network and phone interfacing equipment, comprise the device known as a digital subscriber loop access multiplexer (DSLAM).
Phone systems employ tricks to extend the usable range of their system. One such trick is to employ load coils in loops that potentially extend far from the CO (more than 3 miles or 16 kft). The copper wire will attenuate frequencies in the voice band at these distances, degrading operation. However, the revised resistance design (RRD) distance may not have been reached yet, so switching operation is still possible.
In order to maintain voice service operation, load coils are added in series with the line at periodic distance intervals. These coils are used to compensate for the effect of the cable capacitance through the voice band region on the line. As a result, the frequency response above the voice band (ADSL spectrum) rolls off at an accelerated rate. Load coils typically need to be removed from phone lines that will carry ADSL services. In some cases the plant must be re-engineered to compensate for the missing load coils.
ADSL has evolved to the point of receiving attention at a standards level. In 1998, agreement was reached on a set of standards for ADSL. G.992.1 is part of a suite of standards (the G.99x.x series) covering several DSL systems as well as protocols and tests. Key PHY layer specifications are outlined in Table 1.
|Table 1: G.992.1 PHY Layer Parameters
Overall symbol rate||4 kHz|
Number of carriers per DMT
Cyclic prefix length
Operational modes||FDM or echo cancelled
FDM mode frequency range||64 to 1100 kHz
Echo cancelled mode frequency range
||13 to 1100 kHz
Number of bits assigned per subcarrier||0 to 15 (no bits assigned to 64k QAM)
Synchronization||Pilot tone at subcarrier 64, f = 276 kHz
Number of subcarriers per DMT symbol||32
Cyclic prefix length||4 samples
FDM mode frequency range||11 to 43 kHz
Echo cancelled mode frequency range||11 to 275 kHz
Synchronization Pilot tone at subcarrier||16, f = 69 kHz
The PHY layer specification outlines power levels and spectral masks that must be maintained.
The lower three to six subcarriers are set to a gain of "0" (turned off) to permit operation of POTS, with the addition of a splitter at the home entry point of the phone line.
It is important to note that these standards encompass only a framework for operation. Individual networks and providers are free to adapt their system within that framework. The standards provide a set of boundaries for equipment manufacturers.
In the beginning
When an ADSL termination unit-residence (ATU-R), or modem, is first connected to a DSL network, it goes through a fairly extensive initialization process. This process identifies and qualifies both the capabilities of the network equipment and of the underlying physical infrastructure.
The initialization process consists of four major phases. The first phase is a handshake using the G.994.1 or G.hs protocol. (G.hs is a precursor to the G.992.1 specific initialization and is used by other DSL and telecommunication devices.). The remaining three phases - transceiver training, channel analysis, and exchange - are covered directly in the G.992 standard and apply specifically to standards-based ADSL networks.
The G.994.1 handshake is used to determine the nature and capabilities of the endpoints (such as an ADSL modem) and to indicate which protocol will be used for the remainder of the initialization. The signaling method used for the handshake interchange is designed to be robust to address channel characteristics that could be atrocious. Biphase shift keying (BPSK) modulation is used to modulate multiple single-tone subcarriers, all carrying the same data.
The subcarriers used are selected based on the typical impairments likely to be present in a given global region. The handshake has several possible variants, but, fundamentally, the two endpoints exchange a message which contains information about the endpoint type, and a number of related subparameters such as the frequency range and number of DMT subcarriers supported.
The second phase of initialization is transceiver training. Receivers at each end of the line acquire the DMT symbol stream, adjust receiver gain, perform symbol timing recovery, and train any equalizers. There is an optional echo cancellation training step that can also be performed during this phase, but the specification does not define the training signal to be used.
The transmitter power at each end of the line is set to a predetermined level at this phase, allowing a preliminary estimate of loop attenuation by the receivers. The received upstream power level is reported back to the ATU-R transmitter to allow limited power level adjustment (attenuation), if needed, to meet spectral mask requirements. The training phase is conducted with all available upstream and downstream subcarriers modulated, using two of the four constellation points of a QPSK constellation.
In the third phase, the transceivers exchange capability information and perform detailed channel characterization. For example, the ADSL termination unit-CO (ATU-C) specifies the minimum SNR margin for the system and whether it can support functions such as trellis coding and echo cancellation. Similar information is exchanged about the ATU-R. Although some of these same parameters were exchanged during the G.994 handshake, the handshake message parameters are used to gather information only and are not necessarily used for the connection.
During this third phase, both transceivers attempt to measure specific channel characteristics such as unusable subcarriers, loop attenuation on a per subcarrier basis, SNRs, and any other channel impairments that would affect the potential transmitted bit rates. Based on the discovered channel characteristics, the ATU-C makes the first offer of the overall bit rates and coding overhead that will be used for the connection.
Four possible rates are offered, in decreasing order of preference. In the current release of the ADSL standard, the ATU-C completely controls the final bit rate. All subcarriers are modulated simultaneously with the same information. The primary tool for channel measurement is a pseudo-random bit sequence.
Setting the rates
The last phase of the initialization sets the final overall transmission rates in both the upstream and downstream directions for the connection. These final rates are determined based on calculated channel parameters measured during the channel analysis phase, and are not necessarily the same as the preliminary rates offered during that phase.
As the ATU-C controls data rates, if the ATU-R cannot support any of the offered rates, both terminals will return to the beginning of the initialization process. Otherwise the ATU-R responds with the rate it can support.
Since ADSL uses multiple orthogonal subcarriers, each subcarrier can be assigned a modulation format (number of bits per subcarrier) and relative gain independently. The ATU-C assigns bits and gains for the downstream direction while the ATU-R assigns the upstream parameters.
The resulting assignment maximizes the amount of traffic that can be carried over the PHY layer. The last part of the exchange phase is a synchronized transition from the highly robust BPSK and QPSK modulations used during the initialization to the full traffic rate modulations (such as higher-order QAM) assigned during the exchange phase. At the conclusion of the initialization steps, the system is ready to pass higher-layer traffic.
ADSL and ATM
Along with TCP/IP, ATM, with its key layers and its relationship with ADSL, is one of the key enabling technologies for broadband applications using ADSL.
ADSL uses a framed transport structure in which frames are encoded and modulated into DMT symbols. The ADSL frames can be further grouped into superframes, each consisting of 68 frames. An ADSL frame is sent every 250 microseconds, therefore a superframe is sent every 17 ms.
In full-rate ADSL, a frame can be broken down further into two parts, each being 125 microseconds. These two parts can be classified as the fast data path and the interleaved data path. The fast data path could have a higher bit error rate (BER) because the interleaver is not used to mitigate the effects of impulse noise. However, the removal of the interleaver significantly reduces the latency of this data path (which is well suited for time-sensitive information such as interactive audio and video).
ATM, which allows data to be sent asynchronously, uses cells consisting of 53 B of information. The cells' small size allows the efficient multiplexing of data from multiple sources. ATM is connection-oriented - once a connection is established, the connection carries traffic that meets the quality of service (QoS) requirements requested by the source and destination.
The ATM and ATM adaptation layer (AAL) are the two most important of the several layers in the ATM protocol stack. The ATM layer is responsible for the definition of logical connections through the network. Logical connections in ATM are known as virtual circuits (VCs).
VCs represent fundamental ways of switching in an ATM network; a VC is established between two end users on the network. Bundles of VCs are called virtual paths (VPs) and they share the same end-point. A single VP carries the cells from multiple VCs, and the cells are subsequently switched together.
The AAL is responsible for inserting higher-layer information into cells to be transported over the network. It consists of two sublayers called the segmentation and re-assembly (SAR) sublayer and the common part convergence sublayer (CPCS). The SAR sublayer segments the upper-layer protocol data units (PDUs) into 48-B cell payloads (SAR PDUs). The SAR PDUs are then passed to the ATM layer to form a complete cell.
The CPCS is responsible for performing functions for different classes of service. These are referred to as AAL1-5. A summary of the types of services is shown in Figure 1.
AAL5 is most often used for connectionless Internet traffic, as it allows the entire 48 B of cell payload to transport data with minimal overhead. A 10% overhead is typical when transporting Internet traffic over ATM.
Figure 2 illustrates a basic ADSL network architecture representing the connection between the service provider and the customer. An ATM connection is set up between the ADSL modem and a termination point in the back-end network. This connection is referred to as a permanent virtual circuit (PVC).
The deployment model for ADSL is based on the current dial-up system, which uses point-to-point protocol (PPP) to support network services such as authentication and client addressing. The ADSL Forum recommends this PPP-over-ATM-over-ADSL model, and it has become the standardized method for accessing data networks over ADSL. For the ever-popular encapsulation method of transmission, the Internet Engineering Task Force (IETF) has defined a method called RFC-2364 for PPP encapsulation over AAL5.
A DSLAM is used in all instances of ADSL deployments to aggregate traffic. With this device, the data from the ADSL modems is statistically multiplexed onto a common upstream link that interfaces to an ATM network. The ATM switch then routes the cells to their destination based on the cell header information. This destination is an IP router that reassembles the data cells into packets for transmission across the Internet.
ADSL is primarily deployed for Internet connectivity. Recently, other services have become available. For example, ADSL allows the service provider to offer multiple voice lines over a single copper pair. Instead of the voice being carried by POTS, it is now packetized directly in ATM and carried over the ADSL link. The voice packets are routed to a point in the service provider's network that interfaces through a gateway into the POTS network.
ATM's QoS capabilities allow the network to deliver acceptable voice-quality services. High-quality audio and video streaming are other potential services that can be offered using ADSL.
The demand for bandwidth and high-value content and services is soaring in both the home and office, and ADSL provides a robust and cost-effective mechanism to meet the demand. Existing infrastructures can often be reused, providing for fast and economical deployment.
Because ADSL was designed to be consumer-friendly, it can coexist with POTS. By using existing voice circuits instead of a shared broadcast medium, ADSL can better individualize services. The roll-out of ADSL will continue to increase as business models are created to exploit the capabilities, both business and technical, that are available.
Rob Rhodes is manager of the Communications Design group at Thomson Multimedia (Indianapolis, IN). The group is responsible for the development of the PHY interfaces for satellite, cable, and DTV communications. He can be reached at email@example.com.
Mike Pugel is a principal member of the technical staff at Thomson Multimedia (Indianapolis, IN). He is currently working on advanced communication receiver concepts. He can be reached at firstname.lastname@example.org.
Louis Litwin is a member of the technical staff at Thomson Multimedia (Princeton, NJ). His focus is the development of wireless communication devices used for digital home networking and mobility applications. He can be reached at email@example.com.
John Richardson is a member of the technical staff at Thomson Multimedia (Princeton, NJ). He is primarily involved in the areas of digital home networking and ADSL system development. He can be reached at firstname.lastname@example.org.
- Busby, M., Demystifying ATM/ADSL, Worldware Publishing, TX, 1998.
- Chen, W., DSL: Simulation Techniques and Standards Development for Digital Subscriber Line Systems, Macmillian, IN, 1998.
- Cioffi, J., Silverman, P., and Starr, T., Understanding Digital Subscriber Line Technology, Prentice Hall, NJ, 1999.
- Goralski, W., ADSL and DSL Technologies, McGraw-Hill, NY, 1998.
- Rhodes, R., Pugel, M., Litwin, L., and Richardson, J., "Digital Subscriber Line Technology Tutorial," International Conference on Consumer Electronics, Los Angeles, CA, June 11, 2000 (invited tutorial).