Copper transmission technologies have made remarkable advancements over the last 20 years, producing mature, nearly optimal architectures for signaling over a single copper pair. Transmission over copper, once seen as antiquated in a world moving toward fiber, has had a rebirth. With its ability to handle broadband speeds, it is central to solving the last-mile access problem.
Currently, copper provides broadband connectivity in the form of asynchronous DSL (ADSL) to tens of millions of U.S. residences and serves millions of business establishments with T1-type connections. Its success has been made possible by advances in communications technology and the signal processing capabilities of silicon chips. Modern DSL modems use sophisticated modulation techniques, advanced coding, flexible spectral band plans and sophisticated frequency division duplexing.
But as data rates and signal bandwidths over copper increase, crosstalk interactions among the binder pairs become a performance bottleneck. To overcome this, next-generation DSL systems must provide substantial performance improvements by coordinating transmission across multiple copper pairs. MIMO (multiple input, multiple output) signal processing techniques can mitigate both self and alien near-end and far-end crosstalk.
Modern DSL modems operate in an environment of major crosstalk interference. For example, near-end crosstalk into an ADSL modem can easily be 20dB or 30dB higher than any other noise source. Controlling the detrimental effects of crosstalk in the network is fundamental to achieving a next generation of performance.
The effects of near-end crosstalk can be controlled with appropriate frequency division (or even time division) duplexing. A plethora of these duplexing spectral plans have been added to the ADSL/ADSL2 and VDSL/VDSL2 standards for various loop conditions and desired symmetry ratios. Further advanced studies have investigated the dynamic allocation of spectral bands and/or regulation of power distribution across frequencies, depending on the interference conditions in the given binder.
These dynamic spectral management (DSM) techniques provide impressive performance gains when all modems in the binder cooperate and implement the DSM rules. Gains are smaller when legacy disturbers are present, which do not adhere to the DSM etiquette.
Impressive performance results are achieved if the crosstalking modems are designed to operate synchronously and coordinate their transmit signals at the waveform level. This type of coordinated multichannel signaling is often called vectored modulation and is well-suited for multichannel media with strong interactions across the channels (crosstalk).
Vectored transmission requires joint processing of the signals of all channels at the receiver and/or the transmitter in order to align amplitudes and phases in a way that counteracts the detrimental effects of the channel cross-couplings. These multichannel signal processing techniques are commonly referred to as MIMO.
Migration of Vectored Transmission
Vectored transmission techniques originated in phased array radar systems and then migrated to multi-antenna wireless links. Now they are being incorporated into a host of standards for wireless LANs and MANs. In the wireline world, MIMO near-end crosstalk cancellers have been incorporated into Gigabit Ethernet transceivers (over four copper pairs), while more advanced vectored schemes are currently developed for the next-generation 10gbps Ethernet modems.
Given the proven record of vectored signaling, it is curious that similar advanced multichannel architectures have not yet found their way into mainstream DSL applications. The reasons are both technical and operational.
The copper plant was originally designed to transport low frequency voice signals with minimal crosstalk. Each individual pair has been seen as a discrete channel, well-isolated from other channels. This is in contrast to wireless, a broadcast medium where interference among users has always been of paramount importance. As the frequency bands used by copper modems keep expanding, crosstalk is more pronounced and electromagnetic pair separation less perfect.
The potential of vectoring techniques in the copper network is significant. For example, there are several applications where high data rates are required, well beyond what a single copper pair can provide. These include high-end business access applications and DSL access multiplexer and digital loop carrier uplink connections. In these instances, MIMO processing technologies offer elegant technical solutions to crosstalk containment.
Far-end crosstalk results from interference from the far-end aggressor signal, generated due to electromagnetic coupling between aggressor and victim pairs. It is generally attenuated as loop length increases. If a 26 AWG loop is longer than three kilometers, far-end crosstalk is generally attenuated under the noise floor. In relatively short loops, it presents a challenging impairment.
The Binder as One Unit
It helps to view the entire binder as a joint transmission medium. With this in mind, and analyzing the relevant mathematical formulations, we find that far-end crosstalk not only does not negatively affect the copper pair capacity when used with a MIMO system, but actually can be a useful signal. The resulting capacity gains are possible only if the vectored system includes all the modems in the binder. If the vectored system includes only a portion of the binder, the remaining interferers add “alien crosstalk” which cannot be completely eliminated.
Interference from near-end transmitters also can couple into a victim pair and is generally stronger than far-end crosstalk interference, especially for long loops. The situation worsens when the upstream and downstream transmission bands overlap. If there is no coordination among pairs, near-end crosstalk can be detrimental on long loops. With coordinated transmission, however, near-end crosstalk may be more straightforward to mitigate than far-end crosstalk.
The receiver has access to the interfering signals and can cancel them by utilizing MIMO near-end crosstalk cancellers. This approach is conceptually a straightforward generalization of the SISO echo canceller currently used in several modems with overlapping upstream and downstream spectra. However, the approach breaks down if there are legacy disturbers in the binder not participating in the vector transmission group.
Traditionally, near-end crosstalk interference has been addressed by separating the upstream and downstream transmission bands. Especially for the higher frequency bands (over 500KHz), frequency division multiplexing is generally a better way to engineer the network. Examples include the ADSL/ADSL2 and VDSL/VDSL2 standards. Even FDM architectures, however, suffer from near-end crosstalk interference from legacy disturbers such as ADSL modems on long loops impaired by HDSL or SHDSL interferers.
A number of specific MIMO architectures counter self and alien far-end and near-end crosstalk. Among them are:
*Linear zero forcing
*QR decision feedback
*QR point to multipoint
Each of these is most effective depending on the environment, assumptions, loop conditions, and the characteristics of the signal itself.
A major distinction among different categories of MIMO architectures is whether MIMO processing is performed at the receiver or the transmitter. Generally, receiver-based MIMO is performed in the upstream direction to improve symmetric services, which typically suffer from an upstream bottleneck. Transmitter-based MIMO is performed in the downstream direction to improve residential services that require more downstream bandwidth (e.g., IPTV services).
Receiver-based MIMO is a function internal to the receiver that does not require explicit cooperation from the transmitter. In contrast, transmitter-based MIMO at the DSLAM requires close cooperation with the CPE receiver to continuously obtain channel feedback information. It therefore faces more network management, standardization and complexity issues. In the case of a bonded copper system (where multiple pair links are multiplexed into a higher bandwidth link), one can implement receiver-based MIMO both upstream and downstream, avoiding the added complexity of transmitter-based MIMO.
In conclusion, vectored MIMO technologies have the potential to provide a step function increase in the performance of DSL systems. By applying these architectures to the DSL space and the copper network, we can pave the way for the next generation of DSL systems.
Michail Tsatsanis is co-founder and chief scientist at Aktino. This article was condensed from a chapter in his upcoming book on multiline DSL system architectures. For more information about MIMO, visit www.aktino.com.