Next-Generation VDSL Networks: Enabling 100 Mbps and Beyond
Low deployment costs and high bandwidth have allowed Fiber-to-the-Node (FTTN) networks using traditional copper twisted pair for the Last Mile link to become the architecture of choice for delivering voice, video, and advanced data services to a bandwidth-hungry world. This topology allows the rapid deployment of lucrative Triple Play services at a fraction of the cost of an all-fiber network. However, broadband equipment vendors must overcome several technical challenges in order to support voice and video services with the carrier-class reliability and Quality of Service (QoS) that consumers demand.
The most critical ingredient for ensuring the consumers’ Quality of Experience (QoE) in these applications is Very High Speed Digital Subscriber Line (VDSL) technology and the new noise-cancelling standards that increase the link’s stability in the challenging line conditions found in legacy copper networks.
While standards from the International Telecommunications Union (ITU) G.993.2 and others play an important role in maintaining reliable, interoperable connections, they must be implemented in a way that delivers the real-world benefits of stability and reliability necessary to support large-scale carrier-deployed multimedia services.
This article describes the common external noise sources which adversely affect VDSL/VDSL2 line performance. It then goes on to describe the cutting-edge, adaptive technologies which mitigate these adverse effects and deliver superior system performance far surpassing the requirements set forth in ITU and other standards.
An Ailing Legacy Plant
FTTN networks combine a fiber backbone with existing copper infrastructure — in particular VDSL/VDSL2 — for Last Mile links allowing service providers to transform their networks for delivering revenue-enhancing Triple Play services. Capturing these opportunities requires much more than simply increasing subscribers’ bandwidth. To successfully deliver IP-based telephony and HD video services, the VDSL technology (and in particular the underlying VDSL silicon) employed in these networks must have carrier-class reliability and stability despite challenging line conditions found in legacy copper networks.
In addition to overcoming the attenuation, reflections, and other channel impairments inherent to twisted pair wire plants, a VDSL silicon chipset must be able to operate reliably in the presence of multiple noise sources from both the outside world and from within the cable bundle itself. While VDSL equipment contains many mechanisms for controlling latency, bandwidth allocation, and other QoS parameters, they are rendered useless if the link is subjected to extreme data rate variations or brief interruptions in service.
Standards bodies such as the ITU have implemented mechanisms which improve the reliability of VDSL/VDSL2 links. The ITU G.993.2 VDSL2 standard was approved in November 2006. Compared with earlier generations of xDSL, VDSL2 adopts a number of improvements which address the bandwidth needs of different operators. For example, VDSL2 supports improved impulse noise protection (INP), impulse noise monitoring (INM), SRA, U0 PSD (power spectral density) shaping, and new service and initialization policies for enhanced Triple Play stability. Nevertheless, the link reliability and rate stability afforded by these standards-based mechanisms cannot guarantee a satisfactory customer QoE under some of the extreme conditions routinely experienced by many copper networks.
Performance Limiters: A Mounting Problem
You see, FTTN network deployments are subjected to a variety of noise sources, both internal and external, which degrade system performance. Some of the most common external noise sources in xDSL applications include 2 kinds of Crosstalk (XT):
1. Near End Crosstalk (NEXT)
2. Far End Crosstalk (FEXT)
In VDSL applications, FEXT is a primary performance limiter. This article describes how noise mitigation technologies are enabling 100 Mbps performance in VDSL applications. Therefore, unless otherwise specified, all crosstalk discussions focus on FEXT.
To begin, we must review 4 performance-limiting sources of noise.
Performance Limiter #1: Crosstalk (FEXT)
Crosstalk is a type of radiated noise inherent to any copper conductor carrying high-speed data signals, including xDSL, T1, etc., which is injected via capacitive or inductive coupling into neighboring copper conductors. In VDSL applications, FEXT, a primary performance limiter, is generated from virtually every copper pair within any cable bundle, and its impact is especially high on shorter loops.
Figure 1 illustrates a typical 400-pair copper cable structure. This structure is comprised of four 100-pair superbinders. Each superbinder is comprised of 4 smaller cable binders, each containing 25 copper cable pairs.
In typical xDSL deployments, crosstalk between cable pairs inevitably drives down network performance. To illustrate the effects of crosstalk, consider a typical 50-pair binder shown in Figure 2. Each copper pair within this binder interferes with every other copper pair, which dramatically reduces performance.
Figure 3 illustrates the measured crosstalk coupling for 25 cable pairs spanning 2 binders.
• The white squares along the diagonal from lower-left to upper-right represent the crosstalk measured between a cable pair and itself.
• Squares immediately adjacent to the white diagonal (i.e., lower-left corner to upper-right corner) represent cable pairs within the same binder and are in close physical proximity with one another.
• Squares located in the upper-left corner and lower-right corner represent cable pairs within adjacent binders.
Depending on the geometry of the individual binders and the superbinder, cable pairs within separate binders may or may not be in close physical proximity with one another.
A color-based gradient scale (see Figure 3) has been applied to the measured data to illustrate the relative level of crosstalk coupling between any 2 cable pairs:
• Blue squares represent low-crosstalk coupling between the corresponding Disturber-Victim cable pair.
• Red represents high-crosstalk coupling between the corresponding Disturber-Victim cable pair.
• White squares along the diagonal represent self-coupling between a pair and itself. This is extremely high.
• Additional colors represent varying levels of crosstalk coupling between the corresponding Disturber-Victim cable pair relative to their position along the gradient scale.
Data indicate that the highest levels of crosstalk experienced by a cable pair generally originate from within its own binder:
• Squares adjacent to the white diagonal (i.e., lower-left corner to upper-right corner) represent cable pairs within the same binder that are in physical proximity with one another. As shown, there is a greater concentration of red squares along this diagonal, indicating increased levels of crosstalk due to these adjacent/neighboring disturbers.
• Squares located in the upper-left corner and lower-right corner represent cable pairs located in a different cable binder and are (relatively) physically separated from one another. As expected, there is a greater concentration of blue and green squares, indicating lower measured crosstalk levels due to these remote disturbers.
• Although crosstalk coupling is typically lower between binders, coupling can still be higher in some situations as represented by the yellow boxes in the upper-left and lower-right corners of Figure 3. Since binders are not usually insulated, crosstalk coupling between physically adjacent cable pairs spanning 2 binders is possible. In addition, it is common for telco personnel to reorganize 1 or more cables in the field at splice points, thus introducing additional means of crosstalk being introduced to the system.
In summary, it is extremely difficult to accurately predict which cable pairs will interfere with each other, and in order to be effective, crosstalk cancellation techniques must take the entire cable into account, including any splice points that may have been introduced by the telco itself.
Performance Limiter #2: Radio Frequency Interference (RFI)
RFI is a sinusoidal type of radiated interference generated by electric equipment ranging from broadcast transmitters to industrial microwave equipment. RFI interference has been known to impair, or completely eliminate, multiple tones within the DMT spectrum, resulting in a significant reduction in a channel’s overall data rate. Fortunately, RFI is typically restricted to a relatively narrow and predictable part of the frequency spectrum, making it easy to create RFI-blocking filters. However, RFI typically has a much longer duration than an impulse noise source.
Performance Limiter #3: Impulse Noise
As its name implies, Impulse Noise is an instantaneous disturbance typically manifested as a voltage surge/spike on a line. Impulse Noise can be caused by nearby lightning strikes, transformer surges, or simply by turning home appliances or light switches on and off. Although rare, depending on the source of a disturbance, (e.g., a lightning strike) power surges have been known to completely destroy electronic equipment.
Repetitive Impulse Noise (REIN) is similar to Impulse Noise, but is periodic in nature and persists for an extended length of time. The periodic characteristic of REIN can often be traced back to the frequency of the local line voltage (i.e., 50 Hz or 60 Hz.) In audio applications, REIN is often referred to as hum, and for analog over-the-air broadcast video, REIN can be blamed for that annoying horizontal strip that travels from the bottom of the screen to the top, and then starts over.
In xDSL applications, both Impulse Noise and REIN can cause excessive Cyclical Redundancy Check (CRC) errors on the line which may cause a retrain of the DSL modem.
Performance Limiter #4: Additive White Gaussian Noise (Noise Floor)
Additive White Gaussian Noise, often referred to as the Noise Floor, is typically at a level of -140 dBm/Hz or below, and is the fundamental limiting factor on xDSL data rate in the absence of other noise sources.
Quieting the NOISE
Although most DSL noise mitigation techniques are based on the assumption that the interference emanates from random sources within a Gaussian distribution, this is an incomplete model which underestimates the impact of other types of interference. In many cases, the discrepancy between the Gaussian model and a line’s real-world behavior results in an excessive error rate and/or significantly reduced data capacity.
In addition, traditional channel estimation procedures have been designed to optimize performance in the presence of stationary impairments. As a result, they do not perform well when estimating non-stationary, quasi-stationary, or time-varying interference sources, and are ineffective in helping a modem’s receiver cope with impairments such as impulse noise.
Figure 4 shows the frequency versus Power Spectral Density (PSD) curves for a variety of noise sources that are commonly encountered in current VDSL field deployments.
Ikanos has developed 2 noise-mitigating products that work together to cancel the deliterious effects of crosstalk and noise on copper wires:
1. G.Vector to eliminate far-end crosstalk.
2. G.inp, Rapid Rate Adaptation for narrowband and impulse noise.
Understanding G.Vector
In April 2010, the International Telecommunication Union (ITU) ratified G.993.5, commonly referred to as the G.Vector standard. This standard creates a common framework for performing crosstalk cancellation technology, or vectoring. Vectored DSL technology, also known as Dynamic Spectrum Management Level 3, mitigates the crosstalk effects that form the most serious performance bottlenecks for dense deployments of DSL lines operating at very high speed.
Vectoring is an exciting and increasingly critical addition to a carrier’s arsenal when combating FEXT in FTTN deployments. Using vectoring, carriers can cancel the effects of crosstalk in the field, thus allowing them to reliably deliver 100 Mbps with a minimum of effort.
One example of the implementation of vectoring is called NodeScale Vectoring. It incorporates data compression algorithms to reduce the memory overhead required for raw coefficient data storage. For example, a standard vectoring approach would require 3.6 Gbits of raw coefficient data in order to vector 96 ports. NodeScale Vectoring dramatically reduces this overhead to only 30 Mbits for the same 96 ports. NodeScale Vectoring deployments are easily and economically scalable for applications that have very large port counts (i.e., up to 384 ports).
Figure 6 compares achievable data rates versus loop lengths for VDSL2 systems in the presence of crosstalk. You can see there are 3 VDSL2 systems with 100 ports or more compared in Figure 6:
1. NonVectored VDSL2: The bottom curve (shown in red) shows the achievable data rate for a VDSL2 system without applying any vectoring technology.
2. Line Card Vectored VDSL2: The middle curve (shown in blue) represents the achievable data rate when Line Card vectoring is employed. Line Card vectoring applies vectoring techniques to each line card individually without any coordination between the other Line Cards in the cabinet. As illustrated, minimal data rate gains are achieved since a majority of the crosstalk experienced is originating from neighboring Line Cards in the cabinet.
3. NodeScale Vectored VDSL2: The top curve (shown in green) represents the achievable data rate when NodeScale Vectoring is applied.
With NodeScale Vectoring technology, full system performance is achievable as if there were no crosstalk. For example, at 500 meters, the rates would increase from approximately 50 Mbps to 100 Mbps, and higher rates are achievable on shorter loops.
Figure 7 illustrates a node simulation where the benefits of NodeScale Vectoring are shown for a typical VDSL system of 100 ports or more.
The following curves are shown:
• The expected system performance using a 99% worstcase crosstalk model.
• The expected performance in a FEXT/crosstalkfree environment.
• Non-vectored data rates that have been calculated using the crosstalk characteristics inherent to the specific cable deployed in the field.
As shown, there is a wide range of achievable rates when crosstalk cancellation techniques are not used. Unfortunately, it is virtually impossible to predict what data rate will be achieved by any individual customer. Therefore, service providers will typically deploy service rates based on a worstcase scenario, such as the 99% worstcase model.
The data rate curves shown by red circles represent achievable data rates when a vectoring product such as NodeScale is used. As seen in Figure 7, data rates approaching those for a FEXTfree environment are achievable even on short loops.
It is clear that when coefficient compression technology is used, data rates approaching those for crosstalkfree environments are achievable with minimum added system complexity. Downstream data rates of 100 Mbps or more are achievable at 500 meters, while 50 Mbps can be reliably obtained at 900 meters, roughly triple the previously observed loop length.
Increased SignaltoNoise Ratio
Figure 8 shows theoretical system performance curves for a 16line FEXTfree system:
• The top plot represents SNR per Tone.
• The bottom plot represents data rate per port index.
• Theoretical FEXTFree SNR and Achievable Data Rate for a 16Line VDSL2 System.
Figure 9 shows the effects on these same performance curves after FEXT has been introduced to the system.
The top plot shows that the Signal-to-Noise ratio has been reduced by 20 dB or more.
The achievable data rate plot shows a corresponding reduction in user throughput when FEXT is present. The green bars represent the achievable data rate when vectoring technology is not incorporated.
Finally Figure 10 shows how NodeScale Vectoring virtually eliminates the negative effects of FEXT on SNR and achievable data rates.
The top plot illustrates how the Signal-to-Noise ratio has been dramatically increased, approaching the theoretical FEXT-free levels.
In the lower plot, the red bars represent the data rate throughput that is achievable when NodeScale Vectoring technology is utilized. These data rates again approach those for FEXTfree operation.
Narrowband and Impulse Noise Mitigation
When coping with narrowband and impulse noise sources, additional ITU standards have been developed. These include ITUT G.998.4 (G.inp). The first layer of G.inp’s protection uses a combination of a strong Reed-Solomon Forward Error Correction (FEC) encoding scheme that hardens packets against multiplebit errors and a retransmission scheme to recover packets that are unrecoverable with FEC coding. When a VDSL transceiver encounters persistent, severe impulse noise, interleaving techniques are also employed to further harden the channel against data loss.
Caveat: G.inp is not a standalone standard, and must be implemented in conjunction with the G.993.2 VDSL2 standard.
Figure 11 illustrates the retransmission scheme used by G.inp.
Figure 11. Basic Steps in Retransmission
G.inp relies upon the receiver to verify each packet for integrity. If an unrecoverable error is detected, the receiver sends a retransmission request to the transmitter. The receiver is also responsible for buffering any outofsequence blocks until a retransmitted block is received.
Similarly, the transmitter is responsible for storing any blocks that have not been acknowledged by the receiver and which may be eligible for retransmission requests.
In environments with lowtomoderate levels of impulse noise, retransmission provides an efficient means of recovering lost packets without unnecessary fixed transmission overhead. As packet error levels increase, however, the jitter, latency, and bandwidth consumed by retransmission requires that additional techniques be used to harden the link against data loss.
For environments rich in frequent and short impulses that cause too many retransmit events, again this vectoring solution innovatively implements the standard. It uses a combination of interleaving and Reed-Solomon encoding to minimize retransmission of lost packets. Since these techniques add latency and reduce the channels’ usable bandwidth, this particular solution employs control algorithms that have been tuned to activate only when the overhead they impose causes less impact than the retransmission losses they are preventing.
Since the G.inp algorithms’ effectiveness varies significantly according to how they are implemented, companies such as ours have invested heavily in the development of optimized algorithms and firmware that deliver performance that significantly exceeds the ITU’s requirements while remaining completely standardsbased. It is able to withstand Repetitive Electrical Impulse Noise (REIN) rates of 120 PPS at up to 110 mVpp. When combined with Erasure Decoding and dedicated INP memory, the Ikanos solution delivers twice (200%) the impulse noise protection required by the ITU G.inp standard.
Copper: Alive and More-Than-Kicking
VDSL has the potential to deliver broadband data and Triple Play services at a fraction of the cost of an allfiber solution but it must also deliver carrierclass reliability and quality of service under challenging realworld conditions. This requires hardening the VDSL link against impulse, broadband, and other noise sources to levels that are significantly higher than specified in standards like ITU G.993.2. Advanced implementations of these technologies — in the form of silicon products from suppliers like Ikanos — are the key to providing maximum link reliability while maintaining maximum overall system performance. These solutions allows designers to build standardsbased VDSL equipment that enjoys longer reach, higher capacity, and the superior quality of service (QoS) required by premium voice and HD video services.