DSL SERIES: Basics: The DSL Trick

The PSTN and supporting local access networks were designed with guidelines
that limited transmissions to a 3,400 Hz analog voice channel. For example,
telephones, dial modems, fax modems and private line modems limited their
transmission over the local access phone lines to be in a frequency spectrum
that exists between 0 Hz and 3,400 Hz. The highest achievable information rate
using that 3,400 Hz frequency spectrum is less than 56 Kbps.

So, how does DSL technology achieve information rates in the millions of bits
per second over those same copper loops?

The
answer is simple — eliminate the 3,400 Hz boundary. DSL, much like traditional
T1 or E1, uses a much broader range of frequencies than the voice channel. Such
an implementation requires transmission of information over a wide range of
frequencies — from one end of the copper wire loop to another complementary
device, which receives the wide frequency signal at the far end of the copper
loop.

Now, recognizing that we can choose to eliminate the 3,400 Hz frequency
boundary and dramatically increase the information rates supported on copper
wires, you may be asking, "Why don’t we just ignore POTS transmission
guidelines and use the higher frequencies?" The answer can get far more
complex than we want to cover here, so we will consider the three dominant
issues associated with this question:

• Attenuation: The dissipation of
  the power of a transmitted signal as it travels over the copper wire line.
  In-home wiring also contributes to attenuation.

• Bridged taps:
  These are unterminated extensions of the loop that cause additional loop
  loss, with loss peaks surrounding the frequency of the quarter wavelength of
  the extension length.

• Crosstalk: The
  interference between two wires in the same bundle, caused by the electrical
  energy carried by each.

Attenuation and Resulting Distance Limitations

One
might compare the transmission of an electric signal to driving a car. The
faster you go, the more energy you burn over a given distance and the sooner you
have to refuel. With electrical signals transmitted over a copper wire line, the
use of higher frequencies to support higher-speed services also results in
shorter loop reach. This is because high-frequency signals transmitted over
metallic loops attenuate energy faster than the lower-frequency signals.

One way to minimize attenuation is to use lower-resistance wire. Thick wires
have less resistance than thin wires, which in turn means less signal
attenuation and thus, the signal can travel a longer distance. Of course,
thicker-gauge wire means more copper, which translates into higher per-foot
plant costs. Therefore, telephone companies have designed their cable plant
using the thinnest gauge wire that could support the required services.

In the US, wire thickness is represented by the denominator composed of the
fraction of an inch in wire size, assuming a numerator of 1. Therefore, a wire
that is 1/24 inch in diameter is referred to as 24 American Wire Gauge (AWG).
Wire gauges of 24, and more often 26, are present in most North American cable
plants. The design rules used by nearly all telephone companies provided for a
change in wire gauge, with a thinner gauge used near the entrance of a central
office to minimize physical space requirements and changing to thicker gauges
over long loops to maximize loop reach.

In most markets outside North America, wire gauges are referred to by their
diameter in millimeters. For example, 0.4 mm, which is comparable to 26 gauge,
and 0.5 mm, which is comparable to 24 gauge, are the most common; although in
many developing countries, heavy gauges of 0.6 mm to 0.9 mm can be found in
newly urbanized areas. This variation in wire gauge adds to the challenge of
determining a particular DSL system’s performance over a particular loop.

Advanced Modulation Techniques minimize Attenuation

In
the early 1980s, equipment vendors were working aggressively to develop Basic
Rate ISDN, which would provide up to two 64 Kbps B-channels, plus a 16 Kbps
D-channel used for signaling and packet data. The information payload, plus
other overheads associated with implementation, resulted in 160 Kbps in total
transmitted information. A key requirement of ISDN was that it had to reach
customers over the existing non-loaded copper wire loops, equating to 18,000
feet. However, an AMI implementation of Basic Rate ISDN would require use of the
lower 160,000 Hz, which resulted in too much signal attenuation and would fall
short of the required 18,000 feet loop reach of 26 gauge wire.

By 1988, advancements in signal processing and line coding doubled the
effectiveness of the legacy AMI code by sending two bits of information with
each cycle of an analog waveform or baud. The line code was called 2 Binary, 1
Quaternary (2B1Q). A 2B1Q implementation of Basic Rate ISDN used frequencies
ranging from 0 to approximately 80,000 Hz, which has less attenuation and
results in the desired 18,000-foot loop reach.

HDSL enters the Scene

In the early 1990s, some vendors encouraged the use of 2B1Q at higher speeds
as an alternate way to provisioning T1 and E1 services, without repeaters. The
technique consisted of splitting the 1,544,000 bps service into two pairs (four
wires), which each ran at 784,000 bps. By splitting the service across two lines
and increasing the bits per baud, the per-line speed and resulting need for
frequency spectrum could be reduced to allow longer loop reach. This technique
was referred to as High-bit-rate Digital Subscriber Line or HDSL. The result was
that an HDSL-based DS-1 service could be implemented over Carrier Serving Area
(CSA) specified loops of up to 12,000 feet long (assuming 24 gauge; or 9,000
feet with 26 gauge wire), with no repeaters.

The early 2B1Q-based E1 HDSL initiatives split the 2.048 Mbps service across
three wire pairs (a total of six wires) in an effort to achieve the targeted
loop reach. As the technology matured and performance improved, E1 HDSL
implementations migrated to a two-pair (four-wire) implementation, each
operating at 1.168 Mbps, which was similar to the T1 implementation.

In
parallel with the 2B1Q initiative, Paradyne began development of a similar HDSL
transceiver using a line code called Carrierless Amplitude and Phase (CAP)
modulation. Like 2B1Q, CAP was an advanced line-coding technique allowing
multiple bits of information to be represented by a single frequency cycle or
baud. However, CAP could be designed to transmit multiple bits ranging from two
to nine bits per baud. This enabled CAP-based transceivers to transmit the same
amount of information using a lower range of the frequency spectrum than 2B1Q,
equating to less signal attenuation and greater loop reach. As a result of 2B1Q’s
proven market acceptance with ISDN and CAP’s performance benefits, both line
codes were endorsed with technical reports by both the American National
Standards Institute (ANSI) and European Telecommunications Standardization
Institute (ETSI) standards committees for HDSL.

There are some instances where vendors have developed HDSL products using
line codes other than 2B1Q or CAP.

However, these examples are isolated and alternative line codes are not
recognized by the standards organizations.

The higher-frequency signals associated with the AMI implementation get weak
sooner than the HDSL transmissions. As a result, the CAP and similarly 2B1Q HDSL
systems have substantially longer loop reach than AMI or HDB-3 based T1 or E1
systems, respectively.

Bridged Taps

Bridged taps are unterminated extensions for the loop that cause additional
loop loss, with loss peaks surrounding the frequency of the quarter wavelength
of the extension length. Since wavelength and frequency have an inverse
relationship, short bridged taps have the greatest impact on wideband services,
while long bridged taps have a greater impact on narrowband services. Most loops
contain at least one bridged tap, and the effect of multiple taps is cumulative.
Premises wiring contain additional bridged taps. The additional loss created is
greatest on short bridged taps; consequently, technologies that operate at lower
frequencies are less impacted.

The Effects of Crosstalk

The electrical energy transmitted across the copper wire-line as a modulated
signal also radiates energy onto adjacent copper wire loops that are located in
the same cable bundle. This cross coupling of electromagnetic energy is called
crosstalk.

In the telephone network, multiple insulated copper pairs are bundled
together into a cable called cable binder. Adjacent systems within a cable
binder that transmit or receive information in the same range of frequencies can
create significant crosstalk interference. This is because crosstalk-induced
signals combine with the signals that were originally intended for transmission
over the copper wire loop. The result is a waveform shaped differently than the
one originally transmitted.

Crosstalk can be categorized in one of two forms. Near End Crosstalk,
commonly referred to as NEXT, is the most significant because the high-energy
signal from an adjacent system can induce relatively significant crosstalk into
the primary signal. The other form is Far End Crosstalk or FEXT, which is
typically less of an issue because the Far End interfering signal is attenuated
as it traverses the loop.

Crosstalk
is a dominant factor in the performance of many systems. As a result, DSL system
performance is often stated relative to the presence of other systems, which may
introduce crosstalk. For example, the loop reach of a DSL system may be stated
as being in the presence of 49 ISDN disturbers or 24 HDSL disturbers. As you can
imagine, it is rather unlikely that you will deploy a DSL service in a 50-pair
cable that happens to have 49 (two-wire) ISDN circuits or 24 (four-wire) HDSL
circuits concurrently running in the same bundle. Therefore, these performance
parameters typically represent a conservative performance outlook.

Transmitting and receiving information using the same frequency spectrum
creates interference within the single loop system itself. This interference
differs from Crosstalk because the offending transmit waveform is known to the
receiver and can effectively be subtracted from the attenuated receive signals.
Eliminating the effects of the transmitter is referred to as echo cancellation.

Minimizing Crosstalk

If the effects of the attenuation and Crosstalk are not too significant, the
DSL system can accurately reconstruct the signal back into a digital format.
However, when the effect of these phenomena becomes too significant, the signals
are misinterpreted at the far end and bit errors occur.

Some DSL systems use different frequency spectra for the transmit and receive
signals. This frequency-separated implementation is referred to as Frequency
Division Multiplexing (FDM). The advantage of FDM-based systems over
echo-canceled systems is that NEXT is eliminated. This is because the system is
not receiving in the same range of frequencies in which the adjacent system is
transmitting. FEXT is present, and the FEXT signal is substantially attenuated
and less of an interferer because the origin of the FEXT signal is at the
distant end of the loop. Therefore, FDM-based systems often provide better
performance than echo-canceled systems, in terms of crosstalk from similar
adjacent systems.

One interesting phenomenon that should be considered is that echo-canceled
systems of a like type, introduce what is called Self Next. Self Next introduces
significant interference to other like-type echo-canceled systems in the same
cable binder. As a result, the deployment of multiple like-type echo-canceled
systems will degrade the performance of all other like-type systems within the
cable binder. For example, a single CAP or 2B1Q-based T1 HDSL system may achieve
the targeted 12 kft (kilofeet) loop reach. However, as additional CAP or
eB1Q-based systems are added to the cable bundle, the loop reach of the first
system and the subsequent systems may be reduced to 9 kft or less. This same
phenomenon is true of nearly all echo-canceled systems, such as 2B1Q in general,
echo-canceled CAP HDSL and SDSL, and echo-canceled DMT ADSL systems. Therefore,
when selecting a DSL technology, service providers should examine the system
performance in the presence of Self Next, which is certain to exist as more
services are deployed.

The engineering compromise of FDM systems is that the separated upstream and
downstream signals occupy a greater range of frequencies than echo-canceled
systems, which overlap the transmit and receive signals resulting in less reach.
In some cases, attenuation becomes the most significant factor in performance.
In other cases, Crosstalk is the most significant factor in performance.
Therefore, the optimal implementation varies as a function of the environment.
In deployments where Crosstalking systems are expected to be limited and NEXT is
moderate to low, an echo-canceled system may perform better. In other cases,
where deployments of Crosstalking systems are expected to be significant and
NEXT is likely to be more dominant, an FDM system may perform better.

About the only sure way to manage the issues of Crosstalk, is to first
research the services that are deployed within a given cable bundle and avoid
those services that will provide substantial Crosstalk. One example of this is
the traditional T1 or E1 services. The spectral placement of T1 AMI and
similarly the E1 HDB3-based services provides extensive Crosstalk to almost all
DSL-based services. As a result, most service providers follow design rules that
do not allow the use of T1 or E1 services in the same cable bundles with DSL-based
services. You should expect reductions in loop reach in scenarios where T1 or E1
is provisioned in the same cable bundle as DSL-based services.

The DSL Series is brought to you in association with Paradyne Corp.

Next month: Basics–The Varieties of DSL