TELECOM Digest OnLine - Sorted: Re: DSL Speed


Re: DSL Speed


Neal McLain (nmclain@annsgarden.com)
Mon, 27 Jun 2005 05:26:56 -0500

Choreboy <choreboyREMOVE@localnet.com> wrote:

> A relative has a farm. His phone service comes in on 700 yards
> of ordinary telephone cable buried along his driveway. Last
> week he got Bellsouth DSL. It comes in on the same conductors
> as before, but I've seen speeds fifty times faster than dialup

And in a subsequent post, wrote:

> Between the CO and the customer, isn't voice service just bare
> wire?

Not necessarily. But let's clarify some terminology first.

I assume that:

- By "between the CO and the customer," you mean what's
commonly known as the "local loop."

- By "bare wire" you don't really mean "bare" (as in
uninsulated); you're simply implying that there's nothing in
the wirepair, other than copper conductors, that would affect
the transmission of signals.

Based on those assumptions, here's an attempt to explain "local loop":
it's a pair of metallic (usually copper) conductors between the
customer's premises and the telco's facilities. The conductors are
designated "ring" and "tip." These terms originated from the physical
configuration of the plugs used in old manual switchboards. Photo:
http://tinyurl.com/9pjla . Note that the term "ring," as used here,
does not mean "ringing the telephone."

The two conductors are usually twisted together, and contained inside
a cable along with several other wirepairs. At the customer's
premises, the conductors may run parallel (not twisted) in the drop
cable from the pole (or pedestal) to the building.

At the telco's end, the loop may terminate at the CO, or it may
terminate at a "digital loop carrier remote terminal" (DLCRT, or just
RT). Telcos often deploy RTs to provide POTS service to outlying
areas (e.g., new residential neighborhoods or business parks) in order
to reduce the number and/or length of wirepairs needed to provide
service to additional customers. Photo: http://tinyurl.com/dlj7o .

Each RT is connected to a host CO, and from the point of view of the
customer, it's indistinguishable from the CO. POTS lines served from
the RT are switched at the CO; the RT simply relays signals back and
forth between the customer and the CO. Numbers are part of the same
NPA-NXX blocks as the host CO.

Each RT is connected to its host CO by one or more digital circuits.
Depending on the number of POTS lines needed, the digital connection
can be as simple as a single T1 implemented over two copper wirepairs,
or it can be some multiplexed combination of several T1s implemented
over coax, fiber, or microwave. See http://tinyurl.com/9oqru .

Whether or not these digital circuits are part of the "local loop" is
a matter of some confusion: I've heard it both ways. For the purpose
of this explanation, I don't include them.

Now slightly restating the original definition, we can state: the
local loop consists of two copper conductors between the customer's
premises and the telco's CO or RT.

For POTS service, this copper pair carries an amazing number of signals:

- Balanced baseband analog voice signals in the range
300 to 3000 Hz., carried in both directions simultaneously.

- Audio control signals carried in the same 300-3000 voice
passband: DTMF signaling tones, dialtone, ring, busy, fault
tones, etc.

- DC loop current resulting from a DC bias voltage ("battery")
applied at the CO or RT. Originally, this current was
necessary to operate the carbon microphones (or "transmitters"
as they were called) of older telephones. Modern telephones
don't use carbon mikes, but they still need DC operating power
for their transistor or IC circuits. Because this voltage is
applied directly across the talk circuit, it must be an
absolutely pure DC voltage (no noise, no ripple). Typical
battery voltages, applied at the CO or RT, are:
Tip = ground
Ring = -48 volts

- On hook/off hook status, implemented by interrupting the
DC loop current:
Loop open = no current = on hook.
Loop closed = current > 20 ma. = off hook.

- Rotary-dial pulses, implemented by interrupting the DC loop
current at specified intervals:
One pulse = "1"
Two pulses = "2" etc.
Ten pulses = "0"

- Caller ID data, carried as analog data in the voice passband.

- Ring voltage to ring the customer's phone. The typical ring
voltage for a single-party line is 90 volts at 20 Hz,
asserted across the ring and tip conductors. In party-line
service, several alternatives have been used:
Different frequencies (up to about 70 Hz).
Different connections (tip-to-ground; ring-to-ground)
Different ring cadences (one long, two short, etc.)
Combinations of above.

All of the above signals are carried at frequencies below 4000 Hz.
Although the voice passband is limited to 300-3000 Hz, the actual
range of the audio channel extends to 4000 Hz.

The 3000-Hz cutoff represents the highest frequency necessary for good
voice communication. That may not be very good by modern hi-fi
standards, but it's fine for voice.

Dialup modems (data, fax, home-security, whatever) all utilize this
same frequency band. There are several modulation schemes floating
around, but they all do basically the same thing: they modulate the
data signals onto one or more analog audio carriers, which are then
carried over the loop in the 300-3000 Hz voice band.

Every audio signal arriving at the CO (or RT) is digitized at a rate
of 8000 samples per second before any further switching or
transmission takes place. This sampling rate is dictated by the
Nyquist Sampling Theorem, which states that the sampling rate must be
at least twice the highest frequency being sampled. See
http://tinyurl.com/474f9

After sampling, each sample is quantized at one of 256 discrete
levels, and the resulting value is encoded as an 8-bit binary number.
The final result is a PCM data stream of 64,000 bits per second. This
data stream is then transported to the customer's ISP over the PSTN.

Note that dialup-modem data signals carried in the 300-3000 Hz voice
passband are not demodulated at the CO or RT; instead, they are
sampled at 8000 sps just like voice or any other audio signal. This
fact imposes an absolute theoretical maximum dial-up data rate of
64Kbps. As other contributors have noted, it's impossible to attain
even that rate in practice due to synchronization errors between the
user's modem and the sampling rate.

Note further that this 4000-Hz limitation is imposed by the CO (or RT)
equipment, not by the wires themselves. It's possible to use
frequencies above 4000 Hz for other signals. And that's exactly what
DSL does. At the CO, a separate piece of equipment, called a "Digital
Subscriber Line Access Multiplexer" (DSLAM) is connected ahead of the
voice processing equipment so that it can provide an independent path
for the DSL signals. Small DSLAMs can be installed in RTs. The DSLAM
acts as a modem at the telco's end of the loop: it communicates with
the customer's DSL modem using RF carriers in two frequency bands:
Uplink (Modem to DSLAM) 30- 110 KHz Downlink (DSLAM to Modem) 110-1100
KHz

The DSLAM demodulates uplinked data carriers to recover the original
data stream. It then sends that data stream to the customer's ISP
over whatever data link the ISP has installed (which might even be
another DSL link). For downlink data, the DSLAM accepts data from the
ISP and modulates it onto a downlink carrier for transmission to the
customer's DSL modem. The maximum speed is limited by the speed of
the two data links, the equipment involved, and the policies of the
telco and the ISP. Images: Large DSLAM for CO installation:
http://tinyurl.com/7atcq Small DSLAM for RT installation:
http://tinyurl.com/7nobj

NOTE THIS DISTINCTION:

- Dialup modem signals are carried to your ISP over the
PSTN as a 64Kbps digital representation of the analog
signal that your dialup modem originally generated.

- DSL modem signals are carried to your ISP as the actual
data stream your DSL modem started with.

Choreboy also asked or commented:

> Are there inline amps [between the CO and the customer]?

There are no inline amps, but there are plenty of other things that
can impair DSL signals (and, for that matter, POTS):

NOISE. Wirepairs inside a multipair cable are not individually
shielded (although the cable as a whole may be shielded). Each
wirepair is twisted so that inductive crosstalk from neighboring
wirepairs is cancelled out, but some residual crosstalk (particularly
from other DSL-carrying loops) may not be completely cancelled.
External signals, such as power-line transients or AM radio station
carriers, may be inductively coupled into the cable. Drop cables at
customer premises are usually not shielded; these cables are also
vulnerable to external noise sources, particularly from nearby
power-line transients.

All of these noise sources collectively impair the ability of the loop
to carry DSL signals.

Noise can be mitigated by careful testing to track down noise sources,
and then by making appropriate repairs. Several manufacturers make
test equipment for this purpose; see http://tinyurl.com/7pnbz for an
example.

SIGNAL ATTENUATION. Like any other electrical circuit, telco
wirepairs comply with a fundamental law of physics: the higher the
frequency, and/or the longer the wire, the greater the attenuation.
This situation results from the interaction between the interconductor
capacitance and the DC resistance of the conductors themselves. Taken
together, these two parameters cause the wirepair to act like an RC
circuit (textbooks frequently represent a wirepair as series of lumped
RC circuits; see http://tinyurl.com/cm5mn for an example).

This problem can be mitigated by careful selection of transmission
voltages and by judicious consideration of the tradeoff between loop
length and transmission speed. Ultimately, however, this situation is
one reason for the limitation on the length of loops that can be used
for DSL.

LOAD COILS. The frequency-dependent attenuation characteristics of
the loop (as described above) also affect voice band frequencies
(300-3000 Hz), resulting in rolloff of the higher frequencies of voice
signals. To solve this problem, telcos have traditionally installed
"load coils" at 6000-foot intervals on long (typically >18K feet)
loops. A load coil is a small inductor installed across the
conductors to cancel the affects of interconductor capacitance.
Although load coils reduce high-frequency rolloff within the voice
band, they cause severe attenuation above 4000 Hz. See
http://tinyurl.com/8njv3 .

This problem can be resolved by removing the load coils and/or by
restricting DSL service to loops without load coils. Of course,
removing the load coils brings back the original problem: rolloff in
the voice band. Furthermore, any attempt to remove load coils assumes
that the telco actually knows where they are (anyone who has ever
worked with telco outside-plant records will recognize the futility of
that assumption). Appropriate test equipment can be used to determine
if load coils are present, and to indicate their approximate
locations.

BRIDGED TAPS. In a typical telco distribution network, big multipair
"feeder" cables leave of the CO or the RT, and head off throughout the
service territory, often along main streets. Smaller (fewer wirepair)
distribution cables split off from the feeders to serve the customers
in a "serving area." As the distribution cables pass through the
serving area, "drop terminals" are installed at intervals. From these
terminals, drop cables feed individual buildings. A single-family
home is usually connected by a two- or three-pair drop; larger
buildings are connected by appropriately larger drop cables.

In areas where outside plant (OSP) is installed on utility poles,
telco drop terminals are called "aerial terminals" or "boots";
typically, a terminal is installed at each pole. Images:

Aerial terminal: http://tinyurl.com/7qzan
Aerial terminal: http://tinyurl.com/74y7y
Pole with terminal: http://tinyurl.com/7qqru
Drawing of interior: http://tinyurl.com/b62ej page 74 of 77

In areas where OSP is buried, drop terminals are installed in
pedestals. In urban areas, telco peds are usually installed in
easements along rear-property lines. In rural areas, peds are usually
installed along roadways, at the edge of the right-of-way. Telco peds
are often placed in "ped clusters" near CATV peds, power peds, and
power transformers.

Images: Telco ped, closed: http://tinyurl.com/apd3y
Telco ped, open: http://tinyurl.com/8pok7 Ped
cluster: http://tinyurl.com/cqjk7

Each drop terminal has:

- Two cable ports for the distribution cable: input and output.
When a drop terminal is installed, these ports are often
sealed as protection against water intrusion. These seals
make it virtually impossible to gain access to the individual
wirepairs within the distribution cable.

- Several drop ports, one for each wirepair in the distribution
cable. These ports are usually implemented with screw
terminals or punchdown blocks.

Every wirepair appears at every drop terminal. When a drop is
installed, the installer connects it to the assigned drop port at the
nearest terminal; electrically, the drop is bridged across the
wirepair. But the portion of the wirepair downstream from the bridge
remains connected, and unterminated at the far end. These
unterminated downstream wirepairs have come to be known as "bridged
taps."

These unterminated wirepairs act like tuned-stub filters. Since
they're unterminated, arriving signals are reflected back; these
reflected signals interfere with the primary signals. In the extreme
case -- when the reflected signal is 180 degrees out-of-phase with the
primary signal -- the primary signal is severely attenuated.

This problem can only be solved by locating and removing bridged taps.
This can be an exceedingly difficult job if the distribution cable is
sealed at that point where it exits the drop terminal.

Test equipment, such as the Fluke 990 http://tinyurl.com/7pnbz , can be
used to determine if bridged taps are present, and if so, their
severity. If the effect of a bridged tap is "minimal" (Fluke's term,
not mine), it can probably be left in place.

> Is DSL modulated into some sort of analog signal?

DSL signals are modulated onto carriers in two bands:
Uplink (Modem to DSLAM) 30- 110 KHz
Downlink (DSLAM to Modem) 110-1100 KHz

> It's hard to imagine carrying hig-frequency digital pulses on
> copper telephone lines.

Well, T1 circuits do just that. But carrying high-frequency pulses on
a POTS loop would present a different problem: overlap with the voice
passband.

> The farm appears to be 35,000 feet from the central office. My
> browser often shows downloads faster than 1.5 Mb/s (150kB/s).

If the farm is indeed 35,000 feet from the CO, then I'd have to
conclude that the loop between the telco and the farm is actually
connected to a DSLAM-equipped RT, not directly to the CO. Look for a
large metal box somewhere along the road between the farm and the CO.
It will have an electric meter; it will probably be set on a concrete
pad, and it might be surrounded by a security fence.

> On dialup, the farm couldn't negotiate modem speeds quite as
> fast as I could in town. I assumed the limitation was in the
> wire. That's why I was amazed to see that DSL seems to use the
> wire in the same way as dialup. Was I wrong to think the reason
> dialup data rates were slower at the farm was that the wire to
> the CO is longer?

I'm not sure that it is longer. See previous answer.

> I don't understand what kind of signal dsl uses to carry so much
> more data than dialup without needing broadband cable.

I hope I've answered that question.

> Ah, crosstalk! It seems to me that if DSL uses the same wire
> dialup used, the same crosstalk will be present.

Crosstalk is indeed present, but it's usually only a problem when two
DSL-carrying loops crosstalk to each other.

> Does a POTS line from the CO to a house carry multiple
> voices? Anyway, DSL at the farm uses the same line that
> the phones at the farm still use.

In current practice, there's usually just one analog voice channel per
loop.

Historically, telcos have used various "pair gain" schemes. In one
scheme, additional voice channels ride on RF carriers superimposed
across the primary voice channel. See http://tinyurl.com/9oqru . In
another scheme, a "phantom" channel is run on two loops, yielding a
total of three voice channels on two loops. As far as I know, these
schemes have been phased out by now, but I suppose there might be a
few still in service somewhere.

Of course, T1 circuits running on copper are still widely used today.
Drive down country roads, and you'll often see T1 repeaters spaced at
(approximately) one-mile intervals. Each T1 can carry 24 voice channels
over two copper pairs. But a T1 wouldn't normally go to a farm.
Photo: http://tinyurl.com/cgp6p .

> I have trouble understanding on the phone, and I often resort to the
> phonetic alphabet to be understood. I think the problem may be more
> in the typical quality of phones than in bandwidth.

I agree; however, the limited bandwidth is also a factor.

In a previous life, when I worked for a radio station, we sometimes
used phone patches for connections to remote locations. At each end,
we'd connect a "phone patch box" directly to the ring-and-tip of a
phone line. Then we'd dial up a connection with a conventional phone,
switch in resistors to keep the line open, and hang up the phones.
Voice quality wasn't as good as it would have been with a wideband
audio circuit, but it was certainly far better than it would have been
if we'd used the telephones themselves. More than adequate for a
sports or news report.

Of course, making a direct electrical connection to a phone line was
illegal back in those days (late 50s, early 60s). But we were on good
terms with the phone guys, so they just looked the other way.

> If the telco owns the DSLAM, won't their investment
> cost depend on capacity?

Yes. But the equipment doesn't have to be installed all at once.
Once the initial investment in the infrastructure (cabinets, racks,
power supplies, etc.) has been made, circuit cards can be added as
needed (equipment manufacturers call this approach "scalable"). It's
the same approach telcos take to POTS.

> If they contract for the DSLAM service, won't they be charged
> according to traffic?

Telcos don't "contract for DSLAM service"; they contract with other
ISPs (e.g. Covad) who wish to offer their own DSL service over telco
loops. The telco charges them for the use of their loops. Telco's
claim they can't charge enough to recover their costs, but that's a
whole different story -- one that will precipitate a thread even
longer than this one.

> Think what would have happened if RG-59 hadn't been invented.
> Everybody would have used RG-6, which looks nearly the same but
> attenuates uhf much less. With better reception there would have
> been more uhf stations and less demand for cable.

As a former cable guy, I don't agree with that. Many UHF stations
depended on cable TV systems to distribute their signals throughout
their "specified zones" (which, back in the '60s and '70s, was a
35-mile radius around the city of license). This was particularly
true in mountainous areas where cable TV systems carried UHF signals
to specified-zone communities that were beyond the reach of their
transmitters.

Neal McLain

Post Followup Article Use your browser's quoting feature to quote article into reply
Go to Next message: Choreboy: "Re: DSL Speed"
Go to Previous message: Telecom dailyLead from USTA: "Supreme Court Backs Cable in FCC v. Brand X"
TELECOM Digest: Home Page