CA 02615059 2007-12-21
64723-492D
PREDICTING PERFORMANCE OF
TELEPHONE LINES FOR DATA SERVICES
This is a divisional of Application Serial
No. 2,371,230, filed April 20, 2000.
Background of the Invention
This application relates generally to
communications networks, and more particularly, to
predicting the performance of telephone lines when
transmitting data.
Public switched telephone networks, i.e., plain
old telephone systems (POTS), were originally designed for
voice communications having a limited frequency range.
Today, the same POTS lines often carry data transmissions.
Since data transmissions generally have different frequency
properties, a POTS line that works well for transmitting
voice may work poorly for transmitting data. Since POTS
lines may not work well for data transmissions, both
telephone operating companies (TELCO's) and customers want
tests for predicting which lines can transmit data.
In the past, telephone operating companies
(TELCO's) performed pre-qualification and pre-
disqualification tests on POTS lines prior to connecting
data transmitters to them. These tests identified some
situations where the line can or cannot support data
transmissions without remedial actions. But, the pre-
qualification and pre-disqualification tests both produced a
significant number of mispredictions, i.e., false positives
and false negatives.
More critically, current pre-qualification tests
for POTS lines are frequently not automated and consequently
- 1 -
CA 02615059 2007-12-21
64723-492D
labor intensive. Often, they demand skilled interpretations
of high frequency parameters of a line to determine its data
transmission capabilities at high speeds. The tests do not
make full use of automated testing systems, e.g., as
described in U.S. Patent 5,699,402. At a network scale,
such tests would be very expensive to implement.
Furthermore, as data transmission demands
increase, simple pre-qualification or pre-disqualification
is no longer sufficient. Now, customers also want
information enabling them to choose between competing
options for transmitting data. Instead of simple
qualification or disqualification, the customer frequently
wants to know which transmission medium and/or devices will
work better. Simple pre-qualification does not provide
customers with a way to compare the different viable options
for transmitting data.
The present invention is directed to overcoming
or, at least, reducing the affects of one or more of the
problems set forth above.
Summary of the Invention
In a first aspect, the invention provides a method
of determining the attenuation of a customer's telephony
line, comprising: connecting a test unit to the customer's
telephony line through a switch connecting a plurality of
customer telephony lines to a telephone network; performing
a plurality of one-ended measurements through the switch of
frequency dependent admittances of the customer's telephony
line, the measurements being performed at a plurality of
frequencies in a lower frequency range; processing the
measurements by a set of logical decision trees derived by
data mining; and adjusting values of a frequency-dependent
- 2 -
CA 02615059 2007-12-21
- 64723-492D
.
attenuation for an average telephony line to predict an
attenuation of the customer's telephony line in a higher
frequency range, the act of adjusting being responsive to
results from the logical decision trees.
In a second aspect, the invention provides a
method of determining performance of a customer telephone
line, the line having both a tip wire and a ring wire,
comprising: driving one of the two wires with a first
alternating voltage at one end and the other of the two
wires with a second voltage at the same end and measuring
voltages between each wire and ground while driving the two
wires; driving the other of the two wires with a third
alternating voltage at the same end and the one of the two
wires with a fourth voltage at the same end and measuring
voltages between each wire and ground while driving the two
wires; driving both the tip and the ring wires with a fifth
alternating voltage from the same end and measuring voltages
at the tip and ring wires while driving both wires; and
determining admittance Ytg at a plurality of frequencies from
the measured voltages.
In a third aspect, the invention provides a method
of detecting a bridged tap in a customer line, comprising:
making one-ended electrical measurements over a range of
frequencies on the customer line; determining one or more
admittances as a function of frequency of the customer line
from the measurements; and detecting that the customer line
has a bridged tap in response to finding a ratio of the
imaginary part to the real part of a derivative of
admittance as a function of frequency exceeds a threshold.
- 3 -
CA 02615059 2007-12-21
64723-492D
Brief Description of the Drawings
Other objects, features, and advantages of the
invention will be apparent from the following description
taken together with the drawings in which:
FIG. 1 illustrates a system to speed qualify
customer telephone lines for data transmission;
FIG. 2 illustrates a test apparatus for performing
one-ended admittance measurements on twisted-pair telephone
lines;
FIG. 3 graphically represents the frequency
dependent attenuation both for an average twisted wire pair
located in a standard telephony cable and for a particular
customer line;
FIGs. 4A-4D are flow charts illustrating a method
of finding the attenuation of a line from the attenuation
for an average line of FIG. 3 and one-ended measurements;
FIG. 5 is a flow chart illustrating a method for
speed qualifying a customer line for data transmission;
FIG. 6 is a flow chart illustrating a method for
predicting the data rate of a line in the method of FIG. 5;
FIG. 7 is a flow chart illustrating a method for
predicting the data rate from line and modem models;
FIG. 8 is a graphical representation of the method
of FIG. 6 for a modem model in which the data rate depends
on the line=s normalized noise level and average normalized
line length;
- 4 -
CA 02615059 2007-12-21
64723-492D
FIG. 9 is a flow chart illustrating a method of
finding a line model from one-ended measurements;
FIG. 10 is a flow chart illustrating the use of
data mining to derive rules relating the line attenuation to
one-ended measurements; and
FIG. 11 is a flow chart illustrating a method of
marketing telephone lines for data transmission.
Description of the Preferred Embodiments
SPEED QUALIFICATION SYSTEM
FIG. 1 illustrates a portion of a POTS telephone
network 10 for speed qualifying customer telephone
lines 12-14, 19, 21. The network 10 includes the customer
lines 12-14 that connect customer units 16-18, i.e., modems
and/or telephones, to a switch 15 located in a TELCO central
office 20. Each line 12-14 is a standard twisted two-wire
copper line adapted for telephone voice communications. The
two wires are generally referred to as the ring AR- and tip
AT- wires. The switch 15 may be a POTS switch or any other
device for connecting the lines 12-14 to a telephone
network, e.g., a digital subscriber loop access'
multiplexer (DSLAM) (not shown). A very large portion of
the length of each customer line 12-14 is housed in a
standard telephone cable 23 that carries a number of the
customer lines 12-14 i.e., more than a
- 5 -
CA 02615059 2007-12-21
64723-492D
dozen. The telephone cable 23 is an environment, which
changes the electrical and transmission properties of the
individual customer lines 12-14. The standard cable
23 also houses customer lines 19, 21, i.e., standard
twisted pair telephony wires, that are not connected
either to the switch 15 or to the customer units 16-18.
These lines 19, 21 have been fabricated into the cable in
anticipation of increased customer demand at future times.
Some of the unconnected lines 19, 21 go to customer
residences already having a connected POTS llne, e.g., the
line 19 goes to the customer connected to the line 14.
The other unconnected lines- 21 are not routed to a
particular customer=s residence. But, all the lines 12-14,
19, 21, i.e., connected or unconnected, have a very large
portion of their length confined to the telephony cable
23, which similarly influences the transmission properties
of each line 12-14, 19, 21 therein.
A measurement unit 22 couples to the switch 15 in
the central office 20 via a test bus 25. The measurement
unit 22 controls one-ended electrical measurements from
the central office 20, which are used to obtain
admittances and noise levels for the lines 12-14 being
measured. To perform a measurement, the measurement unit
22 signals the switch 15 to disconnect a selected line 12-
14 from the telephone network and to connect the selected
line 12-14 to measurement apparatus (not shown) within the
switch 15. Then, the measurement unit 22 signals the
apparatus to perform selected measurements. The
measurement unit 22 signals the switch 15 to reconnect the
line 12-14 to the network after measurements are
completed. The bus 25 returns results from the
measurements to the measurement unit 22. Such
measurements are described in more detail in U.S.
Application Serial No. 60/106,845.
- 6 -
CA 02615059 2007-12-21
64723-492D
The measurement unit 22 is controlled by the
computer 26, which selects the type of measurements to be
performed and the lines 12-14 upon which the measurements
will be performed. The computer 24 sends control signals
to the measurement unit 22 through the line 26 and
receives data the measurement results from the measurement
unit 22 via the same line 26. An executable software
program, encoded on storage medium 28, coordinates the
tests by the measuring unit 22 and the processing of test
data to predict data rates.
The measurement unit 22 and computer 24 speed
qualify and/or disqualify the customer lines 12-14 and
associated modems for selected data transmission speeds.
To speed qualify, the computer 28 must determine, with a
high degree of certainty, that the qualified line and
associated modems will support data transmissions at a
specified data rate without remedial measures. To speed
disqualify, the computer 28 must determine, with a high
degree of certainty, that the disqualified line and
associated modems will not support data transmissions at
the specified data rate without remedial measures.
Various embodiments make speed qualification
determinations either before the line is in service or
while the line is in service. Before a line is
transmitting data, the determinations are speed pre-
qualifications or pre-disqualifications. After a line is
transmitting data, determinations are referred to as speed
path testing.
- 7 -
CA 02615059 2007-12-21
_ 64723-492D
ONE-ENDED MEASUREMENTS ONCUSTOMER LINE
FIG. 2 illustrates an apparatus 27 for performing
one type of one-ended electrical measurement used for
speed qualifying and/or speed disqualifying of the lines
12-14 of FIG. 1. The apparatus 27 measures the
admittances of the tip and ring wires T, R of the selected
customer line under measurement. The tip and ring wires
T, R of the line 12-14 being measured couple to driving
voltage sources V1 and V2, respectively, through known
conductances Gt and Gr. The tip T and ring R wires also
connect to voltmeters Vt and Vr for reading the voltage
between the tip wire T and ground and between the ring
wire R and ground, respectively.- The readings from the
voltmeters Vt and Vr enable the computer 24 to determine
effective admittances Ytg, Ytr, Yrg between the tip wire T,
ring wire R, and ground for the customer line 12-14 being
measured.
To determine the admittances Ytg, Ytr, Yrg, the
switch 15 connects the voltage sources V1 and V2 and the
voltmeters VT and VR to the tip and ring wires T, R as
shown in FIG. 2. After connecting the apparatus 27, the
measurements needed to determine the admittances Ytg, Ytr,
Yrg entail three steps. First, the measurement unit 22
grounds the point 29 and applies voltage V2 while measuring
the voltages across the voltmeters Vr and Vt. Next, the
measurement unit 22 grounds the point 30 and applies
voltage Vi while measuring the volfages across the
voltmeters Vr and Vt. Finally, the unit 22 applies both
voltages V1 and V2 and measures voltages across the
voltmeters Vr and Vt. From these three measurements, the
computer 24 determines the admittances Ytg, Ytr, Yrg at
various frequencies.
- 8 -
CA 02615059 2007-12-21
- 64723-492D
During measurements for the admittances Yt., Ytr,
Yrg, the apparatus 27 may apply complex driving voltages Vl
and V2 that superimpose several frequencies. For example,
the driving voltages Vl, V2 may take the form: V(N) = AEi=1_
45 cos (211_f;,NT "FPMj) . The frequencies fi, sampling cycle
values N (at 152.6 Hz), and phases PMi are shown in Appendix
1. The computer 24 Fourier transforms both the driving
and measured voltages Vl, V2, Vt, V= to separate frequency
components. From the Fourier transform, the computer 24
finds the real and imaginary parts of the admittances Ytg,
Yt=, Yrg by well-known circuit-analysis techniques.
From the admittances Yty, Ytr, Yrg, several derived
properties of the lines 12-14 may be determined. First, a
line length can be derived from the capacitances Ctg and Crq
of the tip wire T to ground and of the ring wire R to
ground. For standard bundled telephony cables with
twisted tip and ring wire T, R pairs, both capacitances
are about 17.47x10-9 Farads per 1,000 feet regardless of
the gauge. Thus, the one-ended measurement of capacitances
gives a measure of the apparent length of the measured
line 12-14. Second, the existence of a bridged tap in one
of the lines 12-14 can be derived from the existence of an
above-threshold peak in the ratio:
IM (MzYtg(f) RE (M2Yts(f) )
Mf 2 ~ Mf2
The presence of a bridged tap substantially effects the
capacative measurement of the length of the line. Third,
the admittances Ytg, Yt=, Yrg can also be used to predict
the gauge mix of the measured lines 12-14. The gauge mix
of a line is the ratio of the sum of lengths of the line,
which are fat wire, over the full length of the line.
Typically, fat wire is 22 and 24 gauge wire, and thin wire
-
_ 9
CA 02615059 2007-12-21
64723-492D
is 26 gauge wire. The customer lines 12-14, 19, 21 of
FIG. 1 may have segments of fat wire and segments of thin
wire. Fourth, a frequency dependent attenuation up to
high frequencies can be derived.
A two step procedure is used to derive the high
frequency attenuation of the measured lines 12-14. First,
the attenuation of the lines is approximated by the
frequency (f) dependent average attenuation, AT(f). AT(f)
is the attenuation of an Aaverage- mixed gauge twisted
copper line in a standard telephony cable. The average
attenuation AT(f) is known to approximately be:
AT(xMHz) = A(xMHz)Ctg with
(A(.1MHz) ,A(.3MHz) ,A(.4MHz),A(.5MHz) )
=(.173,.24,.263,.288)DB/10-9F
(A(.1MHz),A(.3MHz),A(.4MHz),A(.5MHz) )
=(.173,.24,.2 63,.288)DB/10-9F
A solid curve 32, shown in FIG. 3, graphically illustrates
the equation for AT(f) as a function of frequency.
Second, for each customer line, the frequency dependent
values of the AT(f) are adjusted using a method found
through data mining. The second step produces the
attenuation, ATT(f), for each customer line. ATT(f) is
generally an improved value of the line=s attenuation
compared to the AT(f) for an average line.
A solid curve 32, shown in FIG. 3, graphically illustrates
the equation for AT(f) as a function of frequency.
Second, for each customer line, the frequency dependent
values of the AT(f) are adjusted using a method found
through data mining. The second step produces the
attenuation, ATT(f), for each customer line. ATT(f) is
generally an improved value of the line=s attenuation
compared to the AT(f) for an average line.
- 10 -
CA 02615059 2007-12-21
64723-492D
= Data mining produces a set of logical decision
trees, which are used to find ATT(f). For each customer
line, the computer 24 of FIG. 1 works through the set of
logical decision trees. Each decision tree determines
whether or not ATT(f), at one frequency, is shifted from
the value of AT(f) at that frequency. At frequencies
between those associated with logical decision trees, the
computer 24 finds the value of ATT(f) by performing a
smooth interpolation. The dashed line 34 of Fig. 3 shows
the ATT(f) of one customer line, which was found by the
logical decision tree analysis (M = 106, K = 103, and DB =
decibels). Data mining produces a set of logical decision
trees, which are used to find ATT(f). For each customer
line, the computer 24 of FIG. 1 works through the set of
logical decision trees. Each decision tree determines
whether or not ATT(f), at one frequency, is shifted from
the value of AT(f) at that frequency. At frequencies
between those associated with logical decision trees, the
computer 24 finds the value of ATT(f) by performing a
smooth interpolation. The dashed line 34 of Fig. 3 shows
the ATT(f) of one customer line, which was found by the
logical decision tree analysis (M = 106, K = 103, and DB =
decibels).
FIGs. 4A, 4B, 4C, and 4D are flow charts showing
the decision trees for finding the values of ATT (.1MHz) ,
ATT(.3MHz), ATT(.4MHz), and ATT(.5MHz), respectively.
FIG. 3 shows the ATT(.1MHz), ATT(.3MHz), ATT(.4MH), and
ATT(.5MHz) (triangles) of one customer line, which were
found from the AT(.1MHz), AT(.3MHz), AT(.4MH), and
AT(.5MHz) values (dots). Each decision tree uses logical
tests based on lower frequency derived quantities, which
are listed in Appendix 2. In Appendix 2, admittances are
given in siemens, capacitances are given in Farads, and
frequencies are given in Hertz unless otherwise indicated.
- 11 -
CA 02615059 2007-12-21
64723-492D
The result from each decision tree provides a value
of ATT(f) at a higher frequency than the frequency used to
measure the admittances Yt., Ytr, and Yrg. Thus, the
logical decision trees enable the computer 24 to improve
ATT(f) for each customer line, at frequencies higher than
the frequencies at which measurements are performed on the
line.
From a line=s attenuation ATT(f), the computer 24
can derive a normalized line length (NLL). NLL ( f) is the
equivalent length of 26 gauge twisted copper telephony
line to produce the attenuation ATT(f). The value of
NLL(f) is approximately: '
NLL(f) = ATT(f)/{E'j=o Pj (1og(f))j} where the Pj are:NLL(f)
= ATT ( f)/( E7 j_o Pj ( log ( f)) j} where the Pj are:
(P0 , . . ., P7) = 103(-1.81718846839515, 2.3122218679438,
-1.25999060284948, .38115981179243, -.06912909837418,
.00751651855434, -.00045366936261, .00001172506721)
Averaging NLL(f) over frequencies between 100KHz and 1Mhz
provides a averaged normalized line length. The averaged
normalized line length and a normalized noise define
properties of a line model for the measured customer line
12-14, which allow the prediction of data transmission
rates.
The one-ended measurements on the selected customer
line 12-14 also include noise power spectra and impulse
noise. Noise power spectra are determined directly
through one-ended measurements using a -spectrum analyzer
(not shown) located in the measurement unit 22. Impulse
noise measurements employ a differential comparator (not
shown) also located in the switch 15. The comparator has
an adjustable threshold and produces a digital output
pulse for each above-threshold spike on the tip or ring T,
R wires. The output digital signal goes to a counter (not
- 12 -
CA 02615059 2007-12-21
64723-492D
shown), which sums the number of counts to produce a rate
for above-threshold noise impulses.
Noise measurements may both disqualify and correct
predicted data rates of the lines 12-14 being qualified.
For high noise levels, synchronization of the line 12-14
for ADSL or ISDN data transmissions becomes impossible,
and the noisy line 12-14 must be disqualified. For
example, impulse noise rates above about five 150
millivolt-counts-per-second disqualify a line for ADSL
transmissions. When noise is not a disqualifier, it still
can lower the predicted data rates for the customer line
in a manner that generally depends on the modem used with
the selected line 12-14.
Referring again to FIG. 1, the customer lines 19,
21 do not connect to the switch 15 and thus, cannot be
automatically tested by the measurement unit 22. Thus,
speed qualification or disqualification of these lines 19,
21 requires indirect measurements henceforth referred to
Aproxy measurements;.
Proxy measurements are one-ended electrical
measurements on a Aproxy= line located in the same cable
23 as the unconnected line 19, 21 to be qualified or
disqualified. The proxy line connects to the switch 15
and thus, can be tested by one-ended electrical
measurements made from the switch 15. For example, the
line 14 is a potential proxy line for the line 19 going to
the same customer.
The proxy line 14 is located in the same cable 23
as the unconnected lines 19, 21 to be qualified. Thus,
both types of lines have undergone the same handling after
fabrication of the cable 23. Similarly, if the cable 23
has more than 12 different customer lines, e.g., a
standard telephony cable, the various lines 12-14, 19, 21
are in very similar cable environments. Then, electrical
- 13 -
CA 02615059 2007-12-21
64723-492D
measurements on the proxy line 14 can, in some cases,
provide a reliable measure of the same electrical
properties for the unconnected lines 19, 21. The
reliability of proxy measurements may further increase if
the proxy line goes to the same residence as the
unconnected line, e.g., lines 14 and 19. But, proxy
measurements may still be reliable if the proxy line is
simply in the same cable 23, e.g., the line 13 as a proxy
for the line 19.
LINE PERFORMANCE PREDICTIONS
FIG. 5 is a flow chart illustrating a method 40 of
speed qualifying or disqualifying a selected one of the
customer lines 12-14 of FIG. 1 for data transmissions.
The method has two parts. In a first part, the computer
24 and measurement unit 22 of FIG. 1 rapidly determine
whether the selected line 12-14 is pre-disqualified for
data transmissions. In the second part, the computer 24
predicts the speed for data transmissions if the selected
line 12-14 is not disqualified in the first part.
To determine whether the selected customer line 12-
14 is disqualified for transmitting data, the computer 24
or an operator selects the type of data service to be
implemented on the selected customer line 12-14 (step 42).
Next, the computer 24 determines the qualification
requirements for the selected type of data service on the
selected line 12-14 (step 44). Next, the computer 24 and
measurement unit 22 perform one-ended electrical
measurements on the selected customer line (step 46).
Then, the computer 24 determines from the one-ended
measurements whether the selected customer line 12-14 is
disqualified for the selected type of data transmissions
(step 48). If the selected customer line 12-14 is
- 14 -
CA 02615059 2007-12-21
64723-492D
disqualified, the computer reports the disqualification
status and stops.
The pre-disqualification part of the method 40 is
generally more rapid than predicting the actual data rates
obtainable. U.S. Patent lVo. 6,385,297 provides detailed
account of some types of measurements and determinations
performed in pre-disqualification steps 42, 44, 46, 48.
These steps may also include further tests specific to the
type of termination at the customer units 16-18. For
example, for ADSL-lite data transmissions the fact that a
customer unit 16-18 attenuates high frequencies could be
used as a disqualifier test.
If the selected customer line 12-14 is not pre-
disqualified at step 48, the computer 24 will predict the
data rate of the selected line 12-14 for data transmissions.
First, the computer 24 creates a line model for the selected
customer line 12-14, e.g., by performing more one-ended
measurements on the line 12-14 and deriving the line model
therefrom (step 52). At substantially the same time, the
computer 24 identifies a modem model to be used with the
selected customer line 12-14 (step 54). The modem model may
correspond to the modem in the central office 20 and/or the
modem at the customer=s residence. Next, the computer 24
uses the line model for the selected customer line 12-14 in
the modem model to predict the line=s performance, e.g., the
data rate. Some modem models are a data file stored in the
computer 24 and indexed by properties of the line model.
Finally, the computer 24 reports the line performance when
used with the identified modem (step 58).
- 15 -
CA 02615059 2007-12-21
= 64723-492D
FIG. 6 is a flow chart illustrating a method 60
for predicting the performance of the selected customer
line 12-14, which was not pre-disqualified for data
- 15a -
CA 02615059 2007-12-21
64723-492D
transmissions at step 48 of FIG. 5. The computer 24 and
measurement unit 24 control one-endec3 electrical
measurements carried out by apparatus 27 on the twisted
pair T and R of the selected customer line 12-14 (step
62). The measurements determine the three admittances Yt9,
Ytr, Yzg of the tip and ring wires T, R and the noise levels
in the selected customer line 12-14. Next, the computer
derives a number of other properties of the selected
customer line 12-14 from the one-ended measurements (step
64). As discussed above, the derived properties may
include a line length, the existence or absence of one or
more bridqed taps, the gauge mix of the line, impulse
noise level, frequency dependent attenuation, normalized
line length, and the noise spectrum.
From these derived properties, the computer 24
calculates a second-level derived' propertyXthe average
normalized line length. The average ndrmalized line
length is the length of 26 gauge paired twisted copper
wires, located in a telephony cable 23 with at least 12
other twisted wire pairs, which would have substantially
the same transmission properties.
The computer 24 also selects a modem, e.g., in
response to a customer=s request or a TELCO=s command to
speed qualify or disqualify the line for a particular
modem type (step 66). Next, the computer 24 looks up a
modem model for the selected modem in a database (step
68). The modem model is a table of -performance data,
i.e., data transmission rates, indexed by the averaged
normalized line length and the line noise level. The
computer 24 may leave the modem model in active memory
while waiting for data on the line model associated with
the selected customer line 12-14. Next, the' computer uses
the line model data in the modem model to find a predicted
data rate for the selected modem in association with the
- 16 -
CA 02615059 2007-12-21
64723-492D
selected customer line 12-14 (step 70). Finally, the
computer 24 reports the predicted data rate to the
customer or to a readable storage device (step 72)
FIG. 7 is a flow chart illustrating one method for
predicting the data rate of the selected customer line 12-
14 as shown in step 70 of FIG. 6. The line model is
either a set of rules or a file for the properties
characterizing the model. From the line model, the
computer 24 reads the average normalized line length (step
82). Similarly, the line model or one-ended measurements
determine a normalized noise level associated with the
selected customer line 12-14 (step 84). Finally, the
computer 24 performs a look up of a predicted data rate in
a table defining the modem model (step 86). The modem
model=s table is indexed by the averaged normalized line
length and the normalized noise level. The table is a
tabular form representing the modem model for the modem to
be used with the selected customer line 12-14.
FIG. 8 graphically illustrates one modem model 90
as a set of curves 92-95 for the predicted data rate. The
values from the curves 92-95 depend on, i.e., are indexed
by, a line=s normalized noise level and averaged normalized
line length. The separate curves 92-95 give the predicted
data rate for four values of the normalized noise level of
the line model. Each curve 92-95 is also dependent on the
averaged normalized line length, which is plotted along
the horizontal axis.
The predicted data rate can be obtained from the
modem model 90 of FIG. 8 by performing a look up with the
parameters of the line model. To predict the data rate,
the computer 24 looks up one of the curves 92-95 using the
normalized noise value from the line model, e.g.,
normalized noise value 2. Next the computer 24 finds the
predicted value of the data rate by looking up the
- 17 -
CA 02615059 2007-12-21
64723-492D
averaged normalized line length, given by the line model,
on the horizontal axis, e.g., value 97. The value 101 of
curve 93 at the intersection 99 with the value 97 of the
averaged normalized line length is the predicted data
rate. Of course, the computer does the look ups in a data
base indexed by the normalized noise level and the average
normalized line length instead of graphically.
Some modem models also depend on parameters such as
impulse noise compensation, noise floor, echo compensation
and phase instability compensation. The impulse noise
compensation is the ability of the modem to resychronize
or to remain synchronized in'the presence of impulse noise
on the customer line. The noise floor is the noise level
below which the modem does not resolve data signals. The
echo compensation is the ability of the modem to
compensate for reflected signals in the customer line.
The phase instability compensation is the ability of the
modem to compensate for time-dependent imbalances in the
customer line, e.g., time-dependent reflections.
Using the values olff each of these parameters, the
computer 24 of FIG. 1 adjusts the predicted data rate from
the rate predicted by FIG. 8. The modem models attach a
figure-of-merit or quality rating to each of the above
parameters. For each parameter, the quality rating may,
for example, be excellent, good, or bad. The quality
ratings determine whether the predicted data rate, e.g.,
the rate from FIG. 8, is adjusted up, down or not adjusted
by the computer 24 to obtain a final predicted data rate.
For example, some embodiments adjust the predicted data
rate' from FIG. 8 up by 10 percent and down by 10 percent
for quality ratings of excellent and bad, respectively.
Similarly, some line models include a gauge mix
parameter, which is given a quality rating, i.e., high,
average, or low. Data mining techniques can be used to
- 18 -
CA 02615059 2007-12-21
64723-492D
infer a test for the gauge mix of a line from the one-
ended electrical measurements. The computer 24 of FIG. 1
adjusts the predicted data rate from the rate predicted by
FIG. 8 according to the quality rating of the gauge mix.
FIG. 9 is a flow chart illustrating a method 110 of
finding a line model for any selected customer line 12-14,
19, 21, i.e. either connected or unconnected to the switch
of FIG. 1. First, the computer 24 determines whether
the selected line is connected to the switch 15 (step
10 112). If the selected line is connected, the computer 24
chooses the selected line itself for one-ended electrical
measurements (step 114). If the selected line is
unconnected, e.g., the lines 19, 21 of FIG. 1, the
computer 24 chooses a proxy line in the same cable 23 for
15 the one-ended electrical measurements (step 116). Next,
the computer 24 and measurement unit 22 perform the one-
ended measurements of the chosen line=s admittances Ytg,
Ytrr Yr9 and noise levels as described above (step 118).
Next, the computer 245 determines the above-described
derived properties for the chosen line from the measured
admittances and noise levels as described above (step
120). The derived properties include the frequency
dependent attenuation, the absence or existence of a
bridged tap, the mix, the frequency-dependent normalized
line length, and the averaged normalized line length. From
the derived properties, the computer 24 determines the
averaged normalized line length using the formula
described below (step 122). Similarly, from the measured
noise levels of the chosen line, the computer 24
determines the chosen line=s normalized noise level. The
computer 24 stores the one-ended measurements, the derived
electrical properties (step 120), normalized noise level
(step 124), and averaged normalized line length (step 122)
as the line model for the originally selected line 12-14,
- 19 -
CA 02615059 2007-12-21
64723-492D
19, 21 (step 126). These stored quantities form a
footprint that characterizes the customer line.
The footprint is stored data on the condition of
the line when operating well. Later, the computer 24 can
call up the footprint to perform speed path testing. When
called up, the footprint is useful for fault detection as
described in U.S. Patent 5,699,402, which is herein
incorporated by reference in its entirety.
The derived properties characterizing the selected
customer line 12-14 and modem models used by the methods
of FIGs. 4A-4D are found through methods referred to as
Adata mining-. Data mining produces derived properties
that are well correlated with the data produced by the
models, e.g., high frequency attenuation.
FIG. 10 illustrates a method 130 for using data
mining to find derived properties correlating well with
the high frequency attenuation. Data mining starts by
selecting a sample line having a known attenuation from a
sample pool (step 132). Next, one-ended measurements are
performed on the selected sample line and a selected set
of derived properties, e.g., low frequency admittances,
are found from the measurements (step 134). Next, the
values of the selected derived properties are stored in a
file indexed by the attenuation of the sample line (step
136). Next, the data mining system determines whether
other sample lines remain (step 138). If sample lines
remain, the system repeats steps 132, 1-34, 136, and 138.
Otherwise, the system compares the values of the derived
properties for the sample lines to determine which
properties or sets of properties correlate well with the
attenuation (step 140). Finally, the system uses the
values of the derived properties correlating well to
formulate a set of.rules, which determine the attenuation
in terms of the well-correlating derived properties (step
- 20 -
CA 02615059 2007-12-21
64723-492D
142). The Arules- are represented by the methods of FIGs.
4A-4D.
FIG. 11 is a flow chart illustrating a method of
marketing customer lines for data transmission. First,
the computer 24 of FIG. 1 speed pre-qualifies a plurality
of the lines 12-14, 19, 21 using one-ended electrical
measurements and speed qualification methods described
above (step 152). The speed pre-qualification, at least,
classifies each line for either high-speed service or low
speed service. Next, the TELCO offers high-speed service
to a portion of the customers who have lines qualified for
the high-speed service (step 154). Next, the TELCO
selectively connects at least a portion of the lines
qualified for high-speed service to customers requesting
the high-speed service (step 156). The TELCO also sets
billing rates for, at least, a portion of the lines at
prices that depend on the speed qualification (step 158).
Other embodiments are within the scope of the
following claims.
- 21 -
CA 02615059 2007-12-21
64723-492D
=
150, 600, 1050. 1500. 1950, 2400, 2850, 3300. 3750. 4200, 4650, 5100. 5550,
6000. 6450, 6900.
7350, 7800, 8750. 8700, 9150, 9600, 10050, 10500. 10950, 11400, 11850. 12300,
12750, 13200,
13650, 14100, 14550. 15000. 15450. 15900. 16350. 16800. 17250. 17700. 18150,
18600, 19050.
19500, 19950.
N; 1, 4, 7. 10, 13, 16. 19, 22. 25. 23, 31, 34, 37, 40, 43, 46, 49, 2, 55, 58,
61, 64,
67, 70. 73, 76, 79. 82. 85, 88, 91. 94, 97, 100, 103. 106, 109, 112. 115, 118.
12.1, 124, 127, 130.
133 respectiveiy.
+1 5.9738, 1.3564. 2.4683, 4.8575. 4.7424. 2.2972, 4.6015. 1.9156. 2.5660.
4.5986,
4.6452. 3.4542, 3.6341. 0.8848, 4.3410. 2.1606, 4.2342. 4.2147, 3.1058.
5.9049, 5.2782, 5.1159.
5.4354, 5.6124, 0.5751, 3.8940, 3.3812. 6.0230, 2.3239. 2.7284, 4.8032.
4.1488, 2-3427, 4.6362,
0.9163, 2.9335, 1.0363, 2.3272. 3.2040. 4.0025. 2.0028, 5.8444, 2.4525.
1.4760. 1.1770
- 22 -
CA 02615059 2007-12-21
64723-492D
3 08 : Raw Massu=amaats :
Y=r(30) - Admictance <<p-Vo-ring measured at 30Hz
Ytg(30) - Admittance ti?-co-ground measured at 30Hz
Yrg(30) - Adtni.ttance r_r.g-to-grour.d measured at 30Hz
3 08z Dnrived Moasurfmaats :
30Gtr - Con.ductance tip-to-ring measured at 30$z = real(Ytr(30))
3OStr - Suscepcance tip-to-risg measured at 30Hz = imag(Y-.s(30))
30Gtg - Conduczance tip-to-ground measured at 30Hz = real(Ytg(30')
30Stg - Sasceptance t:p-co-ground measured ac 30Hz = imag(Yt(30).
30Ctr - Capacitance tip-to-ring measured at 30Hz = Sts(30)/(2=pi=30)
30Ctg - Capacitance tip-to-ground measured at 30Hz = St(30)/(26p_-=30)
Lm.eas - LenQth i.-i kft measured at 30Hz = 30CtgJ17.47
1508x-201C8z Raw Meosuramrats:
Admittance cip-to-r_ :g a'cnere ?=y 5CHz. 5O0Hz. 1C50Hz. -5QOH::. ... 19950Hz
Ytg('f) - Adms.ttance tip-to-grour_d .rhere
=:50Hz.600Hz.1C50Hz.15Q0:iz....19950H2
Yrgtf) - Admittance r:::g-co-ground +rhere f=150Hz.600Hz.105aHz.1530Hz....
19950Hz
I50Hz-201C8z Derived Measnrnmonns:
150Gtr - Conductance tip-co-rir.g measured at 150Hz = real(Ytr(15]))
600Gtr - Conductance tip-to-ring measured ac 600Hz = real(Ytr(600))
19950Gtr - CondLctance c_p-co-rin3 measured at I9950Hz = real(Y=r(1995a))
150Str - Susceptance t_p-to-ring measured at 150Hz = i:aq(Y:z(150))
600Str - Susceptance t'_p-cc-ring measured at 600Hz = imaq(Ytr(600))
:9950Str - Susceptance t?p-co-ri.~.g r..easurcd at 1 950Hz = imag(Ytg(19950))
150Gtg - Conductance :'_a-cc-grour.d measured at 15CHz = real (Y;g (150 ) )
60CGtg - Ccnductance :_p-co- ground measured ac 60CHz reai(Ytg(600i)
i99S0Gtg - Conduczance t_p-cc- ground measured at 19950Hz = reai(Ytg(19950))
150Stg - Suscepcance t'_p-co- ground measured at 154Hz = imaq(Ytc(150))
600Stg - Suscepcance :_p-co- ground :-easured at 600Hz = 7.mag(Ytc(600)
19950Stg - Susceocance c-p-to- ground measured at 19950Hz = imac(Ytg(19950))
?50Ctr - Capacitar.ce cip-co-ring measured at 150Hz = 150Str/
600Ctr - Capacizance cim-to-ring measured at 6(IQHz = 600S tr/(2'pi'600)
19950Ct_ - Capacitance c-p-to-ring measured at 19950Iiz =
19950St.r/(2=pi=19950)
150Ctg - Capaci=ance cip-co-grouna measured at 150Hz =?505tg/(2*pi=150)
600Ctg - Capacitance cip-to-ground measued at 600Hz = 6QOStg/(2*p'_=600)
19950Ctg - Capacir.ance =:p-co-ground measured at 19950Hz =
1995()Stg/(2*pi'19950)
-23-
CA 02615059 2007-12-21
64723-492D
1508z-Z0A8z S.coadasy Dsriv.d M.asurem.ats:
C30/C4K - Ratio of t_p-to-grourd Capacitance at 30Hz to 4200Hz
C4K/C10K - Ratio of t:p-to-ground Capacitance at 4200Hz to I0050112
Cslope - Tip-co-ground Capacitance ratio slope =(C4K/C10K)/(C30/C4Y.)
C30-C4X - Difference iZ tip-to-ground Capacitance at 30Hz and 4200Hz
C4K-ClOK - Difference ?z tip-co-ground Capacitance at 4200Hz and 10050Hz
Cdelta - Tip-co-grour.ci Capacitance di.::erence delta =(C4X-ClOK)/(C30-C4K)
G4X/G30 - Ratio of tip-to-ground Conductance at 4200Hz to 30Iiz
G20K/G4K - Ratio of t'_p-ta-ground Coaductance at 10050Hz to 4200:iz
Gslope - "'ip-co-ground Conductance ratio slope = (GIOK/G4K)/(G4R/G30)
G4X-G30 -?ifference in tip-to-grouad Conductance at 30Eiz and 4270Hz
G1OI{-G4K -Difference in tip-co-ground Conductance at 4200Hz and 10050Hz
Gdalta - Tip-to-ground Conductance dif;erence delta =(G10K-G4K)/(G4K-(;30)
C30/G30 - Ratio of Tip-to-ground Capacitance to Conductance at 30Hz
C30/G4K - Ratio of Tip-co-ground Capacitanee at 30Hz to Conductance at 4200Hz
C4K/G4K -Ratioz of Tip-co-ground C3pacitance to Conciuctance at 4200'siz
Gtr_dmax -'4ax:..:.um posicive slope of Gtrtf) = max(derivat--;ve(Gtr;f)/df))
Gtr_fs.ax - Frequency at which Gt _d.:ax occurs
Gtr_diaia - Maximum negative slope of Gtr(f) = min(derivative(Gtr(f)/df))
Gtz_fmin - Frequency at which Gtr_dmin occurs
Gtr_fpk - Frequency of first peak (local maxima)iz Gtr(f)
Gtr_fval - Frequency of first valley(local miniiaa)in Gtr(f)
Gtr_d_delta - Gtr Max/Min Derivative difference = Gtr_dmax-Gtr_dair_
Gtr_pk_delta - Gts peak/valley f_equency difference = Gtr_fval-Ct__fpk
Gtr -2k - Value of Gtr(f) at f_equency Gtr_fpk
Gtr_val - Value of Ctr(f) ac fre4uercy Gtr_fval
Gtr delta - Gtr peak/val:ey difference = Gtr-pk-Gtr val
Gtg dmax - Maxitaum positive sl.ope of Gtg(f) = max(derivative(Gtq(-.-)/df))
Gtg_=aiax - Frequency at which Gtg_dmax occurs
Gtg_dmin - Maximum negative slope of Gtg(f) = min(derivative(Gtc(:)/df))
Gtg_fmin - Frequency at which Gtg_d:m:.n occurs
Gtg_d_delta - Gtg Max/Mir_ Qerivative diL:erence = Gtg_dmax-Gtg cimin
Ctr_dmax - Maximum posit:ve slope a= Ctr(f) = taax(derivacive(C=(f) /df) )
Ctr_fmax - Frequency ac which Ctr_dmax occurs
Ctr_dmi_z - Maximum negative slope of Ctr(f) = tain(derivative(Ctr(f)/df))
Ctr_f=in - Frequency at which Ctr_dmin occurs
Ctr_fpk - Frequency of f:rst peak (local :naxima)in Ctr(f)
Ctr_fval - Frequency of first valley ( local mi ni õLa )in Ctx( f)
Ctr_d_delta - Ctr Max/Min Derivative difference = Ctr_dmax-Ct=_dxaia
Ctr_pk_delta - Ctr peak/valley frequency difference = Ctr_fval--:tr_fpk
Ctr val - Value of Cts(f) ac frequer_cy Ctr_fval
Ctg_dmax - Maximua positive slope o= Ctq(f) = max(derivac_ve(Ctg(f)/df))
Ctg_fmax - Frequency at which Ctc; daax occurs
Ctg dtain - Maxi:avm negative slope or Ctg(f) = min(der:vative(Ctg(f)/df))
Ctg fain - Frequency at which Ctg_.:.=in occurs
Ctg_d_delta - Ctg Max/Min Derivac'_ve difference = ctg dmax-Ctg_drnin
Str_dmax - Maximum positive slope o= Str(f) = max(der:vative(Str(f)/df))
Str_imax - Frequency at whica Str_d.:.ax occurs
gtr_~ - MaY1mu+++ negative slope o: Str(F) = min(derivative(Str(f)/df))
Str_fmin - F:equency at which Str_::=i.n occurs
-24-
CA 02615059 2007-12-21
64723-492D
1508z-ZC2C8z s.aoadssY aarivrd Measurrr=Mnts:
Str_tpk - Frequesicy of firsc peak (local maxima)in St:r(f)
Str_fval - Frequency of fi.ssc valley(local m7T1ma)in Str(f)
Scz_d_delta - Str Max/Min Derivative differencz = Str dmax-Str_das.n
Str_pk_delta - Str peak/valley f_equency difference = Str_fval-Str_fpk
Stx_pk - Value of Str(f) ac 'requency Str_fpk
Str val - Value of Str(f) ac :requency Str_fval
Str_delca - Str peak/valley d:._ : erence = St,.,pk-Str_val
Stg_cmax - M.aYi~+um pcsitive s;.ope af Stg(f) = max(derivative(Stgt f) /df) )
Stg_,f=ax - Frequency at which Stq dmax occurs
Stg dmin - Maxiututn negative slope of Stg(f) = min(derivative(Stgcf)/df))
Stg fnin - Frequency at which Stg_d-i.n occurs
Stg_fpk - Frequency of f'-rst peak (local maYima)in Stg(f)
Stg_fval - Frequency of first valley(local mini.ma)in Stg(f)
Stg,_d_delta - Stg Max/iKin Derivazive difference = Stg dmax-Stg driizi
Stg_pk_delta - Stg peak/valley =requency d:f;erence = Stg,fval-St:g 'pk
Gtg2Ok/GtgBk - Ratio of Gtg at 19950Hz and 8250Hz
Gtq20k/Gtg4k - Ratio of Gtg at 19950Hz and 42008z
Cg=30/Cgt20k - Ratio of Ctg at 30Hz and 19950FIz
Cgt30/CgtBk - Ratio of Ctg at 30Hz and 8250Hz
-25-
