Field of the Invention This invention relates to a method of estimating the length of a telephone line, in particular a method of estimating a compensated line length that takes into account the effects of additional equipment on the line.

Background to the Invention

In a public switched telephone network (PSTN), a telephone line made of a pair of wires (typically copper) runs from a telephone exchange to a customer premises. The pair of wires can be used to provide telephony as well as data services to the customer premises.

Telephone lines are prone to faults, thus adversely affecting the telephony or data services that may be running over them. Some of these faults are easily identified and rectified, others less so. Examples of faults include disconnections and short circuit between the wire pairs. Various methods have been developed to help identify the faults and propose solutions.

One approach is to routinely test a sample of lines every night. This is often referred to as overnight routine testing (ONRT), and captures a variety of electrical parameters for each line tested. The measured values can be interpreted by mapping onto those associated with known fault conditions. Regular testing can also help identify gradually degrading lines. The electrical parameters measured include voltages, resistances as well as capacitances. It is important that these measurements are accurate, as they can significantly affect fault diagnoses as well as other tests. Measured capacitances can be used for estimating the length of a telephone line as well as helping to diagnose faults. An estimate of the line length has many uses, including use for predicting the bit rate of data services that can be supported on the line, as well as localising faults (distance to fault). Therefore, accurate estimation of line length is very important. US patent 5,937,033 describes a simplified drop test system particularly suited for communicating over fibre optic cable that can selectively deliver either the conventional pass-fail results or detailed drop test data useful to telephone craftsmen. A testing system is described which is capable of testing various line characteristics.

US patent application US2002/0146094 describes a telephone test system that protects a data line from unintended interference from a test set, a test being the device used by engineers in the field to verify proper line operation and troubleshoot problems. US patent 7,573,824 describes a fault management system for an access network of a communications network, which performs a series of tests on each of the lines which extend from a local switch through a series of modes to a user terminal equipment. The results of the tests are analysed with respect to a set of parameters to identify characteristics indicating that a fault is likely to occur within a predetermined period.

Summary of the Invention

According to one aspect of the present invention, there is provided a method of estimating a compensated line length of a telephone line, said method comprising:

identifying an initial line length and capacitance associated with each of a set of telephone lines connected to an exchange;

generating a linear model using the initial line lengths and capacitances, wherein the linear model models the line length as a function of capacitance for the plurality of telephone lines;

determining the gradient and offset associated with the modified linear model; measuring the capacitance of a test telephone line connected to the exchange; and

applying the determined gradient and offset to the measured capacitance using the linear model to estimate a compensated line length of the test telephone line.

The method may further comprise modifying the generated linear model by removing data associated with telephone lines that do not fit the generated linear model, and repeating the generating step. As such, an iterative approach may be adopted, which improves the accuracy of the generated linear model. The set of telephone lines may operate according to a common service type, and the test telephone line may operate with the same service type as the set of telephone lines. Examples of a service type include ADSL2 and ADSL2+. Each telephone line comprises a pair or wires, and the capacitance of a telephone line may be the maximum of capacitance of each of the wires of that telephone line with respect to earth.

According to a second aspect of the invention, there is provided a control module for estimating a compensated line length of a telephone line, said control module comprising processor adapted to:

identify an initial line length and capacitance associated with each of a set of telephone lines connected to an exchange;

generate a linear model using the initial line lengths and capacitances, wherein the linear model models the line length as a function of capacitance for the plurality of telephone lines;

determine the gradient and offset associated with the modified linear model; measure the capacitance of a test telephone line connected to the exchange; and

apply the determined gradient and offset to the measured capacitance using the linear model to estimate a compensated line length of the test telephone line.

Brief Description of the Drawings For a better understanding of the present invention reference will now be made by way of example only to the accompanying drawings, in which:

Figure 1 is a system diagram showing a telephone exchange and a telephone connecting to a customer premises;

Figure 2 is a flow chart illustrating the steps of an example of the invention; Figure 3 is a scatter plot of the capacitance and initial length estimate of a set of lines in an example of the invention;

Figure 4 is a quantile-quantile plot used to illustrate which data points to prune in an example if the invention;

Figure 5 is a table showing the determined gradient and intercept over a number of iterations of linear modelling and pruning of outlying data points; Figure 6 is a further scatter plot of the capacitance and initial line length estimate of a set of lines after several iterations of modelling and pruning of outlying data points in an example of the invention. Description of Preferred Embodiments

The present invention is described herein with reference to particular examples. The invention is not, however, limited to such examples. Proposed is a method of estimating a compensated length of a telephone line that takes into account the effects of additional equipment, typically at the exchange, on the capacitance measures used to determine the line length. A linear model is used to represent the relationship between capacitance and line length for a set of lines offering the same service, such as ADSL2+. The gradient and offset determined from the model can be used to adjust capacitance measures from a test tine, to give an estimate of a compensated line length associated with that test line. The compensated line length is extremely useful, as it is more likely to mirror the actual line length and can serve as the basis for other calculations such as estimating of potential sync rates for DSL lines, and estimating of distance to faults.

Figure 1 illustrates a telecommunications network 100 comprising a customer's premises 102 and a telephone exchange 108 (sometimes referred to as the "central office"). Within the customer premises 102 is network terminating equipment NTE 104, which typically is a device such as a telephone, home gateway/router or such like. The exchange 108 houses a main distribution frame MDF 1 10. The MDF 1 10 in the exchange and the NTE in the customer premises are connected to each other by a telephone line 106. The telephone line 106 is typically formed of a pair of copper wires.

Within the exchange 108, the MDF 1 10 is connected to a relay 1 12, which serves to selectively connect a given telephone line to one of PSTN equipment 1 14 or test head equipment 1 18. The PSTN equipment 1 14 is connected to the PSTN network 1 16. Under normal operation, the relay 1 12 operates to connect the telephone line to the PSTN equipment 1 14, thus providing the NTE 104 with PSTN services. The telephone line 106 can also be connected up to other equipment in the exchange, but have been omitted for simplicity, such as a digital subscriber line access multiplexor (DSLAM), which can provide data services, such as xDSL data services. Other telephone lines from other customer premises are also connected to the exchange 108, though for simplicity they have not been shown in Figure 1 . Thus, the exchange 108 will provide PSTN and data services to a large number of customer premises, each with their own NTE, and each having their own telephone line.

In the telephone line 106, the two wires are usually labelled as the A-leg and B-leg.

The telephone line 106 can experience faults, with various causes. For example, the line may become interrupted by a disconnection at a connection point along the line (not shown in Figure 1 ). A roadside cabinet and junction box on a telegraph pole are examples of connection points. Water ingress at a connection point can also cause problems, resulting in short circuits between the A-leg and B-leg wires, or earth leakage problems. Measuring parameters associated with the telephone line can help with diagnosis of faults.

The test head equipment 1 18 comprises a group of electrical testing units that apply a sequence of known AC and DC voltages to the line and measure the resultant currents and voltages. The test head 1 18 can be brought in circuit with the telephone line 106 by the switching of the relay equipment 1 12. The test head equipment 1 16 and relay 1 12 used to switch to it can be controlled by the control module 120. A data store 122 is also provided, which can store data generated by the test head equipment 1 18 or control module 120. A number of parameters associated with the telephone line are measured by the test head equipment 1 18. The parameters are based around various physical properties associated with the line: resistances, DC voltages, and capacitance. The capacitance of a line can be used to determine an associated line length by applying a conversion factor, typically around 60nF/km. Thus, if the capacitance for a line is measured as 180nF, then the line length is estimated at 3km. However, if the line length estimated is to be accurate using this method, the capacitance needs to be measured accurately. This is particularly important, as the estimated line length is used for other purposes, such as estimating the bit rate that a data service (such as ADSL2+) running on the line can be expected to achieve, or using the estimated line length as the basis for estimating a distance to a fault on the line. The inventors have noted that one cause of inaccurate measurements is that the test head equipment 1 18 is not just measuring the line 106 itself. There are additional components along the path as well as the physical wiring that makes up the line 106 that need to be properly accounted for. These components include internal exchange wiring, the relay equipment 1 12, the MDF 1 10, and even the test head equipment 1 18. Examples of the present invention propose a method of estimated a compensated line length that takes into account the effect of these additional components.

Figure 2 illustrates a flow chart summarising the steps of an example of the invention.

In step 200, line parameters associated with a population of lines are measured by the test head equipment 1 18. This is done under control of the control module 120, with the results stored in the data store 122. The measurements can be done as part of overnight routine testing (ONRT). These measurements include the capacitances for each line, and specifically the capacitance of each of the metallic pairs (A-leg and B-leg) of each line measured with respect to earth, also known as A-to-earth and B-to-earth. The measurements can be repeated, or multiple measurements taken over a period of time, say every few days. In step 202, the control module 120 selects a subset of the population of lines measured. The resulting set of lines comprises all the lines from the population that provide a given service, such as ADSL2+ for example. This is important, as the additional components that we are trying to account for may be limited to equipment unique to a certain service. Then for each of these lines in the set, an associated capacitance measure is determined, which in this example is done by taking the maximum of the A-to-earth and B-to-earth measurements for each line. If the measurements were repeated in step 200, then the average of a number maximums of A-to-earth and B-to-earth can be used instead as the capacitance measure. Each line in the set also has an initial line length estimate. The initial line length estimate can be obtained from cable records that are populated with cable length data when line was originally provisioned. However, these initial line length estimates may not be accurate. Thus, a data set associated with the set of lines is generated, which comprises the capacitance measure and corresponding initial line length estimate for each of the lines. Figure 3 shows a scatter plot of the data set, with each data point representing a line, with the capacitance measure plotted on the y-axis, and against the initial line length estimate on the x-axis. The relationship between line length and capacitance can be modelled with the following linear model: y = ax + £> (1 ) where y is the capacitance (in nF), and x is the initial line length (in km), a is the gradient or coefficient (used interchangeably in this description) representing the capacitance per distance (usually expressed in nF per km or nF/km). The gradient a will depend in part on the materials used for the wires of the lines, b is the intercept or offset (used interchangeably in this description) representing the error in the capacitance measure, which can be attributed to the additional components as described earlier.

In step 204, the control module 120 performs linear regression on the data set, resulting in an initial estimate of the linear relationship (1 ), including values for the gradient a and the offset b. The initial estimate is shown as the line plotted in the graph of Figure 3.

In step 206, the control module 120 identifies and removes the outliers from the data set. As can be seen from Figure 3, there are some lines that have a line length/capacitance that does not fit particularly well with the linear relationship from step 204. These lines, or outliers, may have some existing condition or fault. Some faults will cause a change in the relationship between capacitance and line length. For example a high capacitance reading for the recorded length could be due to contact with a non-working pair. Lower than expected capacitance values might be due to a high resistance connection. Outliers tend to exert a disproportionate amount of influence on the model, and further are not well predicted by the model, and are thus removed for modelling purposes.

There are several ways to identify outliers. In regression, an outlier is an observation or data point that an exceptional value of the Y-variable (the dependent variable, or the capacitance measure in this example) for the value of its X-variable (the independent variable, the initial length estimate here). Outliers occur frequently in real data and can make a large difference in the results of regression analysis. There are several plots that can be used to visualize outliers, one of these is the quantile-quantile plot (Q-Q plot) as shown in Figure 4. It is a probability plot of the standardised residuals against the values that would be expected under the normality assumption. If the normality assumption has been met, then the points of the graph should be close to the line y = x. Figure 4 shows that a number of points the numbered points deviate significantly from this line, they are the outliers. The data point (numbered) identifiers of the most extreme outliers are also shown on the plot.

It is also possible use a test to see if there are outliers present. One of these is Bonferroni Outlier Test which reports the Bonferroni p-values for Studentized residuals in linear and generalized linear models. It checks to see if the Bonferroni p-value is less than 0.05. That is, the probability is so low that the data point is considered to be an outlier. In step 208, linear regression is repeated for the remaining lines after removal of the outliers from the data set.

In step 210, a check is made by the control module 120 to determine if the revised data set falls within thresholds of the linear regression model, or if some other condition is met. For example an R ² value can be determined from the linear relationship and data points. In regression analysis, R ² is a measure of the goodness-of-fit of a linear model i.e. how close the data are to the regression line. If R ² is 0.86, then 86% of the total variance in the dependent variable (capacitance) is explained by the independent variables (length estimate). Thus, a threshold can be set for R ² , for example 0.9 or even 0.95.

If the revised data set does not meet the thresholds or constraints, then processing returns to step 206, and further outliers are removed before performing linear regression again, and so on until the thresholds are met. This iterative approach improves the accuracy of the linear model by successively removing outliers that may exhibit the most anomalous capacitance measure and fit the linear model least well.

Figure 5 illustrates a table showing the gradient and offset determined using the linear model in an example of the invention for an initial analysis and 16 further iterations. The table also identifies potential stopping points for the iterations. For example, the method can be stopped after the 3 ^rd iteration, where the determined offset is less than 0.5 compared to the previous step. Alternatively, the method can be stopped after the 9 ^th iteration, where R ² is greater than 0.95, which represents an excellent correlation.

Whilst an iterative approach has been described, the method can alternatively simply model the linear relationship once, and not iterate after removing outliers.

Thus, if the thresholds are met at step 210, then in step 212 the final linear model and associated gradient and offset are stored in the data store 122. Note, the process can be repeated for other telephone lines that provide other services, such as ADSL2 instead of ADSL2+. The result will be a linear model and associated gradient and offset for each service type.

Figure 6 shows a final scatter plot after 16 iterations and removal of outliers. In this example, the final linear model arrived at is: y = 62x + 94 (2)

or

capacitance = 62 x length + 94 (3)

Thus, the gradient or coefficient a representing the capacitance per distance is 62nF/km, and the intercept or offset a representing the error in the capacitance measure is 94nF.

Equation (2) can be applied to a capacitance measure for a test line to estimate a compensated line length.

In step 214, the test head equipment 1 14 measures parameters associated with a test line. The measurements can be done on demand, or taken from the overnight routine testing for example. The test line is of the same service type as the et of lines from step 202 (ADSL2+ in this example). The control unit determines a capacitance measure for the test line, as the maximum of the A-to-earth and B-to-earth capacitance measurements.

In step 216, the control unit 120 uses the final linear model, with the associated gradient and offset, to estimate a compensated line length of the test line. Rearranging equation 1 , we can see that the line length, or compensated line length, x can be represented by: x = (y - b)/a (4)

Thus, if the capacitance measure for the test line is 220nF, the gradient a is 62, and the offset b is 94, then the compensated line length is 1 .71 km ( (200-94)/62 ).

The steps of generating the linear model for a set of lines from steps 200 to 212 can be repeated over time to ensure that any changes in exchange equipment and wiring are reflected in an up to date linear model.

Exemplary embodiments of the invention are realised, at least in part, by executable computer program code which may be embodied in an application program data. When such computer program code is loaded into the memory of a CPU in the control module 120 for execution, it provides a computer program code structure which is capable of performing at least part of the methods in accordance with the above described exemplary embodiments of the invention.

A person skilled in the art will appreciate that the computer program structure referred can correspond to the flow charts shown in Figure 2, where each step of the flow chart can correspond to at least one line of computer program code and that such, in combination with the CPU in the control module 120, provides apparatus for effecting the described process.

In general, it is noted herein that while the above describes examples of the invention, there are several variations and modifications which may be made to the described examples without departing from the scope of the present invention as defined in the appended claims. One skilled in the art will recognise modifications to the described examples.