WO2000057571A1

WO2000057571A1 - Coaxial cable fault locator

Info

Publication number: WO2000057571A1
Application number: PCT/AU2000/000235
Authority: WO
Inventors: Rodney Karl Eastment
Original assignee: Cable & Wireless Optus Ltd.
Priority date: 1999-03-24
Filing date: 2000-03-23
Publication date: 2000-09-28
Also published as: AUPP941099A0; TW474071B

Abstract

A method and apparatus are disclosed for locating faults in a bi-directional cable network, where separate frequency bands are used for upstream and downstream propagation of signals. The method is particularly suitable for locating common path distortion or passive intermodulation faults. Signals are propagated in the downstream band with defined modulations and at selected frequencies, so that a network fault will return a signal which includes a selected mixing product of the downstream signals. The downstream signals are selected so that the mixing product falls within the upstream band, so that the returned signal can propagate back to the source and be correlated with the downstream signals. An apparatus is also disclosed.

Description

COAXIAL CABLE FAULT LOCATOR

Technical Field

The present invention relates to the field of detecting hardware faults in a bi-direction network. It has particular but not exclusive application to the location of common path distortion (CPD) faults in a bi-direction cable network. Background Art

Bi-directional coaxial cable networks and hybrid fibre coaxial cable networks (HFC) provide services such as cable television and bi-directional telephony and data services. Such networks provide bi-directionality on the basis of frequency multiplexing, with one band, for example, 5 - 65 MHz being used for upstream signals - that is to the exchange - and a further band, for example 85 - 750 MHz being used for downstream and broadcast signals. In such networks, filters are used to prevent the upstream band from propagating downstream, and vice versa. Thus, a signal having a frequency within the downstream band cannot propagate to the end of the network and return, as filter diplexers within the amplifiers would prevent this.

One form of interference, which is of particular concern to network operators, arises due to unintentional non-linear elements occurring in the plant equipment. By plant equipment is meant the cables, gateways, amplifiers, customer taps, and other physical elements of the network. The most common type of non-linear element results from corrosion occurring within connectors and joints, producing metal - metal oxide - metal junctions where signal rectification and intermodulation products arise. These undesired intermodulation products substantially raise the system noise and degrade the performance of the plant in providing services to customers. The noise may affect not only the branch of the network in which it occurs, but also neighbouring branches by decreasing the signal to noise ratio, for example, during upstream transmissions. This type of noise degradation is known as common path distortion (CPD) or passive intermodulation (PIM). The faults that give rise to CPD are often difficult to locate because of network branching, the amount of cable in a particular node (which may be 12 to 15 kilometres) and the multiplicity of joints (typically greater than a thousand) in a given node. At present, there is no instrument or process commercially available to accurately locate such faults. The mere presence of CPD of some sort in the branch of a network can be detected using a spectrum analyser - however, this does not directly assist in locating the fault. Generally, current best practice is to utilise high quality components, carefully controlled installation procedures and perform regular routine maintenance on all branches of the network. If a particular branch has a significant problem, then the state of the art solution is to dispatch technicians to check, tighten, etc each connection and joint in the branch of the network or node. This is clearly a time consuming and inefficient process. The absence of a better fault locating system also has the potential to significantly affect the quality of service which can be promised and delivered via such bi-directional networks.

It is accordingly an object of the present invention to provide an instrument which is able to assist in the location of faults which produce CPD in a network.

Summary of Invention

According to one aspect, the present invention provides a method for detecting propagation time delay associated with one or more faults in a bidirectional network, including the steps of: a) introducing signals F1 and F2 into the downstream band of the network, F1 and F2 each having a defined modulation and specific frequency band, so that said signals propagate in the downstream band of said network; b) monitoring the upstream band to detect signal F3, where F3 is a selected mixing product of F1 and F2; c) correlate the detected F3 signal with F1 and F2 so as to determine the time delay associated with one or more CPD faults.

Preferably, the network has been previously calibrated so as to determine the time delay associated with each node at a number of reference points relative to the test point. In this case, a fourth step allows the range determination of the node in question, or with a suitably sophisticated mapping system, the location of the fault. In a preferred arrangement, where the upstream band occupies lower frequencies than the downstream band, frequency F3 is the difference between frequencies F1 and F2. Whilst the present invention may be employed from an exchange or cable head end, it may equally be deployed in a field unit so as to detect a fault in the field. It will be appreciated that it is not uncommon for there to be multiple faults in a node, and the present invention is capable of detecting separately located sources of CPD.

The principle of the present invention relies on the CPD acting as a mixing element of the transmitted signals. This produces a family of intermodulation products at each fault, some of which will fall within the upstream band. In a preferred arrangement the second order difference product F1 - F2 is detected within the upstream band at frequency F3. It is preferred that the frequencies selected are within reserved bands so as to not interfere with other services. Brief Description of the Figures

Figure 1 shows a schematic representation of an exemplary bidirectional information network. Figure 2 is a simplified block diagram of a fault locating device according to a preferred form of the invention. Figure 3 shows the spectral layout of signals in the network of figure 1.

Figure 4 is a block diagram of the up converter as used in the device of figure 2. Figure 5 is a block diagram of the down converter as used in the device of figure 2. Figure 6 is a frequency versus time plot of a frequency ramp chirp signal generated by the device of figure 2. Figure 7 is a frequency versus time plot of the chirp signal of figure 6 to show the effect of temperature on the accuracy of the range measurement derived. Detailed Description of the Preferred Embodiment

The bidirectional network 100 of figure 1 can provide cable television and bidirectional telephony and data services between an exchange 101 and many end users 102. The network may consist of several sub-networks to suit its particular purpose. In the network of figure 1 , network 100 consists of an optical fibre network (region A), and a coaxial cable network (region B). This is known as a Hybrid Fibre Coaxial (HFC) network. Of course this need not be the case in practice with this invention, which may be used with any bidirectional network having any combination of sub-networks as would be understood by the person skilled in the art.

Exchange 101 includes a transmitter 101 a and a receiver 101 b. Optical signals are sent down fibre 103 through gateway 104 where the signals are converted to radio frequency signals for transmission over the coaxial cable network (region B).

Disposed over a stretch of cable 105 are multiple tap junctions 106 which "tap off" the signal from the main cable 105 to connect to an end user 102. Also disposed over a stretch of cable are multiple line amplifiers 107 which continually amplify the signal being transmitted to counter its degradation due to energy losses occurring during its propagation along cable 105.

Since this is a bidirectional network, end-users can also send information to exchange 101 through the network 100. The return signal propagates through the same cable 105, and is separated in gateway 104 to return fibre 108 to be received at receiver 101 b of exchange 101.

To allow bidirectional transmission of signals on the same cable 105, the forward (downstream) signals are spectrally separated from the return (upstream) signals. This spectral separation is represented in figure 3, where the region denoted 300 is the bandwidth allocated to the downstream signals, and the region denoted 305 is the bandwidth reserved for upstream signals. The region 301 denotes the bandwidth typically taken up by the video, telephone and data signals distributed over the network. Region 301 may also be interspersed between f-| and f₂.

Filter diplexes (not shown) within the line amplifiers 107 are also distributed throughout network 100 to prevent signals having downstream frequencies from propagating upstream and vice versa, enabling bi-directional amplification. As discussed above, faults can occur in any portion of network 100 where there are many connectors or joints. If corrosion occurs within a connector or joint, a non-linear element is produced, which can result in signal rectification and intermodulation products. Intermodulation products substantially raise the system noise and degrades the performance of services. Due to the literally thousands of junctions 106, in a typical network 100, a fault arising in one junction, for example 106a can be extremely difficult to locate.

Using the method and device of the present invention, the location of the fault is greatly facilitated. Common Path Distortion (CPD) fault locator 200 is attached to network

100 as shown in figure 1. Locator 200 generates two test signals at frequencies f-i and f respectively, both having frequencies within the bandwidth allocated for downstream signals. These two test signals are represented in figure 3 as 302 and 303 respectively. Note that frequencies fi and f₂ are spectrally separated from the bandwidth 301 normally used by the normal traffic to avoid interference.

Test signals f| and f₂ propagate down network 100 through fibre 103, are converted to radio frequency signals in gateway 104, and are distributed throughout cable network 105. Upon reaching tap junction 106a, which exhibits non-linear characteristics due to a fault, signals and f₂ produce intermodulation products of various orders including a second order difference product f₃. Signals fi and f₂ are selected so that second order difference product f₃ is centred about a frequency falling within the bandwidth allocated for upstream signals. Second order difference product f₃ (return signal) is denoted in figure 3 as 304. It can be seen that it lies within upstream bandwidth 305. Return signal f₃ propagates upstream towards exchange 101 where it is detected by fault locator 200. Locator 200 essentially measures the time difference in the modulation of return signal f₃ with respect to the modulation of original test signals fi and f₂. A distance between exchange 101 and a fault at junction 106a is determined by this time difference, thus facilitating the location of the fault, or the source of the CPD.

An actual position and distance is achieved by comparing the measured time with a database containing calibrated times together with their corresponding node positions. This database is built up by using a Range Calibrator, which is used to measure and calibrate the return delay to specific points in the coaxial cable portion of the HFC network 100. The Calibrator is a field mobile test box which is connected to customer tap points, preferably close to the downstream outputs of line amplifiers. A log book is maintained for each node of the return calibration delays to known points. These serve as markers to produce a "range map" of the network, and will enable CPD faults to be reasonably accurately located to within limited portions of the coaxial cable when they are detected. At a desired point in the network to be "range-mapped", the range calibrator mixes the 2 ranging signals (at 681.5 MHz and 741 MHz) and injects the 2-carrier mixed product at 59.5 MHz into the upstream spectrum. This return product is band pass filtered to limit its spectrum to 57 - 62 MHz. The Calibrator includes a line phase timing control which enables the return 59.5 MHz upstream signal to be pulsed on (for 2.5 ms) and adjusted to lie in a suitable position within the 20 ms line period. By doing this, it enables the return signal to be moved away in "phase time" from a real CPD fault enabling the range to each to be independently measured should they be at the same spatial location. Due to the network's branching topology, a ranging measurement will not uniquely locate the fault. The faulty branch or leg will have to be located by signal tracing. However, the ranging resolution of the order of a few metres assists in CPD location, since faults mostly occur at tap junctions 106 or line amplifiers 107 and not in mid-span. This accuracy enables candidate branches 105 to be rejected where faults would be placed in mid-span sections.

It will be appreciated that it is only preferred in this case that the return signal f₃ be a difference product of f-i and f₂. In networks where the upstream (return) bandwidth is at a higher frequency band than the downstream band, return signal f₃ may be the sum product of f-i and f₂, or any other suitable product.

The function of locator 200 is described in more detail below with reference to figure 2. Fault locator 200 uses a batch sample signal processing software. The program interfaces through an internal computer bus 202 to an arbitrary waveform generator (AWG) 201 and a high speed analog to digital sampling card (A to D) 208. AWG 201 becomes the signal generator and A to D 208 the receiver. Custom designed and built frequency conversion equipment 213 (described in more detail later with reference to figures 4 and 5) interfaces between the AWG / A to D and the network 100 to shift signals into the correct spectrums.

Measurement signals are generated by software and loaded into AWG 201 which is set up to generate signals in the band 3.75 to 6.25 MHz.

The fault location, procedure involves two steps. There is firstly an initialisation step, followed by the actual measurement step.

The initialisation step is used to determine the timing of triggering signals to A to D 208 such that it will only take samples when the CPD pulse is present. It has been found that the CPD is amplitude modulated at the supply line rate, pulsed on, lasting around 2 milliseconds during each cycle or half mains cycle.

To perform this initialisation step, a continuous wave (CW) signal is output at 5.35015 MHz. It is sent as a series of 184 bursts lasting for 301.4 ms.

It will be appreciated that this initialisation step is only necessary in the case where the CPD is in fact modulated by an external signal. In this case, 90V AC, 50Hz power is supplied through the coaxial cable 105, which modulates any CPD signals. In other cases, a different initialisation step may be required to suit the particular type of modulation of the CPD pulses. In other cases still, there may be no external modulation effect on the CPD pulse, and no initialisation step will be required at all.

When performing the actual measurement step, a frequency ramp chirp signal which linearly ramps from 3.75 to 6.25 MHz is generated. This signal is synthesised in 32768 points clocked at 20 MHz. One complete cycle, called a burst, lasts for 1.6384 milliseconds. Multiple bursts are utilised. The AWG 201 signal is sent to up converter 500 which produces two output signals fi and f₂ separated by nominally 59.5 MHz. One signal is spectrally erect, while the other is spectrally inverted. These signals are injected (218) into network 100 and propagate through the network until they reach fault 106a, producing a family of intermodulation products. Amongst them is a second order difference product f-i - f₂ (element 304 in figure 3). This second order product (f₃) will carry double frequency modulation of the f| , f₂ signals since one is spectrally inverted. Since signal f₃ has a frequency in the bandwidth (305) allocated to up stream signals, it propagates through network 100 to return to exchange 101 and enters (219) frequency conversion equipment 213.

Down converter 400 converts a 57 to 62 MHz band (containing f₃ at 59.5 MHz) to 7.5 to 12.5 MHz and sends this to A to D 208. Low pass filter (6.5 MHz) 203 suppresses alias clock products from AWG 201. Hybrid transformers 204, 206 and 207 provide for signal splitting or summing.

The sampling of this signal by the A to D is controlled by software and employs a 2-pulse interleaved sampling technique (published in IREE Vol. 18, June 1998) using 25 nanosecond delay line 212. The two interleaved samples are down loaded to PC 209, where the signal processor software converts the sampled 10 MHz signal into a complex baseband spectrum.

The A to D sampling is initiated with the external trigger signal from the timing box 216. An internal delayed trigger equal to one burst (1.6384 ms) is used to allow for the network propagation delay. When sampling chirps using a 5 MHz sample clock, this results in the tail end of burst 1 being followed by the first part of burst 2. Since all bursts are identical, this equates to a time-wrapped alias sample signal. If the delayed trigger were not used, the leading portion of the sample would have no signal. When the signal generated by AWG 201 is CW, A to D 208 samples the

10.7 MHz CW at 25 k sample pairs/sec and the signal processing converts it to a complex baseband centred at DC with 25 kHz frequency span lasting 327 milliseconds. Software frequency shifts the sample down 3 kHz and then filters it (2 kHz gaussian). The 3 kHz offset avoids a residual clock alias product from the sampling process. The software then overlays 15 20 ms samples from the 327 ms A to D sample to calculate a composite 20 ms magnitude average. The CW signal in the period between 302 ms and 327 ms is muted and software measures residual noise during this period. Software further determines the CPD Phase Time peak within the 20 ms averaged sample. This timing is used to set the external trigger timing for the AWG/ A to D for the chirp measurement signal when the CPD pulse is maximum. When the signal generated by AWG 201 is the frequency chirp, A to D

208 samples the 7.5 to 12.5 MHz down converter 214 output and software computes the 5 MHz wide complex baseband spectrum centred on DC. The software then "de-chirps" the signal by multiplying it by its complex conjugate 5 MHz frequency chirp. In figure 6, there is shown a representation of the linear- frequency ramp chirp signal, whose function is given by: W(t) = at + b

The chirp signal may be represented by Re^jW(t)t, where R indicates the real part.

The chirp is then: R_e ^{j(atΛ2+b )} Consider a chirp transmitted and received with delay τ:

Received chirp =

To "de-chirp" the returned signal, it is multiplied by its complex conjugate; ie:

Demodulated chirp = e^jW(t)(t"τ). e^"jW(t)t

= e^"jW(t)τ _e-j(aτt+bτ)

The first component in the exponent is a frequency shift proportional to the delay τ. The second component is a fixed phase roll dependent on the delay τ. The "de-chirped" signal is then Fast Fourier Transformed (FFT) to reveal

CPD spectral peaks which correspond to the range delay. Where there are multiple return signals with different delays, i.e. τ_, τ₂, τ₃, ... , these can be separately identified by the FFT producing frequency peaks at f1. f2. f3- ...

Software then locates spectral peaks derived from the FFT which are above a set threshold value, and performs narrow bandwidth (2 kHz) gaussian filtering, followed by phase roll demodulation. The exact filter bandwidth is set to be commensurate with the sampling, i.e. an integer multiple of the resolution bandwidth. The phase roll is measured using a least squares method. This provides accurate sub-FFT element frequency resolution which permits range determinations to better than 1 metre, dependent on the CPD return C/No (carrier to Noise density per hertz).

A to D 208 is synchronously triggered with AWG 201 via an external trigger signal from timing box 216. This synchronous triggering ensures that chirp samples occur when the CPD signal is maximum. It also enables coherent vector summation of multiple chirp signals, improving the signal to noise ratio (S/N) of CPD returns. Signal to noise improvement is directly related to the number of coherent detections. Ten coherent detections improves S/N by 10 dB and 100 detections improves S/N by 20 dB.

In cases where the sampler introduces timing jitter and sampling does not commence instantaneously on the external trigger strobe, a frequency doubled reference 205 (figure 2) derived from the AWG signal is added to the down converter 214 output signal. This reference is equivalent to a zero delay CPD return. The software filters, phase demodulates and measures its phase roll. This is used to reference the absolute timing for the whole sample and removes external trigger timing jitter introduced by the sampler. This reference is only applied during the chirp range measurement.

The signal processing software used by locator 200 generates the ranging test signals through AWG 201 and processes the return samples to compute the return delay. Once the test signal is generated and loaded into AWG 201 , no further signal processing is required for signal transmission.

The return signals are sampled by a digital sampler and passed on to the processor 209 via bus 21 1 where they are batch processed. When processing chirp ranging signals, successive samples are complex voltage summed, producing a coherent summation with improved signal to noise which improves ranging accuracy.

The software also provides graphing and printing, filing, and retrieval of measurement samples.

A distance calculator is also provided to assist the operator to convert the difference between return calibrations and CPD ranges into a distance with allowances for delay through line amplifiers. This facility greatly assists the operator in choosing the most likely candidate branch from several options due to network branching.

The details of frequency conversion equipment 213 are now described in more detail with reference to figures 4 and 5. Up-converter 400 of frequency conversion equipment 213 is shown in figure 4. It is designed to convert a 5 MHz signal to an erect and an inverted spectrum pair of signals, at the desired frequencies. In this case the respective frequencies will be at 681.5 and 741 MHz, which lie at the upper spectrum end of the HFC downstream spectrum 300.

Presented at input 404 from signal divider 204 (figure 2), is a 5 MHz centred signal. Local oscillator 401 is at 24.75 MHz, which mixes the input signal up to 29.75 MHz. Bandpass filter 402 filters out the lower component and passes the 29.75 MHz signal onto local oscillator 403 which is at 711.25 MHz. This mixes with the 29.75 MHz signal to produce an upper and a lower sideband pair of signals, the lower sideband at 681.5 MHz and the upper sideband at 741 MHz. Any frequency errors in the second local oscillator 403 are inconsequential since the difference frequency between the upper and lower sideband signals remains constant at 59.5 MHz. As discussed above, this difference frequency will result in the CPD signal having a frequency of 59.5 MHz, allowing it to propagate through the network as a return signal.

The upper sideband U (or f₂ as previously referred to) may be represented as:

The lower sideband L (or fi as previously referred to) as:

|_ _{= R e}J[(W_|02-W,₀₁-Win)t] where R indicates the real part. fi and f₂ then propagate down the cable, taking time τ dn to reach a CPD fault. The upper and lower sidebands, U and L respectively, can be represented at the CPD fault as:

U = R _eK(W_l02+W_l01 +Win)(t-τdn)]

|_ _{= R e}J[(W_l02-W_|01-Win)(t-τdn)]

The intermodulation difference product D (or f₃ as previously referred to) is then: n - R _θJ[(2W_l01t+2Wint-2W_l01.τdn-2Win.τdn)]

_{= R e}i[(2W,₀₁₊2Win)(t-τdn)] _{&χ χhe CpD føu|t}_

Down converter 500 receives this return signal D (or f₃) after the upstream delay τup at 59.5 MHz. The 2-carrier mixing action of the CPD fault causes the frequency chirp to have twice the original AWG bandwidth input to the up converter 400. Down converter 500 mixes signal f₃ with its local oscillator 502, at 49.5 MHz. Oscillator 502 is derived by frequency doubling up converter 400's first local oscillator 401 and is coherently locked to it. Thus, any frequency errors in local oscillator 401 will be effectively removed in the return signal f₃ as it is frequency doubled by the same amount. The proviso is that the local oscillator frequency remains phase coherent to within a few degrees over a period of up to 500 microseconds which is the order of the maximum 2-way return delay to the CPD fault. This places tolerance limits on the phase noise spectrum of local oscillator 401/local oscillator 502 for the single sideband spectrum to be no greater than -80dBc/Hz at 1 kHz offset and -90dBc/Hz at 10 kHz offset. Down converter 500 has two selectable IF filter bandwidths, one for each measurement signal. For the CW signal, the filter is 25 kHz bandwidth centred at 10.7 MHz. For the chirp signal, the filter is 5 MHz bandwidth centred at 10.0 MHz i.e. 7.5 to 12.5 MHz. Thus, at down converter 500 input 504, signal D (or f₃) is received.

Signal D has now travelled from the CPD fault to the locator 200, taking time τup. thus signal D at input 504 is:

R _eJ[(2W_l01+2Win)(t-τdn-τup)]

The local oscillator after network propagation delay will have phase roll.

Down converter 500 subtracts the local oscillator, hence the negative frequency component is used.

The signal out of the down converter is then: _{R e}J[(2W_|01+2Win)(t-τdn-τup)] „. _{R Θ}J[(-2W_l01)(t-τdn-τup)]

_{R e}J[(2Win)(t-τdn-τup)]

Hence the down converter output is precisely frequency doubled the input signal to the up converter with the network propagation delay.

The effect of temperature on the accuracy of the range measurements is now discussed with reference to figure 7.

Theoretical work on signal to noise analysis indicates the thermal noise limit or ranging accuracy to be in the range of 1 to 2 nanoseconds, being an order of magnitude finer than required. It is believed that thermal expansion factors in the optical fibre and coaxial cable will change path length and cause greater uncertainties in the order of several metres for 10-30° C temperature variations. From the CSM digital filter appendix, the frequency measurement uncertainty due to CNo was shown to be:

1

Δf =

2π L CNo.Tspan^"

if we substitute for Tspan:

N

Tspan , then

Fspan

As can be seen from figure 6, showing a frequency sweep of Fchirp lasting Tspan, the slope of the sweep from figure 7 is:

Fchirp ΔF Tspan ΔT

thus

ΔF.Tspan

ΔT = Fchirp

where ΔF is the frequency measurement uncertainty due to noise, and ΔT is the time measurement uncertainty.

For typical values, Fspan = 5 MHz

CNo = 10⁶(60dB)

N = 8192

Tspan = 1.6384 ms then:

ΔT = 1.36 ns and ΔF = 4.16 Hz With typical 2 way cable propagation at 7.5 ns/m, a ΔT of 1.36 ns equates to 0.18 metres.

It will be appreciated that the above describes only a preferred embodiment of the invention. It may be carried out in many other ways as would be understood by the person skilled in the art.

Claims

1. A method for determining a propagation delay of a signal in a bidirectional network having a downstream bandwidth and an upstream bandwidth, the method including the steps of: a) introducing signals F1 and F2 into the downstream bandwidth of the network, signals F1 and F2 each having a defined modulation and specific frequency band such that said signals propagate in the downstream band of said network; b) monitoring the upstream band to detect signal F3, where F3 is a selected mixing product of F1 and F2; c) correlating the detected signal F3 with signals F1 and F2 so as to determine a time delay associated with one or more faults in the bidirectional network.

2. A method according to claim 1 wherein said bi-directional network has been mapped to correlate a determined time delay with a corresponding range value.

3. A method according to any one of claims 1 or 2 wherein said signal F3 is formed as a result of signals F1 and F2 encountering a non-linear element in said bi-directional network.

4. A method according to claim 3 wherein said non-linear element is a CPD fault in said bi-directional network.

5. A method according to any one of claims 1 to 4 wherein said upstream band occupies lower frequencies than said downstream band.

6. A method according to claim 5 wherein said signal F3 is the second order difference product of signals F1 and F2.

7. A method according to any one of claims 1 to 6 wherein at least a portion of said bi-directional network is a coaxial cable network.

8. A method according to any one of claims 1 to 7 wherein said method further includes an initialisation step to accommodate for the effects of external modulation on said signal F3.

9. A method according to any one of claims 1 to 8 wherein mains power is transmitted through said coaxial cable network.

10. A method according to claim 8 wherein said initialisation step provides timing signals to accommodate for the effects of modulation of said signal F3 by transmitted mains power.

11. An apparatus for determining a propagation delay of a signal in a bidirectional network having a downstream bandwidth and an upstream bandwidth, the apparatus including: a signal generator section for generating signals F1 and F2 having a defined modulation and specific frequency band lying within said downstream bandwidth; a signal detector for detecting signal F3, where F3 is a selected mixing product of signals F1 and F2; and a signal correlator for correlating the detected signal F3 with signals F1 and F2 to determine a time delay associated with one or more faults in the bidirectional network.

12. An apparatus according to claim 1 1 wherein said signal generator section includes an arbitrary waveform generator (AWG) and a first frequency shifting device to produce signals F1 and F2.

13. An apparatus according to claim 12 wherein said first frequency shifting device is an up-converter.

14. An apparatus according to claim 13 wherein said AWG provides a 5 MHz centred signal to an input of said up-converter.

15. An apparatus according to claim 14 wherein said up-converter converts said 5 MHz signal to produce signals F1 and F2, separated by 59.5 MHz.

16. An apparatus according to any one of claims 13 to 15 wherein said up- converter includes a first local oscillator and a second local oscillator.

17. An apparatus according to claim 16 wherein said first local oscillator oscillates at 24.75 MHz, and said second local oscillator oscillates at 71 1.25 MHz.

18. An apparatus according to any one of claims 1 1 to 17, wherein said signal detector includes a second frequency shifting device to receive signal F3, and a sampling device to sample received signal F3.

19. An apparatus according to claim 18 wherein said second frequency shifting device is a down-converter.

20. An apparatus according to claim 19 wherein said down-converter includes a third local oscillator.

21. An apparatus according to claim 20 wherein said third local oscillator oscillates at twice the frequency of said first local oscillator and is coherently locked thereto.

22. An apparatus according to claim 21 wherein said third local oscillator oscillates at 49.5 MHz.

23. An apparatus according to any one of claims 11 to 22 wherein said signal generator section generates a first initialisation signal which is used to provide a triggering signal for said signal detector.

24. An apparatus according to claim 23 wherein said AWG generates said first initialisation signal as a continuous wave signal.

25. An apparatus according to any one of claims 1 1 to 24 wherein said generator section generates a second measurement signal from which is derived the time delay associated with one or more faults.

26. An apparatus according to claim 25 wherein said AWG generates said second measurement signal as a frequency ramp chirp signal.