The present invention relates to a leak noise logger. US Patent No US5974862 discloses a "Method for detecting leaks in pipelines". Its abstract states that the method relates to:

"detecting leaks in a pipeline using enhanced methods of data sensing, digitally encoded transmission, and cross-correlation. A plurality of acoustic sensors are applied to the pipeline to sense a combination of signals from the leak and much greater quantities of noise. The sensed data is digitized at the sensor, encoded and digitally transmitted to a computerized base station. The encoded data from a plurality of sensors received at the base station is decoded, digitally filtered and cross- correlated. Enhanced methods of cross-correlation are performed to estimate the likely presence of a leak and the location and size of any leaks present. Participation in the leak detection procedure by an expert not present at the pipeline is facilitated by communication of data between the base station and a distant supervisory station."

For such a method of determining leak position by correlation, we supply leak noise loggers, having acoustic sensors, for application to a pipeline and logging of the leak noise, subsequent correlation and location of the leak typically via a remote server. Primarily, they are used for detecting leaks from water mains.

Normally such a leak noise logger will comprise:

an audio transducer for detecting leak noise;

a memory for recording a detected leak noise;

a clock for timing recording; and

means for downloading the recording of the detected noise for detecting leak position by comparison with another recording of detected noise from another leak noise logger.

Usually the comparison will be by cross-correlation. Timing of the recording is important because it is conventional to record at times of minimal usage, namely after midnight when water usage and noise inducing from its flow in the mains is at a minimum. In so far as the correlation relies on timing of respect leak noises at the respective loggers, it is particularly important that the relative timing of the respective leak noises in real time should be determinable. For instance in a simplistic example if two loggers are separated by a distance equivalent to two seconds for noise to travel from one to the next and a leak noise is detected as requiring half a second to travel to one and one & a half seconds to the other; the leak is known to be 25% of their separation from the one and 75% from the other. Nevertheless if due to an

accumulated clock error, the one logger's signal appears to be arriving one second later, i.e. at the same time as the other logger's signal, the leak will appear to be midway between the loggers. In practical terms a 1 part per million clock drift in each logger's clock can result after 12 hours in an 86 milliseconds timing error. Such a timing error can give rise to a 100 metre error in leak location or more with typical sound speeds (which can vary in pipelines with pipe material and dimensions).

The object of the present invention is to provide an improved leak noise logger.

According to one aspect of the invention there is provided a leak noise logger comprising:

• an audio transducer, or a cable connector therefor, for detecting leak noise;

• a memory for recording a detected leak noise signal;

• a clock for timing recording; and

• means for downloading the recording of the detected noise for detecting leak position by comparison with another recording of detected noise from another leak noise logger and

further comprising:

• a radio receiver for receiving a radio signal having a modulation,

o the radio receiver having a demodulator for extracting the modulation and • means for passing the modulation to the memory for recording and

the download means being adapted to download recordings of both the modulation and the detected leak noise, whereby:

• the timing of the recording of detected leak noise can be compared with that of another leak noise logger for leak position estimation and

• the timing of the recording of modulation can be compared with that of

another leak noise logger for logger clock time difference and correction of the leak position estimation accordingly.

The radio receiver can be an amplitude modulated - AM - receiver or a frequency modulated - FM - receiver or both, with the demodulator being

correspondingly AM or FM or both. Whilst we prefer AM or FM modulation for below ground reception; we also envisage that the radio receiver may also be digitally modulated, particularly where, unusually, an above ground antenna is practical. Other forms of modulation may also be useable.

Whilst it is envisaged that the radio signal could be received and its modulation memorised synchronously with the detected noise, in the preferred embodiment, the logger is adapted to record the radio signal prior to, or subsequently to, the detected noise.

Preferably, the logger is adapted to memorise in the memory the modulation of the radio signal immediately or within a minute prior to, or similarly subsequently to, with the detected noise. Alternatively, the logger can be adapted to record in the memory the modulation of the radio twice, once a defined period of time prior to leak noise recording and a second time a defined period of time after leak noise recording.

For this a switch is provided for switching either modulation of a radio signal or a signal from the audio transducer to memory for recordal. Normally the switch will be provided in conjunction with an analogue to digital (A to D) converter for converting analogue audio signals to digital signals, which are suitable for comparison by an analysis computer. Alternatively, two A to D converters are provided, one dedicated to the pipe leak noise and the other to the radio demodulation signal. Normally, the logger will have:

• an antenna or connection therefor for the radio receiver and an antenna or connection therefor for the download means and/or means for wired or infra- red transmission download and/or

• a magnet or other means for securement thereof to a potentially leaking pipe.

The logger is preferably adapted to make the recordings in the time domain, with individual samples including information representing the instantaneous amplitude of the modulation or the leak noise. Alternatively, it can be adapted to make the recordings in the frequency domain, with individual samples including information representing the frequency of the modulation or the leak noise.

Preferably, the download means is adapted to download the recordings as: · a continuous stream of data including both the radio modulation data and the leak noise data prefaced with an identification header; or

• a single stream of recordings prefaced with a gap separating the radio

modulation data and the leak noise data, the single stream prefaced with an identification header; or

· a first stream of data including the radio modulation data prefaced with a first identification header and a second stream of data including the leak noise data prefaced with a second identification header

According to another aspect of the invention, there is provided a method of analysing of recorded signals from two leak noise loggers of the one aspect of the invention for estimation of leak position, the method comprising:

• receiving from a first of the leak noise loggers recordings of (i) a modulation from a radio signal and (ii) a leak noise from a pipe on which the first logger is installed at a first position, the recordings having been initiated at a clock time of the clock of the first logger;

• receiving from a second of the leak noise loggers recordings of (iii) a

modulation from a radio signal and (iv) a leak noise from the pipe on which the second logger is installed at a second position, the recordings having been initiated at a clock time of the clock of the second logger, the clock times of the two loggers being nominally the same;

• comparing similarities of the modulation recordings(i,iii) from the two leak noise loggers at differing comparative timings, determining the comparative timing at which the similarity is greatest and generating a clock time difference signal corresponding to the greatest similarity comparative timing, this signal indicating the time interval by which these recordings are out of synchronism due to difference in clock times of the two loggers;

• comparing similarities of the leak noise recordings(ii,iv) from the two leak noise loggers at differing comparative timings, determining the comparative timing at which the similarity is greatest and generating a leak noise time difference signal indicative of leak position between the installed positions of the two loggers;

• performing a correlation on the leak noise recordings from the two leak noise loggers, generating a leak noise time difference signal indicative of leak position between the installed positions of the two loggers and

• adjusting the leak noise time difference signal and/or a leak position signal generated therefrom to take account of the clock time difference signal and estimate the leak position as the loggers' clocks were in synchronism.

The recordings could be in the frequency domain, with individual samples including information representing the frequency of the radio-audio/modulation or the leak noise. However it will be appreciated that providing frequency information for multiple samples requires much data to be memorised in the loggers originally and transmitted for reception. We prefer the recordings to be in the time-domain, with each sample containing a single piece of data, namely the amplitude of the radio audio or the leak noise at the time of the sample.

Whilst we envisage that simpler forms of comparison may be able to be employed, such as subtractive comparison, we prefer to use cross-correlation for both the leak noise comparison and the leak noise comparison, in like manner to conventional correlation of data from pairs of leak noise loggers. Whilst we can envisage performing the comparison of the leak noise recordings first to produce a preliminary estimate of leak position and adjusting this in accordance with the clock time difference, it is likely that, taking account of long term clock drift, the adjusted leak noise time difference will usually be smaller than the clock time difference. Accordingly we prefer to perform the modulation comparison first and then apply the clock time difference to the leak noise data, subsequently comparing the adjusted leak noise data to obtain the adjusted leak noise time difference and thence the estimated leak position. To help understanding of the invention, a specific embodiment thereof will now be described by way of example and with reference to the accompanying drawings, in which:

Figure 1 is a side view of a known leak noise logger;

Figure 2 is a block diagram of the known logger's circuitry;

Figure 3 is a diagram showing two known leak noise loggers, placed on a pipeline with a leak between them;

Figure 4 is a side view, similar to Figure 1 , of a leak noise logger in accordance with the invention;

Figure 5 is a block diagram, similar to Figure 2, of the leak noise logger of Figure 2 in accordance with the invention;

Figure 6 is a flow chart showing correlation of leak data from two loggers in accordance with the invention; and

Figure 7 is a flow chart similar to Figure 6 showing a variant in the correlation method shown Figure 6.

Referring first to Figures 1, 2 and 3, a known leak-noise-detecting data logger 1 has a case 2 with an acoustic transducer 3 positioned in use on a pipeline P. Where the line is iron or steel, or has such a fitting, the logger can be held on magnetically. Extending from the case is a communications antenna 4. Internally of the casing, it comprises an analogue to digital converter 5 to which the transducer is connected. A microprocessor 6, having a clock 7, and a memory 8 is arranged to pass leak noise signals from the transducer via the converter to the memory to be logged as a time- domain, digital data stream for a period, typically of 10 seconds, at a low use time of night. A communications module 9 is provided to transmit the data stream, together with a header of identification data, automatically or on demand on a daily basis.

In use of the known leak-noise-detecting data loggers, two of them are placed on the pipe at a known separation S at the bottom of respective pits 11 beneath manhole covers 12. It should be noted that the loggers are below ground level G. The pipeline has a leak L between the pits.

Each night, typically at 2.00am, the loggers record 10 seconds of leak noise. The recordings are initiated at the same time in accordance with the loggers' own clocks. The relative timing of their logged recordings indicates the position of the leak. For analysing them, they are transmitted at appointed times to a remote server in a protocol identifying at least the individual logger (together with the date, time, duration and logger position of the data recording being transmitted), where the recordings are compared by cross-correlation and the position of the leak between the loggers is estimated. As mentioned above the correlation is liable to be subject to a level of uncertainty due to clock error.

Referring now to Figures 4 and 5, a leak-noise-detecting data logger 101 according to the invention is shown. It includes all of:

• a case 102,

• an acoustic transducer 103,

• a communications antenna 104,

• an analogue to digital converter 105,

• a microprocessor 106,

• a clock 107,

• a memory 108 and

• a communications module 109, which can be a GPRS, GSM or 3G modem.

Other forms of wireless communication may be suitable, such as future cellular telephone, Bluetooth and Automatic Meter Reading technologies.

These components are similar to the corresponding components in the known leak- noise-detecting data logger 1 of Figure 1, although the microprocessor has additional programming. In addition, the logger 101 has:

a radio signal reception antenna 121 separate from the communication antenna;

a radio receiver 122 connected to the antenna and arranged to be switched on and off immediately before and after logging by the microprocessor, a demodulation circuit 123 in the receiver and arranged to output an audio signal from the receiver,

an electronic switch 124 again switched by the microprocessor for switching the modulation from the radio receiver to the A-to-D converter 105 or the transducer output to the A-to-D converter.

In use, again a pair of the loggers 101 is placed on the pipeline with their transducers in acoustic contact with it, a manner analogous to that of Figure 3. Again the loggers are programmed to record at 2,00am each night for instance. Immediately before 2.00am, the radio receiver is switched on. From 2.00am, in accordance with the clock of each logger, 10 seconds of pipe noise from the transducer is logged in the memory 108 with the switch connecting the transducer to the memory via the A-to-D converter. After the ten seconds, the switch 124 is switched to open circuit for 5 seconds and then switched to pass the radio demodulation signal from the receiver 122 and the demodulator circuit 123 to the memory for a further 10 seconds. After this recordal the radio receiver is switched off.

Bearing in mind that the loggers are below ground, in their pits having the pipe at the bottom, the radio signal may not penetrate as far as their installed position. To improve radio reception and indeed data transmission, the antennas may be detachable and reconnectable to the circuitry in the casing via coaxial cables. In this way, the antennas can be secured to the wall of the pit immediately below the pit's manhole cover.

Despite the loggers being underground, our tests have shown reliable radio reception at the loggers, with the radio receiver being both an FM and a medium wave AM receiver. Again at its appointed time, each logger transmits its logged data. Its transmission protocol differs in that each transmission is in two parts. The first part of the transmission is prefaced by a header identifying the logger, the date, time and duration of its recording of pipe noise. The end of the pipe leak noise transmission is positively identified with reference to the 5 seconds break in the recording. The second part of the data transmission is of the demodulated radio audio made with a similar header, including in addition details of the radio signal received in particular its transmission station.

The actual memorisation is conventional, in that several seconds typically between 5 and 15 and usually 10 seconds, of leak noise is memorised in as an amplitude sample at a 5kHz rate. Other frequencies can be used typically from 4kHz to 11kHz. The radio signal is memorised in exactly the same way. 10 seconds of each type of sample results in twice ten seconds worth of samples at 5kHZ, i.e.

100,000 audio samples, including both the radio demodulation samples and the pipe noise samples.

After sampling of the radio signal modulation, the logger returns to its quiescent state.

The logged data can be downloaded either by direct, wired or wireless - e.g. IR optical, connection via the interface 118 in due course and the loggers removed for use elsewhere. Alternatively, where regular monitoring is required the loggers can be left in place and their data downloaded via the modem 109, suitably at a time when an analysis server is expected to be online to receive and correlate the data.

Alternatively, where the data transmission is in a GPRS or GSM transmission in HTTP format, it can be transmitted at any time convenient to a service provider's server and forwarded to the user's server for analysis when the latter is on line.

The method of correlation is shown in Figure 6.

i. At the first stage, the radio and leak noise sample data from the two loggers is separated into respective pairs of radio modulation and leak noise data sample streams; ii. The respective data streams are correlated separately but in the same way. It should be noted that this is possible because the radio modulation and leak noise data streams are effectively of the same sort and can be correlated in the same way as is known in the art. They are both essentially audio data;

iii. The result of correlating the radio modulation data is to obtain a difference in the timing, or clock difference, at which the recording of data - both radio modulation and leak noise - was started by the two loggers. Each starts at the same nominal time determined by the loggers' clocks but actually at different real times due to clock drift. The result of correlating the leak noise data is to obtain an apparent time difference between leak noise being detected at the respective loggers, with the clock drift superimposed on the apparent time difference;

iv. The clock difference can be applied to the leak noise difference to obtain a

corrected leak noise difference;

v. The leak position is calculated, taking account of the physical characteristics of the pipe, in particular its length, material and velocity of sound in it - the "pipe model" of Figure 6.

As an alternative to steps iv. and v., shown in Figure 7, the clock time difference can be applied to the pipe noise data to delay one other of the leak noise data streams, to bring them into synchronism. In other words, with relative timing such that a sample noise generated at the exact midpoint between the loggers would arrive synchronously at the loggers should they have no clock time difference. This has the advantage of being able to correlate over a period equivalent to the time taken for sound to travel along the pipe between the two loggers, as opposed to that time plus the full range of times by which the clocks are expected to be out of

synchronism, which is necessary if the pipe leak noise date is correlated first. Equally the radio modulation data needs to be correlated only over the out of clock synchronism range. The invention is not intended to be restricted to the details of the above described embodiment. For instance, whereas the radio modulation signal was described above as being recorded immediately after the leak noise, a suitable radio signal may be available only during the day. In this case, a radio modulation signal can be recorded exactly 12 hours - in accordance with the loggers' clocks - before and again after the leak noise recording. The correlation of these recordings will allow a calculation of a mean clock difference at the leak noise recording time. For this the loggers' processors are programmed to make the radio recordings at the appropriate clock times before and after the leak noise recordings. Further, we envisage that the radio receiver may incorporate an analogue to digital converter, whereby the converter for the pipe transducer need not be used to convert the radio's signal.

It should also be noted that we envisage the possibility of the loggers including sufficient data processing capability for one to communicate its data to the other for performance of the correlation by the other. The result can then be displayed at the logger in the field and/or transmitted to a remote server or host machine.

Whilst the transducer will normally be connected the casing of the logger, the casing could have a transducer cable connector with the transducer couple remotely via a cable. The radio antenna could be incorporated with the communication antenna on a single aerial structure.

Further, whilst the data transmission has been described above as separate transmissions for each logger pipe leak noise and radio demodulations, each with its own header, the transmission could be with a single header and a combined stream of data. Since the data is in the time domain the numbering in the sequence of data words representing individual samples positively identifies whether they are from a first 50,000 samples, for instance, of pipe noise or a second 50,000 samples of radio demodulation.

Further whilst it is more efficient to transmit in the time domain, the transmission could be in the frequency domain. Also there is not necessity to correlate in the same domain as the data has been transmitted. Where the analysis computer is able to cross-correlate more efficiently in the frequency domain, the time domain data can be transformed into frequency domain data prior to cross-correlation.