Method for measuring a sound signal propagation delay in a fluid by zero-crossing of the said sound signal

The present invention relates to a method of measuring the propagation time of an acoustic signal in a fluid flow between a first transducer playing the role of transmitter and a second transducer playing the role of receiver and located at a determined distance from the first transducer, the acoustic signal emitted by the first transducer consisting of at least one pulse emitted at a determined acoustic frequency Fa and the acoustic signal received by the second transducer comprising a series of characteristic oscillations whose amplitude is first increasing over several periods , then decreasing over several following periods, the envelope of the characteristic oscillations having the shape of a spindle, the method consisting in sampling the received acoustic signal, at a sampling frequency Fe, in digitizing the sampled received acoustic signal, and in seeking, by analysis of the sampled and digitized received acoustic signal, the first significant zero crossing of the oscillations characteristic of the acoustic signal received.

It has been known for many years to measure the flow rate of a fluid (or its volume) flowing in a pipe by using the propagation of the acoustic signals emitted between two acoustic transducers located at points spaced in the direction of flow fluid. In principle, an acoustic signal emitted from the first transducer to the second transducer is received by this second transducer and the propagation time Td of this acoustic signal is measured. Likewise, the propagation time Tu of an acoustic signal emitted from the second transducer to the first transducer is measured after reception of said signal by this first transducer.

In a fluid meter, the flow can be obtained by combining a measurement of the propagation times of the two acoustic signals emitted between the two points in opposite directions with a measurement of the acoustic phase shifts induced in each acoustic signal by the propagation of each of the signals acoustic in the flow. European patent application No. 0 426 309 describes an example of such a flow measurement system, in which the received signal is sampled and digitally converted, the measurement of the acoustic phase shift being carried out by performing synchronous detection of the digitized signal.

When measuring the flow rate of a gas in a gas meter using two ultrasonic transducers, it is necessary to measure the travel time of the ultrasonic wave between the instant of emission and the instant of reception, as soon as the speed of the ultrasonic waves is dependent on the nature of the gas.

FIG. 2 shows the form of a rectangular impulse signal S₁ of width T emitted at an instant T₀ by a first ultrasonic transducer arranged in the flow of a fluid at a first point, and the shape of the signal S₂ constituting the impulse response received in echo at an instant T₁ by a second ultrasonic transducer arranged in the flow of the fluid at a second point distinct from the first point.

The acoustic signal S₂ received by the second transducer is constituted by a series of characteristic oscillations O_vs which increase in amplitude over several periods then decrease, the envelope of the characteristic oscillations having the shape of a spindle. Characteristic oscillations O_vs of signal S₂ are preceded and followed by parasitic oscillations O_p low amplitude. To determine the instant T₁ at the start of the characteristic oscillations, it is advisable to locate the first significant zero crossing of the characteristic oscillations O_vs of the acoustic signal received S₂.

There is shown in Figure 3, on a larger scale than in Figure 2, an example of an acoustic signal S₂ received in response to a rectangular pulse S₁ emitted at a determined acoustic frequency Fa.

To determine the start of characteristic oscillations O_vs, according to a known method, a threshold voltage V is fixed_S, with respect to which the level of the acoustic signal received S is compared₂, the comparison being carried out on a digital signal obtained after sampling of the received analog acoustic signal, at a sampling frequency Fe which constitutes for example a multiple of the acoustic frequency Fa.

In this case, we identify the instant T₂ at which the amplitude of the received signal crosses the threshold voltage V_S, then we identify the instant of the previous (or next) zero crossing that we consider as the instant T₁ of characteristic oscillations O_vs of the acoustic signal received S₂.

Such a measurement method can lead to errors as soon as the characteristic oscillations O_vs of the acoustic signal received S₂ can be more or less amplified depending on the nature of the gas. Thus, there is shown in Figure 4 a curve S₂₁ which corresponds to the form of an acoustic signal received for nitrogen (N₂) and an S curve₂₂ which corresponds to the form of an acoustic signal received for a mixture of carbon dioxide and methane (CO₂/ CH₄). We see that the curve S₂₁ crosses threshold voltage V_S at a time T₄ which triggers the identification of the previous zero crossing at an instant T₃ which is rightly regarded as the mark of the beginning of the characteristic oscillations O_vs. On the other hand, we can see that the curve S₂₂, which is in phase with the curve S₂₁, crosses threshold voltage V_S at a time T₆ after time T₄ and offset from it by the value of a period T_R of the received signal. The instant T₅ then identified for curve S₂₂ as the zero crossing point immediately preceding the crossing of threshold at time T₆, is then considered as the mark of the beginning of the characteristic oscillations O_vs of the curve S₂₂. However, as can be seen in Figure 4, the curve S₂₂ presents a negative lobe which only comes to be flush with the value of the threshold voltage V_s without reaching or crossing this threshold.

As a result of attenuations or amplifications of the received signal S₂, which vary according to the nature of the gas, the conventional method of measuring the time of the first zero crossing of the oscillations characteristic of the received signal can generate an error of more or less a period which significantly reduces the accuracy of the measurements.

The invention aims to remedy the aforementioned drawbacks and to make it possible to reduce the sensitivity of the method for measuring the time of zero crossing of an acoustic signal received, with respect to variations in waveforms of this signal as well as vis-à-vis any external disturbances capable of being detected by a simple method of monitoring threshold crossing, and which could cause erroneous measurements.

These aims are achieved, in accordance with the invention, by a method of measuring the propagation time of an acoustic signal in a fluid flow between a first transducer playing the role of transmitter and a second transducer playing the role of receiver and located at a determined distance from the first transducer, the acoustic signal emitted by the first transducer consisting of at least one imp impulse emitted at a determined acoustic frequency Fa and the acoustic signal received by the second transducer comprising a series of characteristic oscillations whose amplitude is first increasing over several periods, then decreasing over several following periods, the envelope of the characteristic oscillations having the shape of a spindle, the method consisting in sampling the received acoustic signal, at a sampling frequency Fe, in digitizing the sampled received acoustic signal, and in seeking, by analysis of the sampled and digitized received acoustic signal, the first significant zero crossing of the oscillations characteristic of the acoustic signal received, characterized in that, in order to find the first significant zero crossing of the oscillations characteristic of the acoustic signal received, a first ideal characteristic period is defined beforehand determining the first zero crossing of the characteristic oscillations of the acoustic signal received, by characterizing this ideal period by a theoretical amplitude ratio Δ between the maximum amplitudes Pi- and Pi + of the two lobes of this period , for each period of the acoustic signal received sampled and digitized, the maximum amplitudes P- and P + of the two lobes of the period examined are determined, we compare the ratio of these amplitudes P- and P + to the corresponding theoretical amplitude ratio of the ideal period then, if the result of the comparison is greater than a threshold value G_S, we consider the period examined as a parasitic period corresponding to noise while, if the result of the comparison is less than this threshold value G_S, the period under consideration is considered to be a characteristic period and the passage through zero between two lobes of this characteristic period is then determined, which passage to zero is considered to be the first significant zero crossing of the characteristic oscillations of the acoustic signal received.

The theoretical amplitude ratio Δ between the maximum amplitudes Pi-and Pi + of the two lobes of said ideal period is determined beforehand from an average, for several different gases and at different flow rates, of the ratio between the maximum amplitudes P- and P + of the characteristic period observed from the recording of received acoustic signals.

The amplitude ratio Δ between the maximum amplitudes Pi- and Pi + of the two lobes of the first ideal characteristic period of an acoustic signal received is almost constant as a function of the nature of the gases and is independent of a gain factor. The choice of a comparison criterion based on this ratio Δ makes it possible to very greatly reduce the dependence of the measurement on the nature of the gas, and therefore to improve the precision and reliability of the measurements.

According to a first embodiment of the invention, the comparison between the ratio of the amplitudes P- and P + of the period examined and the corresponding theoretical amplitude ratio Δ of the ideal period is carried out by calculating a similarity criterion G which constitutes the result of the comparison compared to the threshold value G_S, and which is defined as follows:<maths id="math0001" num=""><img file="EP0902883B1_D0001.tif" /></maths>

According to a second embodiment of the invention, the comparison between the ratio of the amplitudes P- and P + of the period examined and the corresponding theoretical amplitude ratio Δ of the ideal period is carried out by calculating a similarity criterion G which constitutes the result of the comparison compared to the threshold value G_s, and which is defined as follows:<maths id="math0002" num=""><img file="EP0902883B1_D0002.tif" /></maths> where z represents an offset term intended to make the likeness criterion G slightly dependent on the maximum amplitude (P-) of the negative lobe of the characteristic period examined.

According to this second embodiment, the risks of being able to detect waveforms resembling the resemblance criterion or template G are limited, which would be drowned in noise.

To further reduce the probability of detecting similar periods of low amplitude, according to a third embodiment of the invention, the comparison between the ratio of the amplitudes P- and P + of the period examined and the corresponding theoretical amplitude ratio of the ideal period is carried out by calculating a similarity criterion G which constitutes the result of the comparison compared with the threshold value G_s, and which is defined as follows:<maths id="math0003" num=""><img file="EP0902883B1_D0003.tif" /></maths> where z represents an offset term intended to make the likeness criterion G slightly dependent on the maximum amplitude (P-) of the negative lobe of the characteristic period examined, and z 'represents an offset term intended to render the resemblance criterion G slightly dependent on the maximum amplitude (P +) of the positive lobe of the characteristic period examined.

In order to avoid errors in determining the maximum amplitudes P- and P + of the two lobes of each period examined, errors which may be due to a phase difference existing between the acoustic signal received and the sampling signal, an algorithm can be used approximation of the maximum (in absolute value) from the points sampled in the vicinity of each maximum in order to obtain better precision on the values ​​P- and P +. Always to avoid these same errors, when the acoustic signal received is repeatable, several successive measurements are carried out respectively on several consecutive acoustic signals, by phase shifting for each acoustic signal the sampling signal with respect to the acoustic signal considered, in order to obtain several sampled points shifted near the maximum.

According to an alternative embodiment also making it possible to reduce the probability of detecting similar periods of small amplitude, two or three calculations of the similarity criterion G are carried out for each period examined, with values ​​of the ratio of different theoretical amplitudes close to the ratio of predetermined theoretical amplitude and not deviating from it by more than 10%, and we consider the period examined as a characteristic period if the calculated similarity criterion G is in all cases lower than the threshold value G_s.

To determine the threshold value G_s of the likeness criterion G, as well as the value of the offset term z, we trace beforehand, for different gases and different flow rates, by varying the phase shift between the sampling frequency Fe and the acoustic signal received and by applying a factor gain varying between 0.45 and 1.5, a network of first curves representing the maximum values ​​of the likeness criterion G for the characteristic periods observed and a network of second curves representing the minimum values ​​of the similarity criterion G for the parasitic periods observed, as a function of different possible values ​​of the shift term z, and we choose the threshold value G_s and that of the offset term z as a function of the possible safety margin in the intermediate zone between the network of first curves and the network of second curves.

For example, the threshold value G_s can be between 0.7 and 1.7 and the value of the offset term z can be between 0.21 and 0.25 V.

The invention is advantageously applicable to a method for measuring the speed of flow of a fluid between two transducers arranged at measurement points spaced in the direction of a flow of the fluid according to which the value of the flow speed to be measured is obtained by combining a measurement of the respective propagation times of each of the two acoustic signals emitted between the two points in opposite directions by the transducers with a measurement of the acoustic phase shifts respectively induced in each acoustic signal by the propagation of each of said acoustic signals in the flow, the measurement of the respective propagation times of each of the two acoustic signals emitted between the two measurement points being carried out in accordance with the method according to the present invention.

Other characteristics and advantages of the invention will emerge from the following description of particular embodiments, made with reference to the appended drawings, in which:<ul id="ul0001" list-style="dash" compact="compact"><li>Figure 1 is a representation, in the form of a block diagram, of the assembly of a known gas metering device to which the present invention is applicable;</li><li>Figure 2 is a diagram showing the principle of a known method for determining the propagation time of an acoustic signal in a fluid medium;</li><li>Figure 3 is an enlarged diagram of part of Figure 2 showing the principle of a known method of determining the propagation time of an acoustic signal by comparison of the received signal with a threshold value;</li><li>Figure 3a is a diagram similar to that of Figure 3 showing the way in which a sampled signal is operated, according to a known method of determining the propagation time of an acoustic signal by crossing to zero;</li><li>Figure 4 is a set of two diagrams similar to that of Figure 3 showing the risk of error in a known method of determining the propagation time of an acoustic signal using a comparison of the received signals with a threshold value ;</li><li>Figure 5 is a diagram showing the principle of implementation of the method according to the invention for determining the propagation time of an acoustic signal in a fluid medium; and</li><li>FIG. 6 is a diagram comprising families of curves making it possible to choose values ​​of parameters useful for the implementation of the method according to the invention.</li><li>Figure 7 is a flowchart showing an example of the different steps of the method according to the invention.</li></ul>

Different types of electronic circuits can be used to ensure fluid counting by an ultrasonic method of measuring the flow speed of the fluid. Figure 1 shows the block diagram of an example of such electronic circuits which are associated with two ultrasonic transducers 1, 2 arranged at a distance from each other in a pipe 3 in which a fluid such as a gas. The two transducers 1, 2 are connected to a switching block 4 which includes two switches 5, 6 and allows the use of each transducer alternately as transmitter and receiver. A transmission module 14 and a reception module 17 are respectively connected to the switches 5, 6 of the switching block 4. The transmission module 14 comprises an operational amplifier 16 and a digital-analog converter 15. The reception module 17 comprises at least one amplifier 18 and an analog-digital converter 19 which simultaneously digitizes and samples the received signal. A source of electrical energy 7 and a module 8 for managing the electrical power supply are connected in particular to the transmission 14 and reception 17 modules as well as to the switching block 4 and to a microcontroller 10. The microcontroller 10 comprises in particular a quartz clock 9, an arithmetic and logic unit, random access memory and read only memory circuits, and can cooperate with display circuits 13, a rewritable read only memory 12 and a serial link 11 of the type RS 232.

The gas meters intended to equip each subscriber with a distribution network must be at the same time precise, reliable and the cheapest possible. These constraints make it necessary to avoid the use of expensive components, to use small long-life but medium-capacity power supply batteries for safety reasons and to carry out calculations according to a measurement method ensuring both high precision and reliability while remaining simple enough to be energy efficient.

The measurement method according to the invention which will be explained below is thus advantageously applied to a device for measuring the flow rate of a gaseous fluid, making it possible to count the consumption of this fluid, by putting in place operates two ultrasonic transducers arranged at a distance from each other inside a pipe, in the direction of flow of the fluid. By way of example, the ultrasonic transducers can operate at an acoustic frequency Fa of the order of 40 kHz and the sampling frequency Fe of the acoustic signals received is advantageously 320 kHz, that is to say eight times more higher than the frequency Fa.

The sampling frequency Fe is advantageously equal to a multiple of the acoustic frequency Fa. However, the frequency Fe can also take other values. When the values ​​(multiples of Fa or not multiples) of Fe are too low, it is preferable to use for example an approximation algorithm in the vicinity of the maximum (in absolute value) of each period of the acoustic signal, or even to perform several successive measurements by phase-shifting with each new measurement the sampling signal with respect to the acoustic signal considered, this in order to obtain greater precision on the values ​​of the maximum amplitudes (in absolute value). These methods will be detailed later.

When the Fe values ​​are sufficiently high the use of such methods is not necessary.

The acoustic signal emitted by each transducer is constituted for example by a pulse emitted at the determined acoustic frequency Fa. This pulse is for example rectangular. Several pulses could also be emitted.

Generally, the fluid flow in a pipe can be written:<maths id="math0004" num=""><math display="block"><mrow><mtext>Φ = (SL / 2) (Tu - Td) / Tu.Td</mtext></mrow></math><img file="EP0902883B1_D0004.tif" /></maths> or<dl id="dl0001" compact="compact"><dt>S</dt><dd>designates the average passage section offered to the flow between the two acoustic transducers,</dd><dt>L</dt><dd>denotes the distance between the transducers,</dd><dt>Td</dt><dd>is the propagation time of the acoustic signal emitted in the direction of flow by the first transducer (upstream) until the reception of this acoustic signal by the second transducer (downstream),</dd><dt>You</dt><dd>is the propagation time of the acoustic signal emitted in the opposite direction of the flow by the second transducer, until the reception of this acoustic signal by the first transducer (upstream).</dd></dl>

If the measurement of this flow rate also involves a measurement of the acoustic phase shifts induced in each acoustic signal by the propagation of each of the acoustic signals emitted by the transducers in the flow, the fluid flow rate in a pipe can be written:<maths id="math0005" num=""><math display="block"><mrow><mtext>Φ = (SL / 4πFa) (2π [Fa (Tu - Td)] + (ϕd - ϕu)) / Tu.Td</mtext></mrow></math><img file="EP0902883B1_D0005.tif" /></maths> where S, L, Td, Tu have the meaning indicated above, and ϕd and ϕu respectively represent the acoustic phase shifts induced in each of the acoustic signals due to the propagation of these signals and Fa represents the frequency of said acoustic signals.

More particularly, the propagation times Tu and Td can be written respectively:<maths id="math0006" num=""><math display="block"><mrow><mtext>Tu = L / (cv)</mtext></mrow></math><img file="EP0902883B1_D0006.tif" /></maths><maths id="math0007" num=""><math display="block"><mrow><mtext>Td = L / (c + v)</mtext></mrow></math><img file="EP0902883B1_D0007.tif" /></maths> where c and v represent respectively the speed of propagation of the acoustic signal and the speed of the gas.

For a given gas composition, as the propagation speed c depends mainly on the temperature which varies little during flow measurements, the term Tu.Td also varies little and this term can be recalculated with a relatively low frequency. However, as soon as it is an absolute measurement of the propagation times, the precision obtained for this measurement is directly transferred to the measurement accuracy of the flow rate. Thus, it is essential to avoid having a lag on this measurement.

For example, if the precision sought on the bit rate is 1%, the precision on the measurement of the propagation times must be close to 0.5%, which corresponds to an error of less than 2 µs on the propagation times with the common gases and the usual lengths L between two transducers (example: L = 150 mm and the gas being methane).

According to a method of measuring the propagation times Tu and Td, known as the zero crossing method, the propagation times (Tu pz) and (Td pz) of the acoustic signals emitted by each of the transducers are measured up to detection by zero crossing and the real propagation times Tu and Td are deduced therefrom, with Tu = (Tu pz - To) and Td = (Td pz - To) where To denotes a constant to be subtracted to obtain the propagation time in the gas, To depends only on the nature of the transducers and electronic circuits.

Figure 3a illustrates the zero crossing measurement method performed on each ultrasonic signal received according to a known method, in conjunction with a device such as that of Figure 1.

The received signal is amplified by the amplifier 18 having a given gain, then converted to digital and sampled simultaneously in the converter 19. In the example considered, the sampling is carried out at a frequency Fe for example equal to 8 Fa, ie 320 kHz if Fa is 40 kHz.

The instructions for carrying out the measurements, contained in a read-only memory of the microcontroller 10, make it possible to compare the values ​​of the amplitude of the signal received at the sampled points and which are stored in a random access memory, with a predetermined value corresponding to a threshold V_s, and one searches among the amplitude values ​​of the sampled points for the first point following A, which has an amplitude value greater than the threshold.

From this point, the next zero crossing of the curve representative of the ultrasonic signal is sought. To do this, we determine two consecutive sampled points B and C which surround the point where the amplitude of the curve passes through the value zero and we proceed by linear interpolation between B and C to determine the point D and therefore measure the zero crossing time Tpz.

According to an alternative embodiment, illustrated in Figures 3 and 4 discussed above, it is also possible to search for the last zero crossing of the curve representative of the ultrasonic signal before said curve passes through the threshold.

The present invention aims to improve this method of determining a signal propagation time by measuring the time of zero crossing.

As mentioned previously with reference to Figures 3 and 4, the simple comparison between the voltage amplitude values ​​at the sampled points of a received acoustic signal and a threshold voltage value V_s can lead to errors in the determination of the time of passage to zero Tpz and therefore in the propagation time of the acoustic signal. If in Figure 4 the zero crossing time T₃ for curve S₂₁ relative to the signal corresponding to a gas constituted by nitrogen is correct, on the other hand the time of passage to zero T₅ for curve S₂₂ relating to the signal corresponding to a gas consisting of a mixture of carbon dioxide and methane, is incorrect and excessive.

The method according to the invention makes it possible to avoid this kind of error and, in general, to reduce the sensitivity of the method for measuring the zero crossing time to variations in waveforms as well as to possible external disturbances which, if detected, would cause measurement errors.

According to the invention, to find the first significant zero crossing of characteristic oscillations O_vs of the acoustic signal received S₂, a first ideal characteristic period, distinct from the parasitic oscillations of low amplitude constituting noise, is defined beforehand, which determines the first zero crossing of the characteristic oscillations O_vs of the signal received S₂. This ideal period is characterized by an amplitude ratio Δ between the amplitudes of the maximum values ​​Pi- and Pi + of the negative and positive lobes of this ideal period.

When receiving an acoustic signal S₂ to be analyzed, which is sampled and digitized in a conventional manner, the amplitudes of the maximum values ​​P- and P + of the negative and positive lobes of the period examined are determined for each period of the acoustic signal sampled and digitized. For this, it is not necessary to examine the set of sampled points of the period, and the determination of the maximum values ​​(in absolute value) P- and P + of the negative and positive lobes of the period examined can be made by a search in the simple vicinity of these extremums.

However, it should be noted that there is a phase shift between the acoustic signal received and the sampling signal. Thus, depending on the existing phase shift, the sampled point will more or less faithfully represent the maximum values ​​(in absolute value) P- and P + of the negative and positive lobes.

To remedy this problem, it is possible to use an algorithm for approximating the maximum (in absolute value) from the sampled points in the vicinity of the maximum in order to obtain an estimate of the real value of this maximum whatever the phase difference between the acoustic signal received and the sampling signal. For example, one can use a polynomial approximation algorithm known per se. The previously mentioned problem could also be solved by increasing the sampling frequency.

However, when the process according to the invention is implemented in a fluid meter which operates on batteries, it is essential to reduce the energy consumption of the process as much as possible and this requirement is not compatible with an increase in the frequency sampling.

In addition, a high sampling frequency requires the use of a more expensive clock. An interesting solution to solve the aforementioned problem while respecting the constraints of energy consumption and cost would consist (when the acoustic signal is repeatable over time on several successive measurements) to carry out several successive measurements respectively on several consecutive acoustic signals emitted in the same direction of propagation by shifting the sampling signal with respect to the acoustic signal when a new measurement is made, in order to obtain after these various measurements several sampled points offset in the vicinity of the maximum amplitudes P- and P + of the negative and positive lobes. For example, the first measurement is carried out by normally sampling the received acoustic signal, then the second measurement is carried out on the second acoustic signal by shifting the sampling signal with respect to said second acoustic signal by a determined value 1_{/ 2Fe}.

It is thus possible to carry out several successive phase shifts (of value 1_{/ nFe} for n phase shifts) on the second acoustic signal and the following acoustic signals to obtain more sampled points in the vicinity of the maxima and thus have an even greater precision on these maximum amplitudes P- and P +. It is also advantageous to combine the use of a maximum approximation algorithm (in absolute value) with this solution in order to further increase the precision.

A comparison is then made of the ratio of these amplitudes of the maximum values ​​P- and P + to the corresponding Δ ratio of the ideal period.

If the result of the comparison is greater than a threshold value G_s, we consider the period examined as a parasitic period corresponding to noise, and we continue the analysis on the following period.

If the result of the comparison is less than the threshold value G_s, we consider the period examined as a characteristic period.

We can then reliably determine the zero crossing associated with this period examined. This zero crossing considered to be the first significant zero crossing of the characteristic oscillations of the acoustic signal received can advantageously be the zero crossing located between the two positive and negative lobes of the characteristic period.

Figure 5 shows a first characteristic oscillation O_it of period ΔT with a positive lobe whose amplitude has a maximum value P + and a negative lobe whose amplitude has a maximum value in absolute value P-. The zero crossing associated with this period ΔT of a first characteristic oscillation O_it defines a zero crossing time corresponding to a time T₇.

The amplitude ratio between the amplitudes of the maximum values ​​P- and P + of the two lobes of an examined period of the acoustic signal received constitutes a criterion of comparison independent of the nature of the gas. Thus, the aging of the electronics can lead to variations in gain which will not affect this amplitude ratio between the maximum amplitudes P- and P +. The shape recognition of the acoustic signal received can thus be carried out in a simple manner, which limits the energy consumption, since the comparison is made between the only maximum amplitudes P- and P + of the lobes and a ratio of amplitudes theoretical Δ between the maximum amplitudes Pi- and Pi + of the two lobes of an ideal characteristic period.

The theoretical amplitude ratio Δ is determined beforehand from an average, for several different gases and at different flow rates, of the ratio between the maximum amplitudes P- and P + of the characteristic period observed from the recording of signals acoustic received.

According to a particular embodiment, the comparison between the ratio of the amplitudes P- and P + of the period examined and the corresponding amplitude ratio of the ideal period is carried out by calculating a similarity criterion G which constitutes the result of the comparison compared to the threshold value G_s, and which is defined as follows:<maths id="math0008" num=""><img file="EP0902883B1_D0008.tif" /></maths>

The fact of making comparisons taking into account the ratios of maximum amplitudes of the lobes of a period of the acoustic signal received, the theoretical reference amplitudes ratio Δ is not affected by the nature of the gas or the gain of the amplification chain of the received signal. This helps to facilitate the measurements.

However, among the parasitic periods of the signal which constitute noise, there could be waveforms similar to that defined by the comparison criterion G. To make the detection of the characteristic periods less sensitive to the risk of taking into account resembling waveforms drowned in noise and which would not constitute a first period O_it characteristic oscillations O_vs of the received signal, the resemblance criterion can be made slightly dependent on the amplitude P-.

Thus, according to another particular embodiment, a comparison between the ratio of the amplitudes P- and P + of the period examined and the corresponding ratio of amplitudes of the ideal period is carried out by calculating a similarity criterion G which constitutes the result of the comparison compared to the threshold value G_s, and which is defined as follows:<maths id="math0009" num=""><img file="EP0902883B1_D0009.tif" /></maths> where z represents an offset term intended to make the likeness criterion G slightly dependent on the amplitude of the maximum value (P-) of the negative lobe of the characteristic period examined.

In this case, the denominator of the criterion G is greatly reduced when the amplitude P- is close to the shift term z. If the offset term z has a value slightly greater than the maximum amplitude of the noise (for example of the order of 0.20 to 0.25 Volt), this makes it possible to more quickly perform rejection of the parasitic periods of the signal received.

To determine the value of G_s to which the likeness criterion G must be compared, and the value of the offset term z, we draw a network of first curves representing the maximum values ​​of the likeness criterion G for the characteristic periods observed and a network of second curves representing the values minimum of the likeness criterion G for the parasitic periods observed, as a function of different possible values ​​of the shift term z, and the threshold value G is chosen_s and that of the shift term z, and we choose the threshold value G_s and that of the offset term z as a function of the possible safety margin in the intermediate zone between the network of first curves and the network of second curves.

By way of example, the families of curves 111 and 121 in FIG. 6 were obtained by varying the phase shift between the sampling frequency and the signal and by applying a gain factor varying between 0.45 and 1.5 , the gases taken into account being methane and a mixture of 20% carbon dioxide and 80% nitrogen, the measurements having been made at zero flow rate and at 7 m³/hour.

The networks of curves 111 and 121 make it possible to choose the values ​​of the shift term z and of the threshold value G_s depending on the desired safety margin on these values. If we consider Figure 6, we see that we can choose for G_s for example a value between approximately 0.7 and 1.7 V and, for the term of offset z, for example a value between 0.21 and 0.25 V.

If we choose for example G_s = 2 and z = 0.22 V; the safety margin on z is between 0.20 and 0.25 V while the safety margin for G_s is between 1 and 4. In practice, with this example, a period examined will be considered characteristic if G &lt;G_s = 1,5.

Other alternative embodiments make it possible to further increase the safety and reliability of the process.

Thus, according to a particular embodiment, the comparison between the ratio of the amplitudes P- and P + of the period examined and the corresponding ratio of amplitudes of the ideal period is carried out by calculating a criterion of resemblance G which constitutes the result of the comparison compared to the threshold value G_s, and which is defined as follows:<maths id="math0010" num=""><img file="EP0902883B1_D0010.tif" /></maths> where z represents an offset term intended to make the likeness criterion G slightly dependent on the amplitude of the maximum value (P-) of the negative lobe of the characteristic period examined, and z 'represents an offset term intended to render the similarity criterion G slightly dependent on the amplitude of the maximum value (P +) of the positive lobe of the characteristic period examined.

This makes it possible to reduce the sensitivity to similar periods of low amplitude and, without significantly increasing the complexity of the calculations, to be able to reduce the gain of the amplification chain to values ​​as low as 0.35, which saves energy.

According to another variant, and whatever the formula chosen for the likeness criterion G, two or three calculations of the likeness criterion G are carried out for each period examined, with values ​​of the ratio of amplitudes Δ different close to the ratio d '' predetermined theoretical amplitude and not deviating from it by more than 10%, in that the period examined is considered to be a characteristic period if the calculated likeness criterion G is in all cases lower than the threshold value G_s.

Figure 7 is a flowchart showing an example of the process for implementing the method according to the invention.

If we consider a sampling at a frequency Fe, for example of 320 kHz, which represents eight times the acoustic frequency Fa of the transducers, the sampled signal will comprise 8 points per period of the signal received.

Insofar as the acoustic signal received has the form of periodic oscillations, we know where the successive minimums and maximums are, which avoids examining all the sampled points.

We therefore begin, after the initialization phase 201, to seek during a step 202, the next local minimum (of amplitude P-). During this step 202, if we apply a similarity criterion G involving an offset term z, we only retain the first local minimum whose amplitude P- is greater in absolute value than the offset term z. When this first local minimum has been found, we go to step 203 which consists in seeking, in the vicinity of a previous half-period T / 2, the value of the maximum amplitude P +.

In the next step 204, the value of the criterion of resemblance G is calculated with the previously found values ​​of the minimum P- and the maximum P + of the period examined which has at least a minimum whose absolute value is greater than the shift term z .

In step 205, a comparison test is carried out between the value of the criterion G calculated and the threshold value G_s.

If G ≥ G_s for the period examined, it is considered to be a parasitic period, and there is a return to step 202 for searching for a next local minimum.

If G &lt;G_s for the period examined, this is considered to be a characteristic period and we pass to a step 206 of determining the next zero crossing of the amplitude of the received signal which will constitute the zero crossing situated between the local maximum P + and the minimum local P- and will therefore be the true starting point of the characteristic oscillations O_vs of the acoustic signal received.

The same process will be repeated a few moments later for a new measurement for determining the propagation time of an ultrasonic acoustic signal.