GB2441554A - Ultrasonic anemometer - Google Patents

Abstract

An ultrasonic anemometer capable of providing fluid (e.g. air or gas) velocity and direction data comprises at least one pair of ultrasonic transducers 5,6 which operate at a frequency less than 200 kHz (e.g. between 20 kHz and 150 kHz) and which are separated 7 by up to about 20 metres (e.g. between about 0.5 m and 10 m). The anemometer operates on the transit time principle. The anemometer allows velocity measurements in environments such as restricted spaces since the spacing between the transducers allows the transducers to be placed at either end of the space thus avoiding physical obstruction of the space.

Description

1 2441554

ANEMOMETER

This invention relates to anemometers and particularly ultrasonic anemometers capable of providing fluid velocity and direction data in restricted spaces and/or over distances of up to about 20 m, and methods of use thereof.

Anemometers are used for measuring the velocity and direction of fluid (typically air) flows.

Types of anemometer include ultrasonic, hot wire, hot bead and rotating vane. The measurement of fluid velocities by the use of ultrasonic techniques has been used increasingly in the last two decades or so. Advantages of ultrasonic techniques include the absence of moving parts, and hence low maintenance and high reliability, an ability to resolve the full vector flow and fast response. This type of instrument has been extensively : .. used in meteorological research, particularly in the boundary layer, where much valuable information on the structure of atmospheric turbulence has been obtained. The instruments * : :::* typically comprise one or more pairs of transducers operating at acoustic frequencies in the *:5 region of 200 to 500 kHz at fixed separations of between 0.1 and 0.2 m. Thus, all *** anemometers to date require placement in the flow and resolution of the velocity and * direction within a narrow and fixed distance (maximum 0.2 m) at the location of the instrument.

Ultrasonic anemometers can however be intrusive, particularly in restricted spaces such as passageways and tunnels, and especially where collisions with traffic (people, vehicles) may be a major problem. To date optimal measurement of flow in restricted spaces has required the instrument to be positioned in, or close to, the centre of a space, for example tunnel or passageway. This is neither safe nor practical. However, measurements made by instruments placed at the edge of a tunnel or passageway are heavily compromised with the flow being affected by the presence of the boundary wall. Ultrasonic anemometers also provide only limited data points for calculating fluid velocity and direction, unless multiple anemometers are utilised.

There is therefore a requirement for an ultrasonic anemometer capable of calculating fluid velocity and direction in a non-intrusive manner, particularly avoiding physical obstruction within restricted spaces, and also an ultrasonic anemometer capable of interrogation and resolution of fluid velocity and direction for a distance in excess of 0. 2 m, and preferably in excess of about 5 m.

Ultrasonic anemometers typically comprise one or more pairs of acoustic transducers :: : :* separated by a fixed distance. Each transducer is capable of transmitting and receiving ....* acoustic energy, or ultrasonic signals, and thus they are also sometimes referred to as * * transceivers. Calculation of fluid velocity is possible only when a contribution of the * 5 velocity is in a direction parallel to the path length, and thus velocity perpendicular to the ** path length would register as zero. For this reason most anemometers comprise two or more * pairs of transducers and record in more than one direction. A fluid flow in the direction of the ultrasonic signal leads to an increase in the velocity of the signal, whereas fluid flow against the direction of the ultrasonic signal results in a reduction of the signal velocity. As the speed of sound is dependent on fluid temperature, the velocity of an ultrasonic signal is commonly measured in both directions, thus cancelling the effect of temperature.

The present invention generally aims to provide an ultrasonic anemometer comprising at least one pair of transducers and signal processing means capable of providing fluid velocity and direction data from a transducer separation in excess of 0.2 m, and preferably from any transducer separation of up to about 20 m.

As used herein, transducer separation is the distance between an aligned pair of transducers, which equates to the propagation distance of an ultrasonic signal between said pair of transducers, also commonly referred to as the path length.

Accordingly, in a first aspect, the present invention provides an ultrasonic anemometer comprising i. at least one pair of ultrasonic transducers of frequency less than 200 kHz; and ii. signal processing means capable of providing fluid veloóity and direction data from a : .. transducer separation of up to about 20 m. * .

. The ultrasonic transducers in a pair of ultrasonic transducers may be at a fixed transducer **** :15 separation however, in a preferred embodiment the transducer separation is not fixed, and * more preferably the ultrasonic transducers in a pair of ultrasonic transducers are * : * independently moveable, enabling any transducer separation to be achieved. The ultrasonic anemometer, comprising said signal processing means and at least one pair of transducers of frequency less than 200 kHz, whereby each pair of ultrasonic transducers is not restricted to a fixed transducer separation, is capable of measuring fluid and velocity data over any and all distances between 0 m and about 20m. The ultrasonic anemometer particularly allows velocity measurements in environments that do not lend themselves to other integrated flow measurements, such as a restricted space, as the transducers can be placed either end of the restricted space, thus avoiding physical obstruction of the space. The transducer separation can be customized to the particular situation. Passing objects, such as vehicles or people, cause momentary interruption of the measurement, but with no physical damage to the anemometer or the passing object. On the contrary, such an anemometer is capable of utilising the interruption to count objects and furthermore calculate the velocity of the object.

An ultrasonic anemometer comprising two or more pairs of transducers, at known separations, is capable of deducing object length and whether the object is accelerating or decelerating.

As used herein, at least one pair of ultrasonic transducers includes, but is not limited to, one, two or three pairs of ultrasonic transducers. In a preferred embodiment, every pair of ultrasonic transducers in the ultrasonic anemometer are of frequency less than 200 kHz.

: * :: The skilled person will be acquainted with the additional features and combinations of S...

features essential to an ultrasonic anemometer. An example of one embodiment of the * : :::* present invention further comprises a power supply, a switching circuit, a pulse generator, * .15 and an oscilloscope.

S * . S * *

The theory behind ultrasonic anemometry is that a fluid velocity component in the direction of propagation of an ultrasonic signal between a pair of transducers will lead to an increase in the velocity of propagation of said ultrasonic signal, and vice versa. If the transducer separation is known, the fluid velocity component can be calculated.

The transducer separation may be calculated from the time taken for an ultrasonic signal to propagate between a pair of ultrasonic transducers, i.e. the transit time, at a constant and known temperature. The transit time is dependent on the velocity of an ultrasonic signal in the fluid, which is thereby temperature dependent. Thus, the ultrasonic anemometer optionally comprises temperature measurement means to enable accurate calculation of the velocity of an ultrasonic signal in the fluid, and thereby the transducer separation. The temperature measurement means is preferably a temperature sensor, and more preferably a thermistor temperature sensor. This approach to calculation of transducer separation is however dependent on the absence of a fluid flow.

In the presence of a fluid flow, the transducer separation may be calculated from the transit time of a first ultrasonic signal in the forward direction between a pair of ultrasonic transducers and the transit time of a second ultrasonic signal in the reverse direction between the pair of ultrasonic transducers. This allows the velocity of the fluid flow to be eliminated from the calculation. The transducer separation can then be calculated from the velocity of * : ::* the ultrasonic signal in the fluid and the pair of transit times. This method additionally has *.* the advantage of cancelling out any temperature fluctuations which may have occurred * **. during the measurement of the transit times. **.

S

***.** * .15 * It will also be understood by the skilled person that for precise measurement of fluid velocity * and distance data from one or more ultrasonic signals propagating between said at least one pair of ultrasonic transducers the transducers in a pair of transducers must be aligned. This is actually quite difficult to achieve visually and thus the ultrasonic anemometer preferably further comprises means for alignment of the ultrasonic transducers in a pair of ultrasonic transducers. The means for alignment is preferably achieved by using a computational process capable of accurately identifying when the amplitude of an ultrasonic signal arriving at a transducer attains a maximum value. Alternatively alignment may be achieved by using a narrow beam light source, preferably a low power laser light source, attached to one transducer of a pair and a reflective target attached to the other transducer of the pair. The position of the transducers with the light source attached can then be adjusted until the narrow beam of light is aligned with the reflective target.

The ultrasonic transducers are preferably capable of emitting an ultrasonic signal in the form of a narrow beam. Currently available transducers have full beam widths in the region of 7 .

With a broad beam multi-path effects are likely to occur, particularly if the apparatus were to be used in a confined space, or close to an obstacle such as a wall. The presence of such spurious signals would at the very least confuse the measurement, and very likely cause significant and unpredictable errors.

The ultrasonic transducers in an ultrasonic anemometer are capable of both transmitting and : .. receiving an ultrasonic signal. The ability for the ultrasonic transducers in a pair of S.. S...

..... transducers in the present invention to transmit and receive an ultrasonic signal is preferably * : :::* controlled by transmitting/receiving switching means such that at any one time one :": ultrasonic transducer of a pair of ultrasonic transducers is only capable of transmitting an .*. ultrasonic signal and the other transducer of the pair is only capable of receiving an ultrasonic signal. This switching is preferably by application of a high speed changeover reed relay or a solid state device. The transducers in a pair are preferably identical, i.e. they are an identical make and model of transducer.

The signal processing means is preferably capable of providing fluid velocity and direction data from a transducer separation of between about 0.5 m and about 10 m, and more preferably from a transducer separation of between about 2 m and about 6m.

The ultrasonic transducers are preferably of frequency between about 20 kHz and about 150 kHz, more preferably of between about 50 kHz and about 100 kHz, and most preferably of a frequency of about 75 kI-Iz.

The ultrasonic transducers in a pair of transducers are preferably in communication with each other, and with the signal processing means. The communication is preferably by means of an electronic or wireless link. The electronic link is preferably by means of cables or wires.

Signal processing means preferably comprises digital processing means as this provides a high degree of flexibility in terms of handling the complex signal received and enables : * : several different methods for characterising signal transit time to be explored. **.. * S ****

**** . . . . Signal processing means preferably compnses ultrasonic signal amplification means, and * : i more preferably further comprises a means capable of automatically adjusting signal * threshold to account for or reflect signal amplitude variation within an ultrasonic signal. It * has been shown that even in moderately turbulent conditions large signal amplitude fluctuations occur.

The ultrasonic signal amplification means preferably comprises an analogue amplifier, preferably of about 65 dB gain at 75 kHz which is capable of increasing the received signal to a few volts.

Signal processing means preferably further comprises an analogue to digital converter (ADC), and more preferably in combination with a first-in first-out (FIFO) memory, or alternatively in combination with a static RAM.

Signal processing means preferably further comprises a computer programme for processing the received amplified signal into fluid velocity and direction data.

Signal processing means also preferably comprises means for measurement of the transit time of an ultrasonic signal, which preferably comprises a computerised system and programme for reducing the complexity required for fast data handling. There is a direct relationship between temporal resolution and the minimum velocity than can be resolved.

: ,. Preferably the means for measurement of the transit time of an ultrasonic signal is capable of I...

recording one or more pairs of measurements, in both forward and reverse directions ::::* between the transducers in a pair of transducers, and is preferably capable of producing data at ten times per second, averaged each 1 second interval. The signal processing means is . preferably capable of producing data within a 3 ms window.

S

I.'...

S

The fluid is preferably a gas, and more preferably air.

In a second aspect, the present invention provides a method for determining fluid velocity and direction data comprising i. providing a pair of ultrasonic transducers of frequency less than 200 kHz; ii. positioning and aligning said pair of ultrasonic transducers at a transducer separation of up to about 20 m; r iii. propagating at least one ultrasonic signal between said pair of ultrasonic transducers; and iv. processing of the received at least one ultrasonic signal into fluid velocity and direction data.

The method preferably further comprises determining the transducer separation. The transducer separation may be derived from the length of time taken for an ultrasonic signal to propagate between the pair of ultrasonic transducers, i.e. the transit time, in absence of a fluid flow. Alternatively, in the presence of a fluid flow the transducer separation may be derived from the transit time of a first ultrasonic signal in the forward direction between said pair of ultrasonic transducers and the transit time of a second ultrasonic signal in the reverse direction between said pair of ultrasonic transducers; the velocity component of the fluid see.* flow thereby being cancelled out. Since velocity of an ultrasonic signal is temperature dependent, calculating the transducer separation preferably further comprises measuring the fluid temperature.

C S. * * .

The method preferably further comprises calculating the transit time of the at least one ultrasonic signal. Transit time measurement should be accurate to within approximately I ps; the transit time for a transducer separation of 4 m is about 12 ms and would therefore require the transit time to be calculated from 12,000 data points. This is a huge amount of data to process. Thus, in a preferred embodiment processing of the received at least one ultrasonic signal comprises collecting and/or analysing a fraction of the data produced. Thus in a more preferred embodiment processing of the received at least one ultrasonic signal comprises providing a time period window for collecting and/or analysing a fraction of the data produced.

As used herein, the time period window is a fixed length of time for which data is collected andlor analysed.

As used herein, at least one ultrasonic signal includes, but is not limited to, one, two, three, or four ultrasonic signals.

Processing of the received at least one ultrasonic signal is preferably by means of digital processing, and preferably comprises amplifying the at least one ultrasonic signal. The processing preferably further comprises automatically adjusting signal threshold to account for or reflect amplitude variation in the received at least one ultrasonic signal. Signal ** processing is preferably facilitated by use of a computer programme. * .

* ** Processing of the received at least one ultrasonic signal preferably further comprises storing the data produced in a first-in first-out (FIFO) memory to enable efficient retrieval and processing of the data.

The time period window need not occur at any one particular point in time during the method. However, the time period window preferably overlaps with arrival of the at least one ultrasonic signal at one ultrasonic transducer in a pair of ultrasonic transducers. Thus, in a preferred embodiment the method further comprises selecting the time period window to overlap with the point in time at which the at least one ultrasonic signal from one transducer in a pair is received at the other transducer in the pair, i.e. the transit time of the received at least one ultrasonic signal.

Selecting the time period window to overlap with the transit time of the received at least one ultrasonic signal preferably comprises estimating the transit time of the at least one ultrasonic signal, and thereby providing the time period window to overlap with an estimated transit time. The time period window is preferably a fraction of the estimated transit time, and more preferably about 3 ms.

In a more preferred embodiment selecting the time period window to overlap with the transit time of the received at least one ultrasonic signal comprises i. estimating the transit time of the at least one ultrasonic signal, thereby providing the time period window to overlap with an estimated transit time; : : ii. propagating the at least one ultrasonic signal between said pair of ultrasonic transducers; S... * .

S

iii. processing of the fraction of data produced within the time period window; *.S'S' * . :. iv. adjusting the point in time for providing the time period window; and : : .: v. repeating step ii to step iv until the time period window overlaps with the transit time of the received at least one ultrasonic signal.

In a preferred embodiment the time period window is about 3 ms and adjusting the point in time is by about 1 ms. In a further preferred embodiment repeating step ii to step iv is until the time period window overlaps with the transit time such that the transit time coincides with the middle 60 % of the time period window.

Calculating the transit time preferably comprises calculating the mean of the data collected within the time period window, calculating the sum of the absolute deviation from the mean and calculating a fixed percentage of the absolute deviation from the mean. The transit time corresponds to the point at which the sum of the absolute deviation of the means exceeds the fixed percentage. The fixed percentage is preferably about 5%.

Alternatively, the method for determining fluid velocity and direction data comprises providing multiple pairs of ultrasonic transducers of frequency less than 200 kHz. The method further comprises positioning and aligning the ultrasonic transducers in each pair of ultrasonic transducers at a transducer separation of up to about 20 m, whereby the transducer separation for each pair may be identical, similar or dissimilar. Each pair of ultrasonic transducers preferably functions independently, though signal processing is preferably : performed concurrently to provide a more detailed and accurate description of fluid flow.

The method preferably comprises two or three pairs of ultrasonic transducers, enabling * *.* determination in two or three dimensions, thereby providing much more detailed information *S* * p S..

about both the transverse and longitudinal components of fluid velocity and direction to be *I''' * * obtained. *. * * * w * *

****.* * S The transducer separation is preferably of between about 0.5 m and about 10 m, and more preferably of between about 2 m and about 6m.

The ultrasonic transducers are preferably of frequency between about 20 kHz and about 150 kHz, more preferably between about 50 kHz and about 100 kHz, and most preferably of frequency of about 75 kHz.

Since the velocity of an ultrasonic signal is dependent on fluid temperature, the method preferably comprises propagating at least one pair of ultrasonic signals in both forward and reverse directions between said pair of ultrasonic transducers, thus cancelling out any temperature fluctuations along either the whole path length or during the period of analysis.

The fluid is preferably a gas, and more preferably is air.

In a third aspect, the present invention provides a computer programme for processing the received at least one ultrasonic signal from a transducer separation of up to about 20 m into fluid velocity and direction data.

: ** 10 The present invention will now be described with reference to the following examples and drawings in which S... * S

Figure 1 is a schematic of the set-up for one embodiment of the present invention; S..... * . * * . .

: : ..: 35 Figure 2 is a graphical representation of the most efficient double pulse drive waveform used; Figure 3 is an actual sequence of a low amplitude signal obtained with one embodiment of the ultrasonic anemometer of the present invention in moderate turbulence conditions; Figure 4 is an actual sequence of a high amplitude signal obtained with one embodiment of the ultrasonic anemometer of the present invention in moderate turbulence conditions; Figure 5 is data produced outdoors over a 20 minute period by an ultrasonic anemometer of the present application comprising a pair of ultrasonic transducers of 75 kI-Iz placed at a transducer separation of 3.93 m, 0.5 m from a wall and at a height of 0.4 m from the ground.

Mean wind speed was +0.048 ms'; Figure 6 is data produced indoors with high ventilation over a ten minute period by an ultrasonic anemometer comprising a pair of ultrasonic transducers of 75 kHz p1aced at a transducer separation of 2.72 m, and a height of 0.4 m from the ground, with a window a quarter open, a door fuily open and away from any walls. Mean wind speed was -0.027 ms'; Figure 7 is data produced indoors with low ventilation over a ten minute period by an : ultrasonic anemometer comprising a pair of ultrasonic transducers of 75 kHz placed at a transducer separation of 2.75 m, and a height of 0.4 m from the ground, with all windows and doors fully shut, and away from any walls; * S ::. Referring now to Figure 1, a low impedance 100 V DC supply 1 is connected to a simple :...;t transistor switching circuit 2, designed around a high voltage transistor (MPSA42). The * * pulse generator 3 (TGP 110, Thuriby Thandar Instruments) is used to apply base drive to 2 and hence a high voltage pulse to the ultrasonic transducer 5. The pair of ultrasonic transducers 5 and 6 are separated at a distance 7, corresponding to the transducer separation or path length. In addition, a signal from 3 is used to trigger the oscilloscope 4. An ultrasonic wave is thus propagated over the distance 7 between 5 and 6. A signal to 5 from 2 is applied via a capacitor (0.1 jF) to prevent damage of piezo-electric elements of the transducer which may be caused by use of a direct DC voltage. The receiver circuit consists simply of 6 connected directly to 4.

Examples

Example 1 -Transducer Specification

Three types of transducer were investigated for suitability for propagating ultrasonic signals over transducer separations of up to about 20 m, and especially transducer separations of between about 0.5 m and about 10 m. All transducers investigated manufactured by Murata, and brief details are given below.

Type Nominal Attenuation Wavelength Capacitance Max Directivity Mass Ref Frequency (dBm') (mm) (pF) Applied (deg) (g) (kHz) Voltage : .. MA8OA1 75 5 1.8 4.55 980 120 7 93 S...

MA200AI 200 10 7.0 1.70 340 120 7 6 MA400A1 400 20 27.0 0.85 170 120 7 2 * S **. ______________ ______________ _____________ ____________ ____________ ______ *****S * . * . * . . * o The fact that the transducer output is largely independent of the input drive waveform to SS**S * . some extent simplifies circuit design. In particular it turns out to be feasible to use a simple pulse drive which is convenient as it lends itself to operation with digitally-based circuitry.

Now having regard to Figure 2, extensive experiments showed that the most efficient drive waveform was in the form of a double pulse. The voltage applied was in the region of 90V with voltage on (T0) and voltage off (Toff) of equal duration. The sum, i.e. T0 + Ton' was adjusted to coincide with the period corresponding to the resonant frequency of the particular transducer under investigation. These were 13.33.ts, 5 j.ts and 2.5 s for the 75, and 400 kHz transducers, respectively.

The range, or maximum path length, of the 400 kHz transducer was limited. Signals deteriorated very rapidly with increasing distance, due to the high attenuation rate, and were in the region of I to 2 mV at a separation of 0.5 m. Even with very substantial amplification there was clearly no prospect of operation at 400 kHz for the transducer separations required.

Experiments showed that the received signals from a 200 kHz transducer were about 3 mV at a separation of 1 m. The attenuation data would suggest that this value would be reduced by a factor of roughly 25 at a separation of 5 m, and when coupled with the effects of beam divergence would result in a total reduction in the region of a factor of 625. Hence the : resultant signal presented to the input amplifier would be -5 tV. Theoretically it is possible to amplify microvolt signals to useable levels (this is a typical aerial signal level in a radio S..

system) however the presence of electronic noise was thought likely to cause significant I. * problems in the determination of transit time. * I ** S * S S

: :..: 45 The 75 kHz transducer generated a considerably greater signal. At a separation of 1.1 m typical received voltages approached 50 mV. Allowing for the 1.8 dB m1 attenuation and beam divergence this would produce a theoretical reduction factor in the region of 45 at a separation of 5 m, and thus a receiver voltage in the range of 1 mV, a figure subsequently confirmed by experiment. A system based on the use of a 75 kHz transducer is thus capable of providing path lengths in excess of 5 m, and theoretically capable of the desired transducer separations of up to about 20 m.

Example 2-Transmitting/Receiving Switching Means An ultrasonic anemometer was devised comprising a pair of identical transducers capable of both transmitting and receiving ultrasonic signals. This results in a more compact unit and also helps to reduce costs. In order for each transducer to be switched between transmitter and receiver a switching mechanism was however required as it would not be possible to leave both connected as the high voltages involved in generating the transmission pulse would immediately destroy the receiver input circuitry. Various approaches to addressing the problem were considered but ultimately the best solution was to empioy a high speed changeover reed relay. The particular relay selected, RS 291-9704 (RS components), is capable of at least I 8 operations before replacement.

Example 3 -Signal Processing -Digital Processing of Signal Amplitude Variation : Experiments were conducted at 75 kHz using moderate indoor turbulence generated by a fan.

The results indicated that relatively large signal amplitude fluctuations occur frequently. As * *** might be expected turbulence almost invariably acts in a manner so as to scatter energy out S... * *.*

of the beam thereby causing a reduction in amplitude. Typically amplitudes varied by as * * much as 3:1 among the individual ultrasonic signals and it was very clear that this would : ..: 45 present asignificant challenge in the development of a robust method for transit time measurements.

The minimum speed change discrimination obtained in an ultrasonic anemometer depends directly upon the time resolution with which the transit time of the acoustic energy pulse between the pair of transducers can be determined. Whilst it is possible to measure even very small time intervals with extreme accuracy using electronics i.e. to within 1 or 2 ns the unambiguous definition of the transitlarrival time of a complex and variable amplitude waveform is by no means straightforward. Now having regard to Figure 3 and Figure 4, an ultrasonic signal can be complex even in conditions of moderate (indoor) turbulence generated by a fan. The signals were digitally processed at 9 bit resolution following -65 dB of amplification.

Clearly in qualitative terms there is no difficultly in defining an arrival time associated with each of the two ultrasonic signals, despite their peak amplitudes differing by a factor of 3, as the eye readily recognises the pattern of an ultrasonic signal and is capable of automatically compensating for the difference in amplitudes. However, though automatic pattern recognition techniques exist these are generally software intensive and thus not suitable for this type of application. * .*

A simple electronic technique utilising the points at which a set threshold is exceeded was *.** devised. Such a technique could be implemented by using a high speed comparator. *.* * S

However, suppose the threshold of the technique was set at 256 bits. The data from Figure 3 I..... * *

would then trigger an event after the third well-defined peak, whereas data from Figure 4 * : would trigger an event after the first well-defined peak. The amplitude difference in this case would thus effectively result in a two cycle timing discrepancy, which for a 75 kHz ultrasonic signal would correspond to a timing error of approximately 27.ts which was entirely unacceptable as the concomitant velocity error would be in the region of 0.3 ms* A more sophisticated system in which the threshold is set automatically to reflect amplitude changes was thus devised. Such a technique was unlikely however to be employed by use of an analogue system only, as the system presupposes that the peak amplitude is known before the waveform arrives at the receiver. A high performance digital processing system was thus developed to capture and interpret the received waveforms. This system would clearly offer maximum flexibility as compared with any analogue system in that, provided the initial digital design were adequate, all analysis could be carried out by suitable software. In addition to the purely Digital Signal Processing considerations a digitally based system could also be used to generate the transducer transmitter pulses, store data and write information to an LCD Panel.

Example 4 -Signal processing -Digitisation and Storage of Data Considerable effort was devoted to developing a concept which would enable all the various digitally-based options to be evaluated without significant development of hardware. The concept devised comprised storage of data in non-volatile random access memory : (NVRAM), for subsequent downloading to a computer, with simultaneous displaying of the S...

information on an LCD panel during acquisition. The digital system also enabled operation *SS* of a high speed Analogue to Digital converter (ADC) to digitise the Received ultrasonic *5S* * S S..

* : signals. The system also comprised a means for generating ultrasonic pulses and circuitry for switching between transmitting and receiving from each transducer.

* ..* S * S Example 5 -Signal Processing -Ultrasonic Signal Amplification For the 75 kHz amplifier, a two stage amplification process utilised an amplification circuit comprising a LM6364 high-speed operational amplifier (National Semiconductor). The high frequency response was limited by feedback loop components i.e. 100 k = = resistor in parallel with a 10 pF capacitor, while the low frequency cut-off was controlled by a 22 nF capacitor in series with a 1.0 k = = resistor. Additional bandwidth control was provided by parallel LC filter elements in the output circuits, i.e. a 2.0 nF capacitor in parallel with a 2.2 mH inductor. This amplification process resulted is a reasonably narrow band response centred around a frequency of 75 kHz, reducing noise whilst at the same time avoiding introducing too much additional ringing' from the low frequency signal. Separate gain and DC level trimmer potentiometers were also provided.

Example 6-Transducer Transmission Circuit The high voltage transducer transmitter circuit was based on an inverter manufactured by Hilek Power Corporation, model GMAI2-200P. It is rated at 1.5 W and produces 200 V output with an input of 12 V. The design employed in the ultrasonic anemometer employs a standard 5 V input voltage which is raised to 10 V using a MAX66O switched capacitor voltage converter (National Semiconductor). The inverter output is fed to a 1 j.tF reservoir : capacitor via a 5.6 k) resistor. This combination has a time constant of 5 ms and thus the capacitor is virtually filly replenished in 10 ms, this being more than sufficiently rapid, S. producing up to 20 pulses per second. * . *.*.

I... S. * . :. * The transmitted pulses are derived by means of a switching transistor. The collector load : : *t resistance (4.7 k) is placed across the reservoir capacitor. When quiescent, the capacitor is charged to approximately 170 V. On applying base drive to the transistor, derived from the microcontroller, the transistor conducts and the output pulses generated are fed to the transducer via a 0.1.tF capacitor to block any DC component. Examination of the output waveform from the transducer confirmed that the circuit produced pulse amplitudes in the region of 90 V at the transducer input. This was adequate, though remained well within the maximum voltage rating of 120 V. Example 7 -Signal Processing -Pulse Transit Time Measurement and Fluid Velocity Calculation In addition to its non-intrusive nature a major advantage that ultrasonic anemometry possesses as compared with most other types of instrumentation is its ability to both resolve and respond to very low speed changes. However, obtaining good performance in this respect requires the measurement of pulse transit times to accuracies approaching a microsecond. Consequently, for a digitally based system, the analogue to digital converter (ADC) must digitise the received ultrasonic signals at a correspondingly fast rate. Whilst running an ADC at this speed poses few technical difficulties the ensuing data that are generated present significant problems for direct handling with a microcontroller. The determination of transit time requires that the received ultrasonic signals be digitised at : approximatelyl p.s intervals, however for a transducer separation of 4 m the total interval S...

would be approximately 12 ms. This is a huge amount of data to process. *SS* *5** * . *...

* : * It proved possible to reduce the theoretical amount of data for analysis by adopting a : software scheme which processes the received ultrasonic signals within a 3 ms window, the : *5 position of the window being defined using a special subroutine. Nevertheless this still required approximately 3,000 data points to be examined in order to determine the transit time. Two options were considered as possible solutions, either to write the ADC data to a static RAM, or write the ADC data to a FIFO (first in first out) Memory. The stored data could then be read from the memory and analysed subsequently by the microcontroller.

Previous experience with microcontrollers suggested that, of the two options, the latter would almost certainly be simpler to implement and accordingly was investigated first. The ADC/FIFO combination selected for evaluation was that of an ADS82O ADC (Burr-Brown) coupled to an 1DT7206 (16k x 9) FIFO (IDT). The ADC had a 20 MHz capability, a single clock conversion input and a 10 bit parallel data interface.

A I V sinusoidal 75 kHz test signal was applied to the analogue input provided by a Tektronix Function Generator, model AFG3O2 1. The ADC was configured using its internal references which provided a range of 0.25 V to 4.25 V for the lower and upper limits, respectively. The value of 0.25 V corresponds to 000h and that of 4.25V to 3FFh, each bit being 4/1024 V i.e. 3.91 mV. Noting that the FIFO interface was only 9 bits it was decided to discard the Least Significant Bit (LSB) from the ADC, this therefore reduced the resolution to 7.82 mV. A particular advantage of the ADCIFIFO combination selected was that it proved possible to operate the ADC and load the FIFO using the same microcontroller : 10 port, i.e. the inputloutput (110) line. Following data acquisition the FIFO address pointer can * *,, be reset and the contents read to the microcontroller for subsequent analysis. S... *. * S

As this scheme functioned extremely reliably the system based on the use of a Static RAM S.... * .

was not investigated, however this approach is just as applicable. * S

The primary reason for adopting a digitally based system is the flexibility it confers on the development of signal processing algorithms. The principal difficulty with ultrasonic anemometry is that of characterising the transit time of the acoustic pulse. The received pulse, which takes the form of a train of waves of increasing and then decreasing amplitude, occupies a total time of milliseconds. Since the period of the 75 kHz transmitted signal is just 13.33 ps it is clear the entire train will consist of many cycles, typically several hundred.

In addition to this the overall amplitude of the signal also varies, mostly as a result of turbulence. In practice this may result in fluctuations as large as 3:1 in successive received signals.

It is possible to carry out quite sophisticated mathematical operations with modem microcontrollers e.g. Fast Fourier Transforms however these generally make significant demands on the processor. Whilst this may be acceptable for many applications, ultrasonic anemometry necessarily involves high data rates if it is to provide satisfactory measurements and this immediately precludes the use of the more complex processing techniques. In the case in point the microcontroller is required to carry out all of the following functions before it can devote any time to digital processing, send transmitted signal pulses, set change over relay, wait for specified time interval and then run the ADCIFIFO, download the FIFO memory to the microcontroller, evaluation of transit time to generate the velocity, update the : LCD Screen, write data to the NVRAM, send data to the RS232 Port, and check for inputs * *.. from the operator. Note that the first four operations need to be carried out twice in each *SSI measurement cycle i.e. for the forward and reverse transit time determination. Overall these *S** * S *S operations occupy typically 60 to 70 ms and with the target primary sampling rate of 10 Hz S..... * .

this leaves just 30 to 40 ms to carry out all digital processing. Consequently it is important i that the processing is as efficient as possible. * .

Although the number of RAM Registers available in the microcontroller is fairly generous by most standards (i.e. 16 banks of 256 bytes), almost 4 complete banks were already committed to other uses. This left 12 banks (i.e. 3072 bytes) for digital processing.

The ADC is run at 1 MHz, and the data acquisition window corresponds to a total time of 3.072 ms. Whilst it would be feasible to run the ADC for up to 16.384 ms before the FIFO were full and then transfer the data in segments for subsequent processing this would be computationally very inefficient.

The position of the 3.072 ms window is incremented by 1 ms intervals as the sequence progresses. The Digital processing routine searches for the presence of a received ultrasonic pulse in two successive intervals. This is achieved by setting a threshold of -1 00 mV above the maximum noise amplitude and noting the successive times at which the received signals attain their maximum values. Note that this operation is performed using only the upper 8 bits of FIFO data. If the maximum amplitudes of the received ultrasonic pulses are satisfactory the routine then remembers' the timing and sets the window position accordingly. The test sequence described is carried out twice at 1 s intervals. The results of the two sequences are compared and if in agreement the ultrasonic anemometer is ready to : start data recording. I... * * ****

Once the window timing is set it is maintained throughout the recording. Consequently it is * :: important to consider what may happen if either the ambient temperature or the air speed : change significantly. Clearly if there is substantial variation in these parameters it is : conceivable that the ensuing received ultrasonic pulse could arrive either too early or too late to be correctly sampled within the window. Examination of known equations indicates that for the range of operating conditions envisaged for the anemometer this is most unlikely to occur.

Determination of the transit time is carried out using the full 9 bits available from the FIFO.

During initial analysis of the received pulses it quickly became evident that determination of the transit time depended critically on the information contained just before the start of the signal, until just after the point where the maximum value in the signal was attained.

Accordingly in order to ensure that this criterion was always met the data were processed in the range of -512 to +256 intervals either side of the maximum thus making a total of 768 intervals. The actual sequence of operations is i. re-read the FIFO data over the specified 768 intervals at 9 bit resolution; ii. compute the mean value of the received ultrasonic signal, averaged over the 768 intervals; iii. compute the sum of the arithmetic difference of the signal values over the 768 intervals; iv. compute the exceedance threshold value for determination of the transit time; and v. repeat step iii. noting the time at which the threshold is exceeded. This defines a value for the transit time. * ** * * . ***. S... * S **S*

In detail, the FIFO data is first re-read using the full 9 bits that are available in order to **** * S *SS.

*....: confer maximum accuracy. In theory the 768 data points could be packed' into 768 + * * (768/8) i.e. 864 registers. However the required manipulation of the bits would be time j consuming, consequently the 768 data points actually occupy 1536 registers. A second bank of 1536 registers is used for the reverse direction of the received ultrasonic pulse.

The mean value of the received ultrasonic signal is next computed. This requires 3 registers for accumulating the individual byte values, and long division by 768 is accomplished using a specially written routine.

The third operation in the sequence is the critical one in the sense that it quantitatively defines the received ultrasonic pulse intensity and thus provides a stable normalisation value in order to remove the effect of amplitude fluctuations. Mathematically the operation is defined: n=768 Dev51(Tot)= wherein V4 represents the individual byte (effectively voltage) values and V the previously computed mean. Consequently the quantity denoted by Dev (Tot) is effectively an integral representation of the total intensity of the received signal and is thus far less susceptible to being compromised by electronic or other noise sources than a simple instantaneous amplitude value.

: *e The penultimate step involves computing an arbitrarily set fraction of Dev (Tot) in order a.e4 * ,ro to specif' an exceedance criterion for determining the transit time. After some * experimentation this value, denoted by Dev, (F) was set to * . O.O5xDev5(Tot).

The final step requires re-computing of the sum in step iii up to the point where the increasing sum first exceeds Dev (F). This then defines the transit time.

All ultrasonic anemometers effectively measure a spatially averaged velocity component along the line joining the transducers. In this case this will be several metres in length and, in the applications envisaged, typical air speeds are likely to be quite low probably less than 2 ms1. Consequently the associated time scales will be of the order of one second and this suggests that an appropriate sampling rate should be 1 Hz.

As mentioned above the design was to carry out a speed measurement every 0.1 s, and this was achieved in practice at a transducer separation of 6 m. The scheme permitted an average of 10 individual pairs of transit time measurements to be taken to compute the 1 Hz data points. In addition to reducing any random' noise (by a factor of IiT in theory) it also provided an opportunity to check for and possibly remove any errors, such as outlying points, within a received signal. The steps involved are: i. compute the average values of the 10 pairs of forward and reverse transit times; ii. examine the individual values in relation to their respective averages. Note which of them, if any, should be removed as outlying points. This is achieved by subtraction, the value computed then being compared with a threshold value. This . was set at 10 units i.e. 10 ps; iii. discard the outlying points and recompute the average; and * S

SS

iv. use the values obtained to compute the air velocity. S. I * 4

As experience was gained with the computation procedure it emerged that it could be improved in the following way. Examination of the data produced revealed that the scheme worked very effectively in low-turbulence conditions, i.e. when signal amplitude was not subject to very much variation. However in moderate to high turbulence levels as many as 5 or more transit time values were often being discarded as outlying points. Clearly this was unsatisfactory as it was probable that the majority of discarded values were in fact valid.

Accordingly the computational scheme was adjusted so that the threshold automatically adapted itself to the ambient conditions. Every time 5 or more outlying points were found the threshold was increased by 5 units i.e. 5 ps.

Note this has some consequences for the operation of the anemometer. In particular the instrument needs to adapt itself to the ambient turbulence levels to operate optimally. The simplest way of achieving this is to run the instrument for a few minutes so that it can adapt to the current environment.

Example 8 -Signal Processing -Conversion to Fluid Velocity One embodiment of the present invention employs an analogue amplifier providing about 65 dB of gain at 75 kHz to raise the received signals to levels of a few volts. The incoming signals are then digitised at a rate of 1 MHz at 9 bit resolution and then stored in a FIFO memory before being transferred to the microcontroller for analysis. I S. S * I *1**

The waveform of the received signal is complex and requires considerable processing before the transit time of the acoustic pulse can be accurately determined. In one embodiment a pair * of transit time readings (i.e. forward and reverse along the path length) are taken 10 times per second and then averaged over one second intervals. This information is then converted *5*SS * into air velocity data, written to the screen and also stored in the memory for subsequent downloading. A simple adaptive error correction routine is also included in the microcontroller.

This embodiment of the present invention operates satisfactorily in wind speeds up to -3 to 4 ms, and is capable of resolving air speeds as low as 0.02 ms'.

Example 9-Measurement of Fluid Velocity An ultrasonic anemometer of the present application was tested over a range of conditions, both indoors and outdoors. Having regard to Figure 5, the transducers were placed parallel to the wall of a domestic property. The nature of the topography was such that considerable turbulence, ducting and flow acceleration occurred, and hence the relatively high maximum speeds observed and frequent flow reversals. Having regard to Figure 6, data was recorded in draughty indoor conditions. Surprisingly, though the operator was present in the room throughout the measurement period, there was only an occasional perception of air movement. Having regard to Figure 7, data was recorded immediately following that in Figure 6, and thus shows the initial decay of the remaining air motion before stabilisation of calm conditions with typical air speeds of less than 0.05 ms* 0 Example 10-Measurement of Temperature A thermistor temperature sensor was fitted to the Control/Base Unit of the anemometer in order to estimate the ambient temperature, and hence enable a value for the transducer separation to be determined. The sensor, nominally 30 k at 25 C, was placed in series with I.....

* a 33 k resistor and the combination placed across an electricity supply. The output voltage of the thermistor circuit was then fed to a microcontroller port, configured as an analogue * input. Using a simple algorithm it was possible to derive a value for the temperature accurate to within 1.5 C.

Claims (1)

1. An ultrasonic anemometer comprising i. at least one pair of ultrasonic transducers of frequency less than 200 kHz; and ii. signal processing means capable of providing fluid velocity and direction data from a transducer separation of up to about 20 m.

2. An ultrasonic anemometer according to Claim 1, further comprising means for alignment of the ultrasonic transducers in a pair of ultrasonic transducers.

4. An ultrasonic anemometer according to Claim 1 or Claim 2, in which the ultrasonic I.:::. transducers in a pair of ultrasonic transducers are independently moveable. S... S * S... 5..

* .. 5. An ultrasonic anemometer of Claims I to 4, in which the at least one pair of ultrasonic I.....

* transducers are of frequency between about 20 kHz and about 150 kHz. . . * S * 15 *

* 6. An ultrasonic anemometer of Claims 1 to 5, in which signal processing means comprises digital processing means.

7. An ultrasonic anemometer of Claims 1 to 6, in which signal processing means comprises signal amplification means.

8. An ultrasonic anemometer of Claims 6 to 7, in which signal processing means comprises a means capable of automatically adjusting signal threshold to account for signal amplitude variation within an ultrasonic signal.

9. An ultrasonic anemometer of Claims 6 to 8, in which signal processing comprises use of an analogue to digital converter in combination with a First-in First-out (FIFO) memory.

10. An ultrasonic anemometer according to Claims 1 to 9 in which the fluid is a gas.

11. An ultrasonic anemometer according to Claims 1 to 10 further comprising temperature measurement means.

12. A method for determining fluid velocity and direction data comprising i. providing a pair of ultrasonic transducers of frequency less than 200 kHz; :.:::. ii. positioning and aligning said pair of ultrasonic transducers at a transducer separation of up to about 20 m; iii. propagating at least one ultrasonic signal between said pair of ultrasonic transducers;

S * and

iv. processing the received at least one ultrasonic signal into fluid velocity and direction ***** * data.

13. A method according to Claim 12, in which the frequency is between about 20 kHz and about 150 kl-Iz.

14. A method according to Claim 12 or Claim 13, in which the transducer separation is between about 0.5 m and about 10 m.

15. A method according to Claims 12 to 14, further comprising determining the transducer separation.

16. A method according to Claims 12 to 15, in which processing comprises digital processing.

17. A method according to Claims 12 to 16, in which processing comprises amplifying the at least one ultrasonic signal.

18. A method according to Claims 12 to 17, in which processing comprises automatically adjusting signal threshold to account for amplitude variation in the received at least one ultrasonic signal. a

19. A method according to Claims 12 to 18, in which processing comprises storing the data *** produced in a first-in first-out (F1FO) memory.

S * *

* 15 20. A method according to Claims 12 to 19, in which processing comprises providing a time * period window for collecting a fraction of the data produced.

21. A method according to Claim 20, in which the time period window is about 3 ms.

22. A method according to Claims 20 to 21, further comprising selecting the time period window to overlap with the transit time of the received at least one ultrasonic signal.

23. A method according to Claim 22, in which selecting the time period window to overlap with the transit time of the received at least one ultrasonic signal comprises i. estimating the transit time of the at least one ultrasonic signal, thereby providing the time period window to overlap with an estimated transit time; ii. propagating the at least one ultrasonic signal between said pair of ultrasonic transducers; iii. processing of the fraction of data produced within the time period window; iv. adjusting the point in time for providing the time period window; and v. repeating step ii to step iv until the time period window overlaps with the transit time of the received at least one ultrasonic signal.

24. A method according to Claims 12 to 23, further comprising calculating the transit time of : *. the at least one ultrasonic signal. S...

25. A method according to Claim 24, in which calculating the transit time comprises i. calculating the mean of the data collected within the time period window; . 15 ii. calculating the sum of the absolute deviation from the mean; and p.....

* iii. calculating a fixed percentage of the absolute deviation from the mean; whereby the transit time corresponds to the point at which the sum of the absolute deviation of the means exceeds the fixed percentage.

26. A method according to Claim 25, in which the fixed percentage is about 5 %.

27. A computer programme for processing the received at least one ultrasonic signal from a transducer separation of up to about 20 m into fluid velocity and direction data.