ACCURATE DISPLACEMENT MEASUREMENT BASED ON THE FREQUENCY VARIATION MONITORING OF ULTRASONIC SIGNALS

Transcription

1 XVII IMEKO World Congress Metrology in 3rd Millenniu June 22 27, 2003, Dubrovnik, Croatia ACCURATE DISPLACEMENT MEASUREMENT BASED ON THE FREQUENCY VARIATION MONITORING OF ULTRASONIC SIGNALS Ch. Papageorgiou and Th. Laopoulos Electronics Lab. Physics Dept., Aristotle University of Thessaloniki, Thessaloniki, 54124, Greece. Abstract - This paper introduces a different approach to the easureent of the tie-of-flight of ultrasonic signals. Frequency variation onitoring and recording is used to deterine accurately the arrival tie of the ultrasonic signal. A high speed Digital Signal Processor (D.S.P.) is used for both: transission and direct easureent of the frequency of the incoing signal in every single period and with an accuracy of about 0.1%. The proposed configuration offers sall size and low cost solution to displaceent easureents with a rearkable perforance in ters of accuracy, range and easureent tie. Keywords: ultrasonic transducers, frequency variation, displaceent easureent 1. INTRODUCTION Ultrasonic transducers are widely used in a variety of applications and in technological areas with significant differences in perforance characteristics, operating environent, specifications, etc. Ultrasonic instruentation and easureent systes can be found in applications varying fro distance easureents to non-destructive testing of aterials, edical iaging and robot navigation. In a large nuber of these applications the easureent technique is facing actually a tie easureent proble. The ain effort is to estiate accurately the propagation delay of the ultrasonic wave-front as it travels fro the transitter to the receiver, or, as it is usually described the "tie-of-flight" of the ultrasonic wave. Ultrasound based easureent of distance, etc, has been recognized for soe years now as the ost siple and inexpensive answer to typical applications of non-contact distance easureent. There are actually only two ain drawbacks for this ethod; the poor resolution and the rather strong sensitivity to teperature (0.17%/ C in the air). The variety of applications of the ultrasonic tie-of-flight easureent has been increased over the years including nowadays various high perforance areas like robot navigation, cheical analysis and non-destructive testing. As a result of the stronger requireents for the perforance of the systes used in these applications, certain new techniques have been proposed, aiing to iprove accuracy and reliability of easureents and to diinish the easureent tie. These techniques include, besides the classical threshold coparison (known as pulse echo threshold ethod), other, ore coplicated, approaches like frequency odulation, ultiple frequencies transission, cross-correlation and neural coputing [1-5, 7, 8, 10]. Most of the are based on coplicated calculations or procedures perfored by autoated coputerized systes. Although ore accurate these techniques are also ore expensive and certainly less flexible due to the coplicated hardware and/or software involved. Hardware configuration is usually based on PCs or on single chip icrocoputers (icro-controllers) since they offer versatility in ipleenting different easureent techniques and low cost. Yet, in ost cases the icrocontroller is used rather to ipleent a certain processing algorith for increased accuracy, than to easure the actual tie delay accurately. This is ainly because of the relatively low frequencies of the clock of icro-controllers. A different approach is reported in this work; A high speed DSP icro-controller is used to easure directly the frequency of the incoing signal and this can be done in a fast and accurate way (within one single period and with an error of 0.1%). The proposed syste is presented along with illustrative exaples of the possible capabilities for analysis of the incoing signal and accurate easureent of the tie-of-flight of the ultrasonic waves. 2. THE PROBLEM The classical ethod for the easureent of the propagation delay (tie-of-flight) of the ultrasonic signals is the well-known pulse echo threshold ethod. A short burst of pulses is driven to the transitter and then the received signal (echo) is copared to a certain aplitude level. The distance between the transitter and the object is evaluated fro the equation: d 1 = υ sound t flight (1) 2 This is the ost siple and low cost ipleentation of an ultrasonic ranging syste suffering of course fro poor resolution. There is a nuber of good reasons for this drawback, besides the expected (ore or less) noise probles, and the inherent delay in the response of the transducers. A ajor liitation is caused by the fact that if one single period of the wave-front is not detected (issed due to aplitude attenuation or for any other reason), this

2 results to an error of about 8.5. A siple and coon iproveent is to eploy a dual coparator (detecting on both positive and negative parts of the incoing signal) in order to reduce the possible iniu error to a sei period (or about 4.5). This error is ainly caused by the variable attenuation of the aplitude of the incoing signal (depended on the distance to be easured). This is an inherent proble of the ethod, since the detection technique is based on aplitude coparison. In fact, the received echoes reach the threshold level soe tie after the exact beginning (added delay), aking the target to appear slightly farther away than it actually is. This error could be easily avoided if the added delay was constant, but aplitude changes produce deviations. To quantify this error it is useful to odel the echo wavefor as a daped sinusoid: n ht V ( t) V0t e sin( ω t + ϕ) (2) In this atheatical odel [4], n, h, ϕ, ω are transducer dependent paraeters, ω is the undaped angular resonance frequency of the transducer. On the other hand, echo aplitude change with distance x due to bea spreading and attenuation: αx e V ( x) = V 0 (3) x where α is the coefficient of attenuation in air. It has been suggested that, if the threshold level was ade variable atching (3) with x = υ t, the echo sound produced by a target at different distances would give constant added delays. However, there are other causes of echo aplitude variations which cannot be easily odeled, such as the size and orientation of targets. Therefore, when large echo aplitude variations are to be expected, the error due to the added delay for which no copensation can be ade, is half the rising tie Tr of the wavefor. This error is usually larger than a wavelength and for typical transducers operating in the range of 40 KHz it can reach 1c. Nuerous other techniques have been reported in the literature, aiing to iprove the reliability and accuracy of the tie-of-flight easureent. Our approach is that with the increased capabilities of odern high-speed icrocontrollers, it is possible to onitor accurately the desired characteristics of the incoing signal, and therefore there is no longer need to eploy coplicated ethodologies. In order to present this approach we shall describe first briefly the basics of the proble; the task is to easure the propagation tie of the ultrasonic wave. Now the ain proble is, that although we know exactly the starting tie (transission), we do not have a clear (reliable) way to identify the arrival tie. This is the critical point of the whole easureent procedure; to eploy a ethod that offers a clear triggering based on a characteristic (any characteristic) of the incoing signal. The identification of the arrival needs not to be done within a given period (i.e. the first or the second of the received signal), what is really needed is the reliable identification of the event. One way to iprove detection reliability is for exaple to try to drive the transitter-receiver syste with such a wavefor so that a clear peak appears at the output. This ethod, proposed by Grialdi and Parvis [6], results to a considerable iproveent in the identification procedure and consequently to an iproved accuracy when there are no large variations of the easured distance. Yet, there are two ain drawbacks in this ethod: a) it is also based on aplitude coparison (so it suffers fro the attenuation effects) and b) there is a need for a specialized (custo designed - arbitrary) driving signal, which requires a certain hardware. The aplitude coparison ethod should be avoided if possible, since there is a strong aplitude attenuation effect present in the syste. Another very interesting approach is to ignore practically the aplitude of the incoing signal and to try to identify the arrival event by detecting a frequency change. This is a proising technique since it is insensitive to aplitude attenuation and noise probles. Moreover, as it will be shown in the experiental results, the transducers are responding faster (or at least with the sae speed) to frequency changes than to aplitude changes. The driving signal is a coonly used short burst of pulses of two different frequencies in a Binary Frequency Shift-Keyed ode (BFSK). Frequency changes are detected by onitoring the phase of the incoing (received) signal versus tie. This ethod was originally used in the ultrasonics field by Webster and is known as phase digitizing [5]. All ajor drawbacks of the other ethods do not appear in this one, with the exception of the tie needed to identify the frequency change. Our approach is essentially the sae as the previous one, with one significant difference: the considerable reduction of the easureent tie. The proposed configuration is ipleenting a technique for fast easureent of the frequency value (within one single period). It is onitoring the frequency changes in a fast and accurate way, and consequently capable of identifying the arrival event in the sae way. Moreover, it can be used as a tool for experiental analysis of the overall syste (transducers, etc) in the frequency doain. The DSP icro-controller is used to easure directly each period of the received signal and convert the period value to frequency value. Frequency onitoring of the incoing (received) signal is used to identify the arrival event and to ark the arrival tie accurately. 3. THE SYSTEM The configuration of the proposed syste is based on the capabilities of accurate tie easureent of odern icrocontrollers. The usual series of icrocontrollers can not be used in this application ainly because of their relatively low frequency of operation (clock frequency) which affects the accuracy of tie easureent within one single period. They can not offer the required fast and accurate frequency easureent. A high perforance syste ay therefore be built only on a ore powerful icrocontroller. Larger systes (personal coputer type, etc) are avoided for practical reasons; the overall easureent syste should be cost-effective and sall sized. The PSP56002 single chip

3 digital signal processor sees to offer an excellent solution. This unit is a 24 bit, 80MHz DSP icro-controller with internal RAM (adequate for this application). The tiing easureent technique eployed by the DSP56002 is based on a free running 24bit counter driven by a 40MHz clock (counter's clock is half of the icrocontroller's noinal frequency of 80MHz) as shown in fig. 1. At the beginning of each period a 24 bit Tier Control/Status Register (TCSR) generates an interrupt which loads the content of the free-running counter to a 24bit Tier Count Register (TCR). ultrasonic signal. The proposed practical configuration of the easureent syste is shown in fig. 3. Fig.3. Experiental configuration of the proposed easureent syste Fig. 1. DSP56002 tier syste The user's progra can read the TCR and subtract consecutive values of the counter to deterine the tie between two events (i.e. fro the beginning of a period of the received signal till the beginning of the next one, that is the exact tie of a single period). As we can see in fig. 2, if T IN is the value of period, f the value of frequency of this period, C the nuber of the counted pulses by the counter of the tier and T clock the period of the tier clock, then: T = C T (4) IN clock and the frequency value ay then be derived fro the nuber of clock pulses counted within one period: f = f = f = fclock (5) T C T C IN clock Fig 2. Measureent process of the signal period The accuracy of this easureent is depended on the clock frequency (or period) value, which for our case is 40MHz (or 25nsec). This value results to an accuracy of 0.1% for the easureent of the frequency of the 40 khz The icrocontroller is driving the ultrasonic transitter with a precisely deterined pulse train. This eans that the frequency (or actually the period) of each consecutive pulse sent to the transitter can be controlled, along with the total nuber of the pulses and the aplitude of the pulse train. At the other end, an analog interfacing circuit is eployed to provide the sae icrocontroller with the appropriate input inforation (about the received ultrasonic signal). At the input, right after the ultrasonic receiver there is an analog signal conditioning section, which is fored by a controlled aplifier and a high pass filter (used to reduce possible low-frequency interference). Then the input signal enters two different paths: a frequency easureent path with a specially designed high-speed coparator, and an aplitude easureent path with a axiu value circuit. Fro the first path the icrocontroller gets the inforation about the duration of each consecutive period and fro the second the digital data corresponding to the value of the aplitude for this period. As shown in fig. 3, the basic analog circuit is a high-speed coparator, which triggers the tiing input of the icrocontroller at the exact tie instants of the beginning of each period of the received signal. The easureent of the aplitude of each period of the incoing signal is not actually used in the described application for the reasons entioned earlier in this work (attenuation). It has been included in the syste though for the following two reasons ainly: the aplitude easureent section is needed for other options of the sae syste like the autoated calibration and self-testing procedure. the inforation about the aplitude is used by this syste to validate the frequency easureents. This eans that this inforation is helpful for the estiation of the tie window when an input signal is detected and therefore frequency analysis is useful. It should be noted that it has no effect on the tie-of-flight easureent, and it is not needed in practical applications where a certain frequency variation pattern (odulation) is applied. It is not necessary, but very useful as an additional tool for the experiental analysis of the transducer-receiver setup.

4 On the aplitude easureent path, a precision half wave rectifier followed by a Saple-and-Hold circuit drives the A/D converter with the axiu value of each period (that is the aplitude value of each period). The accuracy of the aplitude easureent is ainly depended on the A/D converter, which is a 10bits parallel output converter with 900nsecs ax conversion tie and 800KHz sapling rate. The final figure for the accuracy of the easureents of the aplitude of each period of the incoing signal is about 5V (or 0.1%). the two levels), along with the capability of the syste to onitor sudden changes of frequency very quickly. 4. EXPERIMENTAL RESULTS The practical configuration of fig. 3 has been experientally built and operated via a serial interconnection with a personal coputer. The frequency values were teporarily stored in the icrocontroller's eory (along with all other required inforation, i.e. tie) and transferred right after the end of each easureent to the PC for analysis. There is actually a serial bidirectional, counication between the icro-controller and the personal coputer. The PC is not needed for the easureents but it is used as a user-friendly interface to control the easureents sequence and analyze the output inforation. In a practical application all tasks can be perfored by the icro-controller and the output inforation (the value of the tie-of-f light or the distance) can be either directly displayed locally, or transferred to any other coputerized syste. The experiental syste has been tested first with an input signal of 40kHz coing directly (without ultrasonic transducers) fro a high perforance generator (HP3325B). The output inforation as shown in fig. 4 and fig. 5 illustrates syste's frequency easureent capabilities. Fig. 5. Frequency response of a triangular frequency odulated signal fro a generator. Fig. 6. Frequency variations of a BFSK odulated signal, where the frequency step is below the resonance frequency of the transducers. Fig. 4. Frequency response of a BFSK odulated signal fro a generator. Fig. 4 shows the easureents of a frequency odulated input varying in a binary ode fro 39kHz to 41kHz (BFSK), while fig. 5 shows a triangular frequency odulated input in the sae range. The expected error of 0.1% (40Hz) can be seen in this figure (as the fluctuation of the easured value on Fig. 7. Frequency variations of a BFSK odulated signal, where the frequency step is above the resonance frequency of the transducers.

5 The easureent syste is then tested with a transducer pair placed at a distance of about 1 inside a 12c diaeter anechoic tube (with practically no reflections). Figures 6-8 show three different cases of frequency variations of a BFSK signal onitored by this syste. The frequency step is the sae in all cases (1kHz), but the higher and lower frequency value is different in each of the; in fig. 6 f HIGH is equal to the resonance frequency of the pair (40.9kHz) and f LOW is 1kHz less (39.9kHz), in fig. 7 f LOW is equal to the resonance frequency and f HIGH is 1kHz higher (41.9), and in fig. 8 the resonance frequency is in the iddle of the frequency step. Fig. 9. Frequency variations of a triangular frequency odulated received ultrasonic signal. Fig. 8. Frequency variations of a BFSK odulated signal, where the resonance frequency of the transducers is within the frequency step. Fro these results it is apparent that the ultrasonic transducers do not respond directly to frequency changes and their behavior is strongly non-linear. There are certain phenoena that should be considered, if frequency variations are to be used for tiing (arrival tie easureent) as in the case discussed in this work. Different high and low frequency values of a BFSK signal have been used as an exaple of the different variations that occur, in order to illustrate the iportance of both requireents of the frequency easureent technique: speed and accuracy. The ain drawback of the phase digitizing ethod (large easureent tie) is apparently causing a drastic liitation on the accuracy of the tiing easureents. Fig. 9 shows the easured frequency variations for a triangular (linear) frequency odulation of the driving signal. The non-linear behavior of the transducers can be also seen in this case, but there are certainly soe clearly identifiable points (peaks, slopes, etc). Fast and accurate frequency onitoring offered by the proposed ethod offers therefore the necessary inforation in order to detect a certain event and easure the arrival tie of the ultrasonic wave. The first experiental tests for this kind of easureents (the tie between the transitting and the receiving of a peak of a triangular frequency odulated pulse train) have shown a rearkable accuracy in the easureent of displaceent, which reaches 0.03 under constant teperature conditions, or under teperature copensation. This value reached the accuracy offered also by the echanical displaceent setup of the experiental syste (0.02) and it should be noted here that this result is a directly obtained easured value - not a coputer calculated statistical figure. The above entioned result (of 0.03) that easured approaches the 75ns axiu overall theoretical error of the syste: 50ns for the transission and 25ns for the reception of the ultrasonic signal. This theoretical value corresponds to a displaceent accuracy of CONCLUSION The syste proposed in this work is using the fast easureent of frequency of the received ultrasonic signal as an identification ethod for the accurate estiation of the tie-of-flight. It follows fro the description of the proble that the distance easureent technique based on ultrasonics is actually facing the proble of deterining the arrival tie of the ultrasonic wave-front accurately. Frequency odulation techniques are a good way to use for identification and fast frequency onitoring ust be used for the ipleentation of this ethod. Modern DSP icrocontrollers have now the appropriate clock frequencies to easure ultrasonic signals directly, accurately and within one single period. These systes therefore can easure the tieof-flight (propagation delay of the ultrasonic wave-front) directly and accurately without the need of sophisticated ethodologies. The experiental results presented here indicate the novel capabilities of analysis offered by this syste and the accuracy of the tie-of-flight easureents.

6 REFERENCES [1] W.L. Anderson and C.E. Jensen, "Instruentation for tieresolved easureents of ultrasound velocity deviation" I.E.E.E. Transactions on Instruentation, and Measureents, vol. 38, n.4, pp , Aug [2] M.G. Duncan, "Real-tie analytic signal processor for ultrasonic non-destructive testing", I.E.E.E. Transactions on Instruentation, and Measureents, vol. 39, n.6, pp , Dec [3] D. Marioli, C. Narduzzi, C. Offelli, D. Petri, E. Sardini and A. Taroni, "Digital tie-of-flight easureent for ultrasonic sensors", I, I.E.E.E. Transactions on Instruentation, and Measureents, vol. 41, n.1, pp , Feb [4] Canhui Cai and Paul P. L. Regtien, Accurate Digital Tie of flight Measureent Using Self-Interference, I.E.E.E. Transactions on Instruentation and Measureents, vol. 42, pp , Deceber [5] D. Webster, "A pulsed ultrasonic distance easureent syste based upon phase digitizing", I.E.E.E. Transactions on Instruentation and Measureents, vol. 43, n.4, pp , Aug [6] U. Grialdi and M. Parvis, "Enhancing ultrasonic sensor perforance by optiization of the driving signal", Measureent, vlo.14, pp , 1995 [7] P. Daponte, F. Maceri and R.S. Olivito, "Ultrasonic signal processing techniques for the easureent of daage growth in structural aterials", I.E.E.E. Transactions on Instruentation and Measureents, vol. 44, n.6, pp , Dec [8] A. Carullo, F. Ferraris, S. Graziani, U. Grialdi and M. Parvis, "Ultrasonic distance sensor iproveent using a two level neural network", I.E.E.E. Transactions on Instruentation, and Measureents, vol. 45, n.2, pp , April 1996 [9] Th. Laopoulos and Ch. Papageorgiou, "Microcontroller based easureent of angular position, velocity and acceleration", I.E.E.E. Transactions on Instruentation, and Measureents Technical Conference (IMTC'96), Brussels, Belgiu, 1996 [10] F. Guening, M. Varlan, C. Eugene and P. Dupuis, "Accurate distance easureent by an autonoous ultrasonic syste cobining tie-of-flight and phase-shift ethods", I.E.E.E. Transactions on Instruentation, and Measureents Technical Conference. (IMTC'96), Brussels, Belgiu, 1996 [11] Ch. Papageorgiou, C. Kosatopoulos and Th. Laopoulos, "Autoated characterization and calibration of ultrasonic transducers", Mediterranean Electr. Conference. (MELECON '98), Tel-Aviv, Israel, 1998 [12] Gilles Mauris, Eric Benoit, Laurent Foulloy, Local Measureent Validation for an Intelligent Chirped-FM Ultrasonic Range Sensor, I.E.E.E. Transactions on Instruentation and Measureents, vol. 49 pp , August 2000 [13] Heinrich Ruser, Valentin Magori, Hans-Rolf Trankler Correlated icrowave-ultrasonic ulti-sensor for reliable easureents of velocity and range, Proceedings of the 19th IEEE Instruentation and Measureent Technology Conference, May 2002, vol. 1, pp Authors: Chris Papageorgiou, Electronics Lab. Physics Dept., Aristotle University of Thessaloniki, Thessaloniki, 54124, Greece, phone: , fax: , e-ail: papageorgiou@physics.auth.gr. Theodore Laopoulos, Electronics Lab. Physics Dept., Aristotle University of Thessaloniki, Thessaloniki, 54124, Greece, phone: , fax: , e-ail: laopoulos@physics.auth.gr