B4.2 High Precision Wireless Measurement o Temperature by Using Surace Acoustic Waves Sensors Leonhard Reindl 1), Ismail Shrena 1), Harald Richter 1), Reto Peter 2) 1 ) Clausthal University o Technology, IEI, Leibnizstr. 28, D-38678 Clausthal-Zellereld, Germany 2 ) Baumer Electric AG, Hummelstrasse 17, CH-851 Fraueneld, Switzerland Abstract Surace acoustic wave devices can be used as wireless sensor elements (SAW transponders) or measuring physical quantities such as temperature, that do not need any power supply and may be accessed wirelessly. The complete wireless sensor system consists o such a SAW transponder and a local radar transceiver. An RF burst in the VHF/UHF region transmitted by the radar transceiver is received by the antenna o the SAW transponder. The passive transponder responses with an RF signal - like a radar echo - which can be received by the ront-end o the local transceiver. The amplitude and phase o this RF response signal carry inormation on the positioning Surace acoustic wave (SAW) devices can be used as sensor elements that do not need any power supply and may be read out wirelessly. The reader unit perorms a radar measurement o the impulse response o the SAW transponder via a high-requency electromagnetic radio link. A temperature variation changes the SAW velocity and thereby the response pattern o the SAW device. By analyzing the time delay between backscattered pulses with dierent time delays we get a coarse estimation o the temperature o the SAW transponder. By using this inormation the ambiguity o ±2π in the phase dierences between the pulses can be eliminated, which enables to an overall temperature resolution o ±.2 C. Index terms Temperature sensor, passive sensors, surace acoustic waves, transponder, radio waves Surace acoustic wave (SAW) devices are special micro acoustic components consisting o a piezoelectric substrate with metallic structures such as interdigital transducers (IDTs) and relection or coupling gratings deposited on its plain-polished surace. Due to the piezoelectric eect an RF input signal will stimulate a microacoustic wave propagating on the surace o the substrate. Figure 1 shows a SEM photo o two SAW pulses, which were emitted by an IDT. Vice versa, a SAW wave generates an electrical charge distribution at the receiving IDT and thereore an electrical RF output signal occurs [1]. SAW technology has been exploited or electronic analog signal processing over the past 3 years, with the development o a tremendous amount o devices and systems or consumer, commercial and military applications running now at some billion dollar annual rate. Many piezoelectric single crystal substrates, like LiNbO 3 or LiTaO 3 show a non-vanishing temperature coeicient o delay. Hereby, a temperature variation strains the SAW chip and also changes the SAW velocity by inluencing the elastic constants o the crystal substrate. Thus, by careully analyzing the impulse response o a given SAW device on these materials, one can determine the temperature o the SAW chip. By using a SAW identiication (ID) tag [2], [3] this SAW based sensor technique can be implemented with sensor elements (transponders) that do not need any power supply and which are connected to their reader unit solely by a wireless radio link [4]-[6]. I. INTRODUCTION Bus bar Figure 2 shows the operating principle o such sensor systems: A local radar transceiver (TRx) used as reader unit sends out a radio requency (RF) electromagnetic read-out signal in the UHF/VHF range. This to be published at SENSOR3, 13. 15. Mai 23, Messezentrum Nürnberg 1 IDT SAW Fig. 1 SEM-photo o an interdigital transducer and two SAW pulses. (Photo by Siemens AG) Read out signal Reader unit Antenna Response signal Relectors IDT Piezoelectric single crystal Fig. 2 Schematic o the operating principle o a SAW-based radio-link temperature measurement system.
read-out signal is picked up by the antenna o the passive SAW transponder and conducted to an interdigital transducer (IDT). The IDT converts the received signal into a SAW signal by the converse piezoelectric eect. The SAW propagates towards several relectors distributed in a characteristic pattern. A small part o the wave is relected at each relector. The micro acoustic wave packets now returning to the IDT are re-converted into electrical signals by the IDT and re-transmitted to the radar TRx unit by the transponder antenna (see igure 3). The velocity o the micro acoustic wave and thereby also the time distances o the RF transponder response is aected by the temperature. The evaluation o this response signal in the radar unit thus allows the determination o the environmental temperature o the passive SAW transponder. Since such sensors need neither wiring nor batteries they Time [µs] can be placed advantageously on moving or Fig. 3 Time domain impulse response o the rotating parts and in hazardous environments such investigated SAW tag with ive relectors. as contaminated areas or high voltage plants, chemical or vacuum process chambers, under concrete, etc. On the other hand, SAW transponders cannot be addressed individually. Using time division techniques typically 5 to 1 dierent SAW sensors, each with 3-4 relectors, can be build up. In the ollowing contribution we will discuss in chapter II the local radar TRx and in chapter III the investigated SAW transponder. The used measurement set up will be explained in chapter IV, and in chapter V we will discuss the achieved measurement results. A short conclusion will end this contribution. II. READER UNIT The read out units o wireless SAW sensor systems resemble those used in traditional radar systems [7]. In Europe, only two requency bands suitable or SAW devices are allocated to unlicensed low power devices (LPDs) such as industrial, scientiic, or medical (ISM) apparatus: 433.7 434.77 MHz and 2.4 2.483 GHz. The allowed equivalent isotropically radiated power (EIRP) in these bands is P = 1 mw. For the investigation presented here, we used a commercial reader unit or SAW ID tags rom Baumer Ident (see igure 4). The sweep bandwidth B o 4 MHz o the transmitter Tx can be chosen between 242-247 MHz with a time resolution o 25 ns corresponding to the inverse o the used bandwidth. The system uses a requency modulated continuous wave (FMCW)-principle with a sweep time o 16 ms. Thus, the applied beat requency is given by 2,5 Hz/ns. The radiated output power (EIRP) can be chosen between,1 and 1 mw according standards FCC part 15 / I-ETS 344. The system obtains a read out distance o up to 6 m depending on the tag type and the reader antenna. Two serial interaces (RS 232 and RS 422) allow data transmission to an external PC or host system. In its normal mode the reader gives out the ID number o every tag, which passes the beam o the reader antenna. The reader, however, can also be operated in a service mode, in which it delivers the sampled real base band data o each measurement cycle. Applying a Fourier transormation to this data and scaling the result with the beat requency o the FMCW system gives the impulse response o the received signal. relative amplitude Fig 4 2.45 GHz SAW reader unit by Baumer Ident. III. SAW TRANSPONDER For wireless SAW transponders we used commercial ID-tags rom Baumer Ident. They use a pulse positioning coding scheme with ive relectors. Figure 5 shows the working principle o such a pulse position coded SAW tag. The local position o each individual relector and thus the corresponding time position o the relected signal can be chosen out o several slots. The tag we used utilizes 1 allowed slots or each relector, which gives a code range o 1 4 with our code relectors and one calibration relector. Similar types are currently - 2 -
Antenna Block #1: allowed positions slots or irst relector Block #2: allowed positions slots or second relector Antenna SAW tag inside o a oven Reader unit Relector #1 Relector #2 Fig. 5 Pulse position coding scheme or SAW tags. IDT LiNbO 3 single crystal Temperature controlling Computer or signal analysis Fig. 7 Used set up or the wireless temperature measurement. 12 9 6 15 3 Relector SMD Housing s Fig. 6 Photo o an assembled SAW tag by Baumer Ident. developed which oer a code range o 2 to 64 Bits [8]. The initial delay between the IDT and the irst relector is 1.2 µs, the coding time slots are 25 ns spaced. Figure 6 shows the mounted SAW chip in a SMD package. The Material o the piezoelectric single crystal is black Lithiumniobat, LiNbO 3 ; the propagation direction is rot 128. The corresponding propagation velocity v o the SAW is 396 m/s, which gives or 2.45 GHz a wavelength λ o.8 µm. The metallization is Aluminium, Al. The bidirectional IDT is placed near the centre o the chip with three relectors to the let and two to the right. The two bright lines parallel to the acoustic path are bus bars or the electrical wiring o the IDT to the pins o the SMD package. The package will be hermetically sealed to protect the surace o the chip and then connected to an antenna to inish the SAW transponder. IV. MEASUREMENT SET UP ϕ 18 For the temperature measurements we connected the Baumer reader unit to a RF antenna. The SAW tag was placed in the beam o the antenna inside o a temperature controlled oven. The electromagnetic waves passed through a small glass window o the oven. For temperature control an additional NiCr-Nithermocouple in good thermal contact to the SAW tag was used. The reader unit was connected to a computer or a detailed signal analysis using the RS 232 interace. Thereby the computer program could read out the sampled real base band data o each read out cycle and could also control all settings o the reader unit. For signal analysis we used the numerical tool MATLAB. The interace to the reader unit was established by the use o the MEX ile option o MATLAB, which allows the call o speciically developed C routines [9]. Figure 3 gives the time domain response signal o the investigated SAW tag. The irst and largest response starts at 1.2 µs. Then there are two well separated impulses ollowed by two interering signals. Finally we get several smaller spurious signals due to multi relections between the coding structures and relections by the chip edge. In the irst step we optimised the requency range o the reader system to ensure a perect base band adjustment o the mixed down Rx signals o the SAW tag. This is perormed i the transer unction o the antenna and the SAW relective delay line is symmetrically to the centre o the requency band o the Tx read out signal. In this case the mixed down SAW signal will be arranged symmetrically to zero requency. The requency adjustment can be done most easily by evaluating the time domain response in a polar diagram. Any shit in requency with respect to zero requency will result in a phase modulation within the peaks, which broadens the spikes. Figure 8 shows the obtained time domain data by using a read-out requency range starting with 2.42 GHz, which corresponds to a centre requency 21 24 ϕ 27 Fig. 8 Polar map o the SAW response signal in the base band measured with a FMCW signal starting with 242 MHz. 3 33-3 -
Read out o base band data and temperature; averaging Weighting, zero padding, Fourier transorming Pulse delay time calculation by using a parable approximation Coarse determination o the temperature by a evaluation o the pulse delay time dierences Fine determination o the temperature by a evaluation o the pulse phase dierences Delay time dierence 2-1 [ns] 256 255 254 253 252 251 25 25 5 75 1 125 15 175 2 Temperature output Fig. 9 Flow chart o the data analysis o a wireless passive SAW temperature sensor =2.422 GHz. For all urther investigations this latter requency range was used, because this setting resulted in the smallest spikes. There remained an intersymbol intererence between the signals rom relector 4 and 5. V. MEASUREMENT RESULTS Figure 9 gives a low chart o the used data analysis. The sampled 512 real base band data o each measurement cycle o the Baumer reader unit were red out via the RS 232 interace and transerred to Matlab using the MEX ile. The ambient temperature o the SAW tag was noted und controlled. For improving the signal to noise ratio an averaging over several reading cycles could be perormed. Previous to the Fourier transormation we applied a weighting unction to the data to minimize the spurious level in time domain. Due to its excellent side lobe level perormance we used a Blackman-Harris weighting unction, but this led to a huge pulse spreading o w=1.9 i compared to unweighted. Maybe the intersymbol intererence between the signals rom relector 4 and 5 could have been avoided by using a weighting unction with a smaller spreading, e.g. a Hamming unction. To improve the resolution we applied a zero padding scheme with additional 512 points beore perorming the Fourier transormation. The resulting point spacing ater the Fourier transormation t FFT then results in: 1 t FFT = 2 B To calculate the exact time position o each impulse we constructed a parable approximation through the local maximum o the amplitude and both adjacent points. The locus o the maximum o this parable Fig. 1 Variation o the delay time dierence 2-1 between the irst and second response as a unction o temperature. was taken or the exact time position o this impulse. Using this technique we got an enhancement in resolution o the peak position o about a actor o 1, which results in an accuracy o 1 ns. Because the distance between the reader unit and the SAW tag is unknown or may vary and thereby the time delay to the irst response impulse may be unknown or may vary too, we evaluated only the time dierence between two pulses, e.g. 2-1 = 2-1, as a unction o the temperature. The linear term o a Taylor series expansion usually describes the situation well enough: Delay time dierence 3-1 [ns] 537 535 533 531 529 527 525 25 5 75 1 125 15 175 2 Fig. 11 Variation o the delay time dierence 3-1 between the irst and third response as a unction o temperature Number o samples Dierence in temperature [ C] Fig. 12 Resulting error i the temperature is calculated solely by the delay time dierences. - 4 -
( ϑ) ( ϑ ) = ( ϑ ) 2 1 re TCD ϑ re, (1) where TCD is the well-known linear temperature coeicient o delay: TCD = 1 ϑ Figure 1 and 11 show the measured variation o the delay time dierences 2-1 and 3-1 between the irst and second, and the irst and third impulse as a unction o temperature. The gradient o the lines o best it are with 7,8 and 7, ppm/ C in excellent agreement with the values rom literature. In both igures a periodical variation in the delay time dierences occurs, which can be aected by varying the tag position inside the oven. Thereore, these variations are likely caused by multipath relections inside the metallic chamber o the oven and the signal to noise ratio o the Rx data. Figure 12 shows the resulting inaccuracy when we try to calculate the temperature solely with the help o the delay time dierence. We get an accuracy or the temperature o σ=1 C by evaluating 3-1. The resolution can be increased signiicantly by evaluating the phase dierences precisely: ϕ =. (2) 2 1 2π Again the temperature dependence is given by: ϕ 2 1 ( ϑ) ϕ( ϑre ) = TCDϕ ( ϑ ϑ re ), (3) where TCD ϕ is now the slightly dierent the linear temperature coeicient o phase. The evaluation o the phase dierences, however, induces an ambiguity when the phase shit exceeds 36. Figure 13 shows the variation o the phase dierence ϕ 2-1 between the irst and second response as a unction o temperature. We use the coarse estimation o the temperature resulting rom the evaluation o 2-1 and 3-1 to overcome this equivocation o the phase evaluation. To determine the absolute phase ϕ 2-1 within 2π the accuracy 2-1 o the delay time 2-1 must be: ϕ ϕ 2π = 2π By demanding ϕ 2-1 <2π we get: 2 1 1 < = (4) (5) Situation becomes a little bit easier i we use a long -18 delay time dierence 3-1 (with a small relative error) or the coarse determination o the temperature and a short delay time dierence 2-1 (with a large unambiguousness) or the irst phase evaluation. This leads to a iner determination o the temperature, which allows the phase evaluation o longer delay time dierences. The required accuracy 3-1 then becomes to: 1 < (6) The relative accuracy in terms o the point spacing ater the Fourier transormation t FFT (including the doubling o the points due to the zero padding) then is given by: ϕ 2-1 18 12 6-6 -12-18 5 1 15 Fig. 13 Variation o the phase dierence ϕ 2-1 between the irst and second response as a unction o temperature. ϕ5-ϕ4 18 12 6-6 -12-18 5 1 15 2 Fig. 14 Variation o the phase dierence ϕ 5-4 between the ith and ourth response as a unction o temperature. 18 ϕ3-1-2 ϕ2-1 12 6-6 -12 25 5 75 1 125 15 175 2 Fig. 15 Variation o the weighted dierence o two phase dierences ϕ 3-1-2 ϕ 2-1 as a unction o temperature. - 5 -
2 < t FFT B. (7) I we assume an accuracy o 3-1 = t FFT /5 by using the parable approximation o the maximum, we get: < t FFT 2 B 1 B (8) I we insert the data o the Baumer SAW reader system, we get a minimum ratio between the longer and the shorter delay time dierence o 6. Thereore a very short delay time dierence would be advantageous. On the other hand we do need a minimum spacing to avoid intersymbol intererence. Figure 14 shows the corresponding variation o the phase dierence between the pulses #5 and #4 o the RF response as a unction o temperature. The intersymbol intererence between these two signals destroys the monotonic relation between temperature and phase dierence. A temperature determination using this phase dierence is thereore no more possible. However, we can construct mathematically a very short delay time dierence by analysing ϕ 3-1 -2 ϕ 2-1. Since the delay time dierence between signal #2 and #1 is 25ns and between signal #3 and #1 is 525 ns, we get one single time slot or the weighted dierence: ϕ 3-1 -2 ϕ 2-1 =25ns Figure 15 gives the temperature characteristics or this very short time dierence. The temperature range o unambiguousness is more than 2 C. Adding this temperature inormation to the coarse estimated value rom the delay dierence inormation we can eliminate one ater the other the ambiguity in phase dierences o ϕ 2-1 and ϕ 3-1. Using this technique we get the continuous phase dierence ϕ 3-1 between the irst and third response (see igure 16), which allows us to determine the temperature with a resolution o ±.2 C. Figure 17 gives the resulting temperature error by using the continuous phase dierence ϕ 3-1. IV CONCLUSION By combining the principles o conventionally wired SAW sensors with the radio-request technique, known rom SAW ID tags, passive sensors read out solely by a radio-requency link can be developed. Based on a commercial read out system developed or identiication applications and the corresponding SAW transponder, we build up a wireless measurement system or temperature. Using a precise evaluation o the delay time dierences between the responded signals we could overcome the limited temperature range o the phase evaluation. The over all measurement accuracy is in the order o.2 C within a temperature range o more than 2 C. The read out distance is up to 6 m. ϕ3-1 [ ] 8 6 4 2 25 5 75 1 125 15 175 2 Fig. 16 Variation o the continuous phase dierence ϕ 3-1 between the irst and third response as a unction o temperature. Number o samples Dierence in temperature [ C] Fig. 17 Resulting error i the temperature is calculated using the continuous phase dierences between the pulses. REFERENCES [1] R.M. White, F.W. Voltmer, "Direct piezoelectric coupling to surace elastic waves, " Appl. Phys. Lett., Vol. 7, pp. 314-316, 1965 [2] P. A. Nysen, H. Skeie, D. Armstrong, "System or interrogating a passive transponder carrying phase-encoded inormation," US Patents 4 725 841, 4 625 27, 4 625 28, 1983-1986 [3] L. Reindl, W. Ruile: "Programmable Relectors or SAW-ID-Tags", in Proc. o the 1993 IEEE Ultrasonics Symp., pp. 125-13 [4] X. Q. Bao, W. Burkhard, V. V. Varadan, V. K. Varadan, "SAW Temperature Sensor and Remote Reading System," in Proc. o the 1987 IEEE Ultrasonics Symp., pp. 583-585 [5] L. Reindl, G. Scholl, T. Ostertag, H. Scherr, U. Wol, F. Schmidt, "Theory and application o passive SAW radio transponders as sensors," IEEE Transactions on UFFC, Vol. 45, No. 5, Sep. 1998, pp. 1281-1292 [6] L. Reindl, A. Pohl, G. Scholl, R. Weigel, "SAW-Based Radio Sensor Systems, Sensors Journal, IEEE, Volume: 1 Issue: 1, Jun 21, Page(s): 69-78 [7] M. I. Skolnik, Introduction to Radar Systems. New York etc.: McGraw Hill, 1979. [8] R. Peter and C.S. Hartmann, Passive long range and high temperature ID systems based on SAW technology, Session A 3.2, Sensor 23, Nürnberg, May 23. [9] M. Kunz, S. Schlatter. Passive, Contactless Temperature Sensor, Second term paper, Hochschule Rapperswil. Schweiz. 1999. - 6 -