A BROADBAND SPECTROSCOPIC SENSOR PROBE

Transcription

1 A BROADBAND SPECTROSCOPIC SENSOR PROBE I M Woodhead 1, I Platt, J H Chistie, S Kenek 1 Lincoln Ventues Ltd, Lincoln Univesity Cantebuy, New Zealand. Woodhead@lvl.co.nz Abstact- The electical popeties of many mateials ae closely elated to thei composition and to thei moistue content in paticula. Fo senso development, chaacteising the esponse of a paticula volume of mateial ove a boad fequency ange is desiable, since sepaate measuements could geneate eos due to spatial vaiability. A coaxial pobe has been designed fo measuement of the pemittivity of smooth and flat, solid o liquid samples ove the fequency ange fom 1 Hz to 6 GHz. Although the pobe is capable of a vey wide fequency ange, sepaate instuments ae geneally equied, and hee we focus on measuements above 1MHz. We demonstate measuements in the fequency domain using a vecto netwok analyse, and in the time domain using a boadband oscilloscope. Fo switching, we employed a coaxial switch and demonstate how that is included within the instument calibation. Calibation of the pobe used thee efeences: an open cicuit, shot cicuit (indium foil) and a efeence liquid, ideally chosen fo a pemittivity simila to that of the sample. The sample complex pemittivity was calculated by a numeical model which used as inputs, the measued eflection coefficient and physical measuements of the pobe geomety. Index tems: coaxial pobe, spectoscopy, calibation, boadband, netwok analyse. I INTRODUCTION The daiy, timbe, wool, and many othe pimay pocessing industies demand apid on-line measuement of poduct popeties, paticulaly moistue. The chief advantage of the dielectic measuing technique is its elatively low cost compaed with neuton backscatte, X-ay analysis, and magnetic esonance imaging. Dielectic techniques ae also non-ionising, and offe clean, apid, on-line measuements of bulk mateial, suited to many divese poducts. Fo a paticula poduct, thee will be one o moe fequency anges whee the dielectic constant has the highest sensitivity to changes in moistue content (o othe 459

2 I M WOODHEAD, I PLATT, J H CHRISTIE, S KRENEK, A BROADBAND SPECTROSCOPIC SENSOR PROBE measuable paametes). Thee ae vaious dielectic sensos aleady available, but in many cases the sensitivity of the senso is not matched to the dielectic popeties of the mateial being measued. It is also possible that if pemittivity specta wee measued fo seveal simila samples of known composition, the paametes descibing the dielectic specta could be coelated with the sample composition. Hee we descibe methods to measue the dielectic popeties of vaious mateials ove a boad fequency ange, which may be used to investigate appopiate fequency bands to enable custom dielectic sensos to be tailoed to the popeties of the mateials being measued. Changes in moistue content and the composition of vaious components within composite mateials such as daiy poducts, wood and masony, alte thei dielectic popeties. The electical popeties of the mateials ae paticulaly sensitive to changes in moistue content paticulaly the low fequency, low moistue content conductivity [1] but also the high fequency pemittivity due to the lage diffeence in pemittivity between fee wate, with a static elative pemittivity of appoximately 8 at 2 C, and many othe mateials which ae geneally less than thee [2]. The chosen measuement fequency fo detemining moistue content o composition may lie within a vey boad ange. Reasons fo choosing a paticula fequency will include sensitivity of the eading to changes in moistue content and the dependence of the eading on othe paametes such as density and dielectic loss. Fo a paticula pola molecula species, as the fequency is inceased, the dielectic loss eaches a maximum and is accompanied by a eduction in the pemittivity. The fequency of maximum loss is the elaxation fequency, and fo fee wate begins to significantly affect the dielectic popeties above appoximately 3 GHz. In contast, the elaxation fequency fo bound wate is typically 1 MHz [3]. Electical conductivity is diectly useful at low fequencies, paticulaly fo mateials with low moistue content whee it povides a vey useful measue of moistue content, and at high fequencies whee it influences the dielectic loss. Apat fom the Maxwell-Wagne effect which is caused by conducting inclusions in a dielectic, the pedominant contibution to dielectic loss at fequencies below 1 MHz is fom conductivity. In dielectics without conducting inclusions, it is desiable to detemine the contibution to the dielectic loss made by conductivity, and subtact this contibution fom the loss at highe fequencies. To measue all the above effects, a pobe suitable fo a ange of mateials would ideally include measuements at a numbe of fequencies ove the ange 46

3 fom below 1 Hz to seveal GHz. Vey low fequencies, typically less than 1Hz, povide infomation about pocesses such as cuing development in concete [4] and ion dift [5]. The appoach used hee has been to employ a coaxial pobe and to incopoate a coaxial switch to allow two measuement means, an impedance analyse o automatic bidge fo fequencies up to appoximately 1 MHz, and a vecto netwok analyse o a time domain eflectomete fo highe fequencies. At high fequencies, the complex eflection coefficient fom the open-ended coaxial pobe depends on the electical popeties of the impedance at the end of the pobe. In this case, the sample mateial teminates the line, and its popeties ae mioed in the eflection coefficient. Commonly, measuements ae made at a numbe of discete fequencies within the desied fequency ange, and then afte some mathematical pocessing (e.g. [6]), the complex pemittivity of the sample mateial is obtained. At lowe fequencies, the pobe and connecting cables ae shot in compaison with the wavelength, so the pobe chaacteistics ae also influenced by those of the connecting cable. A convenient change-ove fequency between the two instuments is 1MHz since this typically epesents the uppe fequency limit of an automatic bidge that could be used to measue low fequency conductivity and pemittivity. Fo measuement at low moistue content, e.g. below fibe satuation in textiles, the conductivity is vey small so specialist techniques such as guading ae usually equied. In this pape, we descibe the pobe design and how measuements may be obtained in the fequency o time domain, and show measuement esults. II EXPERIMENTAL DETAILS The pobe was fabicated fom MPC 14 mm igid coaxial ailine (Mauy Micowave Copoation, Ontaio, Califonia, USA) athe than using a solid dielectic, to avoid themal effects induced by the diffeence in themal expansion coefficient between teflon and bass (o coppe). Use of a teflon dielectic ceates difficulties in obtaining a sufficiently flat face at the tip of the pobe [7]. The design of the pobe is shown in Figue 1 whee the inne and oute diametes of the insulato ae 6.24 mm and mm. The adial dimensions of the tansmission line dictate that the tansvese electic (TE) and tansvese magnetic (TM) tansmission modes ae not sustained within the line (the cutoff fequency fo the TM 1 and TE 1 modes is 26 GHz. A gound, stainless steel flange was soldeed to the oute (gold ove coppe) conducto. The 6 mm diamete of the flange was chosen in accodance with the 461

4 I M WOODHEAD, I PLATT, J H CHRISTIE, S KRENEK, A BROADBAND SPECTROSCOPIC SENSOR PROBE ecommendations of [8] and [9] so that it was a sufficiently accuate appoximation to an infinite gound plane. An insulating colla made fom coss-linked polystyene was used to locate the cente conducto, and povide a measuement plane between the conductos. This mateial was chosen fo its stability, machineability, low loss, well known dielectic constant of 2.53, and compatibility with acylic adhesives. The plastic colla was sufficiently long (15 mm) that the highe ode modes induced at the ai-plastic inteface inside the coaxial cable decayed (by a facto 1 8 ) befoe the pobe end, yet was still less than a half wavelength at 6 GHz. The colla was a pess fit, and was glued into place using Loctite 46 cyanoacylate adhesive to ensue a stable, watepoof gound plane. Finally, the gound plane was machined flat. Figue 1. Coaxial pobe fabicated fom ai line. The eflection coefficient of the pobe was measued between 1 MHz and 6 GHz using an HP8753D vecto netwok analyse (VNA). Calibation of the VNA and pobe is a two-step pocess. In pevious wok, we calibated the VNA at the beginning of the pobe using a coaxial open, shot and 5 Ohm coaxial load, then calibated the pobe using the pocess descibed by [6]. Hee we show that calibation of the VNA may be accomplished at the pobe tip by using an open, a shot in the fom of Indium foil pessed fimly against the pobe tip, and thidly a bespoke calibation boad. The boad compised a cicuit boad disc with concentic conductos that matched those of the pobe, with much of the fibeglass substate emoved between the two electodes, and incopoating two adially positioned 1 Ohm 126 size SMD esistos to fom a 5 Ohm load. Ideally, the esistos would be thin film micowave components, but hee we epot on esults using standad SMD components. 462

5 Within the semi-igid coaxial line that connected the pobe to the VNA, we included a Teledyne coaxial elay CCR-33 (5 Ohm SMA, DC-18GHz) as shown in Figue 2. This switch enabled the coaxial pobe to be altenatively connected to anothe instument fo measuing the low fequency (sub 1MHz) impedance of the pobe, and the switch was included in the instument calibation, which took place at the end of the pobe. Reflection coefficients wee measued at 21 fequencies equi-spaced on a logaithmic scale, with each eading being the aithmetic mean of five measuements. The intemediate fequency band width on the HP8753D was set to 1 Hz. Figue 2. Coaxial elay and netwok analyse. The cente common connecto of the elay was connected to the coaxial pobe, and the shot length of igid coaxial line was teminated and connected to low fequency measuing instument. III ANALYSIS The complex eflection coefficient fom a teminated line is defined, e.g. [1], as Γ= Z Z t t Z + Z (1) whee Z t is the tansmission line impedance and Z is the teminating impedance. Reaanging, the teminating impedance becomes: 463

6 I M WOODHEAD, I PLATT, J H CHRISTIE, S KRENEK, A BROADBAND SPECTROSCOPIC SENSOR PROBE Z t = Z (1 ) + Γ (1 Γ) (2) Then fo the case whee the teminating impedance is puely capacitive (e.g. open cicuit) so that Z = 1 jωc whee ω is the angula fequency and C is the teminating capacitance, t ε = (1 Γ) jωz C (1 + Γ) (3) whee ε is the complex elative pemittivity of the mateial teminating the pobe, and C is the open cicuit capacitance teminating the pobe. The teminating capacitance is just that due to the finging field that extends beyond the end of the pobe. If the eflection coefficient is epesented by a+jb, then eaanging and sepaating into eal and imaginay pats gives and ' 2b ε = ωcz + a + b 2 2 ((1 ) ) 2 2 " 1 a b ε = ωcz a b 2 2 ((1 + ) + ) (4) (5) whee ' ε and teminating the pobe. " ε ae the eal and imaginay pats of the elative pemittivity of the mateial IV RESULTS Equation (4) was used to calculate the tace (Figue 3) of the uncalibated eal pemittivity of the open cicuit pobe. We chosec =.13pF to povide a mean ε fo wate (data shown late) of 8, within the fequency ange 1 MHz to 1 GHz. The vaiation in pemittivity of the " open-cicuit pobe below 1 MHz was due to small eos of the ode of.1% in ε, when its value was close to zeo - the quoted uncetainty in magnitude fo eflection coefficients measued with the HP8753D is ±.2 (HP8753D opeating manual). This highlights the difficulty of accuately measuing zeo loss mateials, although fo pactical measuements this situation does not nomally aise. The decease in the value of ε " with inceasing fequency is due to a combination of the capacitance of the 5 Ohm calibation boad and loss by adiation, a mechanism nomally accounted fo by calibation of the coaxial pobe. 464

7 Reflection Coefficient E+6 1.E+7 1.E+8 1.E+9 1.E+1 Fequency (Hz) Relative pemittivity eal imag eal pemittivity Figue 3. Reflection coefficient and uncalibated elative pemittivity of the open cicuit coaxial pobe obtained using (5). Figue 4 shows a typical locus of complex eflection coefficient fo wate. The esult shows close alignment with pevious measuements up to 3 GHz. We attibute the deviation beyond 3 GHZ to the capacitance of the esistos used in the 5 Ohm calibation boad. Nevetheless, the issue is elatively unimpotant fo establishing the esponse and egions of inteest fo a boadband spectoscopic senso fo industial measuements. 465

8 I M WOODHEAD, I PLATT, J H CHRISTIE, S KRENEK, A BROADBAND SPECTROSCOPIC SENSOR PROBE Imaginay pat of eflection coefficient MHz. 6GHz Real pat of eflection coefficient Figue 4. Uncalibated eflection coefficient of wate. Eos above 3GHz wee due to capacitance of the 5 Ohm calibation boad. Figue 5 shows the same eflection coefficient data of Figue 4, but conveted to pemittivity using (4) and (5). As befoe, the pobe was uncalibated except fo the empiical choice C =.13pF, chosen to povide a mean ε fo wate of 8. The esult shows close alignment with the expected value up to appoximately 1 GHz, wheeupon the capacitance of the esistos used in the 5 Ohm calibation boad begins to shift the calibation efeence fom its ideal value of 5 + j and hence advesely affect calibation of the VNA. Inaccuacies also occu at fequencies whee thee is esonance in eithe the efeence liquid o the sample. Fo the geomety and samples used, esonances affect the specta at fequencies above appoximately 1 GHz fo wate. Although accuate pobe calibation such as descibed by [6] is appopiate fo accuate measuement of complex pemittivity, pobe calibation fo measuement of moistue content o composition is unnecessay. Instead, it is only necessay to povide an oveall calibation between the desied poduct attibutes and the boadband complex eflection coefficient. 466

9 3 Pemittivity E+6 1.E+7 1.E+8 1.E+9 1.E+1 Fequency (Hz) Real Imaginay Figue 5 The eflection coefficient and uncalibated elative pemittivity of wate obtained using (5). V MEASURING INSTRUMENT The esults shown above employed an HP8753D vecto netwok analyse, but an altenative appoach is to measue in the time domain using a boadband wavefom such as a step function and measuing the eflected signal using an oscilloscope. The Hewlett Packad HP54121T fo example, includes a synchonised step function geneato with isetime <5 ps. The instument can then be used to measue the eflection coefficient. The pocedue we employed was to ecod wavefoms of the open-cicuit pobe and when in contact with the taget mateial. An automated analysis pocedue then aligned the wavefoms fom time efeences, tuncated the zones beyond the egion of inteest, and then a amp function was used to match the amplitudes at the beginning and end of the tuncated wavefoms as descibed by [11] and then the atio of the Fouie tansfoms of the measued and open wavefoms was used to calculate the eflection coefficient in a manne simila to that of [12]. As with the fequency domain appoach, use of the time domain method also equies caeful calibation of the instument at a efeence plane that is close to the coaxial pobe, to achieve epeatable esults. 467

10 I M WOODHEAD, I PLATT, J H CHRISTIE, S KRENEK, A BROADBAND SPECTROSCOPIC SENSOR PROBE VI CONCLUSIONS In this pape we have descibed a coaxial pobe fo vey boadband measuement of the electical popeties of mateials. Calibation of the instument is ideally pefomed by efeence impedances (open, shot and 5 Ohms) at the measuement plane, and leads to values of complex eflection coefficient that may in tun be calibated against moistue content o othe popeties such as concete cuing. The pocessing of signals measued in the time domain was also descibed to yield the complex eflection coefficient equivalent to that fom fequency domain measuements. Fo both measuement means, calibation of the pobe itself povides the means to achieve accuate measuement of complex pemittivity ove a boad fequency ange. When measuement of a poduct is equied, only instument calibation is necessay since the oveall tansfe function between the moistue content o composition of the poduct and the eflection coefficient will encompass the pobe calibation. REFERENCES [1] J. H Chistie and I. M. Woodhead. A New Model of DC Conductivity of Hygoscopic Solids. Pat 1: Cellulosic Mateials., Textile Res. J. 22, 72, pp [2] T. Kuda, G. S. V. Raghavan, C. Akyel, R. G. Bosisio, F. R. van de Voot, H. S. Ramaswamy, Dielectic popeties of milk and its constituents at 2.45 GHz, Ameican Society of Agicultual Enginees, June 1991, Albuqueque, pape numbe 9135 [3] S. Mashimo, S. Kuwabaa, S. Yagihaa, K. Higasi, Dielectic elaxation time and stuctue of bound wate in biological mateials, Jounal of Physical Chemisty, 1987, 91: pp [5] J. H. Chistie, S. R. Sylvande, I. M. Woodhead, K. Iie, The dielectic popeties of humid cellulose, Jounal of Non-Cystaline Solids, 24, 341:115. [4] S. Yagihaa, M. Asano, N. Shinyashiki, Boadband dielectic spectoscopy study on hydation of cement and some aqueous solution and dispesion systems, Intenational Confeence on Electomagnetic Wave Inteaction with Wate and Moist Substances, 4/15-18, 27, Hamamatsu, Shizuoka. 468

11 [7] B. G. Colpitts, Tempeatue sensitivity of coaxial pobe complex pemittivity measuements: expeimental appoach, IEEE Tansactions on Micowave Theoy and Techniques 1993, 41, pp [8] P. De Langhe, L. Matens, D. De Zutte, Design ules fo an expeimental setup using an open-ended coaxial pobe based on theoetical modelling, IEEE Tansactions on Instumentation and Measuement 1994, 43, pp 81 7 [9] A. P. Gegoy, R. N. Clak, T. E. Hodgets, G. T. Symm, RF and micowave dielectic measuements upon layeed mateials using eflectometic coaxial senso, National Physical Laboatoy Repot, 1993, DES 125. [6] N. I. Sheen, I. M. Woodhead, An open-ended coaxial pobe fo boad-band pemittivity measuement of agicultual poducts, J agic Engng Res, 1999, 74, pp [1] S. Ramo, J. R. Whinney, T. van Duze, Fields and Waves in Communication Electonics, 1993, New Yok: Wiley [11] A. M. Nicolson, Foming the fast Fouie tansfom of a step esponse in time-domain metology, Electonics lettes, 1973, 9(14), pp [12] W. L. Bellamy, S. O. Nelson, R. G. Leffle, R G, Development of a time domain eflectomety system fo dielectic popeties measuement, Tansactions of the ASAE, 1985, 28, 4, pp