Chapter 3. Development of UWB GPR antennas

Transcription

1 Chpte 3. Development of UWB GPR ntenns 3.1. Intoduction Antenns my be the most citicl point in n impulse GPR system, hving diect influence on its system pefomnce. The ntenns hve to be specilly designed to dite pulses with given popeties into the gound nd tnsduce the bcksctteed signls fom subsufce objects into useful signl without too much ffecting thei shpe. They diffe fom conventionl ntenns in mny spects. Fist of ll GPR ntenns opete ne the gound. The pesence of the gound in the ective field o ne-field egion of the ntenn will influence the ntenn chcteistics. Secondly the ntenn often opetes in poximity to second ntenn, which gives n dditionl poblem of ntenn coss-coupling. Futhemoe, the ntenn tnsmits nd eceives fst tnsient electomgnetic signls with lge fctionl bndwidth. A typicl GPR could equie n ntenn with fctionl bndwidth of 1%. GPR ntenns lso should hve line phse chcteistic ove this bnd, to limit the distotion of the emitted (o eceived) signl, nd constnt polistion with fequency. All these supplementy equiements mke tht the numbe of ntenn clsses, which cn be used, is limited. In some cses it lso led to the design of new o modified ntenns, mtching the chcteistics of the medium of popgtion, the dited signl nd the system equiements.

2 Chpte 3 In this chpte n oveview of existing GPR ntenns nd ult-widebnd ntenns will be given. Futhemoe we will discuss some design gols of GPR ntenns dpted fo the ppliction of demining. The design gols e minly poduct of field tils descibed in Chpte 2. Finlly the development of TEM hon ntenns fo UWB GPR will be descibed Oveview of existing GPR ntenns Conventionl ntenns Nowdys, thee min types of ntenns e used in GPR pplictions: the element ntenns, the fequency independent ntenns nd the hon ntenns. A good oveview of GPR ntenns is given in [1]. Element ntenns Element ntenns e the most widely used out of the thee min types of ntenns. Some exmples of element ntenns e the monopole, dipole nd bow-tie ntenn. This clss of ntenns is non-dispesive, nd is chcteised by line polistion, low diectivity nd some of them hve eltively limited bndwidth. The dition chcteistics of element ntenns e well undestood. The clcultion of the dited field is bsed on the ppoximtion of the cuent distibution on the ntenn by numbe of elementy cuents. In the system configution, the tnsmitting nd eceiving ntenn cn hve eithe the sme polity (co-polised) o n othogonl polity (coss-polised), poviding lowe ntenn coss-coupling nd eventully bette discimintion ginst cetin mn-mde objects. Due to the low diectivity, element ntenns e used in close contct to the gound in ode to couple s much enegy s possible into the gound. 3-2

3 Development of UWB GPR ntenns The ltte hs the disdvntge tht the ntenn impednce is influenced by the gound popeties. Recently some esech is done by TU-Delft to ty to dpt the ntenn to the gound popeties [2]. Much effot hs gone into techniques to extend the bndwidth of the element ntenns. One possible solution is esistively loding the dipoles. The loding cn be done eithe by n end loding, by distibuted loding o by tpeed esistive loding. Unloded dipoles hve intenl eflections t the open end of the dipole. The esistive loding is essentil in cusing pid decy of cuent long the ntenn to educe the eflections t the ntenn open ends, which esults in less distotion o ltetime inging in the til of the dited tnsients. Hence lge bndwidth is chieved. In 1965, Wu nd King wee the fist to mke study on the tpeed loding of dipole ntenn [3]. The pofile fo this esistive loding is known s the Wu-King pofile. An oveview of element ntenns with esistive loding is given in [4]. Studies comping numeicl finite diffeence time domin (FDTD) nlysis with expeimentl mesuements cn be found in [5] nd [6]. A disdvntge of the loding is the eduction of gin nd dition efficiency, lthough the ltte is often consideed less disdvntgeous thn ntenn inging. The dition efficiency is defined s enegy enegy dited dited + enegy dissipted (3. 1) nd cn in some cses dop down to 29% [6]. Anothe ntenn tht cn be used s widebnd element ntenn is the bow-tie ntenn [7] (Fig. 3-1). This ntenn is in fct pln vesion of finite biconicl ntenn, detemined by length L nd n ngle α. Bow-tie ntenns, togethe with its mny vints like shielded bow-tie, loded bow-tie, tpeed bow-tie etc., e pobbly the most popul mong the element ntenns. They cn ech fctionl bndwidths of 1% nd moe. 3-3

4 Chpte 3 L α Blnced feed point Fig. 3-1 : The bow-tie ntenn. Fequency independent ntenns A second min type of ntenns used in GPR systems is the fequency independent ntenn. An ntenn is clled fequency independent if it stisfies two pinciples: the scle pinciple nd the tunction pinciple. It is known tht ntenn chcteistics like ptten, impednce, etc., e invint to chnge of scle tht is in popotion to the chnge in wvelength. If the shpe of n ntenn is entiely detemined by ngles, it would be invint to chnge of scle nd hence, the pefomnce of such n ntenn would be independent of fequency. Futhemoe the ntenn hs to stisfy the tunction pinciple, which implies tht the cuent ppoches zeo t the end of n ntenn. Othewise tunction of the ntenn hs n effect on the ptten. In pctise howeve, the dimensions of the ntenn limit the lowe fequency bound. A good oveview of the subject is found in [8]. Exmples of fequency independent ntenns e the log-peiodic ntenns nd the spil ntenns. An exmple of spil ntenn is shown in Fig The fomul fo the spil cuve in pol coodintes is [8] φ = φ + tn A ln (3. 2) which shows tht the ntenn is detemined only by the ngles φ nd A. 3-4

5 Development of UWB GPR ntenns A ϕ Fig. 3-2: Spil ntenn A new membe of the clss of fequency independent ntenns is the Vivldi ntenn. The Vivldi ntenn, intoduced by Gibson in 1979 [9], consists of n exponentil tpeed slot line etched on dielectic substte (Fig. 3-3). The stuctue is usully fed by mico stip, etched on the othe side of the substte (in dotted line on Fig. 3-3). The ntenn chcteistics e influenced by numbe of pmetes, like the dimensions nd geomety of the slot line, the thickness nd dielectic constnt of the substte, etc. The Vivldi ntenn is extemely wide-bnded, bndwidths up to 5 octves hve been epoted [9]. The lowe cut-off fequency is limited by the dimensions of the ntenn petue. The ntenn is chcteised by long electicl length fo esonble shot physicl length, nd theefoe is n inteesting ntenn fo GPR pplictions. Fig. 3-3: Vivldi ntenn with mico stip feed 3-5

6 Chpte 3 The fequency independent ntenns howeve hve the popety of being dispesive. The impulse esponse of this clss of ntenns genelly esults in chip-wvefom, which mkes them unsuited fo sending shot pulses, s needed in time domin GPR. In stepped-fequency d, the dispesive behviou of the ntenns cn be compensted fo. Tht is why they e only used in combintion with this clss of GPRs. Hon ntenns The thid min type of ntenn is the hon ntenn. Clssicl hon ntenns hve fctionl bndwidth ound 66%, which is nomlly not enough fo GPR. Lge bndwidths e chieved with double-idged hons. Hon ntenns hve the dvntge of being diective. This mens hon ntenns cn be used off gound. Exmples of commecilly used hon ntenns e the 1GHz nd the 2.5 GHz ntenn of GSSI. They wee oiginlly developed fo od inspection nd cn be mounted on vehicle. The disdvntge of this kind of ntenn is the size. Usully hon ntenns e too lge wht mkes them not use-fiendly in ough tein (e.g. the 1GHz ntenn of GSSI e 12 cm * 22 cm* 32 cm) Non-dispesive Ult-widebnd ntenns Thnks to the consideble pogess in signl genetos nd eceives the lst decdes, the time domin UWB d hs won in inteest. The min gol ws to chieve highe sptil esolution, n esie tget infomtion ecovey nd lowe pobbility of inteception. In this kind of ppliction it is impotnt tht fst electomgnetic tnsients e dited, without too much distotion. Theefoe nondispesive UWB ntenns e needed, with fctionl bndwidth of 12% o gete. In one of the pevious pgphs we ledy mentioned the esistive loded bow-tie ntenn, which indeed cn be clssified in the fmily of UWB ntenns. Howeve if moe dition efficiency nd diectivity e pusued, the solution must be found in othe types of ntenns, mostly tvelling wve ntenns. An ntenn is clled tvelling wve ntenn if the length of the ntenn pltes is sevel times lge thn 3-6

7 Development of UWB GPR ntenns the pulse lengths of the feeding signl. This mens the wve ctully tvels long the ntenn pltes befoe it is dited when eching the end of the pltes. The infinitely long biconicl ntenn is n idel dito fo fst electomgnetic tnsients [1][11][12]. Since its length is infinite nd its shpe is invint to chnge of scle, the infinitely long biconicl ntenn is fequency independent ntenn. The infinite biconicl ntenn is equivlent to tnsmission line, guiding spheicl TEM wve. Its chcteistic impednce nd hence the ntenn impednce is only function of θ (see Fig. 3-4). In pcticl use howeve, the ntenn will hve finite length, so tht eflections t the ntenn end will occu. These unwnted eflections cn, just like with the element ntenns, be educed by including esistive mteil in the ntenn stuctue [13]. Fig. 3-4 shows finite conicl monopole ntenn on gound-plne. θ H ϕ E θ Fig. 3-4 : Conicl monopole ntenn of finite length on gound-plne A second type of UWB ntenn, bsed on the biconicl ntenn is the tvelling wve TEM hon. A tvelling wve TEM hon consists of pi of tingul conductos foming V stuctue (Fig. 3-6), guiding essentilly TEM mode between its ntenn pltes. The TEM hon is studied in detil in Section 3.4. A lst type of UWB ntenn is the Impulse Rditing Antenn (IRA). The IRA ntenn is pbolic ntenn with conicl TEM tnsmission line s pimy 3-7

8 Chpte 3 souce. The conicl tnsmission line will lunch essentilly spheicl TEM wve towds the eflecto. Suppose step voltge is pplied in the feed point of the tnsmission line. An obseve on xis nd in the f field of the ntenn will fist see pe-pulse, due to the diect (bckwd) dition of the feeding tnsmission line. This is followed by n impulse of shot dution nd high mplitude. The impulse is popotionl to the fist tempol deivtive of the step on the feeding tnsmission line. The min impulse is followed by til due to eflections on the tnsmission line end nd ssocited with low fequency components of the feeding signl. These eflections e limited by putting esistos between the tnsmission line nd the eflecto. The IRA ntenn is epesented on Fig This type of ntenn is conceived nd studied by C.E. Bum nd E.G. F. Moe detils cn be found in [14][15][16]. Fig. 3-5: The Impulse Rditing Antenn 3.3. GPR ntenn design gols fo the demining ppliction As ledy mentioned in Chpte 2, most GPR ntenns e designed fo pplictions othe thn the demining ppliction nd do not meet the specific equiements s needed fo the demining ppliction. Fom the expeience obtined by the field tils nd tking into ccount the UWB ppoch, five technicl nd pcticl design gols fo UWB ntenns e set: 3-8

9 Development of UWB GPR ntenns 1. The ntenn must be ble to dite o eceive fst electomgnetic tnsients with spectum between 5 MHz nd 4.5 GHz 2. The ntenn must be usble off gound, not only fo sfety esons but lso to impove the mobility of the detecto. As consequence, the ntenn must hve high diectivity so tht it cn still couple sufficient enegy into the gound to chieve penettion depth of 2 cm in ny soil 3. The ntenn must guntee high degee of mobility, hving n impliction on the dimensions nd weight of the ntenn. Minefields hve often ough sufce nd e coveed with lot of vegettion. Only little ntenns cn guntee sufficient flexibility in such scene. Little ntenns e lso bette fo hnd-held pplictions 4. The ntenn popeties must be independent of the gound popeties. The influence of the gound will be educed when using the ntenn off gound 5. The ntenn must be chep in poduction to limit the ovell cost of the senso. This will lwys be sked fo in the cse of humnitin demining. One of the most pomising ntenns tht cn meet these design gols, is the tvelling wve TEM hon. Theefoe we hve chosen this type of ntenn s the stting point fo ou study Ai-filled TEM hon TEM hon ntenns In n effot to incese the diectivity o the ntenn gin fo bodbnd nd nondispesive ntenn, mny eseches hve consideed TEM hon. A tvelling wve TEM hon consists of pi of tingul conductos foming V stuctue (Fig. 3-6), cpble of diting nd eceiving fst tnsient pulse [4]. It is ssumed tht the TEM hon guides essentilly the TEM mode within the fequency nge of inteest by mintining constnt chcteistic impednce nd tht, by neglecting the edge diffction effect nd finge fields, linely polised spheicl wve is dited. The 3-9

10 Chpte 3 opetion of TEM hon is fily simple. On tnsmission, the TEM hon dites signl tht is popotionl to the fist time deivtive of the incident voltge pulse in the feeding point. On eception, the hon outputs voltge pulse tht hs the sme shpe s the incoming electic field ( detiled explntion fo this is given in Chpte 4, Section 4.3.3). A conventionl TEM hon is completely chcteised by thee pmetes: - L the length of the ntenn pltes, - ϕ the zimuth hlf-ngle, - θ the elevtion hlf-ngle. The chcteistic impednce of n infinite long TEM hon (L= ) is only function of the two ngles ϕ nd θ. Theoeticlly TEM mode does not hve n uppe cut-off fequency. In pctice howeve some highe modes exist nd will intoduce n uppe cut-off fequency. The dimension L of the ntenn minly govens the lowe cut-off fequency. L ϕ Feed Point θ Fig. 3-6: The tvelling wve TEM Hon The conventionl design of the TEM hon is bsed on the infinitely long biconicl ntenn. Mny vints e possible, e.g. esistive loding of the ntenn [17], tpeing the ntenn pltes [18], gdully chnging the seption ngle between the ntenn pltes [19], o plcing dielectic lens t the petue [2]. Thee is until now no exct theoeticl nlysis of this stuctue vilble. Vious ppoximte nlyses nd models hve been mde to llow the ntenn design. 3-1

11 Development of UWB GPR ntenns The ntenn impednce To clculte the ntenn impednce nd the suge impednce long the ntenn pltes, two methods e often found in litetue. Both models e bsed on the ppoximtion tht the ntenn is infinitely long. In this cse, the ntenn is equivlent to conicl tnsmission line. Suppose the oigin of co-odintes in the feed point (pex) of the ntenn. Fo given vlue of the two ngles ϕ nd θ, the chcteistic impednce of the tnsmission line is constnt, independent of the dil co-odinte. Hence, the ntenn impednce (fo ), nd the suge impednce will then be equl to the chcteistic impednce of the tnsmission line. The suge impednce is defined s the impednce mesued in ny point of the ntenn by mens of step excittion. It does not tke into ccount the bckwd wves due to eflections t the ntenns open end, iving lte. The chcteistic impednce of the tnsmission line cn be found by confoml mpping. In fist step, the two pltes e pojected by steeogphic pojection into two pltes of cicul cs. In second step, the chcteisic impednce of the cicul c poblem is solved by stndd method of confoml mpping [1][21]. The esults e usully given in tbles nd cuves s function of the two ngles ϕ nd θ. This method of confoml mpping cn lso be used to clculte the field distibution between the ntenn pltes [22]. A second ppoch used fo the clcultion of the chcteistic impednce of the infinite TEM hon is bsed on cscding sections of stip-lines [18][23]. The ntenn is split into electiclly smll segments long its length (Fig. 3-7). Ech segment of the ntenn is ppoximted by pllel twin line nd the ntenn is modelled s collection of those pllel twin lines in seie. 3-11

12 Chpte 3 w(x) x 2. h( x) i 1 i i +1 () Side view (b) Font view of one section Fig. 3-7 : TEM hon s cscde of stip lines The chcteistic impednce of such twin line segment is clculted s two times the chcteistic impednce of mico-stip of height h nd width w in the bsence of dielectic. The chcteistic impednce of the mico-stip cn be clculted fo given tio w h using the ppoximted expessions of Hmmestd nd Jensen [24]. Fo given vlue of the ngles ϕ nd θ, the tio w h is constnt nd the chcteistic impednce of ech twin line segment will be the sme, hence the suge impednce of the TEM hon will be constnt. The dited f field The evlution of the f field, dited by TEM hon, cn be done in the fequency domin s well s diectly in the time domin. In the fequency domin ppoch the field is fist clculted t the TEM hon petue, s the field of the TEM mode tht would exist in hon of infinite length. Bsed on the ssumption tht the edge diffction nd the finging field effects e neglected, the spheicl field t the hon petue is given in [4] s E ( x', y',) = y x'² + y'² e 2 2 jk + x'² + y'² 2 + y'² (3. 3) 3-12

13 Development of UWB GPR ntenns The oigin of co-odintes is tken in the middle of the ntenn petue (Fig. 3-8). The co-odintes in the ntenn petue e indicted by pimes nd is the distnce between the pex nd the cente of the petue. Huygen s pinciple is used with this field to pedict the dited field outside the petue. y x z h Fig. 3-8: Repesenttion of co-odinte system A simple time domin model fo the TEM hon is given in [25]. In this model, the ntenn is consideed s n open cicuited tnsmission line. It consists of ppoximting the ntenn by succession of electic nd mgnetic dipoles, nd summing thei contibutions. The model is simple but useful to help in the design of TEM hon o to mke compisons with othe ntenns. If step voltge of V is pplied t the ntenn feed, the dited electic field on boesight t distnce z is given by V h z + c z + z + 3 E y ( z, t) = δ ( t ) + u( t ) + u( t ) z 2πcf g c 2 c c (3. 4) with h the distnce between the two ntenn pltes t the petue (Fig. 3-8), c the speed of light, f g = Z c / Z the geometicl impednce fcto, Z the impednce of fee spce, Z c the chcteistic impednce of the TEM hon, the distnce between the ntenn pex nd the oigin of co-odintes nd u (t) the unit step function. Fig. 3-9 shows the gphicl epesenttion of the step esponse given by (3.4). The utho lso suggests coection of the model fo high fequencies [25]. 3-13

14 Chpte 3 E y V h Impulse e = z 4π cf g V h z 8π f g z + = c t 2 / c t Fig. 3-9: Step esponse on boesight of TEM hon Anothe possibility of clculting the ntenn impednce nd the dited field e by numeicl modelling of the TEM hon using the method-of-moments [4] o finitediffeence time-domin (FDTD) code. Some didctic simultions of TEM hons using FDTD e found in [18] Study of the wie Model An ccute model of the TEM Hon is found in the wie method, whee the ntenn pltes e eplced by set of infinitely thin wies. The model ws oiginlly developed fo the electomgnetic pefomnce nlysis of Electo-Mgnetic Pulse (EMP) simultos [26][27], but cn be used s model fo the i-filled TEM hon. The wie method is time-domin-bsed method. The tnsient electomgnetic field emitted by TEM hon ntenn is consideed to be the sum of the tnsient electomgnetic fields emitted by ech individul wie. The dvntge of the wie model is tht it povides t the sme time n nlytic expession fo the ely time dited f field s fo the suge impednce. Futhemoe, the expessions of the f field nd the ntenn impednce e simple which mkes them convenient fo simultions nd design puposes. In the model the following ssumptions e mde. The cuent on the ntenn pltes is stictly dil nd tvels with the speed of light. As we e only inteested in the TEM 3-14

15 Development of UWB GPR ntenns mode of the ntenn, this is n cceptble ppoximtion. Futhemoe we ssume tht the wvefom nd mplitude of the tvelling cuent is constnt long the wie, tht the wies e infinitely thin nd tht the cuent is totlly bsobed t the end of the ntenn. In elity howeve pt of the cuent will bounce bck t the ntenn-end towds the feed point of the ntenn. This phenomenon will cuse lte time inging in the ntenn, which will be omitted in this ppoch. The dition by one wie In fist step the field dited by wie of length L is studied. The geomety of the wie nd the ssocited co-odintes e shown in Fig The co-odintes on the wie e indicted with pime. Obsevtion point L L ' L = Fig. 3-1: Co-odinte system fo the study of the wie Suppose cuent pulse I wie (t) is pplied t the wie t = distibution on the wie is then. The dil cuent I( ', t) = I ( t ' wie c ) (3. 5) 3-15

16 Chpte with ' ' =. The dited electic field cn be witten s t t A t t E = ), ( ), ( ), ( φ (3. 6) with ), ( t A the vecto potentil nd ), ( t φ the scl potentil given by ), (. ), ( 2 t A c t t = φ (3. 7) Fo wie of length L, the vecto potentil is witten s ' ' ' ) / ' ( 4 ), ( L wie d c t I t A = π µ (3. 8) with ' the unit vecto long the wie. Afte substitution of (3.7) nd (3.8) in (3.6), the electic field dited by one wie ), ( t E wie cn be clculted [26] nd is expessed s + + = ) ( ). (1 )..( ) ( ). (1 )..( 4 ) ( ) ( 4 1 ), ( ' ' ' ' ' ' 2 2 c L t I c t I Z c L t q c t q t E L wie L wie L L wie L L L L π πε (3. 9) with =, L L = the distnce between the obsevtion point nd the end of the wie, Z the impednce of fee spce, nd q the chge on the wie given by

17 Development of UWB GPR ntenns q( t) = t Iwie( τ ) dτ (3. 1) Fo bette insight of the dition by wie, we show in Fig some snpshots of the dited field of one wie on gound-plne. The snpshots e clculted on 2mm by 2mm gid using eqution (3.9). The veticl oiented wie, shown in white on the snpshots, hs length L of 1 cm. The feed point of the wie is downwds. In the simultion, the excittion cuent is Gussin pulse with full width t hlf mximum (FWHM) of 4 ps. If τ = L c is the time fo the cuent to tvel towds the end of the wie, the snpshots e given t t =.5τ, t τ =, t = 1.5τ nd t = 2τ. In ech snpshot, the electic field is epesented in mgnitude. mx () (b) (c) (d) Fig. 3-11: Mgnitude of the dited electic field by one wie on gound-plne. Snpshots tken t () t / τ =.5 (b) t / τ = 1 (c) t / τ = 1.5 (d) t / τ =

18 Chpte 3 On the snpshots we see tht fist spheicl wve, cented t the feed point =, is poduced when the cuent pulse entes the wie. A second spheicl wve, cented t L = is poduced when the cuent eches the open end of the wie. Becuse the second wve coesponds with deceletion of chges, it will hve field opposite to the fist wve. This is howeve not visible on the snpshots becuse only the field mgnitude is shown. The electic field concentted ound the wie end in snpshot (c) nd (d) is sttic electic field due to the ccumulted chges t the open end of the ntenn. The chges sty thee becuse in the model the cuent is bsobed t the wie end. In (3.9) it cn be seen tht this sttic field deceses with 2 1/ L. A simil nlysis in the time domin of the dited electic field by wie is found in [28], poviding good physicl undestnding of dition phenomen. The cuent distibution on the ntenn plte In next step the cuent distibution on the ntenn plte is consideed. The geomety of the ntenn plte nd the ssocited co-odintes e shown in Fig Suppose tht totl cuent I ( ) is pplied t the ntenn feed nd tht the ntenn plte is infinitely thin. t ϕ z y ϕ θ = x Fig. 3-12: The geomety of the ntenn plte nd the ssocited co-odintes In this cse, the cuent on the pltes cn be consideed s dil diected cuent sufce density ( ', t). The co-odintes on the plte s sufce S e indicted with J S 3-18

19 Development of UWB GPR ntenns pime. The dil distibution of the cuent sufce density is dictted by I ( t ' ). The zimuth distibution of the cuent sufce density is govened by c function F ϕ, ϕ ). Accoding to the cuent distibution on mico stip, the cuent ( zimuth distibution function is tken [26] 1 F ( ϕ, ϕ ) = (3. 11 ) 2 π ϕ ϕ 2 Note tht the cuent zimuth distibution function is nomlised, so tht ϕ F( ϕ, ϕ ) dϕ = 1 (3. 12) ϕ Fig shows plot of the zimuth distibution function fo ϕ =3. We notice tht the cuent sufce density will be moe impotnt ne the side edges of the ntenn plte ϕ [deg] Fig : Azimuth cuent distibution function on the ntenn plte 3-19

20 Chpte 3 Totl field dited by the TEM hon In the next step, the ntenn plte is eplced by mesh of N dil wies with common pex. Ech wie i is chcteised by n zimuth ngle ϕ 1 2 i = ( i ) ϕ ϕ i 1,2,... N = (3. 13) with 2ϕ ϕ = (3. 14) N nd n elevtion ngle θ i = csin(sinθ cosϕ i ) (3. 15) The cuent in wie i is given by ϕi + ϕ / 2, i ( t) = I ( t). F( ϕ, ϕ ) ϕi ϕ / 2 I wie dϕ (3. 16) Finlly the totl field dited by the TEM hon is obtined by summing the field expession fom ech individul wie with its cuent: E Tot (, t) = N i= 1 E wie, i (, t) (3. 17) Note tht, s ech wie in the TEM hon hs n imge wie with n oppositely diected cuent, ll the tems involving the chges q( t / c) e cncelled out. 3-2

21 Development of UWB GPR ntenns In Fig simultion of TEM hon is shown using the wie model. The ntenn hs length L of 1 cm, ϕ = 25 nd θ = 1.2. In the fou snpshots the xz-plne is epesented (see Fig. 3-12). The intesections of the ntenn pltes with the xzplne e shown in white. At t = s Gussin cuent pulse (the sme s in the pevious simultion with the single wie) is pplied t the ntenn feed point. The snpshots e given t espectively t =.5τ, t τ =, t = 1.5τ nd t = 2τ, with τ the time fo the cuent pulse to ech the end of the ntenn pltes. The mgnitude of the electic field is in ech snpshot e-scled nd plotted on colou scle. mx () (b) (c) (d) Fig : Mgnitude of the dited electic field in the xz-plne by TEM hon. Snpshots tken t () t / τ =.5 (b) t / τ = 1 (c) t / τ = 1.5 (d) t / τ = 2 On snpshot () nd (b) we see the pulse-shped wve tvelling between the ntenn pltes. Note the wve on the outside of the ntenn pltes. This is lso spheicl 3-21

22 Chpte 3 wve cented t the ntenn feed, but with n electicl field opposite to the one inside the ntenn (the sign is not visible on the snpshots becuse the mgnitude of the field is plotted). This wve is impotnt to explin the negtive pe-pulse dited by the dielectic-filled TEM hon (see Section 3.5). It is lso t the oigin of the pe-pulse encounteed by the IRA ntenn. On snpshot (c) nd (d) the sttic electic field t the ntenn end, due to the ccumultion of chges, is visible. The field dited on boesight (in the x-diection) hs the shpe of the fist time deivtive of the cuent pulse (this cn not be seen in pevious figue, s only the mgnitude of the field is epesented). To show this we plot in Fig the nomlised E z component of the electic field s function of time in point on the x- xis t 3 cm fom the ntenn feed. In Fig () the Gussin cuent pulse is epesented. 1.8 Cuent.6.4 Nomlised electic field Time [ns] () Time [ns] (b) Fig. 3-15: () The Gussin cuent pulse, (b) the E z component on the x-xis t 3 cm fom the ntenn feed 3-22

23 Development of UWB GPR ntenns The suge impednce of the TEM hon The wie model cn lso be used to clculte n nlytic expession of the suge impednce of the TEM hon. Theefoe we fist hve to clculte the ely time electic field in the ntenn, which mens L t < (3. 18) c In this cse ll the tems involving L vnish nd the expession of the electic field in the ntenn cn be simplified. By definition, the suge impednce cn be found fom Z = I V ( ) ( t / c) = E ς in ntenn I dl (3. 19) with the integtion pth ς chosen s n c in the xz plne, so Z = θ θ E in ntenn I θ dθ (3. 2) In this model the suge impednce (fo L< ), nd hence the chcteistic impednce long n infinite TEM hon (L= ), is given by [26] Z = Z 4π N ϕi + ϕ 1 cosϕ i cos2θ i ln i = 1 1 cosϕ i ϕ ϕ i / 2 F( ϕ, ϕ ) dϕ / 2 (3. 21) with F ϕ, ϕ ) the zimuth cuent distibution on the ntenn pltes, N the numbe of ( wies of the wie mesh nd ϕ i nd θ i the zimuth nd elevtion ngle of the i th wie. We cn see tht the suge impednce given by the nlyticl expession (3.21) is only 3-23

24 Chpte 3 function of the ngles ϕ nd θ, s we expected. To vlidte the nlyticl expession (3.21), we mesued the suge impednce of one ntenn plte on gound-plne by time domin eflectomety (TDR) nd comped it with the clculted suge impednce. The geometic configution is the sme s shown in Fig The ntenn plte hs length of 1 cm nd fixed zimuth hlf ngle ϕ of 3. The set-up gve the flexibility of chnging the elevtion ngle θ vey esily. A TDR mesuement genetes step with shot ise-time nd mesues the eflected signl fom impednce discontinuities, to detemine the eflection coefficient vesus time. Fom this eflection coefficient, the impednce chcteistics vesus distnce of the device unde test is clculted. In ou TDR mesuements, step with ise time of 45ps is used. Fo diffeent vlues of the elevtion ngle θ the suge impednce t the middle of the ntenn plte ws mesued nd comped to the vlues given by (3.21). Fo the clcultion, the numbe of wies N used in (3.21) ws 4. As the mesuement ws pefomed on hlf TEM hon (one ntenn plte on gound-plne) the mesued suge impednce by TDR will be hlf the suge impednce of complete TEM hon (two ntenn pltes). Theefoe, the mesued suge impednce hs to be multiplied by 2. The esults e plotted in Fig Z [O] Clculted Mesued ? [ Deg] Fig. 3-16: Mesued nd clculted suge impednce of i-filled TEM hon 3-24

25 Development of UWB GPR ntenns Expession (3.21) of the suge impednce given by the wie method tuns out to be vey ccute fo pedicting the suge impednce. The wie method slightly undeestimtes the suge impednce. The mximl discepncy between the theoeticl model nd the expeimentl dt is 5%. A quick compison with the two methods given in Section lens tht this method pefoms bette, nd is thnks to its nlyticl fom, flexible fo design puposes. The smll discepncy is pobbly due to the fct tht the wve guided by the ntenn pltes is ctully not puely TEM. Highe odes cn be intoduced by bd feed t the feed point nd of couse by edge diffction Design of i-filled TEM hon In fist pt of the thesis, some wok ws done on the design of i-filled TEM hons. The pupose ws not to develop no to enhnce such n ntenn, but to vlidte the wie model. Futhe we wnted to study the behviou of the i-filled TEM hon so tht its design could seve s bsis fo the development of the dielectic-filled TEM hon. We will not go into ll the detils of the design, but only discuss spects tht e elevnt fo the dielectic-filled TEM hon. The i-filled TEM hon will nevetheless be used thoughout the wok to compe with the pefomnce of the dielectic-filled TEM hon. Fo the design of TEM hon, 3 pmetes: L, ϕ nd θ hve to be detemined, nd this s function of the use-defined ntenn chcteistics like fctionl bndwidth, ntenn ptten, suge impednce, etc. In this type of ntenn, L will minly influence the lowe cut-off fequency nd the fctionl bndwidth [25][29]. It is not esy to extct vlue fo L using the discussed models. As ule of thumb, the length L must be t lest hlf wvelength of the lowest fequency [4]. We opted fo smll ntenn with length L =1 cm. This mens lowe cut-off fequency of 1.5 GHz, which is too high ccoding to the design gol stted in Section

26 Chpte 3 The fist fundmentl question is on the impednce of the ntenn. In the pst, mny ntenn designes hve tied to mtch the ntenn suge impednce t the ntenn end to the wve impednce of fee spce Z = 12 π. In thei design they gdully chnged the suge impednce fom 5 Ω t the ntenn feed to 377 Ω t the ntenn end by continuously chnging long the pltes the ngles ϕ (fling out the ntenn pltes) nd/o θ (tpeing the ntenn pltes). The im of the mtching is to void eflections t the open end of the ntenn. In [18] it is shown by simultion nd mesuement tht the concept of mtching the suge impednce to tht of fee spce to eliminte eflections is eoneous nd tht eflections will occu. Futhe, tpeing nd fling out the ntenn pltes will distub the TEM mode in the ntenn [19] nd hence cete highe modes. Lte on in this wok we wnt to model the ntenns in the time domin. In the ppliction of lndmine detection, the tgets e eltively close to the ntenns (± 3cm), nd these highe modes will mke the model less ccute ne the ntenns. Fo these two esons we hve chosen to keep the suge impednce constnt ove the whole ntenn. This mens tht no tpeing o fling of the ntenn pltes is used. In the design of the i-filled TEM hon we opted fo n ntenn impednce of 8 Ω nd not fo the chcteistic impednce of the feed cble, i.e. 5Ω, s is usully done fo this kind of ntenn. The esons fo this will be explined in Section An infinite numbe of couples (ϕ,θ ) tht yield suge impednce of 8 Ω e possible, so we hve to detemine one moe pmete. In the design phse of the ifilled TEM hon, we used the wie model to optimise the ngle ϕ fo given L nd suge impednce of the ntenn. As citeion fo the optimistion we looked t the pek-to-pek vlue of the dited E-field nd the hlf-powe bemwidth. In Tble 3-1 set of couples (ϕ,θ ) leding to suge impednce of 8 Ω is given in steps of 5 fo the zimuth ngle. The couples e clculted using eqution (3.21) in n implicit wy. ϕ [ ] θ [ ] Tble 3-1: Couples (ϕ,θ ) leding to suge impednce of 8 Ω 3-26

27 Development of UWB GPR ntenns Fo ech couple, the dited tnsient electomgnetic field is clculted in the H- plne (i.e. the xy-plne in Fig. 3-12) t dius of one mete fom the ntenn feed point by eqution (3.17). The cuent pulse used in (3.17) hs Gussin shpe with Full Width t Hlf Mximum (FWHM) of 8 ps. This mens tht the dited field will be monocycle s shown in Fig (b). Fo ech of the couples (ϕ,θ ) the pek-to-pek vlue of the dited E-field (in [V/m]) is plotted on pol digm to fom kind of non-nomlised dition ptten (Fig. 3-17) phi=15 phi=2 phi=25 phi=3 phi=35 phi=4 phi= Fig. 3-17: Pek-to-pek vlue of the dited E-field in H-plne fo diffeent vlues of ϕ The figue shows clely mximum pek-to-pek mplitude on boesight fo vlues of ϕ between 3 nd 4. The mximum is flt, but fo ϕ >4 the pek-to-pek mplitude deceses gin nd the hlf-powe bemwidth inceses. The diffeent vlues of ϕ nd θ hve little effect on the bndwidth. As conclusion we cn sy tht the optiml ngle ϕ fo suge impednce of 8 Ω, is found to be 3-4 degees. In ou design we opted fo ϕ =3, which leds to n ngle θ =11.2 (see Fig. 3-16, mesued dt). A pictue of the i-filled TEM hon is shown in Fig Specil ttention ws pid t the feed-point. The feed point hs to be mechniclly obust but ce must be tken 3-27

28 Chpte 3 tht it does not intoduce too lge impednce discontinuity. A schemtic epesenttion of the feed point is given in Fig Coppe ntenn pltes SMA connecto Scew Fig. 3-18: Pictue of the i-filled TEM hon Fig : Schemtic epesenttion of the feed point To vlidte the wie model, we compe in Fig. 3-2 the mesued ntenn ptten in the H-plne t 5 GHz with the ntenn ptten, clculted by the wie model. The fequency domin mesuements wee pefomed in n nechoic chmbe in the Univesity of Leuven. Fo the clculted ptten, we fist clculted the dited E- field in the time domin nd then conveted to the fequency domin using n FFT. The wie model tuns out to be lso vey ccute in the clcultion of the dition ptten. The 3-dB bemwidth in the H-plne is ound 45. In Fig the S11 pmete of the i-filled TEM hon is shown (efeence impednce is 5 Ω). Fom the S11 plot, it cn be seen tht the TEM hon hs vey lge bndwidth S11 [db] Clculted Mesued Fequency [GHz] Fig. 3-2: Mesued (ed) nd clculted (blue) ntenn ptten in H- plne t 5 Ghz Fig. 3-21: S11 ntenn pmete of i-filled 3-28

29 Development of UWB GPR ntenns 3.5. Dielectic-filled TEM hon In ode to impove diectivity nd to educe the physicl size of the ntenn without limiting too much the bndwidth, we will fill the TEM hons with dielectic, chcteised by el eltive pemittivity ε nd loss tngent. Thnks to the dielectic, the popgtion speed of the TEM wve between the ntenn pltes will be divided by ε, in othe wods, the electicl length of the ntenn will be extended by fcto ε. Futhe the dielectic filling will educe the suge impednce of the ntenn by ppoximtely fcto of ε. To peseve the sme suge impednce s befoe the filling, one cn incese the ngle θ, which gin mens n impovement of diectivity Influence of the filling The infinite i-filled TEM hon is equivlent to pue TEM tnsmission line. Filling this tnsmission line would esult in n inhomogeneous qusi-tem stuctue. If this inhomogeneous tnsmission line becomes of finite length, the stuctue will be vey complex to model nd numeicl modeling would be moe ppopite. In ou ppoch we bsed the design on the i-filled ntenn nd on some pinciples fom mico stip theoy. To study the influence of the filling, the i-filled TEM hon fom pevious section ( L =1cm, ϕ =3 nd θ =11.2 ) is filled with silicone, chcteised by el eltive pemittivity ε of 2.89 nd loss tngent of.84 t 1 GHz. Influence on the suge impednce In the sttic nlysis ppoximtion, used in stip-line theoy, the chcteistic impednce of stip-line without dielectic deceses with fcto of ε, eff when substte with eltive dielectic constnt ε is dded. The effective eltive 3-29

30 Chpte 3 pemittivity ε is intoduced s coection on ε, eff by the fct tht the stuctue becomes qusi-tem fte dding the substte. An expession fo the effective eltive pemittivity s function of ε, w the width nd h the height of the mico stip is given by Hmmestd nd Jensen [24]:.555 ε + 1 ε 1 1h ε, eff = (3. 22) w The effective eltive pemittivity of the silicone used in the design is clculted using (3.22) s ε, eff = 2,55. Accoding to stip-line theoy, the suge impednce of the TEM hon will be educed by fcto ε, eff =1,6. The dielectic-filled TEM hon ntenn impednce is chosen to mtch the 5 Ω diving cble so tht the pt of the tnsient tvelling cuent tht bounces bck t the ntenn petue towds the excittion souce, will meet no mismtches on its wy bck. In this wy ntenn inging will be voided. If filled TEM hon with suge impednce of 5 Ω is puchsed, n empty TEM hon of 8 Ω is needed, which explins the choice of the suge impednce in Section 3.4. Fig shows the suge impednce of the TEM hon befoe nd fte the filling. The mesuement is done by TDR, using step with ise time of 45 ps. The mismtch t the feed point intoduced by the connecto is smll. The eflection coefficient is less thn 5%, which mens tht only.25% of the instntneous tnsmitted powe t the feeding point is eflected bck towds the souce. The suge impednce long the ntenn vies between 45 nd 55 Ω, so the eduction of suge impednce due to the filling is indeed found to be the expected ε,eff. Note lso tht the electicl length of the ntenn fte filling is extended. 3-3

31 Development of UWB GPR ntenns Zc[Ohm] Ω feed Connecto Antenn petue cm Fig. 3-22: Suge impednce of the TEM hon befoe (dotted) nd fte (solid) the filling Influence on the ntenn ptten nd the bndwidth The filling of the ntenn will cetinly hve its effect on the ntenn ptten nd the fequency bnd of the ntenn. As the electicl length of the dielectic-filled hon is extended, it is expected tht the ntenn will be moe diective nd tht the lowe cutoff fequency will decese. Fig shows the ntenn gin of the i-filled TEM hon nd the dielectic-filled TEM hon. The gin is mesued in n nechoic chmbe using clibted hon ntenns, so losses due to the eflection coefficient of the TEM hon is included in the gin. The whole fequency bnd of the dielectic-filled TEM hon is educed (ppoximtely by fcto ε, eff ) nd hs shifted slightly towds the lowe fequencies. Fig shows the ntenn ptten in the H-plne t 5 GHz. befoe nd fte the filling. As expected, the dielectic-filled hon is moe diective. This lso mens tht the dielectic-filled TEM hon will be less sensitive to extenl EM intefeence, which is n dvntge in this ppliction Gin [db] Feq [GHz] Fig : Antenn gin of the i-filled (dotted) nd the dielecticfilled TEM hon (solid) -9 Dielectic-filled.2.4 Ai-filled Co-pol in H-plne Fig : H-plne ptten t 5 GHz 9 befoe (dotted) nd fte the filling (solid) 3-31

32 Chpte Design of the dielectic-filled TEM hon In pevious section the i-filled TEM hon ws filled with dielectic to study the influence of the filling. Antenn mesuements showed tht even fte filling the lowe cut-off fequency ws still too high to meet ou design gols. So in next (nd finl) design of dielectic-filled TEM hon we hd to incese the length L of the ntenn. On the othe hnd, the ntenn s dimensions nd weight hd to sty limited to guntee good mobility of the ntenns (see design gols). This mens we hd to compomise on the length nd length L of 12 cm ws chosen. Futhe we noticed in pevious design poblem with the tnsition fom the unblnced cuent in the cox feed cble to the blnced cuent on the ntenn pltes. Without pecution n unblnced cuent component will be eflected by the ntenn feed nd ppe on the oute conducto of the cox. This cuent will be the oigin of unwnted TEM mode between the hon conductos nd the cox oute conducto. The cox oute conducto will ct s n ntenn nd mke the setup sensitive to pesons nd mteil in the diect neighbohood of the cox. This sitution cn be emedied by putting chokes (feite cylindes) ound the feeding cble. A choke hs the following equivlent scheme (Fig. 3-25) nd will hence suppess the cuent pulse on the cox exteio. R L Fig. 3-25: Equivlent scheme of choke Anothe solution in ode to void cuents on the cox is found in blun. The poblem hee is tht the blun hs to be bod-bnded. In litetue not mny exmples of bod-bnd bluns e found. In [3] some bod-bnd bluns e poposed, bsed on the use of feite coes. A bndwidth tio of 2:1 is obtined. In [31] n impednce mtching tnsition fom coxl line to blnced two conducto line is ccomplished by tpeing the oute conducto of the cox towds line. This 3-32

33 Development of UWB GPR ntenns esults in blun with bndwidth tio of 1:1, but the blun is too long nd not pcticl to use. In ou design we tied out simil kind of blun. The pinciple of this blun is bsed upon n electosttic esoning [8]. In fist pt of the ntenn, n unblnced configution of one ntenn plte on gound-plne is imitted (see Fig ()). The inne conducto of the cox is connected to the uppe ntenn plte. The cuents on the cox oute conducto cn only ise fom lekge of the field fom the uppe ntenn plte to the cox oute conducto ound the edge of the goundplne. The lge the gound-plne, the smlle the cuent on the cox oute conducto. This cuent would be zeo fo n infinite gound-plne. In the second pt of the ntenn, the blnced configution with the two symmeticl ntenn pltes is chieved (Fig (b)). In between gdul tnsition fom the unblnced towds the blnced configution is obtined by tpeing the ntenn pltes (see Fig. 3-27). ϕ θ 2θ () (b) Fig. 3-26: () unblnced configution (b) blnced configution The blun influences the suge impednce of the ntenn. The tnsition fom the unblnced towds the blnced configution will intoduce slight chnge in suge impednce long the ntenn. Expession (3.21) gives the suge impednce of complete TEM hon, so in the unblnced pt of the ntenn (one ntenn plte on gound-plne), the suge impednce is hlve the vlue given by expession (3.21). But the elevtion hlf-ngle (ngle between one ntenn plte nd the symmety plne of the ntenn) is 2θ nd not θ s in the blnced pt of the ntenn. The suge impednce of the second pt is clculted in noml wy using expession (2.21). In 3-33

34 Chpte 3 totl, the suge impednce of the unblnced pt is found to be little infeio to the suge impednce of the blnced pt of the ntenn. The ntenn impednce of the dielectic-filled TEM hon is chosen to mtch the 5 Ω diving cble. Doing so, the pt of the tnsient tvelling cuent tht bounces bck t the ntenn petue towds the excittion souce, will meet no mismtches on its wy bck nd ntenn inging will be voided. The pinciple seems to wok well fo fequencies in the bnd of the ntenn. Accoding to the optiml pex hlf-ngle of the i-filled TEM hon (see pevious section), the ngle ϕ is chosen to be 3 nd the physicl length L of the ntenn pltes, s ledy discussed, will be 12 cm. In this design, the ntenn pltes e not mde out of coppe sheet, but etched on pinted cicuit bod (PCB). Etching the ntenn pltes limits the weight of the pltes nd inceses the pecision in fbiction. Inspied by the wie model, the ntenn pltes e eplced by set of 41 wies (Fig. 3-27). The distnce between the wies is too smll to influence the ntenn chcteistics, but it foces the cuents to be dil nd it limits the sufce of conducting metl. The ltte is vey impotnt, when using the ntenns in combintion with metl detecto. Etching the ntenn pltes on PCB mkes it moe difficult to cete good nd obust feed-point. A schemtic epesenttion in Fig shows how the implementtion of the SMA connecto in the PCB is done. Coppe lye PCB substte SMA connecto Fig. 3-27: Lowe nd uppe ntenn pltes, etched on PCB Fig. 3-28: Schemtic epesenttion of feed point 3-34

35 Development of UWB GPR ntenns Tking into ccount the eduction of the suge impednce due to the filling nd tking into ccount the influence of the blun, n elevtion hlf-ngle θ (defined in the blnced pt) of 14.5 is clculted to mtch 5 Ω. This mens suge impednce in the blnced pt of 56 Ω. Summised, we hve L=12 cm, ϕ =3 nd θ =14.5, which leds to physicl ntenn petue of 12cm by 6cm Results of the dielectic-filled TEM hon Fig illusttes the suge impednce, mesued by TDR, long the ntenn. Note tht the x-xis is clibted fo speed of light. The gdul tnsition between the unblnced pt (suge impednce ound 55Ω) nd the blnced pt (suge impednce ound 65Ω) is obvious. The eduction of suge impednce due to the filling is gin found to be ε, eff. The discepncy between the mesued suge impednce nd the wnted one is due to the undeestimtion of the suge impednce by eqution (3.21) (see lso Fig. 3-16). Z [O] cm Fig. 3-29: Suge impednce of dielectic-filled TEM hon, mesued by time domin eflectomety Most of the ntenn chcteistics hve been mesued in the time domin in the instlltions of the Univesity of Technology in Delft. In Chpte 4 moe ccute desciption of the ntenn is given unde the fom of its impulse esponse. Fo completeness the S 11 pmete of the ntenn is shown in Fig The ntenn pttens in H- nd E-plne e given in Fig These pttens epesent the pekto-pek vlue of the dited impulse in ech diection, nomlised to one. In this 3-35

36 Chpte 3 definition of ptten, the 3dB bemwidth is 32 in the H-plne nd 65 in the E- plne S11[dB] Fequency [GHz] Fig. 3-3 : S 11 pmete of the dielectic-filled TEM hon Fig : P-t-p ntenn pttens in H-plne (solid) nd E-plne (dotted) Inteesting to see, becuse it is closely elted to the ppliction, is the min-bem esponse of two identicl ntenns. Fo this time domin mesuement the two TEM hon ntenns e ligned on boesight of ech othe. The tnsmitting ntenn is excited with the Gussin impulse. The eceived signl is ecoded by 6 GHz digitising oscilloscope. Fig shows the nomlised eceived voltge s function of time. At t <. 2 ns the ecoded signl shows smll negtive pe-pulse. This pepulse is due to the TEM mode tht goes in the opposite wy ound the ntenn pltes (see simultion in Fig. 3-14). As this wve tvels in fee spce, it will ive soone thn the wve tht tvels between the ntenn pltes. This negtive pe-pulse cn be educed by putting d-bsobing mteil (RAM) t the outside end of the ntenn pltes. The RAM lso educes the low fequency inging of the ntenn s we will see in Section 4.5. Aound t = 1. 6 ns the eflection on the ntenn end is visible. The signl in Fig is tken with the RAM ledy on the ntenn pltes. The fequency bnd of the whole system (tnsmitte ntenns - eceive), obtined by time-fequency tnsfomtion of the plot in Fig. 3-32, is fom 1 GHz up to 5GHz, so the postulted fequency bnd of 5 Mhz to 4.5 GHz is not eched. 3-36

37 Development of UWB GPR ntenns 1 Nomlized Amplitude Time [ns] Fig. 3-32: Nomlised mplitude of the min bem esponse 3.6. Summy In this chpte, we gve n oveview of the exiting GPR ntenns nd UWB ntenns, togethe with some design gols fo GPR ntenns in the demining ppliction. The TEM hon seemed to be good cndidte to meet these design gols. In fist step n i-filled TEM hon ws studied nd developed. An ccute model fo design puposes ws obtined fom the wie method. In ode to educe the physicl size of the ntenn nd to impove the diectivity, the ntenn ws filled with dielectic. Due to the lck of good model, the design ws bsed on the i-filled ntenn, ssuming tht the ntenn guides qusi-tem wve. TDR mesuements of the suge impednce showed slight but cceptble diffeence with the theoy. Antenn mesuements eveled tht the ntenns wee moe diective nd tht the fequency nge moved towds the lowe fequencies. The ntenn pltes wee eplced by set of wies, which mkes them suitble fo opeting in combintion with metl detecto. An ult-widebnd blun ws lso integted in the ntenn pltes. Although we e we tht the ctul ntenns cn still be enhnced, most of the design gols e met. The dimensions of the dielectic-filled TEM hons e smll, s needed fo the ppliction, nd they e cpble of diting nd eceiving vey fst tnsient pulses, without too much inging, which is of couse impotnt fo this ppliction. The clene the pulse, the clene the bcksctteed signl, nd the moe esy it will be to post-pocess nd intepet the dt. The ntenns, tht cn be used off-gound, e edy to be integted in n UWB GPR system. 3-37

38 Chpte 3 REFERENCES [1] D. J. Dniels, D. J. Gunton nd H. F. Scott, Intoduction to subsufce d, IEE Poceedings, vol. 135, Pt. F, no.4, pp , Aug [2] A. A. Lesti, A. G. Yovoy nd L. P. Ligtht, Adptive Antenn Fo Gound Penetting Rd, Eight Intentionl Confeence on Gound Penetting Rd, Gold Cost, Austli, pp , My 2. [3] T. T. Wu nd R. W. P. King, The cylindicl ntenn with noneflecting esistive loding, IEEE Tns. on Antenns Popgt., vol. AP-13, pp , My Coection, p.998, Nov [4] M. Knd, Time-Domin Mesuements in Electomgnetics. E. K. Mille, Ed., New Yok: Vn Nostnd Reinhold., 1986, ch. 5. [5] J. G. Mloney nd G. S. Smith, A study of tnsient dition fom the Wu-King esistive monopole FDTD nlysis nd expeimentl mesuements, IEEE Tns. on Antenns Popgt., vol. AP-41, no. 5, pp , My [6] T. P. Montoy nd G. S. Smith, A study of dition fom sevel bod-bnd loded monopoles, IEEE Tns. on Antenns Popgt., vol. AP-44, no. 8, pp , Aug [7] B. Stockboeck nd A. Vnde Vost, Electomgnetic modes in Conicl tnsmission lines with pplictions to the linely tpeed slot ntenn, IEEE Tns. on Antenns Popgt., vol. AP-48, no. 3, pp , Mch 2. [8] V. Rumsey, Fequency independent ntenns. New Yok: Acdemic Pess, [9] B. Stockboeck nd A. Vnde Vost, Copol nd coss-pol dition of Vivldi ntenns on dielectic substtes, IEEE Tns. on Antenns Popgt., vol. AP-48, no. 1, pp , Jn. 2. [1] R. L. Cel, The Chcteistic Impednce of Two Infinite Cones Of Abity Coss Section, IRE Tns. on Antenns nd Popgtion, vol. AP-6, pp , [11] M. Knd, Tnsients in Resistively Loded Line Antenn Comped with those in Conicl Antenn nd TEM Hon, IEEE Tns. on Antenns Popgt., vol. AP-28, no. 1, pp ,