First Measurements of Electron Temperature Fluctuations by Correlation ECE on Tore Supra

Transcription

1 First Masurmnts of Elctron Tmpratur Fluctuations by Corrlation ECE on Tor Supra V.S. Udintsv, M. Gonich, J.-L. Ségui, G.Y. Antar 1, D. Molina, G. Giruzzi, A. Krämr-Flckn, and th Tor Supra Tam Association Euratom-CEA, CEA/DSM/DRFC, CEA/Cadarach, F St. Paul-lz-Duranc, Franc 1 Cntr for Enrgy Rsarch, UCSD, 9500 Gilman Dr., La Jolla CA 9093, USA Association Euratom-FZJ, IPP Forschungzntrum Jülich GmbH, Grmany ABSTRACT. Elctron tmpratur fluctuation studis can hlp to undrstand th natur of th turbulnt transport in tokamak plasmas. At Tor Supra, a 3-channl htrodyn ECE radiomtr has bn upgradd with two channls of 100 MHz bandwidth and tunabl cntral frquncis allowing th shift of th plasma sampl volum in th radial dirction. With th sufficintly larg vido bandwidth and th long sampling tim, it is possibl to rduc significantly th thrmal nois and to idntify tru high frquncy componnts up to 00 khz from th cross-corrlation btwn ths channls. First rsults of tmpratur fluctuation masurmnts on Tor Supra ar rportd in this papr. 1. Introduction Studis of th plasma turbulnc aid in undrstanding th natur of th transport proprtis in fusion. Two gnral typs of fluctuations, lctrostatic and magntic, can b distinguishd in tokamak plasmas. Magntic fluctuations brak th nstd flux surfacs and, thrfor, nhanc th plasma transport. Elctrostatic fluctuations do not dstroy th nstd magntic topology, howvr, th nhancmnt of transport is du to th E B drifts from th fluctuating lctric filds [1]. In ordr to stablish a rlation btwn th fluctuations and transport, masurmnts of various fluctuating componnts, such as n ~, T ~, B ~ r and E ~ θ, as wll as knowldg of corrlations btwn thm, ar ndd. Masurmnts of lctron tmpratur fluctuations ( T ~ ) by mans of corrlation Elctron Cyclotron Emission (ECE) diagnostics ar a rlativly nw subjct in probing th natur of th turbulnt transport in fusion plasmas. Rsults obtaind at TEXT-U [ - 4], W7-AS [5], RTP [6], TEXTOR [7, 8] and Alcator C-Mod [9] hav yildd som intrsting information on microturbulnc proprtis both of lctrostatic and magntic origins. Howvr, no clar charactristics of th turbulnt fluctuations in diffrnt oprational rgims of tokamak plasmas hav bn obtaind up to prsnt days. Ths masurmnts gnrally rquir long intgration tims. Thrfor, thy will find thir bst application in xprimnts in which long(svral sconds) stationary plasmas ar attaind. On Tor Supra tokamak (R 0 =.40 m, a = 0.7 m, B T < 4 T, circular cross-sction), a 3-channl htrodyn ECE radiomtr has rcntly bn upgradd to includ two channls for tmpratur fluctuation masurmnts with a radial rsolution of about 1 cm. Exprimnts hav bn prformd for about thr wks during th 004 campaign. Th xprimntal stup, as wll as th first rsults, ar prsntd in this papr.

2 . Thortical background For optically thick plasmas (optical thicknss τ n >> 1, n is a harmonic numbr), ECE signal S ECE (t) consists of an avrag S ECE and a fluctuation part S ~ ECE ( t ). In thir own turn, ths quantitis ar proportional to th avrag plasma tmpratur T and to its fluctuating componnt T ~ plus th thrmal (or photon) nois N ~ (s also Eq. 3 latr on in this papr). In gnral, th masurd signal rsults from th intgration ovr thr spatial coordinats for th sampl plasma volum: radial r, poloidal θ and toroidal ϕ : f ( r, θ, ϕ, t) instrum S ~ ECE ( t ) = dr dθ dϕ S ~ ' + S ~ θ ϕ. (1) Hr, S ~ instrum is th instrumntal nois causd by vido dtctors and vido amplifirs. Th intgration on ral fluctuation componnt T ~ givs: T ~ T ~ = dr dθ '( r, θ,t ), () f θ implying no toroidal (ϕ ) dpndnc of T ~. On should b awar that th poloidal and radial intgration filtrs th masurd data. Thrfor, using th raw signals is not impossibl but th intrprtation is rathr difficult. This might hav bn a limitation in th past as most of th analyss wr don using th powr spctrum []. Howvr, th instrumntal nois S ~ instrum producd at th Intrmdiat Frquncy (IF) stag and/or by vido dtctors may b a (non-linar) function of th input signal, which maks th idntification of th ral tmpratur fluctuations much mor complicatd. Thrfor, vrification that th instrumntal nois dos not dominat th tru fluctuation spctra is vry important in corrlation tchniqu. In many corrlation ECE xprimnts [ - 7, 9], th following simplifid analysis of data to rtriv th information on tmpratur fluctuations has bn usd. In this analysis, th instrumntal nois componnt S ~ instrum is assumd to b insignificant and/or totally uncorrlatd. ECE signal S ECE (t) from th plasma sampl volum can b writtn in a simpl form as: T ~ ( t ) N ~ ( t ) S = + = + + = ECE ( t ) S ECE S ~ ECE ( t ) c(t T ~ N ~ ) ct 1+ +, (3) T T whr c is a proportionality (or calibration) factor for a givn ECE signal. From Eq. (3), an xprssion for th normalizd fluctuation componnt can b writtn as: S ( t ) ct T ~ ( t ) N ~ ( t ) =. (3a) T ECE S ~ ECE ( t ) = + ct T Th thrmal nois N ~ has nothing to do with th ral tmpratur fluctuations T ~, thrfor, it is ncssary to rduc its influnc in ordr to dtrmin tru fluctuations can b

3 dtrmind. This can b achivd by cross-corrlation btwn two ECE signals whos tmpratur fluctuations ar corrlatd whil th nois is uncorrlatd, or vn by autocorrlation for a singl ECE signal (if th vido bandwidth B V of th ECE radiomtr is much largr than th spctral width of tmpratur fluctuations) [5, 8]. Th schm in which sparat frquncis ar coming from th sam sampl plasma volum to obsrv cohrnt tmpratur fluctuations and to dcorrlat th thrmal nois, is shown in Fig. 1(a). If two spatial volums ar diffrnt (non-ovrlapping), th cross-corrlation analysis is possibl if fluctuations propagat in th plasma (Fig. 1(b)). Th phas vlocity v ph and th distanc btwn volums r dtrmin th tim dlay τ : r τ r =. (4) v ph r From th cross-corrlation and th cross-phas btwn two spatially sparatd channls, corrlation lngths, th wav-numbrs and, finally, th disprsion rlations k ( f ) can b obtaind [10]: k( f )r = πfτ. (5) r Th cross-corrlation function ovr tim priod T for two normalizd fluctuating componnts, S ~ 1 ( t ) and S ~ ( t ), and for a givn tim lag τ, can b writtn as [11]: R T 1 ( τ ) = S ~ 1( t )S ~ ( t+ τ ). (6) T 1 dt 0 For τ = 0 (zro tim lag), Eq. (6) taks th following form: T ~ ( t )T ~ ( t ) N ~ ( t )N ~ ( t ) T ~ ( t )N ~ ( t ) R 1(0 ) = 1( t ( t ) = T1T T1T T1T S ~ )S ~ T ~ ( t )N ~ 1( t ). (7) T T Th particular cas S ~ 1 ( t ) = S ~ ( t ) = S ~ ( t ) dfins th autocorrlation function. Whnvr both signals ar coming from th sam plasma volum, on can assum T ~ 1 T ~ = T ~. Bcaus th thrmal nois is not corrlatd, th last thr trms in th Eq. (7) can b nglctd [5]: 1 T T ~ ( t ) R ( 0 ) =. (7a) 1 From Eq. (7a), an xprssion for th root man squar (rms) valu of th normalizd tmpratur fluctuations can b obtaind: T ~ ( t ) = R1(0 ). (8) T

4 S ~ 1 S ~ r a) b) Figur 1. (a) - th thrmal nois dcorrlation with a singl lin of sight ECE systm; (b) cross-corrlation btwn two spatially sparatd plasma volums aids in dtrmination of th disprsion rlation k(f). Th cross-corrlation mthod dos not rquir ECE signals to b absolutly calibratd. In many cass, bsids th rms stimation of th fluctuations amplitud, it is usful to calculat th cross-spctral dnsity spctrum for two discrt signals, which is th invrs Fourir transform of th cross-corrlation function. Th on-sidd cross-spctral dnsity for two ECE signals can b dfind as follows: CSD jϕ( π f ) 1 CSD1( πf ) =, (9) whr CSD 1 ( π f ) is th amplitud of th cross-spctral dnsity, and ϕ ( π f ) is th cross-phas btwn two signals. Ths quantitis can b usd to stimat th wavnumbr of fluctuations and to driv th disprsion rlation, as it is dscribd by Eq. (5). If broadband mods with a bandwidth B BB xist in th plasma, thy caus a pak in th cross-corrlation function that dcays according th following tim scal [5]: τ = 1 ln dcay B π. (10) BB 3. Diagnostic st-up Th Tor Supra htrodyn radiomtr [1] has bn rcntly upgradd to hav 1GHz spacd, 500 MHz bandwidth 3 masuring channls (Fig. ). It is bing usd on th Tor Supra tokamak to masur th lctron cyclotron mi s si o n i n th frquncy rang GHz for th first harmonic ordinary (O) mod ( E B, k B ) and GHz for th scond harmonic xtraordinary (X) mod ( E B, k B ). Th radial rsolution is ssntially limitd by ECE rlativistic ffcts rlatd to lctron tmpratur and dnsity

5 and not by th channls frquncy spacing. Th radiomtr can act simultanously in two mods: 1. slow acquisition mod during all th plasma duration: 3 channls 1 ms sampling without aliasing (bandwidth B V1 = 400 Hz);. fast acquisition mod during tim plasma windows triggrd by plasma phnomnon: 3 channls 10 µs sampling without aliasing (bandwidth B V = 40 khz). A prcis absolut spctral calibration is prformd outsid th tokamak vacuum vssl by using a 600 C black body sourc. Using analytical formulas, post-puls data procssing taks routinly into account th total magntic fild and th Maxwllian rlativistic radial shift to improv radial location stimat. Ths formulas ar compatibl with ral tim procssing in ordr to us ECE data in fdback control loops. To prform masurmnts of th lctron tmpratur fluctuations, on radiomtr channl is split into two (Fig. 3). On ach of ths two channls, an (IF) YIG filtr with bandwidth around 100 MHz is introducd. Its cntral frquncy is rmotly monitord by a drivr (dsignd at th IPP Forschungszntrum Jülich, Grmany) btwn 6 and 18 GHz, allowing to shift th obsrvation volum in th plasma radially. Th IF filtrs, a Schottky diod dtctor and a vido amplifir with 00 khz bandwidth ar placd insid th isolation box. Th acquisition is don without aliasing ffcts. It is important to mntion that th rippl in Tor Supra rachs 7% at th dg lading to a mismatch btwn th iso-b lins, which dtrmin th localization of th invstigatd volum, and th fild lins to which turbulnc is prpndicular (Fig. 4). Consquntly, for small distanc btwn th two channls th sam fild lin would cross th two volums. Th minimum distanc that can b rachd without including this spurious ffct is dtrmind as: R> wece tan( α ). (11) Th bam waist w ECE is qual to 5.4 cm and is takn to b th sam for th two channls. Th angl α is th diffrnc btwn th curvatur of th iso-b and th fild lins. Th angl of th fild lins is ngligibl with rspct to that of th iso-b lins and is thus nglctd lading to n xprssion of α of th form: tan( α ) = N sin( N φ ) B. (1) c c ϕ Hr, N c is th numbr of toroidal fild coils (qual to 18), φ is th ECE radiomtr viwing angl with rspct to th port axis (φ = 3.5 dgrs) and B ϕ is th magntic rippl that varis btwn 0.18 (in th cntr) and 0.93% (at r/a = 0.4), dpnding on th radial position. This lads to α btwn 1.6 and 8.5 dgrs, rspctivly. Consquntly, th contribution of turbulnt fluctuations on th sam flux surfac to th two channls can b nglctd for distancs gratr than 1.5 mm for masurmnts in th plasma cntr and 8 mm at r/a = 0.4.

6 30 db - 18 GHz db X mod Ghz fc=110.5 GHz GHz Pin Switch f IF = Ghz Stp= Ghz IF B -3db = 500Mhz F band * 30 db 5db / db W band 30 db O mod GHz * fc=114 GHz fc=94 GHz 9 GHz Pin Switch f IF = Ghz Stp= Ghz IF B -3db = 500Mhz Gola Antnna fc=114 GHz fc=94 GHz 9 GHz fc=78 GHz 30 db - 18 GHz 5 db / db f IF = Ghz Stp= Ghz IF B -3db = 500Mhz db 76 GHz * Gyrotron frquncy rjction (118 Ghz) f IF = Ghz Stp= Ghz IF B -3db = 500Mhz RF front nd IF and Vidéo parts Figur. A principl schm of th 3-channl htrodyn radiomtr on Tor Supra. This figur is takn from [1] with kind prmission of th authors. Figur 3. Th principal lctronic schm to masur lctron tmpratur fluctuations on Tor Supra.

7 Figur 4. A schmatic viw of th intgration volum in th quatorial plan as wll as two iso-b lins and on fild lin. As th distanc btwn th two dcrass, th mixing of th information coms from not only th width of th bams but also from th diffrnc btwn th iso-b and th fild lins. 4. Estimation of th masurmnt rror 4.1 Signal to nois ratio for th Tor Supra radiomtr ECE radiation coming from th plasma is attnuatd by th wavguids losss and millimtric attnuator for total A = 4 db + 17 db = 1 db. Th sourc tmpratur T sn by th radiomtr can b dfind as follows: T = TN, (13) A' T + whr A is about 100 (for 1 db attnuation), and T N is th quivalnt radiomtr tmpratur nois. Th nois quivalnt powr (NEP) for on polarisation dirction du to fluctuations of th intnsity of th thrmal radiation quals to [13]: 1 NEP= kt B B IF ), (14) ( v whr B IF, B v ar IF (SSB) and vido bandwidths, rspctivly. If on assums T N to b about K (du to th mixr and th first IF amplifir; th Shottky dtctor nois is ngligibl in cas of strong IF amplification), th minimum dtctabl tmpratur diffrnc Tmin is obtaind whn th black body radiatd powr is qual to th NEP (i.. whn signal to nois ratio is unity):

8 T T 1 min k BIF = k( + TN )( BIF Bv ) A' A'. (15) If T / A' >> T N, Eq. (15) can b writtn as follows: T T min = B B v IF. (15a) For B v = 00 khz and B IF = 100 MHz, on gts th minimum dtctabl tmpratur ratio of about 4.4 %. If T = 3 kv, T is stimatd to b 13 V. 4. Statistical nois lvl min In ordr to rduc th statistical rror blow th cohrnt tmpratur fluctuation amplitud, a long sampling tim is rquird. According to [11], th diffrnc of R 1 (s Eq. (6)) (in cas S ~ 1 ( t ) and S ~ ( t ) consist of whit (Gaussian) nois with a larg bandwidth B) can b writtn as: Var 1 = R ], (16) BT [ R ( τ )] [ (0 )R (0 ) + R ( ) 1 τ whr T is th total intgration tim. For th prsnt corrlation ECE diagnostic on Tor Supra, th following ratio btwn th sampling frquncy M/T and th vido band B V is valid: B V T M/3, whr M is th total numbr of sampls. For th cross-corrlation function, on can writ: ( T ~ ) + N ~ T ~ T ~ + ( t ) 3 Var[ R1(0 )] = = ; (17) T M T 1/ ~ T ~ ( t ) 3 T ~ ( t ) N 1 1 T M T T ~ = + + ; (18) or, for fluctuations with th wak amplitud: T ~ ( t ) T ~ ( t ) = T ~ ( t ). (19) T T T From Eqs. (18) and (19), on finally gts: 1/

9 T ~ T ~ ( t ) ~ 3 ( t ) N 1 1 T M T T ~ = + +. (0) It can b sn that th statistical rror dpnds on 1/ M and not on 1 / 4 M, as can b found lswhr in th litratur [, 5]. For xampl, to rsolv th fluctuation amplitud of 0.%, it is ncssary to hav M = 10 6 sampls to gt th rror lvl of 0.1%. For th 6 fluctuation amplitud of 0.1%, 3 10 sampls ar rquird for th sam rror lvl of 0.1%. Error stimation (standard dviation) for ral signal squncs is shown in Fig. 5 and dviats slightly from both scaling laws. 1/ a) b) Figur 5. Th numbr of sampls vrsus rlativ tmpratur fluctuations amplitud to rsolv th dsird absolut rror (a), and rror stimation (standard dviation) for 1 sris of ral corrlation ECE data with diffrnt numbr of sampls (b). For (b), ECE signals from two succssiv Tor Supra shots hav bn takn. Solid lins rprsnt 1 / M and 1/ 4 M scaling laws for th rror stimation. On additional, though vry important, not has to b mntiond. Th sampling rat of th diagnostic is 1 MHz, howvr, th vido bandwidth is limitd by 00 khz. Thrfor, an ovrsampling ffct is prsnt in corrlation ECE masurmnts on Tor Supra. Bcaus of this, th high-frquncy tail of th spctra (> 00 khz) cannot b usd to analys th spctral charactristics of fluctuations. Prcautions should also b takn whn looking to th cross-corrlation functions at tim lags smallr than 5x10-6 s. Rsampling th input ECE signals at 50 khz (or smallr frquncy) is dsirabl. 5. First masurmnts Dpnding on currnt in th poloidal coils and, hnc, on th toroidal magntic fild in th plasma cntr, a fw masuring scnarios ar possibl for th corrlation ECE on Tor Supra. Figur 6 givs th radial rgion of obsrvation by th corrlation ECE as a

10 Figur 6. A radial rang of corrlation ECE masurmnts for diffrnt valus of th poloidal coil currnt on Tor Supra. function of th currnt in th poloidal coils: I coils = 1100 A corrsponds to th cntral toroidal fild of about 3.4 T, and I coils = 1300 A corrsponds to B T = 4 T. Th valu of th Shafranov shift dpnds on hating rgim and typically varis btwn 5 and 1 cm. 5.1 Low-frquncy MHD tst studis by 3-channl (non-corrlation) ECE radiomtr In ordr to tst th thortical principls dscribd in Sction, dtrmination of th wll known harmonic componnt, such as th m/n = 1/1 prcursor to sawtth, has bn don by mans of cross-corrlation btwn two channls of th 3-channl ECE radiomtr (non-corrlation st-up). Figur 7 shows a simpl xampl of th powr spctral dnsity stimation by FFT applid dirctly to a typical sawtooth signal with th m/n = 1/1 magntohydrodynamic (MHD) prcursor activity. Figur 8 shows th crosscorrlation btwn two nighbouring channls of th prsnt 3-channls htrodyn ECE radiomtr (sampling rat 83 khz) for th sam shot but for filtrd signals, in ordr to gt rid of th sawtooth ris. It can b sn that th shap of th signal can influnc corrlation functions and rsulting spctra. If th amplitud of fluctuations (and, thrfor, corrlation btwn channls) is strong, lik in th xampl shown in Figs. 7 and 8, th ral structur in th plasma bcoms visibl vn without smoothing th input signals. Howvr, to uncovr a prsnc of som broadband high-frquncy mods that may xist in th plasma but hardly dtctabl bcaus of th nois, filtring or smoothing of th input signals is dsirabl, so th low-frquncy stp-lik componnts (such as sawtth) and high-frquncy nois contribution would not influnc th tail of th spctra.

11 a) b) d) c) Figur 7. A powr spctral dnsity stimation (d) for a singl ECE signal during sawtooth activity with th strong m = 1 prcursor (a; slctd signal is indicatd by th black arrow). Vrtical lins in (a) show th tim intrval for which th PSD has bn stimatd. Typical cross-corrlation functions for th whol tim window in (a) ar givn in (b) and (c), and dpict a strong influnc of sawtth. Black solid arrows in (c) indicat th propagation of th m = 1 prcursor btwn two plasma sampl volums. Dashd arrow givs th tim lag btwn two succssiv prcursor oscillation of about s, from which th rotation frquncy of about 3 khz is dducd. 5. Broadband mod idntification by corrlation ECE It is known [3, 14] that th gradint of th lctron tmpratur is on of th nrgy sourc of th turbulnt fluctuations. A radial scan at th HFS on Tor Supra has bn prformd btwn r/a = (-0.6) (-0.1) for th cntral magntic fild B T = 3.3 T. A radial sparation btwn two corrlation ECE channls of about 1 cm has bn chosn. An intgration tim of s has bn usd. A qualitativ comparison btwn cross-powr spctral dnsitis for two xtrm cass, on insid th sawtooth invrsion radius at r/a = (- 0.18), and on in th T-gradint rgion at r/a = (- 0.43), has shown a noticabl diffrnc in th frquncy rang of khz (s Fig. 9). Th origin of th bump for th CSD in th gradint rgion is not clar yt and may b ithr an ffct of th Dopplr shift du to th poloidal plasma rotation transposing spctral componnts to th highr frquncis, or

12 Figur 8. Exampl of cross-corrlation function and cross-powr spctral dnsity for sawtooth-rmovd ECE signals. Black arrows on th CSDs giv th prcis rotation frquncy of about 3 khz for th prcursor. Not that th CSD slop at frquncis blow khz is absnt in cas of filtrd ECE signals. a footprint of th broadband turbulnt mod of lctro-magntic natur. Th lattr is mor likly, bcaus th bump fatur has much wakr amplitud at r/a > 0.5, whilst on would xpct th Dopplr ffct to prsist with th highr amplitud with incrasd valu of r/a. Similar obsrvations hav bn don at th LFS, too. From th cross-corrlation functions in Fig. 9, it is sn that th broadband structur with th maximum at 1.9x10-5 s is suprimposd on th low-frquncy MHD componnts with th bandwidth blow 10 khz. From Fig. 10, an stimation of th rlativ fluctuation amplitud for a statistical trustabl tim (and frquncy) intrval can b mad: for r/a = (- 0.43), T ~ T is about 0.35% for th broadband structur. For th shot with r/a = (- 0.18), th fluctuation lvl is much smallr and wll blow th nois lvl (most likly causd by th Schottky diod dtctor). Anothr xampl is shown in Fig. 10(c), in which corrlation ECE monitors th plasma cntr, wll insid th sawtooth invrsion radius. It can b sn that, although th nois componnt is much wakr than for th shot shown in Fig. 9(d), th ral fluctuations ar significantly rducd in thir amplitud, compard to th T -gradint rgion, and stimatd to b about 0.%.

13 a) b) c) d) Figur 9. Cross-spctral dnsitis (in arbitrary units) for corrlation ECE at diffrnt radial location: blu in th T -gradint rgion at th HFS, rd insid th sawtooth invrsion radius at th HFS. Plot (a) shows th CSD for ECE signals hav bn dtrndd bfor FFT calculations, in ordr to prvnt low-frquncy componnts du to MHD and high-frquncy componnts du to th instrumntal nois to influnc th lvl (offst) of th spctra; plot (b) is for th CSD out of high-pass filtrd (< 0 khz) ECE signals (to liminat th low-frquncy MHD ffcts totally). Cross-corrlation function for th cas of r/a = (c) shows a prsnc of high-frquncy componnts that ar nhancd in th amplitud, compard to th nois pak (black arrow) at zro tim lag; for th cas of r/a = (d), th nois componnt is dominant. Grn horizontal lins in (c) and (d) shows an offprint of th low-frquncy mod of MHD origin with a bandwidth blow 10 khz, on which high-frquncy componnts ar suprimposd. From th phas shift of th broadband structur, th radial propagation vlocity is found to b about 500 m/s. Mor xprimnts ddicatd to study broadband highfrquncy mods ar plannd in upcoming campaigns on Tor Supra.

14 a) b) c) Figur 10. A (biasd) cross-corrlation function for high-pass filtrd (0 khz) and mannormalizd ECE signals (1 MHz sampling rat) at r/a = (a), and for th sam signals but rsampld at 50 khz (b). Vrtical dottd lins show a statistically trustabl tim intrval: 5x10-5 s (and abov) corrspond to 0 khz (and lowr) in frquncy fild both in (a) and (b), and 5x10-6 s corrspond to 00 khz (limitation by th vido dtctor bandwidth in (a)). Nois pak at zro tim lag (and, thrfor, abov 00 khz) is shown by black arrows. Plot (c) is for th cas in which corrlation ECE monitors plasma insid th invrsion radius, nar th plasma cntr. For this shot (#33505), th nois componnt (black arrow) is much smallr, compard to th shot prsntd in Fig. 9(d).

15 6. Summary and futur plans Th fasibility of lctron tmpratur fluctuation masurmnts on Tor Supra has bn invstigatd by mans of radial corrlation ECE diagnostic. Though th rang of ths masurmnts was limitd by th toroidal magntic fild (and also by th rstrictd possibility to chang th cntral frquncy of th YIG filtr), and only a fw xprimnts wr conductd so far, first obsrvations of th broadband frquncy structur hav bn don in th T -gradint rgion, compard to th plasma cor insid th sawtooth invrsion radius. From th collctd data, thr is no possibility yt to driv th disprsion rlation and to idntify th origin of this structur. In upcoming xprimntal campaigns, th following studis by mans of corrlation ECE radiomtr ar proposd on Tor Supra: 1. to prform a dtaild radial scan both on th LFS and th HFS to idntify th radial xtnt, corrlation lngth and dcorrlation tim of th broadband mod and, thus, to driv its wavnumbrs;. to study th rspons of turbulnc amplitud at diffrnt valus of th plasma dnsity (.g. at diffrnt collisionality); 3. to vrify th rsults rportd in [7] that turbulnc is sparatd insid and outsid th transport barrir, in particular with rspct to th rational q surfacs; 4. to compar th tmpratur fluctuation masurmnts with th dnsity fluctuation masurmnts to obtain a rathr comprhnsiv pictur of turbulnc in th plasma cor; In th yar 005, a furthr upgrad of th corrlation ECE radiomtr is plannd. Th futur diagnostic will hav svral tunabl channl sts allowing dtaild studis of tmpratur fluctuations ovr a widr radial rang. Rfrncs [1] W. Horton, Drift Wav Turbulnc and Anomalous Transport, publishd in Basic Plasma Physics, Vol., d. by A.A. Galv and R.N. Sudan, Elsvir Scinc Publishrs B.V., p.383 (1984). [] G. Cima t al., Phys. Plasmas, 70 (1995). [3] C. Watts, R.F. Gandy, Phys. Rv. Ltt. 75, 1759 (1995). [4] C. Watts, R.F. Gandy, T.D. Rmpl, G. Cima, Rv. Sci. Instrum. 66, 451 (1995). [5] H.J. Hartfuss, T. Gist and M. Hirsch, Plasma Phys. Control. Fusion 39, 1693 (1997). [6] B.H. Dng t al., Rv. Sci. Instrum. 7, 368 (001). [7] V.S. Udintsv t al., in Proc. of th 30 th EPS Conf. on Control. Fusion and Plasma Phys., St.-Ptrsburg, Russia, Eur. Conf. Abstr. Vol. 7A, P.314 (003). [8] V.S. Udintsv, Elctron Tmpratur Dynamics of TEXTOR Plasmas, PhD Thsis, Univrsity of Utrcht (003), Intrnt: [9] C. Watts t al., Nucl. Fusion 44, 987 (004). [10] C.P. Ritz t al., Rv. Sci. Instrum. 59, 1739 (1988). [11] J.S. Bndat, A.G. Pirsol, Random Data: Analysis and Masurmnt Procdurs, Wily, NY (1986). [1] J.-L. Ségui t al., Upgradd ECE Radiomtr on th Tor Supra Tokamak, in Proc. of th EC-13 Workshop, Nizhny Novgorod, Russia (004). [13] H.J. Hartfuss, Cohrnt Vrsus Incohrnt Dtction Schms, in Proc. of th EC-7 Workshop, Hfi, China, p. 67 (1989). [14] D.E. Nwman t al., Phys. Fluids B 4, 599 (199).