DEEPRA BHATTACHERJEE, ABHRA MAZUMDAR, EEE 06 Project Guide: Ashok R!"s#$%, EEE De&t' CDMA stands for "Code Division Multiple Access." It is a form of spread- spectrum, an advanced digital wireless transmission technique. Instead of using frequencies or time slots, as do traditional technologies, it uses mathematical codes to transmit and distinguish between multiple wireless conversations. Its bandwidth is much wider than that required for simple point-to-point communications at the same data rate because it uses noise-lie carrier waves to spread the information contained in a signal of interest over a much greater bandwidth. !owever, because the conversations taing place are distinguished b" digital codes, man" users can share the same bandwidth simultaneousl". #$ %&A''I'( deals with the air interface between the )A*+ ,#A'*C+I-+# *,A,I.' / the M.)I&+ 0'I,. It also involves designing and optimi1ing the networ in order to provide a high qualit" service to customers. ,he ob2ective of the pro2ect is to plan, design / optimi1e a cell site at BA(GARPET' ,he main purpose of the planning is fewer base stations per networ, greater s"stem efficienc" and capacit", improved voice qualit" and a reduction or elimination of drop outs and multi path problems. All of which result in improved services and greatl" reduced networ cost. THE VARIOUS STAGES INVOLVED IN CDMA RF PLANNING & OPTIMISATION 3. MAR)ET *UR+E,: Depending on the mareting feedbac, a thorough surve" is conducted to define the area of coverage / potential candidate sites are listed out. ,he maret surve" also gives the area morpholog"-terrain and clutter information. ,he purpose of the maret anal"sis is to collect all the information from the maret that affects the design of the #$ 'etwor. ,he following information is gathered4 a5 Classification of various t"pes of morphologies in the area b5 Ma2or points of interest ,he e6act locations of the important places of the town were recorded using a (&.)A& %.*I,I.'I'( *7*,+M 8I, as shown below. -titude -o!"itude-octio! 39.:;<=< >;.3<;<?#ailwa" Crossing I 39.:;<@ >;.3>=<>(as Agenc" 39.:;;3@ >;.3>@>?#ailwa" Crossing II .' C/(T0(U/U* 1A+E TE*T4 Initiall", the signal propagation is studied in the proposed area transmitting a CA signal. ,he signal propagation is done using an omni directional antenna placed at the center of the proposed clutter 2' PRE3A(A-,*0*4 All the data collected b" the CA test is anal"1ed using a special pre-anal"sis software tool. ,,& uses a tool called %&A'+, +- provided b" MA#C.'I. It is used to design / evaluate networ problems. It provides with a comprehensive set of coverage and interference anal"ses of different cell sites. $I(0#+ 3. )A'(A#%+, %#.%.*+D *I,+ C.-+#A(+ CA&C0&A,I.' 0*I'( %&A'+, +- After Clutter classification is done, the location of the cell site is derived based on the information gathered. )ased on the information collected, design process is carried out as shown4 $I(0#+ 9. D+*I(' %#.C+** B%#+ A'A&7*I*5 ,he penetration median losses, standard deviations, and fade margins calculated on this basis below are taen directl" to the lin budgets to be used in later cell planning. C/MP/*0TE PR/BAB0-0T, /4 *ER+0CE CA&C0&A,I'( #+C0I#+D $AD+ MA#(I' +'-I#.'M+', ,7%+ BDM.#%!.&.(7E5 )0I&DI'( %+'+,#A,I.' .0,D..# C.M%.*I,+ ,.,A& M+DIA' &.**,D) *,D. D+-. F, D) *,D. D+-. F, D) A#+A A-AI&I)I&I,7 ,A#(+,,G $AD+ MA#(I' BD)5 D+'*+ 0#)A' )&D(. 9= ; ; :=GH>?GI+D(+ >.< 0#)A' )&D(. 3? ; ; :=GH>?GI+D(+ >.< *0)0#)A' )&D( 3= ; ; :=GH>?GI+D(+ >.< #0#A& )&D(. 3= ; ; :=GH>?GI+D(+ >.< ,7%ICA& -+!IC&+ ; @ ; :=GH>?GI+D(+ >.< Link Engineering (LINK UDGET) (Values are not shown ) !" Re#er$e Link %&ge' V(i)e P*r*+e'er Uni'$ U SU RBS Receiver Sensitivity dBm Subscriber Maximum Transmit Power Watts Subscriber Maximum Transmit Power dBm % MS Power used or R! Pilot % Reverse Pilot "verhead Penalty dB MS #ntenna $ain dBi MS Total %iRP dBm MS T&' %iRP dBm Buildin( Penetration !oss dB Body !oss dB Probability o cell ed(e covera(e % Slow )adin( Std* +eviation dB )ade Mar(in dB RBS #ntenna $ain dBi RBS Rx,Tx &able and &onnector !oss dB Sot-'ando $ain dB #nt +iversity $ain dB &+M# Traic !oadin( %ect % &+M# Traic !oadin( %ect dB Maximum #llowable Path !oss dB CDMA P,*nner-$ MAPL (ant (ain calculated in ant .attern ile) MS total %/RP dBm MS #ntenna $ain dBi )ade Mar(in dB Buildin( Penetration !oss dB Re.%ire& MS EIRP (MEIRP P,(') NOTE/ (&0 ,($$ & SHO 1*)'(r i$ )*,)%,*'e& in $i+%,*'i(n (MEIRP) ,he antenna selection is done from the data base provided b" various vendor companies. ,he antenna is a cross polar antenna with !-beam of JJ degrees / --beam of JJ degrees along with a gain of JJ.JJ d)d operating at the JJJ Mh1 range. ,he antenna plots are as follows4 $I(0#+ K. !.#IL.',A& %&., $I(0#+ @. -+#,ICA& %&., $I(0#+ ?4 )A'(A#%+, A'A&.( %#+DIC,I.' B,#IA&5 CDMA re5erse 6i!k !6%sis: ,he +- reverse lin anal"sis computes the probabilit" that a mobile user , with the abilit" to transmit at the ma6 +#% specified b" "ou, is able to close the reverse lin. #everse lin closure is computed as the probabilit" of receiving a mobile unitMs signal at the serving siteBs5 Aith sufficient mobile unit traffic channelBreverse lin target5 +bH't for acceptable call qualit". ,he reverse lin anal"sis incorporates the statistical nature of the processes and computes the probabilit" of closing the lin for each handoff t"pe. Closing the reverse lin in no handoff is predicted in terms of a single probabilit". Ahile the various DsoftE handoff configurations are computed in terms of 2oint probabilities that at least one of the contributing entities will receive adequate signal. CDMA 4or#rd 6i!k !6%sis4 ,his anal"sis verifies that the specified threshold B,NADD or ,ND#.% specified for each cellHsector5 for pilot channel coverage has been met or e6ceeded before a cellHsector is considered to be a valid contributor. ,he hand-off contributors are based on the strongest pilot signals at each bin. 7' R4 0MP-EME(TAT0/(: .nce the proposed cell site has been finali1ed, #$ implementation team swings into action wherein the proposed cell site is ph"sicall" implemented. $I(0#+ <. ,.A+# C.'*,#0C,+D A, )A'(A#%+, 8' (ET1/R) PER4/RMA(CE M/(0T/R0(G: .nce the base station is set up, its performance is continuousl" monitored. ,his process is called drive testing. ,he tool is called ,+M* I'-+*,I(A,.# from +#IC**.'. ,he drive test s"stem is placed in a vehicle / driven throughout the coverage area of potential site. It actuall" gives a subscriberMs point of view of the networ. It can evaluate call - processing operations, perform call processing functions and signal qualit" of received base station signal. Dri5e test 9or $ode6 c6i:rtio!: A significant number of t"pical sites are evaluated using the test transmitter / receiver to determine signal deca" rates and to get a fairl" accurate understanding of the effects of t"pical clutter in the area. ,ests are also conducted to evaluate the additional attenuation which the signal suffers during penetration of t"pical buildings and vehicles. ,he focus is on developing models generall" applicable to the area, not on performance of specific individual sites. Dri5e test 9or site e56utio!: Although propagation models for an area alread" have been refined coverage of a particular site is so critical, or its environment so variable due to urban clutter that it is essential to actuall" measure the coverage and interference it will produce. ,he focus is on this specific site. CDMA Per9or$!ce 0!dictors: 0!dictor ;<: 4ER =$#AM+ +#A*0#+ #A,+5 .' forward channel Breali1ed at handset5 .' reverse channelBreali1ed at base station5 $+# is e6cellent call qualit" Dsummar"E statistic $+# is the end Oresult of the whole transmission lin. If $+# is good then an" other problems arenMt having much effect If $+# is bad thatMs not the problem-- it is the end result of the problem Ae must investigate other indicators to get a clue what is going on. 0(D0CAT/R ;.: Ec>0o OAhat does it meanP Ah" canMt we 2ust use the handsetMs received power level to guide handoffsP )ecause it is a simple total #$ power measurement, the total of all sectors reaching the mobile. Ae need a wa" to measure the signal strength of each sector individuall", / we must be able to measure it quicl" / simpl". ,he solution is to use each sectorMs pilot BAalsh =5 as a test signal to guide handoffs. At the mobile, if the pilot of a certain sector is ver" strong / clean, that means we also should be able to hear a traffic channel on that sector, so handoff would be a good idea. If the pilot of a certain sector is wea, then we probabl" wonMt be able to get much benefit from using a traffic channel on that sector. 6' P/*T A(A-,*0*4 ,he drive test data is loaded into a post anal"sis tool. It assesses the end-to-end performance of the networ infrastructure. It can be used to anal"1e the benchmar drive data to compare the performance of their own networ with those of competitors. ,,& uses ,+M* D+*8CA, as the post anal"sis tool. Ae get the following parameters4 +cHIo, ,6 power, #6 power and $+#. $I(0#+ >4 CDMA %A#AM+,+#* $#.M ,+M* I'-+*,I(A,.# $I(0#+ ;4 +CHI. %&., Starting optimization on a new system R4 co5er"e co!tro6 ,r" to contain each sectors coverage, avoiding gross spillover into other sectors. Too6s 4 %' plots, !andoff state plots, mobile ,6 plots. (ei"h:or 6ist tu!i!" ,r" to groom each sectorMs neighbor to onl" those necessar" but be alert to special needs due to topograph" / traffic. ,ools4 %*MM data from mobilesQ propagation prediction. *erch #i!do# setti!"s $ind best settings for *#C!NAI'NA,N',N# +speciall" optimi1e *#C!NAI'NA per sector using collected finger separation dataQ has ma2or impact on pilot search speed. ACCE** 9i6ures, dro&&ed c66 !6%sis $inall", iterative corrections until within numeric goals. 35 %erformance monitoringHgrowth management 95 )enchmar e6isting performance K5 Identif" problem cells / clusters @5 &oo for signs of overload ?5 ,raffic trending / pro2ection ,hese steps must be continuousl" applied to guide needed growth. CONCLUSION ,he construction of the cell site at )angarpet was completed. ,he site became full" operational in the first wee of APR0-, .006' ,he pro2ect wor is carried out for ,A,A ,elecom. PARTIAL DISCHARGE MEASUREMENT FOR HIGH VOLTAGE CALE TERMINATIONS A#0' 80MA#, )!A#A,.(, *0MI, 80MA#, +++ =< %ro2ect (uide4 Asho #angaswam", +++ Department Due to recent advances in cable manufacturing technolog" it has been generall" recogni1ed that the %D in cable insulation itself is no longer a ma2or threat. Consequentl", attention has been paid to the cable accessories such as cable 2ointsHterminations where the comple6 structure and the construction can cause potential ha1ard to the whole s"stem. Ahen cable 2oints are sub2ected to !- testing, it is observed that the wea points in the insulation lie void, crac and other imperfections lead to partial discharges in the insulation. %artial discharge is defined as locali1ed discharge process in which the distance between two electrodes is onl" partiall" bridged that is the insulation between the electrodes is partiall" punctured. %artial discharge B%D5 measurements are universall" accepted as a technique giving some indication of the state of the insulation in high- voltage apparatus. Cable end users are een to adopt %D monitoring during assembl" and commissioning of s"stems. ,his pro2ect is aimed in setting up the partial discharge tester with the sensitivit" in the order of pico coloumbs, stud"ing the problems in partial discharge measurement, charting out calibration procedure and the stud" of %D testing for the !- cable industr". P*r'i*, &i$)2*rge in HV )*3,e 'er+in*'i(n Ahen !- cable terminations are sub2ected to !igh voltages in surve"s ,it is observed that the wea points in the insulation such as void, cracs / other material imperfection leads to partial discharge in the insulation which is catastrophic in nature. As per the standard BI+C <=9>=5, the partial discharge in cable terminations are measured in broad band frequenc" range / required sensitivit" level of whole measuring s"stem should be less than or equal to ? %ico coulombs. %artial discharges are, in general, a consequence of local electrical stress concentrations in the insulation or on the surface of the insulation. (enerall", such discharges appears as pulses having a duration of much less than 3 micro secRI+C<=9>=S.R3S %artial discharge measurement in !- cable terminations is an important step for demonstrating the qualit" of electrical insulators. 0ntil the %D measurement of cable terminations are less than %ico coulomb, the" can not put into service. $I(0#+ 3. ,+*, C7C&+ .$ A' +&+C,#ICA& !.-. +C0I%M+', PD Me*$%re+en' Se'4%5 ,he abilit" to measure low levels of partial discharge is referred to as sensitivit". ,()k4Di*gr*+ $I(0#+ 9. )&.C8 DIA(#AM .$ %D ,+*,+# *+,0% Manuacturin( Shi..in( #ssembly Runnin( Time Runnin( Time %nd o !ie-cycle Ty.e Test Routine Test Sam.le Test PD Te$' "n-!ine Monitorin( "-!ine Monitorin( "n-!ine Monitorin( M*n%1*)'%rer On4$i'e !- AC *uppl" AC *uppl" Control %D Detector )locing Capacitor ,est .b2ect Calibration In2ector Capacitor $I(0#+ K4 +C0I-A&+', CI#C0I, $.# %D M+A*0#+M+', ,he above figure shows a simple arrangement in which a gas filled void is present. ,he partial discharge in the void will tae place due to difference in electrical stress. Due to geometr" of the material the various capacitances are formed. $lu6 lines starting from electrode A and terminating at the void will form one capacitance C b3 and similarl" C b9 between electrode ) and the cavit". C c is the capacitance of the void. *imilarl" C a3 and C a9 are the capacitance of the health" portions of the dielectric on the two sides of the void. In the above figure C a TC a3 UC a9 VV.. Beq.35 C b T 9 3 9 W 3 Cb Cb Cb Cb + VV.. Beq.95 Closing of the switch * is equivalent to simulating partial discharge in the void as the voltage - c across the void reaches breadown voltage resulting into a current I c Bt5 to flow. #esistor # c simulates the finite value of current I c Bt5. TE*T *ET3UP It is used for measuring the magnitude of partial discharge $I(0#+ @4 ,+*, *+, 0% R ,+. - ,est ,ransformer, C m - Measuring Capacitor, C
- Coupling Capacitor, % - ,est .b2ect,A8--D - Measuring Impedance, D,M - %D MeterS & a0 V & a & b & c R c / c S & b0 & c & b1 & a0 P & 2 +TM #3V-+ & m T%" Power Su..ly Me*$%re+en' Pr()e&%re $igure K depicts ph"sical model and equivalent circuit of a dielectric with internal void as a source of partial discharges. ,he model comprises of the following elements4 C c O Capacitance of defect C b O Capacitance of health" dielectric in series with the defect C a O Capacitance of the specimen # O #esistance of the discharge path If the alternating voltage uBt5 is applied to the specimen and the gap does not brea down, the voltage u 3= Bt5 appears across the gap u 3= Bt5 T 5 Bt u C C C b c b + VVV Beq.K5 if the gap breas down at constant, polarit" independent voltage u 1 Bdischarge inception voltage5 and the discharge e6tinction voltage is 1ero, a pulse lie voltage u 3 Bt5 appears across the defect. At the same time, a pulse-lie current is developed in the circuit as a result of discharging capacitance C c through the gap resistance #. the amount of charge developed at the void is Xq i T
BC c U C b 5u 1 VVVBeq.@5 and the apparent charge, i.e. the charge measurable at the sample terminal is Xq T C b u 1 VVV Beq.?5 which leads
to Xq i T Xq b b c C C C + T Xq
B3U 5 b c C C VV Beq.<5 *ince the ratio C c H C b is unnown Bthe defect has unnown location and geometr"5 there is alwa"s the uncertaint" about the conclusions drawn from the discharge magnitude and pattern measured at the specimen terminals. 'onetheless, the simple equivalent circuit allows the following observation4 Ba5 %D current pulses appear as pulse train of alternating polarit" Bb5 ,he number of discharges increases with increasing test voltage whereas their amplitude remains nearl" the same Bc5 ,he amount of charge released in a void is not identical with that measurable at the specimen terminals as apparent charge Although the equivalent circuit is well suited to the e6planation of principles of partial discharge development it proves inadequate for modeling practical %D phenomena. ,here are a number of difficulties which arise ,he discharge inception voltage, 0 1 , ma" not be identical for positive and negative polarities of the e6citation voltageQ it ma" not be also identical from one pulse to the ne6t ,he discharge e6tinction voltage ma" not be 1ero, not identical for positive and negative polarities of the e6citation voltage or even from pulse to pulse %h"sics of discharge phenomena at the interface dielectricHdefect are not well understood to mention a few more general ones. As the result, identification of partial discharges in power apparatus is based on empirical methods. %ower suppl" is given to &- side of test transformer and the !- side is connected to the test ob2ect with the help of coupling capacitor and measuring capacitor as shown above. ,he partial discharge meter is connected across the measuring impedance. ,here are several variations of a commonl" adopted %D measuring circuit. .ne of them is shown in the figure below. ,he specimen C 6 is in series with input unit. ,he input unit contains a current transformer whose primar" side is connected to terminals A and + and the secondar" sides to terminals Amp and ) of the unit. ,he current transformer provides means for electrical isolation of the test circuit, e6posed to high voltage, from the electronic circuitr" of the measurement set. ,he input unit also contains an #&C filter. ,he current transformer together with the #&C filter form the detection impedance which shapes the signal sent to the detector amplifier. ,he amplifier can have a bandwidth of 3= 8!1 O ?== 8!1. ,his bandwidth is designed for good re2ection of ?=!1 signal as well as e6ternal high frequenc" interference. Calibration of Partial Discharge Measurement Accurac" of measurement depends on the accurac" of calibrators and we have to carr" out operational test. ,hese tests have to be carried out because of determination and eeping of characteristics of calibrators. It is of the following t"pes4 %D meter calibration Charge in2ection calibration Aindow setting *tandard specimen calibration &oad on %ower suppl" to %D ,ester )efore Calibration After Calibration Aithout light 33.> pc 33.= pc Aith light <.9 pc ?.? pc &alibration P+ Meter &alibration &har(e /n4ection &alibration Window Settin( Standard S.ecimen &alibration ,A)&+ 3. M+A*0#+D DA,A N(i$e in PD D*'* A ma2or bottlenec encountered with %D measurement is the ingress of e6ternal interferences Busuall" of ver" high amplitude comparable to %D signal5 that directl" affects the sensitivit" and reliabilit" of the acquired %D data. Ma2or e6ternal interferences encountered during %D measurement and their sources are4 Discrete spectral interferences BD*I5 from radio transmissions and power line carrier communication s"stems %eriodic pulse shaped interferences from power electronics or other periodic switching *tochastic pulse shaped interferences from arching between ad2acent metallic contacts, %D and corona from the power s"stem which can get coupled to the apparatus under test #andom or white noise from components !armonics from main suppl"R9S. INDPNDN! COMPONN! "N"L#SIS $IC"% ICA is a new class of neural networ algorithm, with actual wor starting onl" as late as late3::=Ms. *ince then it has found its worth in a variet" of applications, mostl" dealing with blind source separation and feature e6traction. %resentl" it is one of the most sought after topic for research in fields lie digital signal processing and neural networs. Its main assumption is that the individual signals are nongausssian. ,his is one of the main reasons for its dela"ed birth, prior to which all random signals with (aussian distribution was taen as a standard. ,he woring principle of ICA is to find the demi6ing matri6 of a set of mi6ed signals, b" estimating the mi6ing matri6 from the observed mi6ture and taing its inverse to get bac the demi6ed signals. It e6ploits the statistical correlation in the observed mi6ed signal to estimate the mi6ing matri6. ,he code was developed and simulated in matlab. *tart C is the Covariance matri6 of mi6ed signals. V is the matri6 of +igen vectors of C. Y is the Diagonal matri6 of +igen values of C. Ahitening4 x V x T / - ~ 9 3 = * +igen Decomposition4 p=3 and k== Choose initial random weight vector 5 = B w p with norm 3. B is the null matri6 of si1e of N N Tnumber of independent components 5 = B 5 = B 5 = B 5 = B 5 = B 5 = B p p p p T p p w w w w BB w w = = ) / (-u ) u ( (u) g ) / (-u u (u) G g(u) 9 e6p 3 9 e6p 9 9 9 = = A A
6 B5 w u , p = { } { } (k) w (u) g E - g( u) x E ) (k w p p = +3 5 3 B w 5 3 B w 5 3 B w + + = + k k k p p p 5 3 B 5 3 B 5 3 B 5 3 B 5 3 B 5 3 B + + = + + + = + k w k w k w k w BB k w k w p p p p T p p 5 B w 5 3 B w k k r p p + = r 3 + = k k Re.lace the .th column o B with wp (k50)* #ter whitenin(6 add wp (k50) T as the .th row o W* Then /ncrement . to .50 *top N p ,rue $alse ,rue $alse F*$' ICA A,g(ri'2+ $igure ?.3 4 *ource signal B*35 and 'oise source B*95 $igure ?.9 Mi6ed signals from *3, *9 B0nnown mi6ing ratio5 $igure ?.K +stimated signals 73,79 from mi6ed signals using $ast ICA %lease note that the non linear function (auss B"We6pB-"Z9H955 is used for the simulation. $rom the simulation, it is observed that $ast ICA algorithm can be considered as potential candidate for the noise removal from %D source signal, which is the long standing problem in the !- engineering industr". $ast ICA algorithm, as e6plained in RKS, was implemented using MA,&ab. $urther stud" of ICAMs application in !- engineering ma" be carried out, after careful observation of the theor" and limitations of Independent Component anal"sis concept, which can be found elsewhere. CONCLUSION ,hus %artial discharge B%D5 measurements are universall" accepted as a technique giving some indication of the state of the insulation in high-voltage apparatus. ,he stud" on partial discharge measurement for !- cable terminations were successfull" carried out at ,7C. electronics and the partial discharge tester has been installed in association with the !- sphere gap testing unit. Calibration on the testing equipment was done as per the calibration procedure, described in RKS. Initial testing was carried out over the samples given at the industr" and was found e6ternal interference Bnoise5 as the prominent bottlenec in the deplo"ment of %D testing in the product test c"cle of !- cable termination assembl". ,he application of filter circuits to the main suppl" of %D tester was of little use to solve this problem. After careful literature surve", we found that wavelet anal"sis of %D data was shown as successful simulation techniques in separation of partial noise from %D signal. Ae also found that a concept nown as Independent component anal"sis found similar applications in other areas such as #adar s"stems, astroph"sics, +C( and M#I. After sufficient understanding of the algorithm, we tried to simulate the popular $ast ICA algorithm over %D lie data and found satisfactor" initial results. !ence, we recommend the further stud" of Independent Component Anal"sis in the area of !- +ngineering, for more interesting results. ,he same was recommended to ,7C. +lectronics officials for further stud" in future. RE4ERE(CE*: 3. D%artial discharge measurementsE- I+C %ublication 9>= B3:;35 M.Massanori et al, *tud" on Application of wavelet anal"sis of degradation diagnosis of partial discharge in -oid, I+++ ? th intern Conf. on Conduction and )readown in solid dielectrics, pp K>3-K>?, 3::?. 9. I+C %ublication <=9>= 3:;3 Partial Dic!arg" #"aur"$"nt %nd "diti&n' K. %ro2ect #eport on D%A#,IA& DI*C!A#(+ M+A*0#+M+', $.# !I(! -.&,A(+ CA)&+ ,+#MI'A,I.'*E, Department of +lectrical and +lectronics +ngineering, CM# Institute of ,echnolog"- )angalore, 9==<.