Scrambling Code and Channelisation Code Parallel Generation for WCDMA on Vector Core

Transcription

1 teratioal oferece o omputer ciece ad ervice ystem ( 2014) cramblig ode ad haelisatio ode Parallel Geeratio for WDMA o Vector ore Yogguag Yu Jiagsu Uiversity of ciece ad Teology Zhejiag, hia @qq.com Hogya You R &D stitute of Microelectroics ZTE orporatio Najig, hia youhy84@163.com Abstract a Widebad ode Divisio Multiple Access (WDMA) eviromet, ea user is assiged a uique complex-valued spreadig codes to ecode its iformatiobearig sigal. The spreadig sequece composed of scramblig codes ad aelisatio codes spread trasmitted symbols ad despread received ips. This essay presets the parallel geeratio method of scramblig code ad aelisatio code for WDMA o vector core. Keywords- scramblig code;aellisatio code;spreadig code;parallel geeratio; WDMA; vector core;. NTRODUTON The DMA teology is widely used i 3G wireless commuicatios. cramblig code is used i WDMA ad DMA2000.haelisatio code is used i WDMA, DMA2000 ad TD-DMA. cramblig code is geerated from Gold sequece while aelisatio code is geerated from OVF(Orthogoal Variable preadig Factor) code[1]. The typical implemetatio methods of 3G system adopt the method of A(Applicatio pecific tegrated ircuit), whi is optimal i the ip area ad power cosumptio, while deficiet i log desig period, high cost, low flexibility, difficult upgradatio ad high desig risk. Thus the advatage of A ip is merged oly i case of large quatity ad regardless of ip cost. With the fast developmet of wireless commuicatio, the 3G system like WDMA, DMA2000 ad TD-DMA requires a multimode coexist with 4G system like TD-LTE ad LTE-FDD i base statios or with WLAN ad Bluetooth i termial statios. Ad it will be a large waste i the ip area ad power cosumptio if every system uses a A ip. Besides, takig it ito cosideratio that the typical evolutio of 4G system requires cotiuous evolutio of base statio ad termial statio ips, it will icrease ip cost by usig A ips. To avoid those demerits, the DR(oftware Defied Radio) seme defied by usig R (Reduced structio et omputer) should ejoy priority i developig base statio ips that require low power cosumptio, rather tha termial statio ips that require high power cosumptio. Yet, with the evolutio ad bledig of those commuicatio systems, aloe with the mature applicatio of VORE(Vector ore), whi is multicore, efficiet i power cosumptio ad fast calculatio, the DR seme of R teology will be a ew tred of itelliget termial ips with its advatage of low desig period, low desig cost, high flexibility, easy upgradatio, low desig risk, smaller ip area ad low power cosumptio due to multi commuicatio systems itegratio. Mature resear ad applicatio have come ito use with the help of large ad simple data processig by 4G system as TD-LTE/LTE-FDD, whi are easily realized by VORE. This essay discusses the teology of realizig 3G DMA o VORE. The trasport ael ecodig & decodig ad physical ael trasmitig & receivig i physical layer have the large amout of computatio. The ovolutioal code ad Turbo code are used to trasport ael ecodig ad decodig, whi ca be kept as hardware acceleratio i vector core. While the physical ael trasmitig ad receivig vary i differet commuicatio systems, it is more reasoable for implemetig i the vector core.. The spreadig codes composed of scramblig codes ad aelisatio codes spread trasmitted symbols ad despread received ips. this paper, it discusses the parallel geeratio method of scramblig code ad aelisatio code for WDMA o vector core.. PEUDO-RANDOM EUENE Pseudo-radom sequeces are sequeces of 1s ad 0s geerated by a algorithm so that the resultig umbers look statistically idepedet ad uiformly distributed. A radom sigal differs from a pseudo-radom sigal i that a radom sigal caot be predicted. A pseudo-radom sigal is ot radom at all; it is a determiistic ad periodic sigal kow to both the trasmitter ad the receiver. Eve though the sigal is determiistic, it appears to have the statistical properties of sampled white oise. To a uauthorized listeer, it appears to be a truly radom sigal. Pseudoradom sequeces are typically geerated usig a system of liear feedback shift registers (LFRs). The LFR geerators produce a sequece that depeds o the umber of stages, the feedback tap coectios, ad the iitial coditios.[2] a DMA seme, all users trasmit o the same frequecy ad are differetiated by their uique spreadig codes composed of scramblig codes ad aelisatio codes. The spreadig codes appear to have the statistical properties of sampled white oise, so the spreadig sigal appear to have the statistical properties of sampled white The authors - Published by Atlatis Press 438

2 oise, whi makes spreadig sigal easier to be trasmitted i the wireless ael. The receiver correlates the received sigal with a syroously geerated replica of the spreadig code to recover the origial iformatio-bearig sigal. [3] The third-geeratio partership project (3GPP) specificatios defie how these dowlik complex scramblig codes ad aellisatio codes are geerated.. PREADNG FOR WDMA WDMA dowlik trasmissios, all dowlik physical aels are subjected to spreadig with a complexvalued spreadig code composed of scramblig code dl, ad aelisatio code.f,.m. Fig 1(a) illustrates the spreadig operatio for all physical aels. Fig 1(b) illustrates the spreadig operatio for someoe physical ael.the spreadig operatio icludes a modulatio mapper stage successively followed by a aelisatio stage, a combiig stage ad a scramblig stage. dowlik physical ael omeoe dowlik physical ael P Modulatio Mapper,F,m j +j (a) preadig for WDMA all dowlik physical aels P Modulatio Mapper j +j dl,,f,m (b) preadig for WDMA someoe dowlik physical ael Fig. 1. preadig for WDMA dowlik physical aels The spreadig codes maitai the orthogoal property ad distictio betwee differet aels i the same ELL i WDMA, so as the WALH ODE. First, we must recogize the differet cells, ad the distiguish differet aels, ad we should also miimize mutual iterferece betwee cells, therefore, the scramblig codes are itroduced. Gold sequeces are used as scramblig codes because Gold sequeces have lower cross-correlatio properties tha M- sequece. OVF odes are used as aelisatio code because their orthogoal property ca distiguish differet physical aels. The legth of OVF odes varies with the data rate. V. RAMBLNG ODE GENERATON FOR WDMA cramblig code Typical Geeratio o Asic The sequece of complex valued ips shall be scrambled (complex ip-wise multiplicatio) by a complexvalued scramblig code dl,. A total of = 262,143 scramblig codes, umbered 0 262,142 ca be geerated. However ot all the scramblig codes are used. Oly scramblig codes k = 0, 1,, 8191 are used. Ea of these codes are associated with a left alterative scramblig code dl, ad a right alterative scramblig code, that may be used for compressed frames. The left alterative scramblig code correspodig to scramblig code k is scramblig code umber k , while the right alterative scramblig code correspodig to scramblig code k is scramblig code umber k [1] The scramblig code sequeces are costructed by combiig two real sequeces ito a complex sequece. Ea of the two real sequeces are costructed as the positio wise modulo 2 sum of ip segmets of two biary m-sequeces geerated by meas of two geerator polyomials of degree 18. The resultig sequeces thus costitute segmets of a set of Gold sequeces. Gold sequeces have low cross-correlatio properties to esure as may users as possible to use the ael with miimum mutual iterferece.the scramblig codes are repeated for every ip. Let x ad y be the two sequeces respectively. The x sequece is costructed usig the primitive (over GF(2)) polyomial 1+X 7 +X 18. The y sequece is costructed usig the polyomial 1+X 5 +X 7 + X 10 +X 18. The sequece depedig o the ose scramblig code umber is deoted z, i the sequel. Furthermore, let x(i), y(i) ad z (i) deote the i:th symbol of the sequece x, y, ad z, respectively. The m-sequeces x ad y are costructed as: itial coditios: - x is costructed with x(0)=1, x(1)= x(2)=...= x (16)= x (17)=0. - y(0)=y(1)= =y(16)= y(17)=1. (1) Recursive defiitio of subsequet symbols: - x(i+18) =x(i+7) + x(i) modulo 2, i=0, (2) - y(i+18) = y(i+10)+y(i+7)+y(i+5)+y(i) modulo 2, i=0,, (3) The :th Gold code sequece z, =0,1, 18-2, is the defied as: - z (i) = x((i+) modulo (2 18-1)) + y(i) modulo 2, i=0,, (4) These biary sequeces are coverted to real valued sequeces Z by the followig trasformatio: 1 Z( i) for i 0,1, Fially, the :th complex scramblig code sequece dl, is defied as: - dl, (i) = Z (i) + j Z ((i ) modulo (2 18-1)), i=0,1,, (6) Note that the patter from phase 0 up to the phase of is repeated. The m-sequeces y iitial value is idepedet of the :th complex scramblig code. Fig 2(a) illustrates the scramblig code typical LFR geerator o A (5) 439

3 (a)dowlik scramblig code typical LFR geerator x y x Ax Ay x + x Ax,4096 Ay,4096 x + (b) Dowlik scramblig code parallel geerator Fig. 2. Dowlik scramblig code geerator cramblig code Parallel Geeratio o Vector ore The scramblig code typical geeratio o asic is a ipby-ip way, whi is ot suitable for fast geeratio ad caot use the fast calculatio of vector core. The followig itroduces a scramblig code parallel gereratio whi ca be used o vector core. Let x m =[x(m*+31),,x(m*+1),x(m*) ] T (7) ad y m =[y(m*+31),,y(m*+1),y(m*) ] T. (8) The m-sequeces x ad y are costructed as: itial coditios: - x 0 =[ ] T =[ ] T. (9) - y 0 =[ ] T =[9c03 ffff] T. (10) Recursive defiitio of subsequet symbols; - x m =A x,m *x 0 mod 2 (11) - y m =A y,m *y 0 mod 2 (12) A x,m is defied as: m - A x,m = A x mod 2. (13) A y,m is defied as: m - A y,m = A y mod 2. (14) itial coditios: A x = , A y = DDDD EEE 000 B7E DB DB2 000 B64D 4000 DBB DD B B AD0A B A A A D A6A 4000 A5A D A A (15) The :th Gold code sequece z is the defied as: - z =x,m +y m mod 2 (16) - z = A x, * x,m +A y, * y m mod 2 (17) A x, ad A y, ca be respectively aieved from A x ad A y. The x,m is defied as: x,m = A x,m *x mod 2, (18) ad x ca be merged from x / ad x /+1. These biary sequeces are coverted to real valued sequeces Z by the followig trasformatio: 1 Z( i) for i 0,1, (19) Fially, the :th complex scramblig code sequece dl, is defied as: - dl, =Z +jz (20) Fig 2(b) illustrates the scramblig code parallel geerator o vector core. V. HANNELATON ODE GENERATON FOR WDMA haelisatio ode Typical Geeratio o A For all physical aels (except H) i the WDMA eviromet the ad braes shall be spread to the ip rate by the same real-valued aelisatio code,f,m, i.e. the output for ea iput symbol o the ad the braes shall be a sequece of F ips correspodig to the aelisatio code ip sequece multiplied by the realvalued symbol. The aelisatio code sequece shall be aliged i time with the symbol boudary.[1] The aelisatio codes i WDMA are Orthogoal Variable preadig Factor (OVF) codes that preserve the orthogoality betwee a user s differet physical aels. The OVF codes ca be defied usig the formula (21).The aelisatio codes are uiquely described as,f,k, where F is the spreadig factor of the code ad k is the code umber, 0 k F-1. The leftmost value i ea aelisatio code word correspods to the ip trasmitted first i time. 440

4 ,1 1,1,1 1 1,1,1, ,1 1,1,1 1,3,1,1 : : : (21) The typical geeratio method for the aelisatio code o asic is defied as Fig 3. ad -braes are further multiplied by complex-valued scramblig code, where ad deote real ad imagiary parts, respectively. The spreadig operatio is defied as F 1 R = =0 ( dl + ()) (25) Where F is preadig Factor, is the trasmittig complex-valued symbol, ad R is the trasmittig complexvalued ip after spreadig operatio. dl is the scramblig codes sequece, ad is the aelisatio code sequece. The dispreadig operatio performs the reverse operatio compared to the spreadig operatio. The despreadig operatio is defied as: F 1 = =0 R coj( dl + ()) (26) Where R is the received complex-valued ip, ad is the received complex-valued symbol after despreadig operatio. f scramblig code ad aelisatio code parallel geeratio are used, the spreadig ad dispreadig operatio are defied as: R = ( dl + ) (27) = R coj( dl + ) (28) V. REULT Fig. 3. Dowlik aelisatio code typical gereratio o A The aelisatio code geerator is composed of a bit biary couter, AND gates ad a XOR gate. Firstly, the AND gates are iitialized by the aelisatio code id i sequece whi the least value i aelisatio code id correspods to the last AND gates. The, bit biary couter couts from 0 to 2-1 i sequece whi the least value i couter correspods to the first AND gates. Fially, the outputs of every AND gates perform XOR operatio bit-by-bit. The output of XOR gate is the sequece correspods to aelisatio code id. haelisatio ode Parallel Geeratio o Vector ode f aelisatio code parallel geeratio is implemeted o vector core, the aelisatio code are defied as:,f =B*codeid mod 2, (22) codeid=[,f,codeid (0),,F,codeid (1),,,F,codeid (F1)] T.(23) Whe F=, B is defied as: B = [ e 1f] T. (24) V. PREADNG ODE GENERATON FOR WDMA WDMA preadig is applied to the physical aels. t cosists of two operatios. The first is the aelizatio operatio, whi trasforms every data symbol ito a umber of ips, thus icreasig the badwidth of the sigal. The umber of ips per data symbol is called the preadig Factor(F).The secod operatio is the scramblig operatio.where a scramblig code is applied to the spread sigal. With the aelisatio data symbols o so-called -ad -braes are idepedetly multiplied with a OVF code. With the scramblig operatio, the resultat sigals o the - (a)the result of real ad image part of scramblig code (b) The result of aelisatio code Fig. 4. The result of typical geeratio ad parallel gereratio The plots i Fig 4(a) show the correspodig scramblig code typical geeratio ad scramblig code parallel geeratio results for the complex scramblig code ip value. As these figures idicate, the scramblig code typical geeratio ad scramblig code parallel geeratio results agree. The plots i Fig 4(b) show the correspodig aelisatio code typical geeratio ad aelisatio code parallel geeratio results for the complex aelisatio code ip value. As these figures idicate, the aelisatio code typical geeratio ad aelisatio code parallel geeratio results agree. Fig 4(a) ad Fig 4(b), the x-axis represets the umber of the ip, ad the y-axis represets the magitude of ea of the ips. The complex scrambled sigal obtaied from the parallel geerator mates the typical geerator. 441

5 (a)pk ostellatio Before omplex preadig Fig 5(a) shows the sigal costellatio for the -/code multiplexed sigal before complex spreadig, ad Fig 5(b) shows the sigal costellatio after the complex spreadig operatios. The -/code multiplexed sigal with complex spreadig results i a rotated PK costellatio. Fig 5(b) shows the resultig costellatio aieved by both the typical spreadig ad the parallel spreadig. REFERENE (b) PK ostellatio Map after omplex preadig [1] 3GPP, T V9.20 ( ): preadig ad Modulatio (FDD), Release [2] "cramblig ode Geeratio for WDMA o the tarore 140/1400 ores", by mra Ahmed, Freescale emicoductor, Rev.1, November, 2004, pp. 2-9 [3] B. klar, DGTAL OMMUNATON Fudametals ad Applicatios. New Jersey: Pretice-Hall, c., Fig. 5. PK ostellatio Before omplex preadig 442