STUDY ON LINK-LEVEL SIMULATION IN MULTI- CELL LTE DOWNLINK SYSTEM

Transcription

1 Proceedngs of IEEE IC-BMT0 TUDY O LIK-LEVEL IMULATIO I MULTI- CELL LTE DOWLIK YTEM Yang Zhang, Ruoyu Jn, Xn Zhang, Dacheng Yang Beng Unversty of Posts and Telecommuncatons, Beng 00876, Chna 05330@bupt.edu.cn Abstract mulaton of wreless communcaton system s usually dvded nto system-level and ln-level. Ln-level smulaton s generally used to smulate the pont-to-pont physcal layer technologes wth propagaton model taen nto account, whch means t s based on sngle-cell and sngle termnal scenaro. Mult-cell model s not often studed n ln-level manly due to ts hgh complexty. o we usually consder the mpact of Gaussan whte nose wth fadng n ln-level smulaton but gnore the mpact of nterference from other cells. Ths paper brngs a new perspectve of ln-level smulaton for long term evoluton (LTE) system under mult-cell scenaro. By analyzng downln nterference n mult-cell, we demonstrate the nterference power s no longer flat n frequency doman, and the correlaton between nterference s affected by cell IDs. o we beleve that ln-level smulaton under mult-cell s more sgnfcant for evaluaton of a practcal system, especally for ts control channel. We tae LTE downln control channel as an example, and propose a new method for ln-level smulaton n mult-cell scenaro. The smulaton results confrm our analyss. Keywords: user equpment, OFDM, MIMO, mult-cell nterference. Introducton The 3rd generaton partnershp proect (3GPP) long term evoluton (LTE) s desgned for hgher data rate, hgher spectral effcency, mproved servce qualty, and lower latency as well as hghcapacty voce support. To satsfy the requrement of hgh-speed transmsson, LTE downln adopts orthogonal frequency dvson multplexng (OFDM) and multple-nput multple-output (MIMO) technologes. The two ey technologes greatly ncrease spectral effcency [-4]. LTE downln control channel s used to transport varous control nformaton. It contans physcal downln control channel (PDCC), physcal control format ndcator channel (PCFIC) and physcal hybrd ARQ ndcator channel (PIC) []. PDCC s used to transport downln control nformaton (DCI), PCFIC transports control format ndcator (CFI) and PIC transports ARQ Indcator (I) [5-6]. Ln-level smulaton s an mportant method for performance analyss. Wreless communcaton smulaton generally ncludes system-level smulaton and ln-level smulaton. ystem-level smulaton s an ndspensable means of wreless networ performance evaluaton n standardzaton and plannng. It has a wde range of applcaton. Ln-level smulaton s manly to test the performance of varous physcal layer technologes. Accordng to the basc module and related algorthms of physcal layer, through approprate channel modelng, we can establsh a pont-topont wreless ln, and get the basc relatonshp between bt error rate/bloc error rate (BER/BLER) and sgnal-to-nose rate (R). In ln-level smulaton we often consder the mpact of Gaussan whte nose besdes fadng because ln smulaton s always bult n a sngle-cell scenaro. owever, the nterference from other cells exsts n practcal systems, so t s necessary to analyze ts mpact on the ln-level smulaton. Ths paper taes ln-level smulaton nto multcell scenaro n LTE system. Through topology setup of a LTE communcaton system, smulaton generates several ndependent channels of adacent cells to transport ndependent data, and supermposes the sgnal on useful sgnal. Both theoretcal analyss and smulaton results show that the nterference s not Gaussan whte nose. But the nterference from mult-cell becomes color nose. We also show that f tradtonal smulaton method s taen for mult-cell, the output curves would be not smooth anymore. In ths context, we propose a new smulaton method that taes model of mult-cell scenaro nto consderaton. Then we fnd the result for mult-cell has a slghtly worse performance. LTE and mult-cell ntroducton. Introducton of OFDM and MIMO LTE adopts OFDM and MIMO as ts ey technologes. OFDM converts seral data streams //$6.00 0IEEE

2 nto parallel blocs of sze after Inverse Fast Fourer Transform (IFFT). Tme-doman samples of OFDM symbols can be obtaned from frequency-doman symbols by (), 0 n, x ( m) IFFT n e n, m () Where n, s the transmtted data symbol at th subcarrer of n th OFDM symbol. s the amount of pont n IFFT. xn ( m ) represents m th sample pont of n th OFDM symbol.. Downln control channel In mult-cell scenaro, resource mappng changes wth the cell ID []. In order to reduce the nterference among cells, resource mappng of all the control channels has relatvely shfts base on ts cell ID. Because ths nd of resource mappng s fxed when cell ID s fxed, correlaton between two sgnals n the frequency doman can be determned. As showed n Fgure. Accordng (), we can construct ndependent mult-path fadng channel models. 3 Analyss of ln smulaton n multcell We wll frst ntroduce ln smulaton n snglecell scenaro, then expand our analyss to multcell. 3. Ln smulaton n sngle-cell When smulatng n sngle-cell model, the nose whch adds to the sgnal s Gaussan whte nose. uppose there are transmttng antennas and M recevng antennas. The receved sgnal n tme-doman can be descrbed as (3). y () t x () t h () t n () t (3) 0 Where y () t s the receved sgnal at th recevng antenna. h () t s the channel mpulse response from th transmttng antenna to th recevng antenna whch can be generated by (). x () t s the sgnal at th transmttng antenna after IFFT n tme-doman. n () t s the Gaussan whte nose at th recevng antenna, whch means En() t n( tt) 0. uppose channel does not change wthn an OFDM symbol. After samplng and FFT, the receved sgnal can be descrbed as (4), Y ( ) ( ) ( ) ( ) (4), 0 Fgure Resource mappng example.3 Cell topology LTE models wth 9 hexagonal macro cells and each cell can be dvded nto 3 sectors. After the establshment of topology, we can get the path loss from any base staton to any UE. The total loss can be determned by the large-scale path loss, shadowng effects and fast fadng n the range of small-scale, descrbed as (), () () n () () Pdt dt dt Kdt () n Where Pdt () represents large-scale path loss. d() t represents shadow effect. K dt () represents fast fadng. Where represents the ndex of subcarrer. ( ) s the sgnal n frequency doman from th transmttng antenna. s the channel fadng from th transmttng antenna to th recevng antenna n frequency doman. To facltate subsequent dscussons, we defne 0 M Y ( ) Y( ),, Y ( ) T ( ) ( ),, ( ) T 0 ( ) 0, ( ),,, ( ) T T ( ) 0 ( ),, ( ) T M. And rewrte (4) as (5),

3 Y( ) ( ) ( ) ( ) (5) It s easy to prove that ( ) s also Gaussan whte nose n frequency doman. Because FFT and IFFT don t change the statstcs of nose. 3. Ln smulaton n mult-cell In the case of mult-cell, the smulaton model would change. As n Fgure. Fgure Ln smulaton model n mult-cell. Where B represents base staton whch transmts the useful sgnal, IB represents base staton whch transmts sgnal as nterference, nt () represents thermal nose. Each channel s ndependent, generated by (). As we see, the useful sgnal s not only dsturbed by nose, but also the sgnal from other cells. o from Fgure, nt () n (3) becomes as (6), 0 CELL m0 0 y () t x () t h () t n () t I ( m,) t (6), Where CELL s the number of nterference cell, n () t s the thermal nose. I, ( m, t ) s the nterference sgnal from transmttng antenna n cell m to recevng antenna. Assume the channels of other cells slowly change, after samplng and FFT, the receved sgnal become as (7). Y ( ) ( ) ( ) ( ) CELL I m0 0 ( ) ( ) ( ) CELL (7) m 0 Where ( m, ) s matrx equals 0,,,, means the channel fadng n frequency-doman matrx at th subcarrer from cell m to th recevng antenna, m (, ) s a matrx represents the symbol at subcarrer of cell m. As dscussed before, t would transport PDCC symbol, PCFIC symbol, PIC symbol, reference sgnal (R) symbol or be reserved. But no matter what t s, when PDCC 00% loads, m (, ) s a determned value whch can be represented by functon of m and. Because obeys complex Gaussan dstrbuton, o CELL I s m0 0 stll a random vector whch obeys Raylegh dstrbuton. The nterference power s as (8), CELL E I m0 0 CELL E m 0 CELL CELL E m0 n0 ( n, ) ( n, ) CELL m 0 E tr m (, ) (, ) CELL m0 0 m (, ) (, ) E m E m (8) Assume the fadng power m s (, ) ndependent n dfferent cell but s the same n dfferent transport antennas and subcarrers. o E ( m, ) s ust a functon of m, defne as P m, and (, ) E m s a functon of, m and. Defned as P m,,, PPDCC when spdccsymbol PPCFIC when spcficsymbol Pm,, PPICwhen spcficsymbol PR when srsymbol 0 when sreserved (9)

4 o, (8) can be rewrte as (0), CELL CELL E I Pm Pm,, m0 0 m 0 0 (0) It s obvous that CELL Pm P s a functon m,, m0 0 of, but t s not related to recevng antenna ndex. o the nterference power s dfferent n dfferent subcarrers, When cell ID s determned, the resource mappng can be determned, Then we can easly determned the functon between (0) and. The correlaton of nterference n frequencydoman can be descrbed as (3), CELL CELL E I I ( m, ) m0 0 m0 0 CELL CELL E m0 m0 0 0 ( m, ) ( m, ) () CELL E m 0 ( m, ) ( m, ) () E ( m, ) m0 0 CELL E m m (3) (, ) (, ) Where from () to () bases on the assumpton that ( m, ) s ndependent between dfferent nterference cells and transmttng antennas. o E ( m, ) s the functon of m, but s not related to and. From [7], t s not related to. o we defne t as Rm,. But from resource mappng t can be easly found that E ( m, ) s related to, we defne t as Rm,,,. Rewrte (3) as (4), CELL CELL E I I ( m, ) m0 0 m0 0 CELL Rm, Rm,,, m0 0 (4) o correlaton of nterference s a functon of and. The same as the power of nterference, the correlaton can also be determned when cell ID s determned. 3.3 mpact of some algorthms Many algorthms whch base on the Gaussan whte nose are not sutable for mult-cell scenaro. Mnmum mean square error (MME) taes the best performance to estmate the channel fadng coeffcent. The formula s showed n (5), ˆ MME P L X ˆ R R (5) Where R s the channel autocorrelaton matrx. ˆ L s the estmated result matrx by least squares (L) method. P X s the power of sgnal. s the power of Gaussan whte nose. When the nterference become color nose. omethng must be change to match t. We rewrte (5) as ˆˆ ˆ R R ˆ MME (6) ˆ Where R ˆˆ s autocorrelaton matrx of estmated value of channel. R s cross correlaton Ĥ between estmated value of channel and true value. 4 mulaton 4. mulaton method n mult-cell From (0) and (4) we now the nterference power and ts correlaton n frequency-doman are both related to cell ID. If we fx UE poston n smulaton, the cells whch cause strong nterference s fxed, so the cell ID s fxed, and t can t reflect the characterstcs of nterference. o we smulate a lot of UEs whch randomly scatter n the cell concerned. For each UE, we get all cell IDs from nterference cells whch have obvous nterference. Then usng resource mappng to generate ndependent OFDM symbols of each nterference cell by usng these cell IDs, as n Fgure. 4. mulaton result In our smulaton, antenna confguraton s, system bandwdth s 0Mhz, and use MME as (6) to estmate channel. otced the mpact of resource mappng, we randomly scatter UE n the area of concerned cell, and select several cells whch have strong L

5 nterference. Accordng to channel fadng of selected cells, we calculate average IR. We smulate PDCC wth CCEs n mult-cell, and compare the result wth sngle cell. If we don t use the method descrbed above, scatter ust one UE n each IR pont. The result s showed n Fgure 3, 0 0 nterference would not be Gaussan whte nose anymore, t obeys Raylegh dstrbuton, and ts power and correlaton are dfferent n dfferent subcarrers, but can be determned by cell ID. The tradtonal smulaton method s not sutable for mult-cell, then we propose a new method for smulaton n the new scenaro. From the smulaton result, the curve becomes smooth and nterference causes a slght loss n performance. Acnowledgements BLER IR Fgure 3 The result whch doesn t use smulaton method we proposed. The curve s not smooth anymore, even fluctuant. o we conclude that because of the nterference, the smulaton method used n sngle-cell s unreasonable n mult-cell. If we use the method proposed, and compare the results between sngle-cell scenaro and mult-cell scenaro. As n Fgure 4, BLER sngle-cell mult-cell IR Fgure 4 Comparson of sngle-cell and mult-cell n LTE downln PDCC. The curve becomes smooth. Because nterference n mult-cell s not Gaussan whte nose anymore, ts performance s slghtly worse than the result n sngle-cell. Ths result confrms our analyss. The wor n ths paper s sponsored by Proect of Chna under Grant o. B References [] 3GPP T 36. V9.0.0 (009-) [] G. L. tuber, J. R. Barry,. W. McLaughln, Y. L, M. A. Ingram and T. G. Pratt, Broadband MIMO-OFDM wreless communcatons, Proceedngs of IEEE, vol. 9, no., pp. 7-94, Feb [3] Y. ongwe, A road to future broadband wreless access: MIMO-OFDM-based ar nterface, IEEE Commun. Mag.. vol. 43, no., pp.53-60, Jan.005. [4] A. J. Paulra, D. A. Gore, R. U. abar, and. Bolcse, An overvew of MIMO communcatons a ey to ggabt wreless, Proceedngs of the IEEE, vol. 9, no., pp. 98-8, Feb [5] R. love, R. Kuchbhotla, A. Ghosh, R. Ratasu, W. Xao, B. Classon, and Y. Blanenshp. Downln control channel desgn for 3GPP LTE, Proc. IEEE WCC, pages 83-88, 008. [6] Jalng Lu, Robert Love, Kenneth tewart, Mchael Eon Bucley, Desgn and analyss of LTE physcal downln control channel, VTC prng 009. IEEE 69 th. [7] Ye (Geoffrey) L, Plot-ymbol-Aded Channel Estmaton for OFDM n Wreless ystems, IEEE TRAACTIO O VEICULAR TECOLOGY, vol.49, o.4, July Conclusons Ths paper manly brngs ln-smulaton nto mult-cell scenaro. By generatng several ndependent channels n mult-cell scenaro, we transport data on these channels and add the sgnal to useful sgnal. Through our analyss the