International Conference on Advances in Mechanical Engineering and Industrial Informatics (AMEII 2015)

Transcription

1 International Conference on Advances in Mechanical Engineering and Indstrial Informati (AMEII 05) An Improved SRS oise Estimation Algorim Based on Interference Cancellation Kewen Li a Weiwei Tian b Mingye Wang c Xianglin Chen d and Jiha Gong e Mobile Commnications Lab Whan University of Technology WhanHbei China a likewen@wht.ed.cn b tianweiwei@63.com c wangmingye@6.com d chenxianglin@6.com e gongjiha@6.com Keywords: LTE;SRS;Interference Cancellation;oise Estimation Abstract. This esis stdies SRS estimation meod wi mlti-sers in e Long Term Evoltion(LTE) and proposes a Discrete Forier Transform () time-domain filtering algorim based on interference cancellation. By calclating signal power in e system protecting band decision reshold of time domain filtering can be denoted. Therefore one can redce e impact of on e sefl signal wiin e window and also obtain a more accrate estimation. Simlation reslts show at eier in e case of single-ser or mlti-ser e proposed algorim can get good performance. At e same time e algorim can accrately reflect e impact in neighborhood interference on each band channel qality so as to achieve e freqency domain resorce selective schedling effectively. Introdction As a new generation of mobile commnication technologylte standard adops carrier freqency division mltiple access (SC-FDMA) for plink orogonal freqency division mltiple access (OFDMA) for downlink transmission []. OFDM technology has high spectral efficiency bandwid scalability anti-mltipa fading freqency domain schedling and adaptive advantage. However when e nmber of independently modlated sbcarriers consistent wi e se OFDM symbols have a very high peak-to-average ratio (PAPR) which increases e cost and power consmption of e transmitter amplifier and is nsitable to be sed for terminal so e plink channel sing e SC-FDMA which has a single carrier property. Throgh e base station schedler UE side implement mlti-ser mltiplexing in e freqency and time domain on PUSCH. And base station side check e channel qality of each ser for resorce schedling [].An important UL reference signal Sonding Reference Signal(SRS) is defined in spport of freqency dependent schedlingpower control and UL synchronization maintenance. By estimating e signal to interference ratio(sir) SRS can reflect e channel qality information of each ser on each RB. At e same time offset can be estimated wiot sending Physical Uplink Shared Channel (PUSCH) and Physical Uplink Control Channel(PUCCH) s e eode achieve resorce schedling and time compensation. Covering e whole system bandwid SRS can get entire channel bandwid channel qality information on each RB. Then choosing e channel qality band as a Shared channel freqency band and next time slot it can be sed to strengen e power control. How to detect e SIR on each RB is e SRS research priorities and e key is to estimate mltiser and interference on each RB. In [3]some SR estimation algorims are compared in AWG in detail. [4] proposes a meod of channel estimation based on IFFT in e time domain to calclate e estimate vale.however at large SR e signal energy leakage significantly leading to smaller signal power estimation and SR estimation error is large. [5] proposes an algorim by comparing MSE of channel coefficients to select a window fnction. Even ogh it can get a better performance in practice is decision conditions is not easy to achieve. In [6] vale can be estimated better bt one cyclic shift is reserved exclsivelylending to redce e nmber of mltiplexed sers. 05. The aors - Pblished by Atlantis Press 63

2 To solve how to accrately estimate e SIR for each ser on each RB problem in a mlti-ser code division mltiplexing case according to e strctral characteristi of e SRS transmission mode and LTE plink is paper presents an improved SRS channel qality estimation algorim based on interference cancellation. The algorim not only improves e SIR estimation accracy in single-ser sitations bt also be able to accrately estimate e mlti-ser environment SIR each RB and provides an accrate jdgment basis for plink freqency selective schedling. SRS Transmission Model As to FDD-LTE frame strctrea radio frame is 0ms long and consists of 0 sbframe of leng ms.a sbframe is defined as two consective slots of 0.5mseach made of seven SC-FDMA symbols(normal cyclic prefix configration) and demodlation reference signal (DMRS)in e middle.when e sbframe is configred for SRS transmissione last symbol is reserved for e SRS[].The SRS sbframe nmber is determined by e cell level and UE level period configration. The FDD-LTE frame strctre is depicted in Fig..SRS transmission of e configred sb-frame is always located in e last SC-FDMA symbol. Demodlation reference signal (DMRS) and sonding reference signal (SRS) located at different SC-FDMA symbols. DMRS is associated wi e transmission of PUSCH and PUCCH. Assigned to e SRS symbol does not allow for PUSCH transmission.however SRS has noing to do wi PUSCH and PUCCH transmission. Mltiple UEs can be mltiplexed in a combination of FDM and CDM rogh high-level configration and schedling. SRS ses CAZAC seqence as an plink reference signal (DMRS / SRS) seqence. SRS is definded according to r ( n) = e r ( n) 0 n< M () ( α) jαn RS v v sc RS RB M sc m sc max UL where = is e leng of e reference signal seqence and m RB. Mltiple reference signal seqences are defined from a single base seqence rogh different vales of α.base seqences rv ( n) are divided into gropswhere {0...9} is e grop nmber and v is e base seqence nmber wiin e grop[]. α is e base seqence of cyclic shift and e same base seqences of different shift will form a different reference signal seqences. The cyclic shift α of e sonding reference signal is given as π n where n is configred for each UE by higher layers and n = Mlti-ser SRS signal can be transmissioned rogh code division and freqency division mltiplexing. The transmission strctre of CDM-based two-ser is shown in Fig.. The revieved SRS can be expressed as: α = () Y= HX + HX + (3) jαin where X X are e transmitter freqency domain signal of UE and UE. Xi( n) = e rv ( n) RS 0 n< M sc n SRS αi = π ; H H are e channel estimation coefficient of UE and UE;and is e signal. oise Estimation Algorim oise estimation based on First we get MMSE (minimm mean sqare estimation) of channel freqency response H ( k ) 64

3 * H ( k) = Y( k) X ( k) = H( k) + V( k)( k ) (4) Where for e leng of SRS signal; Hk ( ) for e actal channel coefficients; V( k) for portions.after Iwe get e time domain implse response h ( n ) h ( n) = IFFT{ H ( k)} = h( n) + v( n)( n ) (5) Sppose L response to e impact of e channel leng.when n> h ( n ) = 0. Generally Since e channel implse response leng is less an e leng of e cyclic prefixit is considered at e channel implse response beyond e cp part consists of e. Ths e power can be expressed as = ˆ σ h ( n) (6) cp n= cp oise estimation based on interference elimination For example two ser code division mltiplexing (CDM). SRS signal received by e base station can be expressed as: Y= HX + HX +. Then e entire freqency band and interference can be expressed as = Y HX HX (7) where H and H are e Channel coefficient of ser and serwhich can be estimated by ; X X Can be generated locally by e base station.so e and interference can be calclated by 6. Improved Algorim De to e first estimate of e channel coefficient error of conventional estimation meod based on interference cancellation is large so it is far from e actal deviation of e and e reslting performance is poor. The traditional estimation meod is jst to calclate e average power on e entire band estimate e SIR nable to show e differences on each RB clearly. Therefore we propose an improved algorim nder interference cancellation meod. In order to obtain accrate channel of each ser estimated coefficients we will fill e estimate of e coefficient of zero transform to time domain en add window to filter ot e sefl signal. Bt ere are still parts of e wiin e windowwhich have a greater impact on e reslting. De to e LTE system left protection band distribtion at bo ends we can ink only we can calclate is part of e power as decision reshold of filtering wiin e window. The specific implementation of e algorim flow chart is shown in figre 3 e implementation steps as follows:.after removing e CP 7.5 KHz freqency offset compensation and FFT freqency domain e received data is recorded as Y which is e sperposition of all ser data in e freqency domain. Calclating e signal power on e gard band wiot effective data transmission as initially ˆ σ I for all sers of e cell is TTI(transmission time interval). RS.Then do sbcarrier demapping and get SRS data corresponding to e sbcarriers of leng M sc. 3.The seqence after demapping do conjgate mltiplication wi ZC base seqence rv ( n ) en zero padding I changes to time domain (zero-padded to prevent edge effects).if CDM sers exist in is band ere will be a corresponding time-domain implse signal by n SRS to distingish.to separate mltiser in time domain windowing is added in time domain. If ere is no time offset ( nsrs) effect implse signal shold be located in I window leng I. To conter 65

4 CP / I e time offset e window fnction on bo sides is extended to. Then remove e FFT time domain cyclic shift. 4. Alogh e time-domain filtering removes otside e window e is still left wiin e window.seting e reshold e can be removed wiin e window. K is set for e selected reshold before we only retain e samples whose power greater an K ˆ σ ( a) (sideband initial ). For example in e ser window i represents e sample I point.if ˆ SRS h ( i a) K srs _ ( a) en ˆ SRS h ( ia ) = 0. Oerwise keep its data.(otice at is kind of can t remove e hidden in e symbol so freqency domain is very important ) 5. Do transformed to e freqency domain and remove excess zeros.the leng of e original signal is restored and we can obtain channel estimation coefficients H... H Hn. 6. Using e channel estimation vale to calclate signal power: E ( a) ( H ) ( a ) () SRS ˆ PRB = i a σ 6 i PRB L s a PRB srs (9) ˆ ( ) _ M Where L for e reserved valid points in time domain. The above calclation is e signal power of a receiving antenna for e case of mltiple antennas.the E ( a ) obtained in each antenna are added and averaged. 7. In order to calclate e power sing a similar meod SICwhen detecting data for each UE we can pt e oer UE s data as interference to eliminate. After e sefl signal are removed e reslting signal is considered as interference and. Specifically if ere are two UE code division mltiplexing.for UE signal after interference cancellation is and UE data. Y ' = Y H X. When detecting UE UE sefl signal is eliminated and signal ' is obtained. = Y HX = Y HX HX where for e signal on freqency domain. is sed to calclate power in each PRB(e addition averaging on Mltiple antennas): = (0) ˆ PRB ( a) ( i a ) 6 i PRB σ PRB. SIR and e average SIR is calclated for each UE on each RB E snr _ PRB _ db 0log ˆ snr PRB = () s PRB E = () n _ PRB PRB _ db 0log( ) n _ PRB i= ˆ s PRB Simlation Simlation scenario and parameter Settings The simlation environment as shown in table : 66

5 Table Simlation Parameters System Bandwid(MHz) 0 mber of antennas Channel EPA5 mber of SRS sers ~ SR range(db) [-0 +0] SRS bandwid(rb) 0 Simlation times 500 Simlation reslt Measre: e average SIR and MSE The average SIR: average of 500 measrements of SR T aver _ snr _ db = snr _ db (3) T i = i Mean sqare error (MSE) is defined: T MSE = ( snr _ dbi SRideal ) (4) T i = Where snr _ db i for e i- estimated average SR SRideal for a real vale T for e nmber of simlations. The following are e algorim performance simlation verification of single UE and UE scenarios. case A:single UE In EPA5e time-domain filtering algorim based on e interference cancellation is compared performance wi [4] and [7]. Fig. 4 for MSE performance comparison. Fig. 4 shows at in e case of a single UE e proposed elimination of interference improved algorim otperforms traditional algorims based IFFT. The new algorim and Qi s are bo relatively accrate bt e MSE of new proposed algorim is always less an Qi's algorim wi higher stability of e estimated SIR. case B:UE CDM UE transmit code division mltiplexed togeer and pass e EPA5 channel. The time-domain filtering algorim based on e interference cancellation is compared performance wi [7]. Fig.5 for MSE performance comparison. Trogh different n(0)code-division mltiplexinge SRS bandwid of two UE are bo 0RB. The figre only shows e performance crve of one UE anoer is similar sitation. From Fig. 5 shows at when e channel environment is poor (SR < 0 db) e proposed algorim is sperior to e Qi s which can better reflect e channel qality information. When e channel environment is better e impact of on SIR measrement is redced and e performance of e two algorims are eqivalent.. case C:UE pls cell interference UE transmit code division mltiplexed togeer and pass e EPA5 channel.at e same time adjacent CellInterference is added to e cell SRS and e inflence of interference to e algorim performance is measred.from e Fig.6e ble part is e average SIR performance crve of new algorim 0dB.Where e interference on e pper half bandsir=0db. The red part is e average SIR performance crve of new algorim 0dB.Where e interference on e pper half bandsir=0db. Fig. 6 shows e performance SIR estimation crve on each RB after adding SRS neighbor cell interference at snr = 0dB and 0dB. The simlation reslt shows at e proposed algorim flly embodies e differerence between adding interference on e low freqency band and no interference on e high freqency band of SIR. Ater adding adjacent CellInterference on e low 67

6 Sideband calclation ' domin filtering Signal Power estimation ' domin filtering Signal Power estimation freqcency bandwheer ere is interference or not on e bande real SIR can be estimated accrately. + + UE time domain signal UE time domain signal Channel oise FFT Y UE Sbcarreier de-mapping ZC compenastion Zeropadded and I UE x h = y Remove time cyclic shift domain Remove time cyclic shift domain Slot(0.5ms) Slot(0.5ms) UE Rx and Zeros and Zeros DMRS Sbframe(ms) DMRS x h X Y SIC Y = Y - H X CFO H X CFO H Compter = Y - H X DMRS DATA SRS UE Fig. FDD-LTE frame strctre Fig.. SRS transmission model Fig.3 Receiver flow chart Compte UEsignal power Compte power on each RB Compte UE signal power IFFT based meod Qi's meod EW meod Qi's meod EW meod 0 5 0dB+cell interference 0dB+cell interference MSE Estimated SR(dB) Estimated SIR(dB) Actal SR(dB) Actal SR(dB) SRS(4RB) Fig.4 Single UE EPA5 Fig.5 UE(CDM) Fig.6 UE(CDM) Adjacent CellInterference Conclsions Based on e featres and fnctions of e SRS signal a novel SIR estimation algorim is proposed. The algorim is based on e idea of interference cancellation combined wi algorim in e time domain filtering. Simlation reslts show at e new proposed algorim not only in single-ser and mlti-ser sitations measred SIR performance accrately bt also in e adjacent cell interference can detect e channel qality information in each band accrately which is benefit for base stations to achieve freqency domain selective schedling. Acknowledgements This work was spported by atral Science Fondation of Hbei Province(O.03CFB349). References [] 3GPP TS 36. V Technical Specification Grop Radio Access etwork; Evolved Universal Terrestrial Radio Access (E-UTRA)[s]; Physical Channels and Modlation (Release ). [] Safa H Tohme K. LTE plink schedling algorims: Performance and challenges[c]telecommnications (ICT) 0 9 International Conference on. IEEE 0: -6. [3] D. Palzzi and. Bealie A comparison of SR estimation techniqes for e AWG channel IEEE Trans. Commn. vol. 4 no. 0 pp [4] Fei J Gangliang R Zhe Z. A new variance and post detection SR estimation meod for MIMO OFDM systems[c]commnication Technology 00. ICCT 00. IEEE International Conference on. IEEE 00: 79-. [5] Tian H Yang L Li S. SR estimation based on sonding reference signal in LTE plink[c] Signal Processing Commnication and Compting (ICSPCC) 03 IEEE International Conference on. IEEE 03: -5. [6] Bertrand P. Channel gain estimation from sonding reference signal in LTE[C] Vehiclar Technology Conference (VTC Spring) 0 IEEE 73rd. IEEE 0: -5. [7] Qi Shi LI Pingan MaTing. The research of estimation algorim on LTE plink sonding reference signal [J]