Prof. Xinyu Zhang Dept. Of Electrical And Computer .

2y ago
10 Views
3 Downloads
2.72 MB
31 Pages
Last View : 15d ago
Last Download : 2m ago
Upload by : Camille Dion
Transcription

Prof. Xinyu ZhangDept. of Electrical and Computer EngineeringUniversity of Wisconsin-Madison1"

Overview of MIMO communicationsSingle-user MIMOMulti-user MIMONetwork MIMO3"

MIMO (Multiple-Input Multiple-Output)Transmitter/receiver can have multiple antennasA modern wireless communication technologyTheory: late 1980’sStandards and products: after 2000’sNow: core feature in WLAN (802.11 WiFi) and cellular (3G LTE,WiMax)Two benefits, simply putImprove link SINRImprove link concurrency4"

MIMO network architecturesSingle-user MIMO (in 802.11n-2009, LTE)One TX, one RX. Either TX or RX or bothcan have multiple antennasMulti-user MIMO (in 802.11ac-2014, LTE-Advanced)One TX, multiple RX. Parallel transmissions.Network MIMO (expected in near-future)Multiple TX, multiple RX. Parallel transmissions.5"

Basic communication ple*data*streams)*

Song"Song1"SISO" "SIMO"MISO"Song2" "MIMO"7"

Diversity gainReceiver diversityTransmit diversityMultiplexing gainSpatial multiplexing8"

Receiver coherently combinessignals received by ut*(Single*data*stream)*Asymptotic gain: Increasing SNR proportionally to Nr (#ofreceive antennas)Intuition: received signal power adds upWhat’s the capacity gain?Logarithmically, according to Shannon’s equation: C B log(1 SNR)When SNR is low,, so gain is almost linear w.r.t. Nr9"

Selection combiningImproves SNR toMaximum Ratio combiningImproves SNR to10"

Multiple receive antennas allow compensation of by nonnotches in the other11"

Transmitter sends multiple versionsof the same signal, throughmultiple le*data*stream)"Two modes of transmit diversityOpen-loop transmit diversityClosed-loop transmit diversity12"

PrincipleSend redundant versions of the same signal (symbol), overmultiple time slots, and through multiple antennasEncode the symbols differently for different time slots and TXantennasSpace-Time Block Code (STBC)13"

Example: 2 TX antenna STBCSend two data symbols,Time slot 1:Time slot 2:TXRXTXRXReceived signals:14"

Example: 2 TX antenna STBCDiversity combiningi.e., signal power is boosted fromtoOpen-loop transmit diversity gain:In general, open-loop transmit diversity increases SNRlinearly with the number of transmit antennasWhat’s the capacity gain?15"

PrincipleSend redundant versions of the same signal (symbol), overthe same time slotEncode the symbols differently for different TX antennasi.e., weight the symbols on different antennas, following a precodingalgorithmPrecoding design requires feedback of channel state information (CSI)16"

Why precoding?Signals from different antennas need to sync (align) their phasesBut the different channels (between TXantennas and RXantenna)distort signals differently, causing phase offsete.g., both TX antennas sends; RX may receiveone TX antenna, butfrom the other, which weaken each other!SendTXSendRX17"

How does precoding help?Precoding: TX compensates the phase offset, and aligns the phasesof signals going through different channelsSendTXSendRXWhy CSI feedback is needed for precoding?TX must know the phase offset, in order to perform compensation18"

Asymptotic gain from closed-loop transmit diversitySignal level combining, also called transmit beamformingSuppose we have 2 transmit antennas, then instead of x, we receive:x x 2x, received power becomes, SNR increases to 4 times!More generally, withTX antennas, SNR increases toWhat’s the capacity gain?19"

Spatial multiplexing conceptForm multiple independent links (on the same spectrumband) between TX and RX, and send data in parallel throughthemUnfortunately, there is cross-talk between antennasCross-talk must be removed by digital signal ut*(Mul.ple*data*streams)*20"

Example 2x2 MIMO spatial multiplexingData to be sent over two TX antennas:Data received on two RX antennas:TXChannel distortions:Only two unknowns:RXcan be estimated by the receiver, easily obtained by solving the equations!21"

Asymptotic gainIn general, capacity gain from spatial multiplexing scaleslinearly withIn practiceSpatial multiplexing gain also depends on channel “condition”If the channels between different antennas are correlated, e.g.,are all the same, then you can’t solve the equations. Spatial multiplexingbecomes infeasible!Channel condition can be profiled using “condition number” (seereference)Practical wireless devices’ multiple antennas are separatedsufficiently far (further than half-wavelength), so the channelis usually uncorrelated22"

Concept of Multi-User MIMO (MU-MIMO)Single-antenna networkMulti-user MIMODesiredDesireddatadataCrosstalkMU-MIMO enables multiple streams of data to be sent to differentusers in parallel, without cross-talk interference23"

MU-MIMO differs from traditional MIMOData to be sent over two TX antennas:RX1Data received on two RX nodes:TXRX2Each RX only has one equation, but two variables; no way tosolve it directlyx2 causes cross-talk interference to x1, and vice versa24"

How to remove cross-talk?Send a weighted mix of x1 and x2TX antenna1 sends:RX1TX antenna2 sends:TXData received on RX1:RX2RX1 only wants x1, so ideally, we should have25"

MU-MIMO precodingTX can obtainto satisfyfrom RXs’ feedback, so it can tuneThis cancels the cross-talk interference from x2 to x1Similarly, we can cancel that from x1 to x2This is called Zero-Forcing Beamforming (ZFBF)How does TX obtain channel state informationSimplest approach in 802.11ac: CSI feedback schedulingAnnounceDATARX1 CSIACK1RX2 CSIACK2time

Asymptotic capacity gainIf the transmitter hasantennas, then it can sendstreams of data simultaneously tousers, increasing capacitytotimes compared with single-antenna transmitterLimitationMU-MIMO is essentially a form of spatial multiplexingSo the channel must be well-conditioned

Limitations of existing MIMO architecturesOnly one transmitter at a timeSimultaneous transmission fromdifferent transmitters causescollision!TransmissionBackoff So network capacity doesn’tscale with transmitter densityTransmissionBackoff 28"

A giant-MIMO comprised of many APsAPWireline backhaulAPs are tightly synchronized and share dataMutual interference can be cancelledAsymptotic gain: Network capacity scales linearly with thenumber of APs, theoretically29"

Is network MIMO practical?A super-giant-MIMO solution?APs need:Full synchronization: carrier phase, frequency, sampling-clockFull data sharing: large volumes of data and CSI exchangeInfeasible!30"

“NEMOx: Scalable Network MIMO for Wireless Networks ”,ACM MobiCom’13, by Xinyu Zhang, Karthikeyan Sundaresan, Mohammad A. (Amir) Khojastepour,Sampath Rangarajan, Kang G. ShinClusterClientmAP (master AP)dAP (distributed AP)A hierarchical architecture to realize scalable network MIMOIntra-cluster: dAPs within each cluster can TX concurrentlyInter-cluster: neighboring clusters contend for channel accessCapacity can scale with #of dAPs within each cluster,and with #of clusters (capacity 6 in this case)31"

Take-home message:(1) What are the various modes of operations inMIMO?(2) How does each MIMO mode scale link/networkcapacity?References:Book1: Fundamentals of LTEBook2: Fundamentals of Wireless Communications32"

A hierarchical architecture to realize scalable network MIMO mAP (master AP) Intra-cluster: dAPs within each cluster can TX concurrently Inter-cluster: neighboring clusters contend for channel access Capacity can scale with #of dAPs within each cluster, and with #of clusters (capacity 6 in this case) Cluster

Related Documents:

Prof. Hassan Soliman Prof. Hisham Abou Grida Prof. Hossam El Wakil Prof. Hossam Tawfik Prof. Ihab Nabil Prof. Mohamed El-Heneidy Prof. Mosaad Soliman Prof. Omar El Kasheif Prof. Sameh Zarad Prof. Sherif Moamen Prof. Tarek Radwan Prof. Usama El Nahas Prof. Wassila Taha Alphabetical Organizin

Address for correspondence: Xinyu Zhang, Academy of Mathematics and Systems Science, Chinese Academy of Sciences, East Zhong-Guan-Cun Road, Beijing 100190, People's Republic of China. . generalized linear models (Zhang et al., 2016) and missing data models (Fang et al., 2019; Zhang, 2013). When covariates are measured with error, we face .

Apr 11, 2007 · State Dept. of Education Dept of Children and Families Judicial Department Higher Education Community and Technical colleges Public Health Dept of Mental Health and Addiction Services Dept of Mental Retardation Dept t of Labor Dept of Transportation Dept of Corrections We do not have a

Dr. Intekhab Ahmed Prof. Bahram H. Arjmandi Prof. David A. Arnall Dr. Vivek Bhalla Prof. Bertrand Cariou Prof. Dmitry A. Chistyakov Prof. Fernando Cordido Prof. Franco Battista Ennio Folli Prof. Paul Albert Fournier Dr. Bruce H. Francis Dr. Xianxin Hua Prof. Joaquin Lado-Abeal Dr. Hongyun Lu Dr. Abbas A

NCERT, New Delhi PG DAV College Dept. of Maths Nehru Nagar, New Delhi Jamia Milia Islamia University, New Delhi Prof. D.P. Shukla Prof. V.P. Gupta Prof. C.P.S. Chauhan Dept. of Maths Dept. of Measurement Dept. of Education Lucknow University & Evaluation, NCE

dr Stanko Boboš (Serbia), Prof. dr Ljiljana Nešić (Serbia), Prof. dr Petar Sekulić (Novi Sad), Prof. dr Mirjana Milošević (Serbia), Prof. dr Cvijan Mekić (Serbia), Prof. MVD Juraj Pivko, DSc. (Slovakia), Prof. dr Šandor Šomođi (Hungary), Prof. dr

Dept of Inorganic Chemistry AU Vizag Dr. Y. R. SATYAJI RAO, Scientist 'F' and Head,NIH,Kakinda Prof DV PRABHU, ACT General Secretary Mumbai Prof D.C.DEKA, Guwahati University,Guwahati,ACT President Prof P.V.S.MACHIRAJU, ACT Vice president Prof M.SWAMINATHAN ACT Dept of Chemistry

GeneArt Strings DNA Fragments 8 Gene assembly 9 Mammalian expression systems 10 Selecting a mammalian expression system 10 . the five protein classes. The selected genes were individually optimized using the GeneOptimizer algorithm [2]. For comparison, the corresponding wild type genes were subcloned using native sequences available from the NCBI database. Each gene was then .