`US 7,742,546 B2
`(10) Patent No.:
`
` Ketchumet al. (45) Date of Patent: Jun. 22, 2010
`
`
`US007742546B2
`
`(54) RECEIVER SPATIAL PROCESSING FOR
`Cer. TRANSMISSION INA MIMO
`
`(56)
`
`References Cited
`U.S. PATENT DOCUMENTS
`
`3,459,679 A *
`
`8/1969 Rosinski et al... 502/65
`
`(75)
`
`73)
`
`Assi
`
`51)
`(51)
`
`Inventors: John W. Ketchum, Harvard, MA (US);
`MarkS. Wallace, Bedford, MA (US); J.
`Rodney Walton, Carlisle, MA (US);
`Steven J. Howard, Ashland, MA (US)
`Assignee Seer ca 8) neorporated,
`(73)
`San
`,
`(*) Notice:
`Subject to any disclaimer, the term of this
`patent is extended or adjusted under 35
`U.S.C. 154(b)by 643 days.
`.
`(21) Appl. No.: 10/682,160
`(22) Filed:
`Oct. 8, 2003
`(65)
`Prior Publication Data
`US 2005/0078762 Al
`Apr. 14, 2005
`Int.cl
`Int.
`Cl.
`(2006.01)
`HOAL 23/02
`(2006.01)
`HO3H 7/30
`(2006.01)
`HOAN 7/12
`(2006.01)
`HO04J 3/00
`(2006.01)
`HO04J 1/00
`(52) US. Ch iceccccccceceeeee 375/341; 375/229; 375/265,
`375/262; 370/480; 370/497: 370/498; 370/529
`.
`z
`,
`.
`/
`(58) Field of Classification Search................. 375/260,
`375/267, 295, 296, 335, 346, 349, 148-149,
`375/152, 136-137, 262, 265, 316, 341, 343,
`375/344, 350; 370/208, 319, 210, 295, 344,
`370/436; 455/63, 92, 101, 296, 702, 701,
`455/703, 73, 91, 150.1
`See application file for complete search history.
`
`:
`
`QUALCOMM I
`
`ted, S
`
`EP
`
`(Continued)
`FOREIGN PATENT DOCUMENTS
`ne
`5/2007
`1786118 Al
`(Continued)
`OTHER PUBLICATIONS
`Edforset al, “An introduction to orthogonal frequency division mul-
`tiplexing”, Sep. 1996, pp. 1-58."
`(Continued)
`Primary Examiner—David C Payne
`Assistant Examiner—Linda Wong
`(74) Attorney, Agent, or Firm—Thien T. Nguyen; Ross L.
`Franks
`57
`67)
`For eigenmode transmission with minimum mean square
`error (MMSE) receiver spatial processing, a transmitter per-
`forms spatial processing on N, data symbol streams with
`steering vectors to transmit the streams onN,spatial channels
`of a MIMO channel. The steering vectors are estimates of
`transmitter steering vectors required to orthogonalize the spa-
`tial channels. A receiver derives a spatial filter based on an
`MMSEcriterion and with an estimate of the MIMO channel
`response and the steering vectors. The receiver (1) obtains Np
` teceived symbolstreams from N; receive antennas,(2) per-
`forms spatial processing on the received symbolstreams with
`the spatial filter to obtain N, filtered symbol streams, (3)
`performs signal scaling onthefiltered symbol streams with a
`scaling matrix to obtain N, recovered symbolstreams, and (4)
`processes the N, recovered symbol streams to obtain N,
`decoded data streams for the N, data streams sent by the
`transmitter.
`
`ABSTRACT
`
`31 Claims, 6 Drawing Sheets
`
`150
`wr
`
`140
`
`TX Data/
`Controller
`RX Spatial
`
`Data
`
`Spatial
`4
`Processor
`Processor
`
`Feedback
`Info
`
`
`
`
`Received
`Pilot
`
`Symbols
`
`
`
`TX Data
` RX Spatial
`
`Processor
`Processor
`
`Ng
`N
`Ng
`Ng
`e
`Ns
`N,
`Ny
`Recelved
`Received
`Recovered
`Decoded
`Data
`Data
`Transmit Modulated
`Signals
`Symbo!
`Symbol
`Data
`Streams
`Symbol
`Symbol
`Signals
`Streams
`Streams
`Streams
`Streams
`Streams
`
`a 4N
`
`Page 1 of 17
`
`SAMSUNG EXHIBIT 1017
`
`Page 1 of 17
`
`SAMSUNG EXHIBIT 1017
`
`
`
`US7,742,546 B2
`
`Page 2
`
`U.S. PATENT DOCUMENTS
`
`WO
`
`W00341300 Al *
`
`5/2003
`
`
`
`4/2002. Raleigh .....sssseusseeee 375/299
`6,377,631 BL
`8/2003 Crilly etal.
`....
`. 342/378
`6,611,231 B2*
`3/2005 Templeetal.
`..
`. 507/119
`6,861,393 B2*
`5/2006 Thoumyetal.
`. 375/275
`7,039,120 BL*
`9/2002 Khatti oo... cece 455/103
`2002/0127978 A1*
`
`.. 375/267
`...........
`2002/0191703 A1* 12/2002 Ling etal.
`455/92
`2003/0003880 Al*
`1/2003 Lingetal.
`.....
`
`........... 375/295
`2003/0108117 A1*
`6/2003 Ketchum et al.
`A
`x
`....
`. 370/208
`2003/0123381 Al*
`7/2003 Zhuang etal.
`
`375/259
`2003/0185310 A1* 10/2003 Ketchum el al
`:
`aan
`*
`
`.......... 375/260
`2004/0042556 Al
`3/2004 _Medvedevetal.
`
`.. 375/267
`2004/0179627 Al*
`9/2004 Ketchum etal.
`
`FOREIGN PATENT DOCUMENTS
`
`JP
`JP
`WO
`
`2002204193
`7/2002
`2003209534
`7/2003
`W0O0278211 A2 * 10/2002
`
`OTHER PUBLICATIONS
`Joonsuk Kimet al., “Transmission Optimization with a Space-Time
`Filter at Low SNR Wireless Environment,” Globecom 1999,vol. 1B,
`Dee. 5, 1999,pp. 889-893,
`_
`Burr A.G.,
`“Adaptive Space-Time Signal Processing and Coding,
`EEF 2000, vol. 2, Oct. 22, 2000,pp. 710-714.
`Nevang “ et al., FepamanceAnalysis of eeNarbon
`Multiuser Receiver tor
`obile Systems with
`Spatial
`Diver-
`sity,” VTC 2001 Spring. IEEE VTS 53”. Vehicular Technology Con-
`?
`we
`ference. Rhodes, Greece. May 6-9, 2001, IEEE Vehicular Technol-
`ogy Conference, NewYork, NY: IEEE, US, vol. | of4. Conf. 53, May
`
`6, 2001, pp. 142-146
`,
`ree
`:
`International Search Report PCT/US04/032 106—-International
`Search Authority—European Patent Office Nov. 1, 2005.
`Written Opinion PCT/US04/032106-ISA-European Patent Office
`Apr. 8, 2006.
`International Preliminary Examination Report PCT/US04/032 106,
`IPEA US, Jan. 30, 2006.
`
`* cited by examiner
`
`Page 2 of 17
`
`Page 2 of 17
`
`
`
`U.S. Patent
`
`Jun. 22, 2010
`
`Sheet 1 of 6
`
`US 7,742,546 B2
`
`4O|ld
`
`
`
`WALSASOWIW
`
`aSJBAOY
`
`OWIAN
`
`jeuUueyD
`
`
`
`seyewnsyjauueuy
`
`peaieosy
`
`1O|ld
`
`eyedXu
`
`JOSS800)d
`
`piemo.j
`
`OWIIN
`
`jouueyy
`
`IOld
`
`XUeWjouueuyD
`
`sjoquiAs
`*seyeunsyIt
`
`jauueyD
`
`Xue
`
`Joyewnsy
`uolyeyndwoy
`
`Oj}U]
`
`uoHe|NPOW
`
`SOUND
`
`ayeyeleg
`
`eyeqXL
`
`1OSS89014
`
`eyeqg
`
`g01No0S
`
`7
`
`
`Oju|Yoeqpss
`10SSS89014jelyeds
`yeveqXL
`
`
`pepoveqpaiaaooeypeaieoey=penieseypeyejnpo-y=suedeyeqeyeg
`
`
`
`
`
`
`
`
`
`eyegjoquiASjoquiASsjeubissjeubisjoquiAsjoquiAsSUIPOIIS
`SWEOISsweasSSUIPANSSLUBAI]SSWBaIIS
`
`
`
`
`SN“N+N“N°NSN
`OstOk
`zeit]Aiowey|ljoyewnsy
`snieis1211/4OlleBuuacrs
`
`
`
`OgksjoquAS0g!lld0%!ait
`
`TealsooudorJo}jiusuelL
`0616hOrt8hOFt
`
`penedspenrsceysjoujUaD
`
`
`
`jouueyy
`
`40SS800/4ereq
`yoeqpssv
`
`frenedsxy
`
`Page 3 of 17
`
`Page 3 of 17
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`U.S. Patent
`
`Jun. 22, 2010
`
`Sheet2 of 6
`
`US 7,742,546 B2
`
`Ng
`Data
`Symbol
`Streams
`
`beeeeeedeweeeeeeeeeeeeeeeee
`
`e
`
`:
`
`216a
`
`
`
`
`Mapping
`
`Nz
`Data
`Streams
`
`
`
`:
`
`
`
` Freeteedaewwwewwwneweeeeeeeeeesnake
`
`120
`
`TX Data Processor
`
`212a
`
`214a
`
`e
`
`:
`212s
`
`e
`
`:
`214s
`
`Encoder
`
`Channel
`Interleaver
`
`®
`
`:
`216s
`
`Symbol
`
`Coding
`Controls
`
`Modulation
`Controls
`
`FIG. 2
`
`Page 4 of 17
`
`Page 4 of 17
`
`
`
`
`sossaooigmeewTRTi3e098eore__ONIN}ggapSloqutgrole|HLL(OlidPOZE
`
`
`
`
`sole}Po]aeXL[1[eeded[VEcay|xeYYiNg}'|rt'LL'J\|pe|WEeT|!ge1__JoHinPoWWdI0ti£1088a1jOAd4al
`
`
`
`
`
`peyeinpoy|joquiAsloquiAs
`
`xnwegt)SHXL|||MueddJory(hgxUIeYYt{'s}'|14:LLiLN)ONOAD44st
`
`
`sjeubigZLSUUBAIISof!sweals
`
`+NWUISUedLeyed
`jenedsXLDnMni_sesgmmeeset
`TTTTTiggmgpba3pi
`
`
`wun')soyer9uexwun|xAWAdanxnweg
`
`
`xnAdninyy pooaaaaaTE“N*NpotasaassssaTEELNNWun||soyesauesywoVs
`°°
`fot10lldCnyB
`+N°N
`
`
`
`US 7,742,546 B2
`
`€5d
`
`sjoquwAs_
`
`U.S. Patent
`
`Jun. 22, 2010
`
`Sheet 3 of 6
`
`Page 5 of 17
`
`Page 5 of 17
`
`
`
`
`4i'
`'!i''
`t{
`|'|i1't11
`'(''1(!
`'‘''ti}‘
`
`'t
`y''
`
`''
`11
`(1
`(!
`(a
`
`DSAIaOOH
`
`
`
`S|OQUIAS}0|Id+5
`
`bpDId
`
`
`
`palsacoayPpealeooy
`
`SWedSoglSWE8)S
`
`joquiAgJoquiAS
`
`SN4N
`eenneeoeth
`20SeCry20bPEOgP
`
`xnweg;ASWCu}
`feqeds
`eMoHrhsjeubis
`r]LadxyerdOAD!|tHxY[I1Peppay}etCNtJoyejnpoweqdWO4OIiPOLPee|{e0zP
`pon--------------.
`~-apnee--=e
`aoaaaaaa7wun
`YADY
`l0zP:®13°e'e
`WerWer|apiZN
`Joye|InpowegWa40tot
`
`jeaowoy[+]wun|}
`yun|}issxyeldOATV|edxy[I
`wun)feaowoy|r}
`
`
`U.S. Patent
`
`Jun. 22, 2010
`
`Sheet 4 of 6
`
`US 7,742,546 B2
`
`'1''''4'i]4''1i1
`
`'i
`t'
`‘‘+
`
`t
`
`('Ng}
`
`><
`2D
`= +
`
`_
`
`wn
`ea &
`
`40S$S990/qjenedsxy
`
`JOuI4
`
`L
`
`ASNW
`
`reeds
`
`10414
`
`~,
`=—
`
`4N
`
`pealaaay
`
`Page 6 of 17
`
`Page 6 of 17
`
`
`
`
`
`
`
`
`U.S. Patent
`
`Jun. 22, 2010
`
`Sheet 5 of 6
`
`US 7,742,546 B2
`
`N
`
`Decoded
`Data
`Streams
`
`
`
`170
`RX Data Processor
`514a
`
`512a
`
`516a
`
`$
`e
`512s
`Symbol
`
`:
`e
`514s
`
`fos
`:
`s
`e
`s1és
`
`N
`S
`Recovered
`Symbol
`Streams
`}
`
`i
`
`|
`!
`
`:
`.
`
`
`Channel
`
`
`
`Demodulation
`Controls
`
`Decoding
`Controls
`
`FIG. 5
`
`Page 7 of 17
`
`Page 7 of 17
`
`
`
`streams from N, transmit antennas
`
`Obtain estimate, Aww, of channel
`response matrix for each subband
`
`Decompose estimated channel
`response matrix H(k) foreach
`subband to obtain matrix V(x) of
`steering vectors that orthogonalize
`N, spatial channels for the subband
`
`Process N, data streams to
`obtain NV, data symbol! streams
`for transmission on the N,, spatial
`channels of the MIMO channel
`
`For each subband, perform spatial
`processing on vector s(k) for N,
`data symbol streamswith the
`steering vectors to obtain vector
`x(k) for N, transmit symbol streams
`
`Transmit N, transmit symbol
`
`FIG, 6
`
`U.S. Patent
`
`Jun. 22, 2010
`
`Sheet6 of 6
`
`US 7,742,546 B2
`
`w
`-—
`
`2
`
`712
`Obtain estimate, Arn, of channel
`response matrix for each subband
`
`714
`
`
`
`Decomposeestimated channel
`response matrix H(k) for each
`subband to obtain matrix
`
`
`V(4 for the subband
`
`716
`
`
`
`For each subband,derive
`spatialfilter matrix W(A) and
`diagonal matrix Do(k) based on
`MMSEcriterion and with matrices
`H(«) and V(4j for the subband
`
`718
`
`
`
`Obtain N, received symbol
`
`
`streams from N,, receive antenna
`for N, data symbof streams
`
`
`transmitted via N, spatiat
`
`channels of the MIMO channel
`
`
`For each subband, perform
`
`
`spatial processing on vector r(k}
`for the N, received symbol
`
`
`streams with the spatial filter matrix
`W(k) to obtain vector 3(k}
`
`
`for N, filtered symbol streams
`722
`
`720
`
`
`
`For each subband, perform signal
`
`
`scaling on vector S(k) with diagonal
`matrix DA(k) to obtain vector 8(k) for
`
`
`N, recovered symbol streams
`
`724
`
`
`
`
`
`Process the N, recovered
`symbo! streams to obtain
`N, decoded data streams
`
`
`
`
`
`End
`
`FIG. 7
`
`Page 8 of 17
`
`Page 8 of 17
`
`
`
`US 7,742,546 B2
`
`2
`ditions and transmits the N;.transmit symbolstreamsfrom the
`N, transmit antennasto the receiver.
`The receiver derives a spatial filter based on a minimum
`mean square error (MMSF)criterion and with the channel
`response estimate and the steering vectors. The receiver also
`derives a scaling matrix. The recciver obtains Nz reccived
`symbol streams from N, receive antennas for the N, data
`symbol streams transmitted on the N, spatial channels. The
`receiver performsspatial processing on the N, received sym-
`bol streams with the spatial filter and obtains N,filtered
`symbol streams. The receiver further performs signal scaling
`on the filtered symbol streams with the scaling matrix to
`obtain N, recovered symbol streams, which are estimates of
`the N, data symbol streams sent by the transmitter. The
`receiver then processes(e.g., demodulates, deinterleaves, and
`decodes) the N, recovered symbol streams to obtain N,
`decoded data streams, which are estimates of the N, data
`streams sent by the transmitter.
`The receiver spatial processing techniques described
`herein may be used for single-carrier and multi-carrier
`MIMOsystems. lor a multi-carrier MIMO system,the spa-
`tial processing at the transmitter and receiver may be per-
`formed for each of multiple subbands.
`Various aspects and embodiments of the invention arc
`described in further detail below.
`
`
`
`BRIEF DI tfSCRIPTION OF THE DRAWINGS
`
`1
`RECEIVER SPATIAL PROCESSING FOR
`EIGENMODE TRANSMISSION IN A MIMO
`SYSTEM
`
`BACKGROUND
`
`I. Field
`
`The present invention relates generally to data communi-
`cation, and more specifically to techniques for performing
`receiver spatial processing in a multiple-input multiple-out-
`put (MIMO) communication system.
`II. Background
`AMIMOsystem employs multiple (N,) transmit antennas
`and multiple (N;) receive antennas for data transmission and
`is denoted as an (N;, Nz) system. A MIMO channel formed
`by the N, transmit and Nz receive antennas may be decom-
`posed into N.,spatial channels, where N,<= min {N,, Nz}).
`The Ng spatial channels may be used to transmit up to Nx
`independent data streams to achieve greater overall through-
`put. Spatial processing may or may not be performed by a
`transmitter and is performed by a receiverin orderto transmit
`multiple data streams on the N, spatial channels.
`The N, spatial channels may or maynot be orthogonal to
`one another. Orthogonal spatial channels can only be
`obtained when both (1) the transmitter performs spatial pro-
`cessing with the proper steering vectors and (2) the receiver
`performsspatial processing with the properspatialfilter. The
`orthogonality of the spatial channels thus depends on (1)
`whetheror not spatial processing was performedatthetrans-
`mitter and (2) whetheror notthe spatial processingat both the
`transmitter and the receiver was successfulin orthogonalizing
`the spatial channels. Each spatial channel is referred to as an
`“eigenmode”of the MIMOchannelifthe N, spatial channels
`are orthogonal to one another. In this case, N, data streams
`may betransmitted orthogonally on the N,, eigenmodes. Per-
`formance is better when the spatial channels are orthogonal.
`However,in a practical system, the N., spatial channels are
`usually not completely orthogonal to one another due to vari-
`ous reasons. For example, the spatial channels would not be
`orthogonal if (1) the transmitter has no knowledge of the
`MIMOchannelor(2) the transmitter and/orthe receiver have
`an imperfect estimate of the MIMO channel. If the spatial
`channels are not orthogonal, then each data streamwill expe-
`rience cross-talk from the other data streamsat the receiver.
`‘The cross-talk acts as additive noise that degrades pertor-
`mance.
`
`There is therefore a need in the art for techniques to miti-
`gate the deleterious effects of cross-talk when transmitting
`data on multiple spatial channels in a MIMOsystem.
`
`SUMMARY
`
`Techniques for performing receiver spatial processing in a
`mannerto mitigate cross-talk and achieve better performance
`are provided herein. Initially, a transmitter and/or a receiver
`estimates the response of a MIMO channel and decomposes
`the channel response estimate to obtain steering vectors,
`whichare estimates ofthe transmitter steering vectors needed
`to orthogonalize the N, spatial channels of the MIMO chan-
`nel. The transmitter is provided with the steering vectors if
`they are derived by the receiver. The transmitter processes
`(e.g., encodes, interleaves, and modulates) N.. data streams to
`obtain N, data symbol streams for transmission on the N,
`spatial channels. The transmitter performsspatial processing
`on the N, data symbol streams with the steering vectors to
`obtain N,-transmit symbolstreams. The transmitter then con-
`
`40
`
`45
`
`Page 9 of 17
`
`The features and nature of the present invention will
`become more apparent from the detailed description set forth
`below whentaken in conjunction with the drawings in which
`like reference characters identify correspondingly through-
`out and wherein:
`FIG. 1 shows a transmitter and a receiver in a MIMO
`system;
`FIG. 2 shows a transmit (TX) data processorat the trans-
`mitter;
`FIG. 3 shows a TX spatial processor and a transmitter unit
`at the transmitter;
`FIG. 4 shows a receiver unit and a receive (RX) spatial
`processorat the receiver;
`FIG. 5 shows an RX data processorat the receiver; and
`FIGS.6 and 7 show processes performedby the transmitter
`and the receiver, respectively, for eigenmode transmission
`with MMSEreceiverspatial processing.
`
`DETAILED DESCRIPTION
`
`The word “exemplary”is used herein to mean “serving, as
`an example, instance, or illustration”? Any embodiment or
`design described herein as “exemplary” is not necessarily to
`be construed as preferred or advantageousover other embodi-
`ments or designs.
`The receiver spatial processing techniques described
`herein may be used in a single-carrier MIMOsystemas well
`as a multi-carrier MIMO system. Multiple carriers may be
`provided by orthogonal frequency division multiplexing
`(OFDM), other multi-carrier modulation techniques, or some
`other constructs. OFDMeffectively partitions the overall sys-
`tem bandwidth into multiple (N,,) orthogonal subbands,
`which are also commonly referred to as tones, bins, or fre-
`quency channels. With OI'DM,each subbandis associated
`with a respective carrier that may be modulated with data. For
`clarity, the receiver spatial processing techniques are specifi-
`cally described below for a MIMO system that implements
`OFDM (i.e.,a MIMO-OFDMsystem).
`
`
`
`Page 9 of 17
`
`
`
`US 7,742,546 B2
`
`3
`A frequency-selective MIMO channel with N,. transmit
`antennas and N, receive antennas may be characterized by N;-
`frequency-domain channel response matrices H(k), for k=
`1...N,, each with dimensions of N,xN,. These channel
`response matrices may be expressed as:
`
`4
`Eigenmode transmission refers to transmission of data on
`the N, eigenmodes of the MIMO channel. Eigenmodetrans-
`mission requires spatial processing by both the transmitter
`and the receiver, as described below.
`The spatial processing at the transmitter for eigenmode
`transmission on subband k maybe expressed as:
`Xideah= M(K)s(K),
`
`Kq (4)
`
`Eq (1)
`hyp(k)
`hya(k)
`Wyk)
`
`Arik)—ha2(k) hawk)
`H(k)=
`.
`.
`:
`>
`10
`wheres(k) is an (N;x1) data vector with N, non-zero entries
`for N, modulation symbols to be transmitted on the N, eigen-
`modes for subband k; and
`X,geak) is a (N;x1) transmit vector with N; entries for N;
`transmit symbols to be sent from the N; transmit anten-
`nas for subband k.
`
`Fink) fing2k) + Pvpp(k)
`
`fork=1... Ne,
`
`15
`
`where entry h,,(k), fori=l...Nxz, j=l ...N,andk=1...N;,
`is the coupling(i.e., complex gain) betweentransmit antenna
`j and receive antennai for subband k.
`The channel response matrix H(k) for each subband may
`be “diagonalized”to obtain the N,, eigenmodesfor that sub-
`band. This diagonalization may be achieved by performing
`either singular value decomposition of the channel response
`matrix H(k) or eigenvalue decomposition of the correlation
`matrix of H(k), which is R(k)=H”(k)H(k).
`The singular value decomposition of the channel response
`matrix H(k) maybe expressedas:
`A(R=UDI&) VA), for k=1... Np,
`
`Eq (2)
`
`N, entries of s(k) can represent N, data streams and the
`remaining entries of s(k), if any, are filled with zeros.
`The received symbols obtained by the receiverfor subband
`k may be expressedas:
`FidealK)-H(R%iaealk)+n (=H(kK) M(R)s(k)+n(8),
`
`Eq (5)
`
`wherer,;.,(K) is an (N,x1) received vector with N, entries
`for N, received symbols obtainedvia the N, receive antennas
`for subband k,; and
`n(k) is a noise vector for subband k.
`The spatial processing or matchedfiltering at the receiver
`to recover the data vector s(k) may be expressedas:
`
`where U(k)is a(N,xN,) unitary matrix ofleft eigenvectors of
`H(k):
`X(k) is an (N,xN,) diagonal matrix of singular values of
`H(k); and
`V(k) is a (N;xN,) unitary matrix of right eigenvectors of
`H(k).
`
`A unitary matrix M is characterized by the property M7M=I,
`wherelis the identity matrix. The columnsof a unitary matrix
`are orthogonalto one another.
`The cigenvalue decomposition ofthe correlation matrix of
`H(k) maybe expressed as:
`A(RHADH(D=V(AK) M0, for K=1 2. Nz
`
`Kq 3)
`
`35
`
`40
`
`45
`
`where A(k) is a (NpxN,) diagonal matrix of eigenvalues of
`R(k). As shown in equations (2) and (3), the columns ofV(k)
`are eigenvectors ofR(k) as well as right eigenvectors ofH(k).
`Singular value decomposition and eigenvalue decomposi-
`tion are described by Gilbert Strang ina bookentitled “Linear
`Algebra and Its Applications,” Second Edition, Academic
`Press, 1980. The receiver spatial processing techniques 5
`described herein may be used in conjunction with either sin-
`gular value decomposition or eigenvalue decomposition. For
`clarity, singular value decompositionis usedforthe following
`description.
`Theright cigenvectors ofH(k) are also referred to as “steer-
`ing” vectors and may be used for spatial processing by a
`transmitter in order to transmit data on the N,, eigenmodesof
`H(k). The left eigenvectors of H(k) may be used for spatial
`processing by a receiverin orderto recoverthe data transmit-
`ted on the N, eigenmodes. The diagonal matrix X(k) contains
`non-negative real values along the diagonal and zeros every-
`whereelse. These diagonal entries are referred to as the sin-
`gular values of11(k) and represent the channel gains for the N.
`eigenmodes of H(k). Singular value decomposition may be
`performed independently on the channel response matrix
`H(k) for each of the N; subbands to determine the N, eigen-
`modes for that subband.
`
`Side(k) = AHA) V4 (VHF (Wrideat lh)»
`
`Eq 6)
`
`= ATVTHYHOO (s(k) + n(h)),
`= 5(k) + Rigeatlk),
`
`where §,44k) is an (N;*1) estimated data vector with up to
`N, recovered data symbols for subband k; and
`fiizeedk) is a vector of post-processed noise for subband k.
`The matchedfilter used by the receiver for subband k may
`be expressedas:
`
`Miaeatk)-VQH
`
`Eq (7)
`
`The multiplication by A7!(k) in equation (6) accounts for the
`(possibly different) gains of the N, spatial channels and nor-
`malizes the output of the matchedfilter so that recovered data
`symbols with the proper magnitude are provided to the sub-
`sequent processing unit. The normalization(i.e., signal scal-
`ing) is based on the following observation:
`VF(TFDITK) ViR)=2F(DE)=A(K).
`
`Eq(8)
`
`Equation (8) indicates that the eigenvalues of H”(k)H(k) in
`the diagonal matrix A(k) are also the squares of the singular
`values of H(k) in the diagonal matrix X(k).
`Equation (6) indicates that the N,, data symbolstreams s(k),
`distorted only bypost-processed channel noise fi,7.,,(k), may
`be obtained with the proper spatial processing at both the
`transmitter and the receiver. However, the result shown in
`equation (6) is ideal in that both the transmitter and the
`receiver are assumed to have perfect information about the
`MIMOchannel. In a practical system, both the transmitter
`and the receiver will have noisy estimates of the MIMO
`channel and/or noisyestimates ofthe eigenvectors and eigen-
`values. In this case, the recovered data symbols for each
`streamwill be corrupted by cross-talk fromthe other streams.
`
`Page 10 of 17
`
`Page 10 of 17
`
`
`
`5
`Thespatial processing at the transmitter in a practical sys-
`tem for subband k maybe expressed as:
`
`6
`and the data vector s(k) is minimized. This MMSEcriterion
`may be expressedas:
`
`US 7,742,546 B2
`
`x)= Fibs),
`
`Eq (9)
`
`where V(k) is a matrix of steering vectors used bythe trans-
`mitter for subband k; and
`x(k) is a transmit vector obtained with V(k).
`
`The matrix V(k) is an estimate ofV(k) and may be obtained,
`for example, by performing singular value decomposition of
`H(k), which is an estimate of H(k).
`The received symbols obtained by the receiver for subband
`k may be expressedas:
`
`HA)=H(DPRs()+n (0).
`
`Eq (10)
`
`The matched filter M(k) for the received symbols may be
`expressed as:
`
`Mi)=)"(DHAK).
`
`Kg (11)
`
`15
`
`20
`
`Similar to the transmitter, the receiver in the practical system
`only has an estimate of this matched filter.
`The spatial processing at the receiver in the practical sys-
`tem for subband k maybe expressed as:
`
`Spractk) = AW (KN (rth),
`= ACOM OHCOVOSH)+ fipacllO,
`= s(k) + c(k) + ftpractk),
`
`Bq U2)
`
`30
`
`35
`
`40
`
`where M(k) is an estimate of M(k) for subband k;
`A(k)=diag [M(k)H(k)V(k)] for subband k; and
`c(k) is a vector of cross-talk terms for subbandk.
`In equation (12), A(k) is a diagonal matrix whose diagonal
`elements are the diagonal elements of M(k)H(k)V(k). The
`cross-talk terms in c(k) are generated by the off-diagonal
`terms of M(k)H(k)V(k), which result from (1) the use of an
`imperfect estimate of V(k) by the transmitter and (2) the use
`of an imperfect estimate of M(k) by the receiver. The cross-
`talk termsact asadditive noise that degrades the quality ofthe
`estimated data vector §,,,(k).
`The powerin the cross-talk vector c(k) may be small rela-
`tive to the signal powerinthe data vectors(k) ifthe transmitter
`has a good estimate of V(k) and the receiver has a good 5
`estimate of M(x), both of which require a good estimate of
`H(k). Good estimates of both V(k) and M(k) are needed to
`orthogonalize the N, spatial channels and to minimize deg-
`radation dueto cross-talk. If the transmitter has a good esti-
`mate of V(k), then a good estimate of M(k) is needed to
`minimize the off-diagonal terms of M(k)H(k)V(k). However,
`if the transmitter has a poor estimate of V(k), then the cross-
`talk terms may havesignificant amounts of power evenif the
`receiver has a perfect estimate of M(k).
`The receiver can use MMSEspatial processing to suppress
`the cross-talk terms and maximize the signal-to-noise-and-
`interference ratio (SNR) ofthe estimated data vector. The
`MMSEreceiverspatial processing can provide improvedper-
`formance when the transmitter has an imperfect estimate of
`
`V(k). An MMSEreceiver utilizes a spatial filter having a
`response ofW(k), whichis derived suchthat the mean square
`error between the estimated data vector fromthespatialfilter
`
`minE[(W(kr(k) —s(k))(Wir) — sk),
`
`Eq (13)
`
`where E[x] is the expected value of x.
`Thesolution to the optimization problem posed in equation
`(13) may be obtained in various manners. One exemplary
`method for deriving the MMSEspatial filter matrix W(k) is
`described below. For this method, the matrix W(k) may be
`expressedas:
`
`WU)=F"0A) [AMPH(+6,|,
`
`Eq (14)
`
`where ,,,,(k) is an autocovariance matrix of the receive noise
`process for subband k, whichis 9,,,,(k)=E[n(k)n(k)].
`The spatial processing by the MMSEreceiver for subband
`k maythen be expressedas:
`
`31K) = Dg (Wk),
`= Dol (kK) WHOOVsth)+ Filk),
`= Do(HOWs(k) + i),
`
`where fi(k) = Dol ow kintk),
`
`Oth) = WIDHE)V
`= 0OH"Go|HuotGot” Gon" (K+ Sontk)]
`H(AV(k),
`
`= OT(HYHH" (h) + GlTHEVA),
`
`Eq (15)
`
`Eq (16)
`
`and Dotk) = diag [W(k)HK)Vk)].
`
`Eq (17)
`
`Do(k)is a diagonal matrix whose diagonal elements are the
`diagonal elements of Q(k). Using the matrix inverse identity,
`equation (16) can be rewritten as:
`
`Q)=9H0,(HHMPRL
`PAGO"mn(OEP.
`
`Eq (18)
`
`If the noise vector n(k) is additive white Gaussian noise
`(AWGN) with zero mean and an autocovariance matrix of
`6,,,(K)=071, where o” is the variance of the noise, then equa-
`tions (14) and (18) may be simplified as:
`
`Wi=""(OHAOIA}MMnH(+077 |, and
`
`Eq (19)
`
`Ob)FARAH) M0 [PFUoAW
`Peery.
`
`The MMSEreceiverspatial processing in equation (15) is
`composed oftwo steps. In the first step, the vector r(k) for the
`Np received symbol streams is multiplied with the MMSE
`spatial filter matrix W(k) to obtain a vector S(k) for N, filtered
`symbolstreams, as follows:
`
`SH)=WR
`
`Eq (20)
`
`TheN,filtered symbolstreamsare unnormalized estimates of
`the N., data symbolstreams. In the secondstep, the vector8(k)
`is multiplied with the scaling matrix Dg'(k) to obtain the
`vectorS(k) for the N, recovered symbolstreams, as follows:
`
`S(K=Dg(HS(K).-
`
`Eq (21)
`
`Page 11 of 17
`
`Page 11 of 17
`
`
`
`US 7,742,546 B2
`
`7
`The N, recovered symbol streams are normalized estimates
`of the N, data symbol streams.
`Asnoted above,the receiver spatial processing techniques
`described herein mayalso be used for a single-carrier MIMO
`system. In this case, the description above applies, albeit
`without the subband index k. The spatial processing at the
`transmitter can be expressedas:
`x=Vs.
`
`Eq (22)
`
`The MMSE spatial processing at
`expressedas:
`
`the receiver can be
`
`§=DgWr
`
`Eq (23)
`
`or §=Wrand s=Dg"'8.
`The MMSEspatial filter response W can be expressed as:
`W=THHF VV4q,,.) >
`Bq (24)
`
`If the noise is AWGN with an autocovariance matrix of
`byn=0'l, then the MMSEspatialfilter response simplifies to:
`Eq (25)
`W=PHHYAPP2p!
`
`
`
`The MMSEspatial filter matrices W and W(k) mayalso be
`derived using, other methods. For example, these matrices
`may bederived using time recursive methods such asa recur-
`sive least square method, a least mean square method, and so
`on, which are knownin theart.
`FIG. 1 showsa block diagram of a transmitter 110 and a
`receiver 150 ina MIMO system 100. At transmitter 110, a TX
`data processor 120 receives N. data streams from a data
`source 112. TX data processor 120 processes (e.g., encodes,
`interleaves, and modulates) each data stream in accordance
`with a rate selected for that data stream to obtain a corre-
`sponding data symbol stream. The selected rate for each data
`stream mayindicatethe data rate, coding schemeorcoderate,
`modulation scheme, and so on,to use for that data stream,all
`of which are indicated bythe various controls provided by a
`controller 140. A TX spatial processor 130 receives N, data
`symbolstreams from TX data processor 120, performsspatial
`processing on these streams with the matrices V(k), for k=
`1...N,, multiplexes in pilot symbols, and provides N;
`transmit symbol streams to a transmitter unit (TMTR) 132.
`The pilot symbols are modulation symbols knowna priori and
`may be used byreceiver 150 for channel estimation.
`Transmitter unit 132 performs OFDM modulation on the
`N;,transmit symbol streams to obtain N; OFDM symbol
`streams. Transmitter unit 132 further conditions (e.g., con-
`verts to analog, frequency upconverts,filters, and amplifies)
`the OFDM symbol streams to obtain N, modulated signals.
`Each modulated signal is transmitted from a respective trans-
`mit antenna (not shown in FIG.1) and via a forward MIMO
`channel to receiver 150. The MIMOchanneldistorts the N;
`transmitted signals with the channel response H(k), for k=
`1...N,, and further degrades the transmitted signals with
`noise and possibly interference from other transmitters.
`At receiver 150, the N, transmitted signals are received by
`each ofNz receive antennas (not shown in FIG. 1), andthe Nz
`received signals from the Nz, receive antennasare provided to
`a receiver unit (RCVR) 154. Receiver unit 154 conditions,
`digitizes, and pre-processes each receive signal to obtain a
`corresponding received chip stream. Receiver unit 154 fur-
`ther performs ODM demodulation on each received chip
`stream to obtain a corresponding received symbol stream.
`Receiver unit 154 provides N, received symbol streams (for
`data) to an RX spatial processor 160 and received pilot sym-
`bols (for pilot) to a channel estimator 172.
`
`Page 12 of 17
`
`10
`
`15
`
`40
`
`45
`
`5
`
`8
`RX spatial processor 160 performs spatial processing on
`the Nz received symbol streamsto obtain N, recovered sym-
`bol streams, which are estimates of the N, data symbol
`streams sent by transmitter 110. An RX data processor 170
`further processes (e.g., demodulates, deinterleaves, and
`decodes) the Ny, recovered symbol streams to obtain No
`decoded data streams, which are estimates of the N, data
`streams sent by transmitter 110. RX data processor 170 also
`provides the status of each decoded packet, which indicates
`whetherthe packet is decoded correctly or in error.
`Channel estimator 172 processes the received pilot sym-
`bols to obtain channel estimates for the forward MIMOchan-
`nel (e.g., estimated channel response matrices H(k), for k=
`1...N,, noise variance estimate, o”, and so on). A matrix
`computation unit 174 receives the channel estimates, com-
`putes the MMSEspatial filter matrices W(k) and the scaling,
`matrices Dg '(k) fork=1 ...N,, and provides these matrices
`to RX spatial processor 160. Matrix computation unit 174
`may also compute the matrices V(k), for k=1 ... Nz, of
`steering vectors for transmitter 110.
`A controller 180 receives the channel estimates from chan-
`nel estimator 172 and the packet status from RX data proces-
`sor 170, selects the rates for the N, data streams, and
`assembles feedback information for transmitter 110. The
`
`feedback information may include the Ng, selected rates,
`acknowledgments (ACKs) and negative acknowledgments
`(NAKs)forthe decodedpackets, the matrices V(k), and soon.
`‘Lhe feedback information andpilot symbols are processed by
`a TX data/spatial processor 190, conditioned by a transmitter
`unit 192, and transmitted via a reverse MIMO channel to
`transmitter 110.
`
`At transmitter 110, the Nz signals transmitted by receiver
`150 are received and conditioned bya receiver unit 146 and
`further processed by an RX spatial/data processor 148 to
`obtain the feedback information sent by receiver 150. Con-
`troller 140 receives the feedback information, uses the ACKs/
`NAKsto control the transmission of data packets to receiver
`150, and uses the N, selected rates to process newpacketsfor
`the N,data streams.
`Controllers 140 and 180 direct the operation at transmitter
`110 and receiver 150, respectively. Memory units 142 and
`182 provide storage for program codes and data used by
`controllers 140 and 180, respectively. Memoryunits 142 and
`182 maybeinternal to controllers 140 and 180, as shown in
`FIG. 1, or external to these controllers. Some of the process-
`ing units shown in FIG.1 are described in detail below.
`Transmitter 110 may be an access point and receiver 150
`may bea userterminal in the MIMO system,in which case the
`forward and reverse MIMOchannels are the downlink and
`uplink, respectively. Alternatively, transmitter 110 may be a
`user terminal and receiver 150 may be an access point, in
`whichcase the forward and reverse MIMO channelsare the
`uplink and downlink, respectively.
`FIG. 2 shows a block diagram of an embodiment of TX
`data processor 120 at transmitter 110. For this embodiment,
`TX data processor 120 includes one set of encoder 212,
`channel interleaver 214, and symbol mapping unit 216 for
`each ofthe Ns data streams. For each data stream {d, }, where
`I=] ...N,, an encoder 212 receives and codesthe data stream
`based ona coding schemeindicated by the coding control and
`provides code bits. The data stream may carry one or more
`data packets, and each data packet is typical