INTERFERENCE CANCELLATION IN A MULTI-USER RECEIVER

TECHNICAL HELD

The present invention generally relates to interference cancellation, and more particularly relates interference cancellation in a multi-user receiver.

BACKGROUND

Mam types of wireless communication devices receive and process signals transmitted by more than one user. For example, a base station in the WCDMA (Wideband CDMA) uplink receives signals transmitted by multiple users. One or more users may transmit more than one signal Dev ices that receive and process signals transmitted by different users are commonly referred to as multi-user receivers because they demodulate and decode a codeword extracted from each of the different signals. Signal transmissions by multiple users often overlap in time and frequency, causing interference at the multi-user receiver.

An optimal muiti-υser receiver demodulates all signals jointly, at the expense of enormous complexity. Suboptimal multi-user structures such as parallel or seπal multiuser receivers are more practical solutions for cancelling multi-user interference in that they can prov ide suitable interference cancellation for certain applications with much lower complexity. Suboptunal multi-user receivers enhance performance by invok ing the individual signal decoders in the multi-user receiver. For example, in the WCDMA uplink, multiple users transmit in a somewhat un coordinated manner. That is, signals arrive at the base station with their slots misaligned, different users may have different spreading factors, etc. Each user individually coordinates with the base station, but without any direct relation to other users.

The base station can boost its performance by cancelling signals it has already processed from the common received signal. Doing so improves the performance of subsequent signals. In a purely illustrative example, two signals may arrive at the multi-user receiver and overlap by T' nis. The span of each signal contains ail the samples corresponding to a respective codeword. In the VVCDMA uplink, this is referred to as a TTi (transmission time interval). A slot duration is 0.67 røs in WCDMA and a TTf may include 3 slots (2ms) or 15 slots (H) ms). As such, the spans of the {wo signals are not necessarily of equal length T he station can cancel the first Signal fiom the combined receiv ed signal to benefit (he second signal

How e\ct , (he teceix ci foϊ the fust signal must wail until all samples in the corresponding span are received, which ends / - ? ^' ras later than the span of the second signal After the receiver for {he fnsj signal is done with decoding it {hen \ econ.su ucts the fust signal for the purposes of cancellation fgnoiing processing delav s foi simplicity, the receπ ci for the second signal must w ait at least an additional dela> / ~ / 'ms to benefit from the cancellation of the first signal fhe dela\ incurred is related to the length of {he codew ord associated with {he fnsj signal and slows {he feedback ftoni die decoder to the cancellation stage This delav is not acceptable for cct tain application such aa applications mav prevent othetvx ise beneficial cancellation from being used Furthermore, it ι$ difficult to run a stable system wheie cancellation ;s turned on and oft depending on iaudom ielatn e offsets benv een signals

SUMM ARY

Aceoidmg to the methods and apparatus disclosed heiein, the deias associated \\ uh pei foi mm" interference cancellation m a multi-user receiv er is reduced bv enabling partial decoding for cancellation I he teim 'coded bus' used herein means the bus produced b\ an encodet, in conuast to infotmanon bits, w inch aie ieά to the encodci In one ernbodirneut, partial decoding ι% peifoiraed Ueating unaunlabie coded hits aa enismes, or punctωed bits The unav ailable coded bits are cairied modulation symbols that arrix e after a certain time threshold, and thus {heir hard or soil \ aiues are not asaϊlable foi decoding These unavailable bits aie seated as effectiv ely punctured bus dining the decoding process b\ setting the corresponding soft -values to zero Standard decoding techniques for block, comoHitional, and turbo codes can all handle pimctuted bits appropriated Ia addition, ev en if too main bits aie absent for the original puipose of decoding mfonnation bits accuiateh . U is still beneficial to use the partial decoding to obtain soft v alues for the coded bits

According to an embodiment, mterfeience is canceled fiom a iecen ed signal hav ing signal contributions ironi multiple users b\ partmll> decoding a fust codewoid obtained from the teceix cd signal and tiansmitted b\ a fast user \ ia a Hist signal The first signal is reconstructed based on the paitial decoding of flic first codeword and canceled fiom (he receiv ed signal to yield a leduced-mjerfeience signal V second codcw oid transmuted b> the fast usci oi a second user v ia a second signal ts at least partialh decoded from the reduced interference signal

f he demodulator output is m the form of log likelihood ration (LLR) of {he coded bits, or an appio\ntιatκm thereof \ positive (negative) so(t value indicates a bit is a 0 { 1 ) A soft thai is equal to zero indicates the demodulator lacks mfoimauon about the corresponding bit The missing bit can be treated as an erasuie. in the sense that the channel did not cam information about that hit As such, zero soft v alises can be used to rcptesent unav ailable coded bits The decoder is capable of processing codcw oids where cciiain bus been punctured at the transmitter, and not sent to the receiv er, b\ setting their soft v alues to zero In the same v\ a} , the deeodei treats unav ailable bits as puocluted bits, aKo setting then soli v alues to /eio

ϊn one embodiment Ui e fust signal is reconstructed b> (a) eϊTectiveh puncturing the unav ailable coded bits of {he first signal, e g bv setting the corresponding soft v alues to /ero. <b) paitialij decoding the litst codew ord piocessmg Uie av ailable soft v alues and the alue soft v alues, to produce piobabiluies of the available and possibh the unav ailable coded bits, (C) reconstructing the soft or hard s> mboK and Ui) spreading and filtering {he i eeottsli ucied symbols to complete die estimate of the lint signal

Of coui&c. the picseiit invention is not limited to the abov e feattues and adv antages Those skilled sn the art will iecogni/c additional fcatδies and adv antages upon reading the follow ing detailed description, and upon v iew ing {he accompanying draw∑ngs

BRIhF DFSCRlP T lON OF THh DRAW INGS

Figute 1 iUustiates a block diagram of an embodiment of a multi-uses iecei\ ei Figuie 2 iUustiates a block diagram of an embodiment o( a receiv er module included in the multi-ubci receive* of Figuie 1

Figure 3 illustrates a signal transmission diagram of ov erlapping signals processed bv the muifi-usei receiver of Figure 1 Figure 4 illustrates a diagram of a bit puncturing technttμie cmpioj ed b> the iiniitt-usei receu er of Ftguie 1

Ftguie 5 illustrates a diagiam of a symbol reconstruction technique emρlo\ cd b\ the multi-user recener of Figure J

Hgure 6 illustrates a block diagram of an embodiment of a signal cancellation module included ui the mulu-uset reeen er oϊ Frgtae i

f igure 7 illustiatcv a signal transmission diagram of a signal processed by the muitϊ-nser receix er of b igυre 1

Figure H illustrates a reconstructed inter fei ing sy mbol ox erlapping sj mboK of another signal processed by die multi-user receiv er of Figute ϊ

Ftguie 9 iUustiates a block diagiarø of an embodiment of a serial mterfetence cancellation multi-user rccei\er

Figute IO jllusualεv a signal transmisMυu diagiam υf o\«t lapping bigαaib puxessεd b> the seπai intetieience canceilation τnulti-uset tecet\ eι of Figure ⁽ >

DL 1 All fcD DFSCRlP \ ION

figui e S illusUatεs mi embodiment oϊ a nιulti-us»ei lecetvei 100 for cancelling mterieicrice from a received signal ha\ mg signal contπbiUions from multiple users 102,

104, 106 fcach signal is transmitted h\ a different user 102, iO4, Wb and include^ a codewoid Ike codeword repiesents uUormation bits encoded bs the eøuespondmg usei . Each codewoid is uuei leaved, modulated uuo one or raoie s>>mboϊs, spiead omo chips and ilkcted, and transmitted to the multi-user receiver 100 o\et a channel allocated to the user Other standard transmitter functions such as frequenc) shifting and amplification and similar recenei functions, which are not of duect reles ance to the present im ention aie not discussed herein In one embodiment, the multi-user receϊxer HM ⁾ is a base station m the WCDMA uplink According to this embodiment, multiple useis 102, HM. 106 tiansuut signals to the base station in a somewhat uucooidmated manner and die base station cancels interletence caused bv overlapping signal contributions The nuilti-usei recen ei 100 can be an> t>pe of ieccn ei sn the uplink or the downlink of WC DM \ and cancel an> t> pe of muiti-usei mterfeience, including othei ceil uiteiference The multi-user iecenei 100 can operate w ith othei an interfaccs such as OM)M (orthogonal frequency ision multiplexing) or 1 IJKf Λ (time diMSioti multiple access)

In mote detail, the niulti-usei tecerv ct H)U cancels signal interference by partialh decoding a first codeword obtained from a common iecened Mgnal The partial!) decoded codeword is transmitted bj a fust user H)2 \ ia a first signal and at least pausally ovedaps Λ\ ith a codeword ttansnrsued b> the same uses 102 or one or more other useis iO-t, J 06 wa a different signal t he ronlti-user tecervet 100 ieeonstiucts the first Signal based on the partial decoding of the first codewoid and cancels the reconstructed first signal from the receπ ed signal to \ ield a reduced- interference signal The multi-user iceen et 100 at least parttalh decodes a second codew oid obtained from the rcduecd-mteifetence signal and transmitted Ma a second signal Ihus. the raulti-usscr receiver KM ⁾ can decode the second codeword w ithout significant mtetfeience fiom the first code even though the enttie fust codeword has> not been tecesved and fuil> decoded

The multi-user recener 100 includes a receiv er module H)H allocated to each one of the uses s 102. H)4 ! (){> for processing signals nansmittcd b> the coπcspoiκ!ing usei The iecen et modules KS8 can be impleinented in hatdwate, e g as an \SfC (application-specific integrated circuit}, softw are, e g as a DSP or a combination of hardware and software, e g partK as an 4SlC and parϊl) as a DSP hach of ihe ieceiv ei modules 108 mcludes a tecen e chain ! IO and a sunial τeconstrucho« cham 1 12. The tecerve cham UO enables the couespondiBg tecen et module 108 to demodulate and decode a Mgnal transmitted the assigned iiset , the transmitted signal including samples conesponding to a rcspectne codewoi d I he codeword includes coded bits corresponding to infomiaisoα bits encoded the user Coded bits are also common!) iefened to as modem bits, the terms betng used herein inlerchangeabh unless otherwise noted Signals transmuted by the isseis 102 104, 106 ma> e diHeient length codewords Foi example, in the WCDNf Λ lipinik, the span of each signal is a TTi w hich ma> include 3 slots ot 15 slots In addition, signals tiansiπuted b\ mo oi iπoie usαs H>2, 104, 106 mas arrive at the multi-user τecei\ eτ 100 at different times As such, signals ing at the multi-user iecenei iOO in tune f he iecenei modules 108 account foi the o\eιlaρρing rtatme of the iecened -f>- sijmais when the signal reconstruction chain 1 12 reconstructs the signals foi mietfesence cancellation

Ftguϊc 2 ilhjsftatcs an embodiment ot the receiver modules K)H of the multiuser receiv er 100 Each rcecrver module iOS includes the receiv e chain i iO and the signal reconstruction chain 1 12 I he receiv e chain I H ⁾ includes a demodulator 200,

202. and decoder 204 In one embodiment the demodulator 2Oy is a GRAKF (general i/cd RAKF) demodulator A GRAKF demodulator ss a s\rnboϊ-b> - linear demodulator which treats all souices of noise and mterfeiertce as a single colored noise, modeled by a eox arance matrix Another !> pc of demodulator can be used in place of the GRAKF such as a RAKF recen er, w hich can be \ ieΛ\ed as a special case of GRAKF wheie noise and interference are treated as white noise The demodulator 200 can also be a linear chip cqaah/er or a more complev nonlinear demodulatut such as a DfE (deαsjon-feedback equalization), MLSC (maximum hkehhood sequence estimation), oi DFSE (Decision Feedback Sequence EstimatoT) modulator

The recener 100 mav include multiple antennas (not shown) Facft antenna can be connected to a complete tecerve chain HO, producing baseband sample? The samples from all antennas are fed to the demodulator 200 hoi the GRAKR embodiment, the processing ss the same rcgaidless of the mimher of antennas That ΪS, a cυvaπance matrix enUi reilects the shou tetm second Older statistics of two samples, w hich mas belong to the sarae ainenna or different anteααas The output of the dcmoduiatoi 200 JS coded bit soft \ alucs m LI R foim, oi a similar fotm as discussed earlier Regardless, the on the soft % allies, and undoes the inteileauny performed pnoi to signal uansmtssion The decoder 204 outputs hard oi soft v alues for the coded hits T he signal reconstruction chain 1 12 of the recen er module 1OH include* a modulator mbol re-constructor 208 and signal te-coasttucioj 210 fot te-modulaung receded signals

Opemtion oi ike muiu-user lecerv ei 100 and the receiv ei modules 108 is descϋbed πe\t w ith iefcience to Figure 3 which illustrates two sigfuls 300, 302 transmuted different users that atme at the receiv er 100 at different times W hile Fϊgiue 3 iliustiates two uansmuted signals, the eiabodnaents described herein readilv extend to any number of transmitted signals ^' 1 he first signal 300 is transmitted bv one of the users 102, 104, 106 and includes a codeword hav ing a length spanning / ms The second signal 102 is ti admitted by the same usei oi a diffeient one of the users 102, 104, 106 and includes a codcvvoϊd, which nun be of a diffetent length, and m thts example tt ss shorter than the firs* codewoid and which arriv es before the first codeword In addition, the first and second signals 100, 302 ov erlap b\ T ms and ail samples of the fast signal 300 ate not ieeerv ed until / ^' P ms a(tet the end ol the second signal 302 l hc ieemei module 108 assigned to the first user 102 does not watt until ail samples of the first signal 3Wl are received and demodulated to decode the first signal Instead, the corresponding decoder 200 partial!} decodes the codeword of the Jflist signal 300 ϊn one embodiment, partial decoding is performed b\ treating later amvmg coded bus of the codeword earned bv modulation sv mboh receiv ed after a particular point m time as erasures, or effectn ch} punctuicd bits This means that their unav ailable suil values aie replaced wnh /eio values

Bv Ueatiug ibe iatet atmuig coded bits uf the fust signal 300 as eliecttveh punctured bits, the decoder 200 assigned to the first signal 100 can decode the codeword even though some of the coded bits aie not \et receiv ed and demoditlated This js tefeπed io hetetn as 'partial decoding ^' The stgtiai ieconistmcuon chajo 1 12 assigned to the first signal 300 uses the partial decoding results to reconstruct the transmitted signal and cancel the reconstructed signal from the common receiv ed signal Doing so i educes the mtetfaerice caused bv the frtsi signal 300, enabling the ieeen ei module 108 associated w ith the second signal 302 to demodulate and decode the second Signal >02 absent significant mtetference from the first signal 100 w ithout hav ing to wait for the entire codeword of the firsϊ signal 300 5o be receiv ed, demodulated and decoded this ΪS particuiash advantageous lbs urøe-sensttiw applications ^ueh as v oicc whei e long delays are intolerable

With reference to the exemplary signal transmission embodiment of Figure ϊ, 7"nis of the total span T of the first signal -U ⁾ O oλalaps the second signal 302 In some embodiments, the multi-usei iecenet 100 attempts to avotd all delav bs tgnoung the paU (F T ^' ) of the ffis>t signal 35H) that does not υv eilap the second signal M ⁾ 2 ϊ et /" denote the numbei of coded bits mapped mto the symbols spread o\ ei the span T Also let / ' denote the nojubei of coded bits mapped into the spsead over pan T' By ignoring the non-o\ei lapping part of the first signal >00 (T - F' ), the /,-/• coded bus of (he first signal 300 are eiϊecmeS) punctured, s e {heir soft \ aloes aie sei io xero as shown in Figui e 4 The miss-mg coded bits that are punctiued aie {tented as /ctoes at the input to the corresponding de-sntet leaver 202, indicating the absence o( information about the missing bits After de-interfeauiig. the punctured bits are distributed throughout the codew ord as n in 1-igurc 4 is is the coded bus of {he codeword whsch are decoded fot use in inter fcsence cancellation, not decoded information bus

In other embodiments, the time span T' can be chosen to exceed the end of the second signal ϊ02. constrained b> the acceptable delay tn processing the second signal >02 That is, a Luger T' allows more information to reach the seccrvet module 108 associated with die first signal 300 Thss enables the signal reconstruction chain i 12 associated with the first signal 300 lo yield a bettei reconstruction of the first signal 30O, benefiting the decoding oi lite second bignai 302 In yet othei enibodϊmeαts, the tune span T' can be chosen to be iess iiian the end of the second signal 302, e g if the delay requirement of the second signal ^02 is \ en strict and processing tune must be accounted for In each embodiment decoding of the first signal 300 is pctfotmcd using less than all of the coded bus of the signal 300 by u eating the imssrng ot bits as punctured Tins enables the mulli-user receiver iOO to reconstruct the first signal 100 and cancel the reconstructed signal from she common receiv ed signal w ithout iiawng to wait for all of the coded bus of the HT si signal 300 to amve at the iecenei 100

Dccodcis arc w ell-eqtnpped to handle punetuted bits Bit puncturing is typieaily initiated at a transmitter and undone at the recen er to prox ide design flexibility An example of bit puncturing is described next w ith iefereoce to a punctured con\oluϋonal code A punctured convoktional code is derrved from a mother code which has no puncturing The mother code has a nominal encoding rate 1 A aad memoiy J) Thete ate B mfoimatton bits to be encoded Without much loss of generality , encoding can be assumed to stan and end to state 0 The B infotoutioπ bits Oi e appended wuh /> tail bits, all .set to 0 The eπcodei toi the mothet code accepts a single infoimafion bit at a time and produces A modem bus at a time, for a total of A[B-P) bits OI those, L modem bits sie aeuiall} pioduced The remaining A(B+Dyl bits are puncttued accoiding to a puncturing table ^' 1 he true rate of the pitnetured code -Λ>_ is H L VVuhoul puncturing, the true rate ts H (4(β+li)) At the leccner, the demodulator produces / soft \ alues foi the modem btfs The> ai e de-mtei ieav ed and accepted bv the dccodct The decodes fust inset is Λ(/? t / ⁾ )-λ /ems at the appropriate locations to represent the bits punctured at the encoder A /cro \ aiuc indicates the absence of know ledge about a punctured bit From {hat point on, the decoder for {he punetuied code is that of the raothet code No further special treatment fot the punctuied bits is needed, as thev are reflected properly through the zero \alues in the decoder metric The state space of the cnm oluuonai code is constant throughout the codeword, w ith special allowance for termination T he state space is of M/C 2 ^;i l he decoder operates met a UeIUs, which desαibes the ptogresston through the state space o\ ei the codew ord length

hi corπ entionai systems, decodei operation is focused on recov ering the oagmal mfotmauon bits ϊlete, the decodei & 204 of the i«cen ei modules 108 focus on the coded bits As such, each decoding stage leptesetvk art J-tuple of coded bits of the mother code When bits are punctured bv the itser encoder, the corresponding at the decodei 204 are assigned soft soft \ alues can be transformed into ptobabilittes, a.s in equation ( iθ) below !n patttcukt foτ /eto soft v alues, the piobabiUty of a U or I is ^! , > ϊ he treiiis has (/*-/» stages At each stage of the trelhs, there are branches connecting starting states to ending states 4 branch is interchangeable w ith the pan (c\C) of its staaing and ending slates, iespecm ei> Each bianeh (c ¹ c) ha^ a label w hich consist of an Ηupie of coded bus of {he mothei code Yt stage A, foi each state pan (c\c) , a probability y _k {c\c) ts computed ftora the 1- tuple and the coded bit piohabiliUes Foi pairs (c\c) w tthout btanches, ,\{c',c) 0

Ia one embodiment, the decoder 204 of the iecenei modules 10^ can be MAP (maximum a postenoii) decoders including foiw turd recursion, backward recntsion, and combining stages Such a MAP decoder can be used for com okmonal codes Λhemam elj , the MAP decodei can be used for tmbo codes whose component codes are eomøiuuonal codes The M \P decoder ma\ aiso be ut-ed (oi block codes, where the state space vanes over the codew oid length Regardless, the forwatd tecutsion is (C) = I^ ₁ (C) ₁ MCc) fO -SO-

T he backward recursion is given by:

The initial conditions are given .

a _o (O) ^ X and a _o {c) ^ O,c≠O O) β _{B o} (0) ~ 1. and /S _β , _π (c) ~ 0,σ * 0 (4)

Suppose the 7-th bit at stage k corresponds to coded bit y ^l1l (/) where the superscript* ¹ ' indicates the first signal. If the Mh bit at stage k corresponds to a punctured bit, the combining step is skipped. The set Ω _o ( Ω^ ) contains the branches (e\c) at stage A- whose ?-th bit label is equal to O ( I ). The following values are computed:

and

The coded bit probabilities are then given b\ ^■

p(y^) ^.. θ)- -A~ (7)

Fhe respective decoder 204 is capable of producing hard and soft values about the coded bus. in addition to the information bits which are not of direct reles ance for the interference cancellation embodiments disclosed herein. In one embodiment, the coded bit soil are in LLR form denoted asx , or an approximation thereof. The terms 'LLR" and 'coded bit probability- are used interchangeably herein and mean the same tiling unless others ise noted. The probability of a coded bit being a 0 is denoted

π \ T he coded bit probability can be expressed as: -S S -

That is, a positive λ indicates a o, A hard bit decision can be obtained by taking the sign of /L For example, a hard decision of -H indicates a 0, and ,τ - 1 . The signal reconstruction chain 1 12 assigned to the first user 102 reconstructs the first signal 300 based on the partial decoding results.

Figure 5 illustrates an embodiment of the modulator• ^■ interleave* ¹ mid symbol re- coastruclor components 206, 2OS of the signal reconstruction chain 1 12. The first signal 300 uses a modulation of si/e 2 ^¥ and a spreading factor F ^i1! , The er- sampiing rate N is ώe same foi both signals in this example. The channel allocated to the first user 102 has dispersion of length D ^iU samples, and is represented by a vector h' ¹ of channel taps wheie die superscript ' indicates the first signal 300. When the receiv er 100 has multiple antennas, each antenna experiences a different channel. Then the delay D ^{V: reflects the maximum dispersion among the multiple channels, Also, the channel tap v ector h ^t1( reflects the multiple channels Referring to Figure 5, the codeword includes de-interleaved coded btta r { /) output by the corresponding decoder 204. An ύuedcaver 500 arranges the coded bits as was initially done by the first user 102, The resulting M'-" coded bits y ^v) \j), for j ^~ V- - ,A?*" are mapped b> a sj iiiboi mapper 502 into a modulation symbol S ^kfi of the firsi signal 300 where l ^{i Λ} ij } denotes the location of y ^<1; (/) in the codeword before interleaving by the interleax er 500. The soft s allies for the coded bits are denoted /t' "(/ ( and the cor responding coded b« probabilities are denoted P{/ ^Ϋ> {j }\

Given the coded bit probabilities, the signal reconstruction chain 1 12 assigned to the first user 102 can reconstruct the soft symbol for S ^m . Particularly , the symbol re-constructor 208 forms a distribution on the constellation, using the mapping from coded bits to symbol. That is, for each point σ on the constellation, the product of the probabilities of the corresponding M' ¹ ' modem bits is assigned, As a result a symbol probability P(s" ^! } is generated. The s>mbol re-constructor 2OK also computes a soft symbol value a,s the expected value of s ^l1 \ denoted s ^[ '\ given by: ~S2-

In case of hard decisions on the coded bus, all except one sj mbol \ aluc sτ ₀ has piobabiht> 0, and S ^l1) will be equal to σ;, If the probabihts distuhution for one or more of the M ^ll) coded hits mapping into s ^{v1 -} is unavailable, then it is replaced by the umicum distribution U /2, 1 /2} Ia the degenerate case whete no coded btt probability is available, then the distribution of s ^{(1 '} also becomes uniform ψ. ^M ,- - -2 ^¥' J

The signal ie-constructoi 210 multiples the soft symbol s ' ¹ ' In the apptopπate spreading sequence of F' ^u chips, denoted by a secto? q ^i1t , to produce a \ ectoτ t' ¹ ' gnen by

t ^{t i!} S ¹¹ N f U2)

The jεconslmcted lust Signal b>

u ^l ° h ^ϋ) * t ^{t i!} U3)

vvheie * indicates convolution

Figuie 0 illustrates an embodiment of the tecetvet modules 10H assigned to the first and second users ! 02. 104 along with a signal cancellation module 600 The

Signal cancellation module 600 cancels the reconstructed first signal u' ¹ ' from the common received signal r to r^ ¹ ' as gn en by.

r ^l1> r u ^;!> (14)

The reduced-interfeience signal r* is relativ e!) fitec from intetfeieiice ftom the first signal ^OO ex en though the first signal 300 was not full) decoded. The same over- samphng N can be used the recener module UW assigned to the second user 104 for demodulating the second signal 302 when the users 102, 104 employ the same cn er- samplmg rate. The second signal 302 uses a modulation of st/e 2' ^Λ and a spteading factor F' ² ' wheie the .superscript ^"' indicates the second signal 302 The channel allocated to the second user 302 has dispctsion of length D ^{U i} samples, and i^ represented by a x ector h' ^{? l} of channel taps

Figure " illustrates an embodiment of wmnow mg scheme of proper st/e for ensuring the channel represented by vectot N ⁷ ' & folly covered. A symboi-by-sj mboi demodulator such as a GRAKE demodulator requires a of data ss/e F ^U Η - D^ samples to fuih co\er h ^{l? i} of the second signal -?02

Figure S ilϊustiates a Signal transmission embodiment where F ^[V - F' ¹ ' and the P ^1! N ÷ D ^* span, of r ^{1' 1} cov ets C ^Δ Ψ ^J) N - D^* samples corresponding to £ ^{^!} symbols of the second user 104 In the absence of cancellation, the effect of s'" would be present in the common icccised signal r , through the stgnai approximated hv u ^! \ w hich would mterfete with the ieception of the second signal 302 This mteifetenee would be accounted ioi m the cov anance matriλ R of a GRΛKC demodulator for the second signal M)2 With the cancellation of u ^v1 sn r ^v11 e g as gn cn bΛ equation { 14), the leceptϊon of the L ¹ ' ⁵ svmbois of the second usei S 04 improves ϊti a GRΛKfc embodiment, the GiL-VKE assigned to the second signal 302 computes a eo\ aπance matrix R ^v which icllccts the absence ol intetierence ftom the first stgnai Η)0 hi that sense, the GRAKE dcmodulatoi can re-allocate its resources to cancel otliei interference In Figure S, the ieft-most sy mbol v ^{~ ;} of the second signal 302 is parth cox ered b> the span of the reconstructed s>mboi I of the lust signal 300

Aceoidinglv, the unco\e*ed samples of the left-most sy mbol f 'are not affected bv the Oi igϋial s>mbo! s' ^1! m thss example Of coin se, the samples ma\ be affected b> anothet sviwbol of the fust signal 300, or bs of other signa!(s»)

Λs explained pie%sousl> herein, effectiveiy ptmctined coded bits (i e latei- mg bits) aie piocessed the same waj as the true punctuieU bits Referring back to Figtne 4, the corresponding demodulator 200 pioduccs A ¹ soft salues ieflecting the partial reception of the fπst signa! MW The missing (AV '} soft \ alues are effectiv e punctured bits, and they are represented with zeros ϊhe total A soft % aloes aie fed to the corresponding de-interiea% ei 202 as pre\ iously desciibed herein No furthei special lieatineni of tiie dlectne puiiLtureϋ bits is needed Also as desutbed alxn e, The A de- ed soft -values are augmented with Λ{β-D)-1 /eios at the appiopnate locations to repiesent the bits oπginaily punctured at the encodei of the first user 102 The tota! A(B* D) soft \ aloes aie fed to the corresponding decode* 2{ ⁾ 4 of the mother code K\ en if coded bity ^{l f} fj) is effective!) punctured, the decoder 204 still pioduces useful tafoimation about the mg bit Λs such, sjmbofc not actually ed can -1.4- still be reconstructed. in turn, the spaa for the reconstructed signal can exceed the received span. The partial decoding embodiments described herein can be readily extended to more than two signals.

Figure 9 illustrates an embodiment of the multi-user receiver HM) implemented as a serial interference cancellation. (SIC) structure with A ^' signals, in some embodiments, larger signals can be processed early and smaller signals processed later. Similar to the receiver module 108 of Figure 1 , the ith receiver module 900' has a receive chain and a signal reconstruction chain for signal i. The modified received signal r ^{ir ! !} generated by the previous receiver module has been obtained from the common received signal f by subtracting the previous /- J reconstructed signals. Jn one embodiment, a demodulator of the ith receiver module 900' is a GRAKE demodulator. The corresponding covariance matrix reflects the interference left in the modified received signal r ^! ' ^"1i . Similar to the signal cancellation module 600 for signal 1 in Figure 6, the ith cancellation block 902' removes reconstructed, signal u ^y - ^! from r" ^" "' ^■ to produce which helps reduce signal interference for the subsequent receiver modules. The issue of overlap time becomes more challenging in SIC environments where several signals are received and processed.

Figure 10 illustrates an embodiment of three signals 910, 920, 930 processed sequentially by the SIC structure. The overlap time of the first signal 910 to cover the second and third signals 920, 930 is T" . In one embodiment, the first signal 910 is partially decoded as previously described herein over time T" in the .first receiver module 900. As a result, the reconstructed, signal u' ^;:l covers T" , and first benefits the second receiver module 900' via the modified received signal and then benefits the third receiver module 900" via the modified received signal T ^U) . In another embodiment, the signals 910, 920, 930 are progressively reconstructed. That is, an early signal in the SIC structure may be reconstructed over increasingly longer overlap times to benefit later signals in the structure. For example, the overlap time of the first signal 910 to benefit the second signal 920 is T' < T" . Thus, it is possible to do a partial decoding and reconstruction in the first receiver module 900 over time T' to benefit the second signal 920. Then, the first receiver module 900 is revisited for a second partial decoding and reconstruction over time T" to benefit the third signal 930. Van nig the cn crlap time has an impact on the piøeesstng m the second iteration of (he fust recen er module 900 In one embodiment, the first iecen et module 900 petfotnts demodulation o\et tiie eκlta time (F" T') Decoding is then begun anew usmg the coded bib fiom 7" and from [T" ~ T ^' } The signal reconstruction pioeess pet formed tn the first module ^! W is also started anew to produce a new reconstructed signal u ^m to K o\ eτ the e\tra time [T" ~ T') and concatenated w ith the old reconstructed signal ox er V to produce a new u' ¹ In j et another embodiment, the decoding state is taken based on the coded bits ox er T' as a starting posnt and adding the coded bits ov er (T" - T') as new sκ1e information Decoding can then be pesfoimed again as a second itetation, for instance using a tut bo decoding apptoach

in one embodiment larger signals are piocessed earhci in the SK ^' In other embodiments. Jt may be more beneficial to consider Mgnal o\ei !aρ and deisn Jn the decision to otder the signals given the impact of ! oτ instance, the powet estimate of a signal, used in ranking the signals for processing m the SfC structure, can be adiυsied based on Hs since the signal interacts w ith {he other vtgπaK m lite o\ eilap tεgion ΛccotdmgK , a iasge signal w ith a small ov erlap ma> get pushed late* in the SIC sttuctute

A second considctation JS the coding πrte, gπcn its impact on patUal decoding A signal with a low coding rate can w ithstand a more extreme paitiai decoding, so it ma> get pushed earliei in the SlC stiuctuie Furtheimore, the coding iate can be adjusted io tellect the overlap, accounting fot pnnctiited bits

Various signal quality metrics are ax aiiable to the multi-user rcccn cr 100 lor deciding whethei a sufficient numbei of bits are a\ aιlable to begin pariial decoding and w hetlie* signal ieconsiruction and cancellation benefits othet signals) SXR ⁽ sιgnai-to- noise lauo) symbol metrics and coded bit ϊ LR values each can be used to make partial decoding and signal reconstruction decisions SNR comes from the paiametei estimation process, which precedes the demodulator 200 of the module 1 OS Symbol metrics come from the dcmodulaiot 200, which precedes the decoder 204 The modem bit LLR ¹ S come from the decoder 204, which precedes anj ( RC (c\chc redundancy code) check which maj be petfoimed ~ S 6-

SNR is the byproduct of the parameter and is !> pϊcall> av ailable for providing feedback of channel quality mlbimatjon (CQI) to the remote transmittei The demodulator 20 ⁽ 1 may also produce mettles as a bx ptoduet of its operation 1« one embodiment the metrics me lode the Euclidean distance between the decided v aie auulable, e g m {he form of the LLR's /. ^{i !} {j S as described preuoush herein, another signal quaiitj metric can be derned as gnen by

)

Ihjs signal quality metuc reflects the least reliable coded bit Other signal quality metrics mav also be dem ed from the LLR's, e g tank information such as the ! θ ^h percentile v alue, or other percentile values Yet other quality metrics nun be deriv ed from the LLR's, e g lmeai metπcs such α weighted or unweighted av erage 4 C ¹ RC check performed at the mulυ-user iecen ei 100 can be used as a signal quality metric, e g w hen the transmission format applies a parity cheek code, tv pieallv a ORC, to the information bits before the enoi control code Lndcr these conditions, the C RC cheek at the receiver K)O is a relatn ^gnal ieceptioii qualify

In some embodiments, it may be beneficwϊ to establish a thteshoϊd for starting the partial decoding ptocess That is, thcte ma\ be a minimum percentage of a signal span sufficient for producing a useful reconstructed signal for cancellation I his uansϊates into a minimum peicentage of the coded bus that need to he av ailable for decoding Know ledge of the code rale and the cuπeut SNR can be used to determine the partial decoding threshold Since the coding formal of the current reeen ed signal is based on a prev ious SXR estimate the current SNR estimate can help predict successful the paHiaS decoding w ill be Without know ledge of the cuiient SNR. the partial dccodmg threshold can be set to a preset \a!»e, such that the teeoiisttucted signal ts useful to cancel l he partial decoding threshold can be irnpiov ed when the cuirent SNR is known if the current SNR is highei than the prev ious, then the ihie^hold can be lowered, *>o that partial decoding can start easliei I! the cut tent SNR is kπvei than the prev ious, then the threshold can be increased, so that partial decoding can start later ϊ he demodulator metπcs can be used m the same w a> For instance, the metrics can be avemged among those of symbols m the pattul signal span to ~S 7- mdieate the effectn e local noise lex el If the noise lex el is low , partial decoding can be started eashei !f the noise iesei i<s high, partial decoding can be staaed Iatet

Sn addition to deciding when to begin partial decoding a (uilhcr decision can be made on whether to reconstruct one or more signals for cancellation In one embodiment, the cntuc codcwotd of <i signal can be evaluated, e g , hv using /, _{m i} to make a iecon.su ueuon decision foi example, /L _n ca« be cotnpated to a pre-set threshold ϊhe A^ _irk parameter cao also be used ni conjunction w ith the CRC check, w herebj both conditions must be satisfied Alternativ ely, eithes condition could be satisfied Sn the context of soil leconstmction, signal teeonsttuchon can ahvav s be performed if desired because soft reconstruction mherenth tends to scale dow n less reliable signals, thus limiting the potential damage of erroneous cancellation The CRC check can be used as side mioimauoo to «npto\ e so U tecooittuetton In o«e erobodiroenL if the CRC fails, -Hc entire soft ieconsti noted signal is scaled dow n to indicate a reduced confidence If the CRC checks the reconstructed can be scaled up to indicate added confidence, iesuUmg m mote complete cancellation

Spatial!} relatrve teims such as "undei", "below ", "lowet". "ovef\ "upper ^" , and the like, are used for ease of description to explain the positioning of one element ielatix e to a second element These teirm aie intended to encompass dilTeient ouenωuous of the ice m addition io diffetent oϊientationjb than those depicted to the figuies Miithei terms such as 'Tn si , "second \ and the like, ate also used to describe \ anous elements, regions, sections, etc and are a!so not intended to be limtϊmg I ike terms refei to like elements tiuoughøut the description

Λs used heteso, the tenns ing ^" "containing", "including", "comprising" and the like arc open ended terms that indicate the presence of stated elements or features, but do not preclude additional elements or features I he articles "a-\ "an" and "the" are intended to include the plural as well as the singular , unless the context cleaiiy indicates oiheiw ise

With the abo\e range of \ anations and applications m mini H should be undei stood that the present inv ention is not limited bv the foiegoing descuptωn not ts u limited bv Uie accompans'mg dmw mgs Instead, the present inv ention ts hnnted onlv b> the following clamis. and then legal equiv alents