Space- ime on -ends o RAKE ecei e s in he FDD mode o UTRA1
Josep Vidal, Ma ga i a Cab e a, Ad ian Agus ín
Signal Theo y and Communica ions Depa men , Uni e si a Poli ècnica de Ca alunya
Ba celona, Spain
{pepe,ma ga}@gps. sc.upc.es
Abs ac
A uni ied and gene al ision o di e en space- ime p ocesso s is p esen ed. Many popula ecei e s can be
accomoda ed, like V-RAKE ecei e s, weigh ed V-RAKE, o spa ial na owband beam o ming. By making
app op ia e assump ions on he space/ ime cha ac e is ic o he in e e ence i is possible o enhance he
pe o mance o he ecei e h ough spa ial/ empo al p e-p ocesso s. These ecei e s will be es ed in he FDD
mode o UTRA.
I. In oduc ion
The ad en o he 3 d gene a ion o mobile communica ions sys ems has been accompanied by he ecogni ion o
he la ge inc ease in sys em capaci y ha can be ob ained om he use o adap i e an enna a ays. Ca e has been
aken in he de ini ion o he s anda d o include capabili ies o space- ime p ocessing o he signals incoming
and adia ed om he base s a ions. Sec ion II e iews he main cha ac e is ics o he Eu opean p oposed ai
in e ace (UMTS). Sec ion III p esen s he signal model. In sec ion IV he di e en space- ime ecei e s a e
p esen ed in a uni ied ision. I will be seen ha he use o mul iple an ennas and he empo al co ela ion
di e si y o mul iple use s allow addi ional deg ees o eedom o cancella ion o mul iple access in e e ence.
Only single use app oaches will be in oduced al hough some ideas a e easily ex endable o he mul iuse case.
All hese ecei e s need a eliable es ima e o he co ela ion p ope ies o he in e e ences, so di e en op ions
a e p esen ed in sec ion V. Pe o mances a e compa ed in e ms o p obabili y o e o in sec ion VI.
II. UTRA FDD uplink ai in e ace
The uplink physical channels o he FDD mode o UTRA a e as ollows: each use gene a es a leas one
Dedica ed Physical Da a Channel (DPDCH) along wi h a single Dedica ed Physical Con ol Channel (DPCCH).
Each channel is sp eaded wi h a di e en OVSF o hogonal code a a chip a e o 4.096 Mchip/s and hen in-
phase and quad a u e mul iplexed in a QPSK modula ion. The exis ence o OVSF codes o di e en leng h
allows he p esence o di e en bi - a e use s in he same cell. These complex symbols a e sc ambled by a
mobile-s a ion speci ic sc ambling code using ei he codes om he Ve y La ge Kasami se o he Gold sequence
o leng h 40960 ( igu e 1 illus a es he signal gene a ion scheme [ETSI-UTRA]).
Figu e 1. Modula ed signal gene a ion Figu e 2. Slo s uc u e o he uplink in he FDD mode
The Kasami codes a e in ended o base s a ions ea u ing mul iuse de ec ion, while in he Gold codes assu e he
sepa a ion o use s on he g ounds o good co ela ion p ope ies o he codes and single use de ec o s ha e o
be used. Fo a e iew o mul iuse de ec o s o a ay obse a ions he eade may eso o [Jung][Ve du].
The UTRA ai in e ace is designed o achie e a ull 1:1 euse, which implies no loss o capaci y due o
equency planning. Since all use s access he channel asynch onously, in e cell and in acell use s sepa a ion
1 This wo k has been done hanks o he AECI, EC (IST-199-10322, SATURN p ojec ), CICyT o Spain (TIC98-0412, TIC98-0703, TIC99-
0849) and CIRIT o Ca alunya (1998SGR-00081).
Channel 1
Channel
i
Channel
i
+1
Channel
Q
β
1
β
i
β
i
+1
β
Q
jComplex
Sc ambling
C
Si
+jC
Sq
p
(
)
x
(
)
... ...
C
c1
C
c
i
C
c
i+1
C
c
Q
Channeliza ion
... ...
DATA (Nda a bi s)
DPDCH
PILOT (Npilo bi s)
DPCCH TFI (NTFI bi s)TPC (NTPC bi s)
0.625 ms, 2560 chips
elies on he good co ela ion p ope ies o he sc ambling codes. Unde his p emises, all channeliza ion OVSF
codes a e a ailable o each use so many DPDCH can be se up o ob ain he desi ed bi - a e (as shown in igu e
1) wi h di e en weigh ing ac o s (no ed wi h β). The esul ing bi - a e g anula i y is eally high.
Due o he se e i y o he mobile channel, he dynamic ange o ecei ed powe s in he BS can be as high as 90
dB. In o de o a oid loss o e iciency due o he nea - a p oblem, a igh con ol o he ansmission powe is
in oduced: e e y 0,625 ms, he base s a ion sends powe con ol in o ma ion h ough he o wa d link, and he
mobiles upda e hei ansmi ed powe a he same equency.
The ame s uc u e is also shown in igu e 2. Each ame o du a ion 10 ms is spli in o 16 slo s o du a ion
0,625 ms, which co esponds o one powe -con ol pe iod. The DPDCH is used o anspo da a symbols, while
he DPCCH is di ided in h ee ields which a e used espec i ely o channel es ima ion (PILOT), ansmission
o powe con ol in o ma ion (TPC) o he downlink and anspo - o ma indica o s (TFI). TPC and TFI occupy
only he las 10% o he slo du a ion, so he ecei e is able o es ima e and ack he channel o he i s 90%.
Usually he channel es ima ion is made using he pilo chips, and di e en ampli udes can be assigned o he
DPDCH and he DPCCH chanels.
III. Signal model
The single-use signal model assumed o he signal ecei ed a M senso s a e ma ched il e ing and chip- ime
sampling can be w i en in column ec o o m as: wHdy
+
=
(1)
whe e each e m is de ined as:
H is he space- ime channel o he desi ed use , d includes he N a ic symbols (d ) and he Np pilo symbols
(dp) o his use and w is he ec o accoun ing o noise plus in e e ences. N is he numbe o chips in a single
slo , Qp is he DPCCH sp eading ac o and Q is he DPDCH sp eading ac o . Ma ix Hi con ains he
con olu ion o he impulse esponse o he channel seen by senso i (compu ed a chip ime) and he sp eading
code. The e ec o he long sc ambling code can be ep esen ed by he ime a ia ion o he sp eading code om
symbol o symbol, which is deno ed wi h he supe sc ip (k):
)()()(
)()()(
)()(,
)(
)(,
ncnhnh
ncnhnh
k
i
k
i
k
pi
k
pi
∗=
∗= (3)
In equa ion (2), 5 symbols ha e been plo ed o he pilo channel and 10 o he a ic channel. I is assumed
ha he empo al leng h o he physical channel is L chips.
IV. A amily o space- ime ecei e s
Wi h his model in mind and modeling he in e e ence-plus-noise as spa ially and empo ally co ela ed
Gaussian noise, i is possible o o mula e he likelihood unc ion which, app opia ely minimized, gi es he
de ec o o he unknown symbols:
=
M
H
H
H
HM
2
1
=
i
H
N
p
N
Qp
1
L+Q
p-1
Q
hp,i
h ,i
(
k
)
(
k
)
[ ]
[ ]
[ ]
)1()1()0(
21
−=
=
=
Nyyy iii
T
i
T
M
TTT
T
T
p
T
L
L
y
yyyy
ddd
)2(
{
}
HdRHdHdRyyRyHdyRHdy1111 Re2)()( −−−− +−=−−= w
HH
w
H
w
H
w
H
J (4)
Some assump ions a e possible so as o simpli y he ecei e :
A.1. The co ela ion ma ix o he noise-plus-in e e ence can be decomposed as he K onecke p oduc o a
spa ial co ela ion ma ix and a ime co ela ion ma ix:
wsww ,, RRR ⊗=(5)
which ag ees wi h he mos ecen channel models [Pede sen], in which he ime and angula sp eads a e
shown o be independen phenomena.
A.2. The physical channel sp ead (L chips) is much sho e han he leng h o he sp eading code (which is he
case when designing a DS/CDMA sys em), so he ma ix HRH1−
w
H is almos diagonal and he las e m in
(4) can be neglec ed in i s minimiza ion. In ac , his is one o he easons why high bi - a e use s canno be
alloca ed in u al o hilly en i onmen s, whe e delay sp eads a e usually long compa ed o he leng h o he
sp eading codes.
The e o e only he middle e m emains in (4) and i cons i u es a su icien s a is ics o he p oblem. I s
maximiza ion leads o he well known Rake ecei e when bo h space and ime co ela ion ma ices a e assumed
whi e:
{
}
( )
( )( )
dHyHdRRRRy
HdRRyHdRy
d
B
H
B
H
wsw wsw
H
wsw
H
w
H
max
=⊗⊗=
=⊗==ℑ
ℑ=
−−−−
−−−
2/1
,
2/1
,
2/1
,
2/1
,
1
,
1
,
1
Rea g
ˆ
(6)
The in oduc ion o he co ela ion ma ix o w implies a p ewhi ening o bo h he signal ec o y (which is no ed
wi h yB) and he desi ed use channel ma ix H (which is no ed wi h HB). This ope a ion can be done sepa a ely
in ime and space (no e ha 2/1
,
−sw
R apply only on he spa ial componen s o y and 2/1
,
− w
R apply on he empo al
componen s). Figu e 3 is a ep esen a ion o he ope a ions pe o med by his ecei e .
Figu e 3. The ecei e in equa ion (6) wi h empo al and spa ial p ewhi ening ma ices. Acco dingly, he spa io-
empo al channel H has o be empo ally and spa ially whi ened be o e being used in he ake combine s.
O cou se his ecei e could be ully implemen ed by using sample es ima es o bo h co ela ion ma ices, bu i
is usually he case ha he complexi y o he esul ing s uc u es does no jus i y he imp o emen ob ained wi h
simpli ied e sions. These di e en ecei e s can be o mula ed om equa ion (6) by doing ce ain
app oxima ions on he co ela ion ma ices.
A. Tempo al co ela ion ma ix
T.1. Tempo ally whi e in e e ence. I is assumed usually and is a easonable assump ion i he numbe o
in e e en use s can be conside ed high.
.
.
.
Tempo al
whi ening
RAKE 1
y
1(nT)
y
2(nT)
y
M(nT) RAKE
M
+
.
.
.
)(
ˆ
nd
Re{.}
Spa ial
whi ening
ma ix
.
.
.
y
B,1(nT)
y
B,M(nT)
T.2. p h o de Ma ko model o he empo al co ela ion. This assump ion is no usually used bu i wo ks well
o a low numbe o use s in a low noise scena io. The simplici y o his model is e lec ed in he s uc u e
o he empo al ma ix o he i s o de case, which akes he o m:
ρρ
ρρ
ρρ
σ=
−−
−
−
1
1
1
21
2
1
2
,
L
MOMM
L
L
NN
N
N
w
R (7)
his ma ix, and hose ob ained o highe o de models, ha e in ac a closed exp ession o i s in e se
[Kay] which could be used in (6) and p e en s om ma ix in e sion. Howe e , i is much mo e in e es ing
and p ac ical o ecognize ha he empo al whi ening ole o 2/1
,
− w
R will be done exac ly by a p h o de
FIR il e . O cou se, high o de s o he model, imply long FIR il e s which in oduce addi ional in e chip
in e e ence and, as a consequence, educes he alidi y o assump ion A.2. Ca e should be aken o use
sho leng hs compa ed o he delay sp ead o he channel impulse esponse.
B. Spa ial co ela ion ma ix app oxima ions
S.1. Spa ially whi e in e e ence. This assump ion is ealis ic only in he case o a high numbe o in e e e s o
in a highly angula dispe si e scena io. Then, he ecei e becomes he well-known VRAKE [VanE en].
S.2. Reduced ank app oxima ion:
H
swBBR11
,−− ≅Σ, whe e RM
C×
/∈B wi h R<M (8)
The spa ial co ela ion ma ix is now educed o a numbe o is componen s and, i empo al whi eness o
he in e e ence is assumed, he o e all ecei e ope a ion can be w i en as:
∑
=
⊗⊗
σ
=ℑR
i
H
ii
H
i
1
))((
1HdIbIby (9)
The eigen ec o associa ed o he beam o me i gi es a measu e o he eliabili y o he in o ma ion
con eyed by he b anch i o he combine .
V. Spa ial on -ends
Some app oxima e spa ial ecei e s will be de eloped in he sequel. No assump ions a e made on empo al
co ela ion ma ix, hanks o he spa ial- empo al uncoupling o he p oblems s a ed in equa ion (5).
A. Noise-plus-in e e ence ma ix in e sion (NIMI) ecei e
This ecei e is based on he educed ank app oxima ion o he in e se o he spa ial co ela ion ma ix gi en by
equa ion (8). I is illus a i e o in e p e equa ion (9) as a cohe en combining (maximum a io combining) o
he ou pu s o R beam o me s (see igu e 4). The na u e o each is easily seen om a simple case: assume he
case o P<M poin in e e e s. I ec o s bi a e aken as he noise eigen ec o s o sw,
Reach one ac s as a spa ial
in e e ence cancelle . Seen in his way, di e en in e e ence cancelle s can be es ima ed acco ding o di e en
c i e ia, as p esen ed below.
B. Spa ial e e ence (SR) ecei e
An al e na e way o de e mine he ec o s in (9) is o compu e R spa ial e e ence beam o me s which cancel he
in e e e s, each associa ed o an incoming di ec ion o he desi ed signal. This app oach elies on an angula
e e ence o he use o in e es [Wid ow]:
Ri osubjec min i
H
iisy
H
i
i
,...,11
,==pbbRb
b (10)
No e ha he numbe o signi ican signal componen s pi will de e mine R, he deg ee o he app oxima ion in
(9) and hence he numbe o ake combine s in igu e 4. Each combine will be gi en by he classical exp ession:
(
)
1
1
,
1
,−
−− =δδ=isy
H
iiisyii pRppRb (11)
Figu e 4. The ecei e in equa ion (6) wi h empo al p ewhi ening (gi en by he FIR il e s a ec ing equally o
each b anch) and spa ial p ewhi ening and educed ank app oxima ion (M>R). The gains a he
ou pu o each beam o me a e gi en by he associa ed eigen alues in equa ion (8).
A possible me hod o de e mine he ec o s pi is as ollows. Le us i s ede ine he signal model o equa ion (1)
as:
WHDyY+== ×+− )(
)1(MLN
un ec (12)
whe e ma ix D is a Toepli z ma ix buil a chip ime om he QPSK complex sp eaded and sc ambled symbols:
LLN
C
LNdNdNd
dLdLd
dLdLd
×+−
/∈
−−−−
−
−−
=)1(
)1()2()1(
)1()1()(
)0()2()1(
L
MOMM
L
L
D(13)
whe e N s ands o he numbe o chips in he pilo , L is he leng h o he es ima ed physical channel, and all
e ms d(n) belong o he se {-1-j, -1+j, 1+j, 1-j}.
H
con ains he esponse o he physical p opaga ion channel a
chip ime o all senso s (no e he di e ence wi h ma ix H in equa ion (2)):
[ ] [ ]
)1()1()0(
)1(
21 −=/∈=×+ LhhhCiii
T
i
ML
MLL hhhhH (14)
The es ima ion o he channel ma ix can be pe o med using leas squa es. Then he ec o s pi may be ob ained
as he p incipal eigen ec o s o
H
H
H, o , al e na i ely, as hose eigen ec o s poin ing o he angula di ec ions
in which he SINR is highe . This la e condi ion means choosing he eigen ec o s pk o
H
H
H ha maximize
he signal o signal-plus-noise-plus-in e e ence a io:
k
sy
H
k
k
k
sy
H
k
k
HH
k
k
SSINR pRppRp
pHHp
,,
λ
== (15)
In p ac ice his choice p o es o be he mos e ec i e. The las e ms o be de ined a e he eliabili y ac o s σk,
ha is, he noise-plus-in e e ence le els a he ou pu o each beam o me . Since he signal powe is equal in all
b anches hanks o he es ic ion in (10), i is easonable (al hough no s ic ) o conside he eliabili y ac o o
be he o al signal powe a he ou pu o each beam o me . This e m u ns ou o be δi compu ed in (11).
As a inal commen , no e ha his app oach applied o equa ion (9) is equi alen o conside ha he noise-plus-
in e e ence co ela ion ma ix is subs i u ed by he exp ession:
.
.
.
x
1(nT)
x
2(nT)
x
M(nT)
+
.
.
.
)(
ˆ
nd
b
R
b
1
.
.
.
RAKE 1
1/
σ
1
RAKE R
Re{.}
.
.
.
.
.
.1/
σ
R
a
a
a
[
]
Rsy
H
syswpppPRPPRR L
21
1
,
1
,
1
,=≈ −−− (16)
C. Ma ched desi ed impulse esponse (MDIR) ecei e
A hi d way o build a spa ial ecei e is o ob ain a combine b ha maximizes he SINR a i s ou pu . This
co esponds o he MDIR app oach de eloped in [Lagunas]:
1
,=bHDDHbbRb
b
HHH
sy
H osubjec min (17)
I is shown he e ha he choice o he beam o me is ob ained as he eigen ec o b o he equa ion:
bHDDHbRHH
syλ=
, (18)
associa ed o he minimum eigen alue λ. Βeing ixed o 1 he signal powe a he ou pu o he beam o me b, λ
akes he alue o he in e se o he signal-plus-noise-plus-in e e ence powe . No e howe e ha he e a e M
eigen ec o s gi en by (18) and each yield a di e en signal wi h di e en quali y. This di e si y can be
cohe en ly combined using he ake in igu e 4, wi h he eigen alues used as eliabili y ac o s in he b anches o
he ake.
F om he exposi ion abo e, he SR ecei e and he MDIR ecei e seem look e y simila . No e howe e ha in
he o me case, each beam o me poin s o a di e en di ec ion, gi en by he ec o s pi while in he MDIR
app oach each beam o me ends o poin o all di ec ions om whe e signals a e incoming. The possibili ies o
he SR ecei e seem be e , since he e a e mo e ee deg ees o eedom pe beam o me o cancel
in e e ences, al hough he ac ual pe o mances may depend g ea ely on he angula and empo al dispe sion o
he channel and ha e o be de e mined expe imen ally.
The mos impo an di e ence is ha , o igu e 4 o be alid, he noise componen s be ween b anches ha e o be
unco ela ed. I is easy o show ha his is he case o he MDIR ecei e and he NIMI ecei e , bu no o he
SR.
Theo em. The noises a he ou pu o he beam o me s ob ained wi h he MDIR app oach a e unco ela ed,
unless he eigen alues associa ed a e equal.
P oo . Le us ake equa ion 18 and ecognise ha he same solu ion o he eigen ec o s can be ob ained by
subs i u ing Ry,s o Rw,s. Now le us ex end he equa ion wi h all he eigen ec o s as:
BSRBSHDDHBRsd
HH
sw,, == (19)
By le -mul iplying wi h he conjuga e anspose o B we ob ain on he le hand side o he equa ion, he
co ela ion ma ix o he noises a he ou pu o he di e en beam o me s:
BSRBBRBsd
H
sw
H,, =
Now, since he le -hand side o he equa ion is an he mi ian ma ix and he eigen alues a e eal we can w i e:
BRSBBRBsd
H
sw
H,, =
Wi h no loss o gene ali y assume ha Σ has a mul iple eigen ec o σ, so i he p oduc abo e is commu a i e i
can be w i en as:
′
σ
=
′
σ
=S0
0I
ED
DC
ED
DC
S0
0I
BRSB H
H
sd
H,(20)
By ope a ing i is easy o see ha D=0 and ha E has o be diagonal. The e o e we can conclude ha :
′′
σ
=S0
0C
BRBsw
H, whe e
S
′
′
is diagonal.
VI. Es ima ion issues
A. Spa ial co ela ion ma ix es ima ion
The NIMI ecei e assumes he knowledge o he ma ix sw,
R. Relying on he ac ha he noise-plus-
in e e ence is ime unco ela ed wi h he desi ed signal we can de i e an exp ession ollowing equa ions (12),
(13) and (14). No e howe e , ha acco ding o he signal s uc u e o he FDD mode o UMTS, we canno
comple ely de e mine ma ix D be o ehand since i con ains he known chips o he pilo channel bu also he
unknown chips o he a ic channel:
pDDD 21 β+β= (21)
whe e β1 is he weigh ing ac o associa ed o he pilo (known) chips and β2 is he one associa ed o he a ic
(unknown) chips, as shown in igu e 1. Fi s o all, i is wo h men ioning ha he channel in equa ion (14) may
be es ima ed consis en ly by appliyng a leas squa e es ima ion:
YDYDDDHH
p
H
pp
H
pLN11
1
11
1)1()2()(
ˆ−−−− +−β≅β= (22)
in which inco ela ion be ween known and unknown chips is assumed. Unde hese p emises, he space
co ela ion ma ix o in e e ence and noise can be compu ed e godically as:
sssy
HHHHH
swLN,,
2
2
2
1,ˆˆ
)1)((2
ˆRRHHYYHDDHYYR µ−=+−β+βµ−≅µ−= (23)
whe e in he las equali y we ha e assumed empo al inco ela ion be ween he in-phase and quad a u e
componen s o he sc ambled chips, and aken in o accoun he di e en ampli udes o he pilo and a ic
channels. The e m µ is included wi h he ollowing pu pose: one o he sho comings o (23) is he ma ix
subs ac ion, an ope a ion ha may lead o non-posi i eness o (23) due o es ima ion e o s. Then, he ac o µ
can be chosen con enien ly so as:
zzRzzRzzRz∀>µ−= 0
ˆˆˆ ,,, ss
H
sy
H
sw
H (24)
I his equa ion has o be posi i e de ini e o e e y possible ec o z, hen he alue o µ has o be smalle han
he minimum alue o he Rayleigh quo ien :
<µzRz
zRz
ss
H
sy
H
min
,
,
ˆ
ˆ (25)
ha is, smalle han he minimum eigen alue o he ma ix pencil
(
)
sssy,,
ˆ
,
ˆ
RR .
B. Tempo al co ela ion ma ix es ima ion
I was s a ed in sec ion IV ha he empo al unco ela ion ope a ion ha is pe o med in equa ion (6) can be
e icien ly done using a empo al whi ening il e . The design o his il e is s aigh o wa d om he linea
p edic ion heo y [Haykin]: he Wiene -Hop equa ions yield he coe icien s o he p coe icien s whi ening
il e , p o ided we dispose o he i s p+1 e ms o he empo al co ela ion o he in e e ence. A compac
o mula ion can be ob ained i equa ion (1) is ew i en as:
(
)
WHDIY+⊗=
~
~
M (26)
whe e ma ix Y is de ined in he ollowing way:
)1()(
2
1
)1()2()1(
)1()()1(
)0()1()(
~
~
~
~
~+×−
/∈
−−−−
+
−
=
=ppN
iii
iii
iii
i
M
C
pNyNyNy
ypypy
ypypy
L
MOMM
L
L
MY
Y
Y
Y
Y (27)
ma ix D is de ined as in equa ion (13) (using he symbols a chip ime) excep o he dimensions:
)()(
)()2()1(
)2()()1(
)1()1()(
LppN
C
pLNdNdNd
Ldpdpd
Ldpdpd
+×−
/∈
−−−−
+−+
+−−
=
L
MOMM
L
L
D(28)
and
H
~
con ains he es ima ed channel a chip ime o all senso s:
Assuming ime inco ela ion be ween he desi ed use and he in e e ence, as well as inco ela ion be ween he
a ic and pilo chips, he ime co ela ion ma ix o he in e e ence esul s in:
(
)
(
)
HHYYHDIDIHYYR
~
~
))(1(2
~
~
~
~
ˆ2
,HH
M
H
M
HH
wpN−β+−≅⊗⊗−= (30)
F om his exp ession i is possible o eco e he coe icien s o he empo al whi ening il e s using Le inson
ecu sion. No e ha he egula iza ion de ice p oposed abo e o he spa ial co ela ion ma ix can also be used
he e.
VII. Expe imen al pe o mance e alua ion
A. P opaga ion channel model
In o de o e alua e he ecei e in a ealis ic mobile scena io, we ha e ca ied ou simula ions based on a
Gaussian s a iona y unco ela ed hypo hesis o he channel, assuming independence be ween angula and
Dopple sp ead, as i has been expe ienced om measu emen s aken in down own S ockholm in he 1,8 GHz
band [Pede sen]. The e, i is empi ically shown ha azimu h spec um ollows a Laplacian law, along wi h
Gaussian dis ibu ion o he di ec ions o a i al (φ) a ound he mean angula posi ion o he use . The angula
sp ead ( ha is he s anda d de ia ion o he Gaussian, φ
σ) is aken 8º. The numbe o ays impinging he a ay is
i ed as a Poisson andom a iable o mean 25. An exponen ial law is ound in [Pede sen] o he powe delay
sp ead, bu ou simula ions will be based on he pedes ian and ehicula models o empo al sp eading
ecommended in he SMG2 documen s o UTRA. The ampli ude associa ed wi h each p opaga ion pa h (α) is a
complex Gaussian andom a iable whose powe dec eases as he ime delay and he angula di ec ion o a i al
wi h espec o he mobile posi ion inc ease.
A classical Cla ke’s ba h-shaped Dopple spec um is ob ained by assuming mul iple e lec ions close a ound he
mobile. The ca ie equency is 2,0 GHz. All senso s ha e la spa ial esponse in a sec o ed a ea o 120º, and
a e linea ly and uni o mly spaced a d/λ=0,5. All plo s shown in he simula ions below a e ep esen a ions o he
pe o mance o he link le el which can be used la e , h ough con enien mapping, o ob ain FER ( ame
e asu e a io) when conside ing channel coding o o he sys em le el ea u es [Hämäläinen].
B. Simula ions
ppL
iC×−+
/∈
=)1(
~
H
p
1
L
hi
=
M
H
H
H
H
~
~
~
~2
1
M
)29(
A se o simula ions has been pe o med using up o 9 use s o sp eading ac o 16. All use s a e assumed o
ha e con olled ansmi ed powe wi h no e o and he p opaga ion channel is he pedes ian, wi h mobiles
mo ing a ound a 3 km/h. A di e en numbe o ecei e s has been es ed: MDIR, SR wi h di e en numbe o
eigen ec o s, VRAKE and NIMI and he p obabili y o e o plo ed in igu e 5. In all cases, he pe o mance
was supe io o he con en ional VRAKE ecei e , so subs an ial gain om he use o spa ial beam o ming is
achie ed. I is ound ha simila pe o mance is ob ained wi h MDIR and SR ecei e s, showing no
imp o emen s when going om 2 o highe numbe o eigen ec o s (al hough MDIR shows he bes
pe o mance when a single eigen ec o is used). This is e i ied in igu e 6, whe e he cumula i e unc ion o he
a io o he inc easing eigen alues o he maximum eigen alue o he MDIR ecei e a e depic ed. I is clea ha
he second eigen alue ( ha is, he SNIR associa ed o he ou pu o he second eigen ec o ) is always signi ican ,
al hough i dec eases sligh ly as he numbe o ac i e use s inc ease. No e ha he hi d ei en alue is only
signi ican o a low numbe o use s, so i can be disca ded.
VIII. Conclusions
Di e en space- ime p ocesso s ha e been p esen ed which boos he powe o spa ial cancelling and cohe en
ake combining. They ha e been es ed in a ealis ic scena io, using UMTS’ FDD mode and up- o-da e models
o spa io- empo al p opaga ion channels. Resul s show a signi ican imp o emen in he p obabili y o e o
wi h espec o con en ional app oaches, ha is, only spa ial beam o ming o only VRAKE combining. Fu he
wo k is in ended o empo ally ack he es ima ed channel and co ela ion ma ices pa ame e s so as o educe
bi e o s when high speed scena ios a e ound.
IX. Re e ences
[ETSI-UTRA] “Submission o P oposed Radio T ansmission Technologies: he ETSI UMTS Te es ial
Radio Access (UTRA) ITU-R RTT Candida e Submission”, ETSI SMG2. Da e o submission:
29/1/1998. A ailable a he ITU WWW h p://www.i u.ch/im /
[Hämäläinen] S. Hämäläinen, P. Slanina, M. Ha man, A. Lappe eläinen, H. Holma, O. Salonaho, “A No el
In e ace be ween Link and Sys em Le el Simula ions”, P oc. o he ACTS Mobile
Telecommunica ions Summi , Aalbo g, Denma k, Oc obe 1997, pp. 599-604.
[Haykin] S. Haykin, Adap i e Fil e Theo y, P en ice Hall, 1996.
[Jung] P. Jung, J. Blanz, “Join de ec ion wi h cohe en ecei e an enna di e si y in CDMA mobile
adio sys ems”, IEEE T ans. On Vehicula Technology, ol. 44, no.2, Feb. 1995.
[Kay] S. Kay, Mode n Spec al Analysis, P en ice-Hall.
[Lagunas] M.A.Lagunas, J. Vidal, A.I. Pe ez, “Join beam o ming and Vi e bi equalize s in wi eless
communica ions”, P oc. 31s Asiloma Con . On Signals, Sys ems and Compu e s, No . 1997.
[Mes e] X. Mes e, J. R. Fonollosa, “Algo i hms o Flexible Mul i-S anda d A ay P ocessing: Pa 3”,
Deli e able D711, AC347/UPC/A72/PI/I007/b1, ACTS 0347 SUNBEAM p ojec .
[Pede sen] K. Pede sen, P. Mogensen, B. Fleu y, “A S ochas ic Model o he Tempo al ans Azimu hal
Dispe sion seen a he Base S a ion in Ou doo P opaga ion En i onmen s”, submi ed o IEEE
T ans. on Vehicula Technology.
[Rup ] M. Rup , F. Ta köy, J. L. Massey, “Use -Sepa a ing Demodula ion o Code-Di ision Mul iple
Access Sys ems”, IEEE Jou n. Selec . A eas in Comm., June 1994, pp. 786-795.
[VanE en] W. Van E en, “Maximum Likelihood Recei e o Mul iple Channel T ansmission Sys ems”,
IEEE T ans. on Communica ions, Feb. 1976, pp. 276-283.
[Ve dú] S. Ve dú, Mul iuse De ec ion, P en ice-Hall, 1998.
[Wid ow] B. Wid ow, P.E. Man ey, L.J. G i i hs, B.B. Goode, “Adap i e An enna Sys ems”, P oc.
IEEE, ol. 55, pp. 2143-2159, 1967.
