Joint Throughput Optimized CQI and PrecodingWeight Calculation for MIMO HSDPA

Christian Mehlfuhrer, Sebastian Caban, Martin Wrulich, and Markus Rupp

Institute of Communications and Radio-Frequency EngineeringVienna University of Technology

Gusshausstrasse 25/389, A-1040 Vienna, AustriaEmail: {chmehl, scaban, mwrulich, mrupp}@nt.tuwien.ac.at

Web: http://www.nt.tuwien.ac.at/rapid-prototyping

Abstract—In MIMO High Speed Downlink Packet Access, thechannel adaptation is performed by means of the Channel QualityIndicator (CQI) and the Precoding Control Indicator (PCI). TheCQI value is utilized to determine the coding rate and modulationalphabet, as well as the number of spatially multiplexed datastreams. The PCI value is associated to a specific precoding vectorthat is applied to the transmit signal at the basestation. In thiswork, we derive analytic expressions for the post equalizationSINR. These SINR values are then evaluated at the receiver tojointly calculate the CQI and the PCI in order to maximize thedata throughput. The SINR expressions are verified not only bysimulations but also by outdoor MIMO HSDPA measurements.

I. INTRODUCTION

In 2007, 3GPP standardized Dual Stream Transmit AntennaArray (D-TxAA) as the 2×2 MIMO extension for HighSpeed Downlink Packet Access (HSDPA) [1, 2]. D-TxAA usesprecoding at the transmitter and can be operated in single ordual stream mode depending on the channel conditions. Likein SISO HSDPA, the Transport Block Size (TBS), that is, thenumber of data bits to be transmitted in one data stream persubframe, is adjusted using Channel Quality Indicator (CQI)signaled by the user.

In SISO HSDPA, the user equipment can calculate the CQIvalue by blindly estimating the post equalization Signal toInterference and Noise Ratio (SINR) at the demapper input(that is after equalization, descrambling, and despreading).Using a lookup table, the post equalization SINR is mappedto a CQI value that allows for a transmission with a BLERsmaller than 10% [1, 3]. The required mapping tables ormapping functions are obtained by AWGN Block Error Ratio(BLER) simulations.

In MIMO HSDPA, the user equipment has to signal, addi-tionally to the CQI value, a so-called Precoding Control Indi-cator (PCI) that determines a precoding vector. The precodingvector is then applied to the user’s downlink chip stream toform the transmit signals of the two transmit antennas. In flatfading channels, the 2×1 Transmit Antenna Array (TxAA)system could achieve the same diversity [4] and array gainas a 1×2 system if the precoding would not be quantized.Nevertheless, due to the quantized precoding and the frequencyselectivity of the wireless channel, some performance loss is

immanent. Note that TxAA was already standardized for SISOHSDPA but was never applied widely in mobile networks.Related WorkThe HSDPA specification [1] proposes a method for evaluatingthe channel dependent feedback based on the channel knowl-edge at the receiver. This method separates the problem ofCQI and PCI calculation by first selecting a precoding vectorin order to maximize the received pre-equalization Signalto Noise Ratio (SNR) [5, 6]. The appropriate CQI value isselected by observing the post equalization SINR. Althoughthis method is of low complexity, it has two drawbacks:

1) It does not provide a decision if single stream or dualstream transmission shall be used.

2) Assume that the user equipment receives an HSDPA sub-frame at time t=1. Then the user equipment determinesthe PCIt=1 from the estimated channel coefficients.In the same subframe, the user equipment also deter-mines the CQIt=1 from the current post equalizationSINR although the received subframe was weightedby the previously signaled precoding PCIt=0. In casethe precoding information often changes over time, theCQI will be underestimated (since the new, updatedprecoding vector usually will lead to larger SINR) andconsequently the data throughput will reduce.

ContributionFirst, we derive analytic expressions for the post equaliza-tion SINR of MIMO HSDPA receivers. More specifically,these SINR expressions cover arbitrary equalizers, interfer-ence cancelation schemes, and channel estimation effects. Wefurthermore utilize the SINR expressions to jointly determinethe precoding vector, the supported TBS, and the numberof streams. For doing this, we need to evaluate the postequalization SINRs for every1 possible precoding vector. Thepost equalization SINRs are then mapped to CQI values usingan SINR-to-CQI mapping table. The throughput optimizedprecoding vector is then obtained by maximizing the totaltransport block size2 associated to the evaluated CQI values.

1In D-TxAA four precoding vectors are standardized in the single streammode and two precoding vectors in the dual stream mode.

2Note that in the dual stream mode the total TBS is given by the sum ofthe TBSs of the two individual streams.

Copyright 2001 SS&C. Published in the Conference Record of the Fourtysecond Asilomar Conference on Signals, Systems and Computers, 2008, October 26-29, 2008, Pacific Grove, CA, USA.

We implemented our method in a physical layer link levelsimulator and performed outdoor-to-indoor measurements withour testbed [7].

Besides the feedback calculation, the derived SINR expres-sions can also be used in system simulations to model thephysical layer [8, 9]. Since we validate the SINR expressionsnot only by physical layer HSDPA simulations but also byoutdoor measurements, these expressions are very well suitedto obtain accurate system simulation results and thus realisticnetwork performance estimates.

II. SYSTEM MODEL

In this section, we present the mathematical description ofthe HSDPA system required to derive analytical expressionsfor the post equalization SINRs. A block diagram of thissystem description is shown in Fig. 1.

Assume that we are transmitting Ns independently codedand modulated data chip streams, each of length Lc = Lh +Lf −1 chips, Lh and Lf corresponding to the channel and theequalizer length, respectively. We define the stacked transmitchip vector as

sk =[s(1)Tk , . . . , s(Ns)T

k

]T

. (1)

The Ns chip streams are weighted by the NT×Ns dimensionalprecoding matrix

W =

w(1,1) . . . w(1,Ns)

.... . .

...w(NT,1) . . . w(NT,Ns)

(2)

forming the data chip streams of the NT transmit antennas.At each transmit antenna, pilot, synchronization, and controlchannels accumulated in

pk =[p(1)T

k , . . . ,p(NT)Tk

]T

(3)

are added. With the Lc×Lc dimensional identity matrix ILc ,the transmit signal is given by

ak = (W ⊗ ILc) sk + pk. (4)

The frequency selective link between the nt-th transmit and thenr-th receive antenna is modeled by the Lf×Lc dimensionalband matrix

H(nr,nt) =

h

(nr,nt)0 . . . h

(nr,nt)Lh−1 0

. . . . . .0 h

(nr,nt)0 . . . h

(nr,nt)Lh−1

, (5)

where the h(nr,nt)i (i = 0, . . . , Lh − 1) represent the channel

impulse response between the nt-th transmit and the nr-threceive antenna. The entire frequency selective MIMO channelis modeled by a block matrix H consisting of NR×NT bandmatrices defined in (5)

H =

H(1,1) . . . H(1,NT)

.... . .

...H(NR,1) . . . H(NR,NT)

. (6)

Fig. 1. Generalized system model for describing the HSDPA physical layer.

At the receiver, noise and out-of-cell interference, their sumdenoted by vk, deteriorates the desired signal

bk = Hak + vk = H (W ⊗ ILc) sk + Hpk + vk. (7)

The signal bk is then processed in an MMSE based [10, 11]or an interference aware [12, 13] equalizer F to obtain anestimate of the transmitted chip stream

sk =[s(1)k−τ , . . . , s

(Ns)k−τ

]T

= Fbk

= FH (W ⊗ ILc) sk + FHpk + Fvk. (8)

The equalizer matrix F here consists of Ns vectors, each oflength NRLf

F =[f (1), . . . , f (Ns)

]T

. (9)

III. POST EQUALIZATION SINR CALCULATION

In this section, we derive analytic expressions for the postequalization interference terms. These expressions can be usedto determine the total post equalization SINR required forcalculating the CQI and PCI feedback information.

Since the true channel matrix H is unknown at the receiverand can only be estimated with a certain Mean Square Error(MSE), we approximate the true channel matrix by the esti-mated channel matrix H and a matrix H∆ representing thechannel estimation error [14]. The matrix H∆ is constructedlike the channel matrix H in (5) and (6). The non-zeroelements of H∆ are assumed to be i.i.d. Gaussian with avariance equal to the MSE of the channel estimator.

A. HS-PDSCH Interference

The High Speed Physical Downlink Shared Channel (HS-PDSCH) consists of the data chip streams in sk. These codedivision multiplexed data chip streams generate interferencedue to non perfect equalization at the receiver. The totalimpulse response seen by all data chip streams in sk to theoutput of the ns-th equalization filter is given by

g(ns)Ts = f (ns)T

(H + H∆

)(W ⊗ ILc) . (10)

The interference power at the output of the equalizer can bedivided into a deterministic part (caused by H) and a stochasticpart (caused by H∆). The deterministic interference can becanceled by a decision feedback equalizer [15], whereas thestochastic interference can only be decreased by reducing theMSE of the channel estimator.

The deterministic HS-PDSCH interference power is cal-culated by accumulating the energies of the total impulse

Copyright 2001 SS&C. Published in the Conference Record of the Fourtysecond Asilomar Conference on Signals, Systems and Computers, 2008, October 26-29, 2008, Pacific Grove, CA, USA.

response at delays k 6= τ . The remaining interference at delayτ (the chosen delay of the transmitted chip stream after thechannel and the equalizer) is irrelevant since it is perfectlyremoved by the despreading operation

γ(ns)

s,H=

PHS−PDSCH

Ns·

Lc∑k=1k 6=τ

∣∣∣(f (ns)TH (W ⊗ ILc))

k

∣∣∣2 .

(11)

Here, PHS−PDSCH corresponds to the transmit power availablefor all Ns data chip streams. The operator (.)k denotes the k-thelement of a vector.

The stochastic HS-PDSCH interference power is obtainedby building the expectation with respect to the unknownchannel estimation error

γ(ns)s,H∆

=PHS−PDSCH

Ns·

· EH∆

Lc∑

k=1k 6=τ

∣∣∣(f (ns)TH∆ (W ⊗ ILc))

k

∣∣∣2 ≈

≈PHS−PDSCH

Ns·MSE · (Lh − 1) NT

∥∥∥f (ns)∥∥∥2

2. (12)

In this derivation, the last step can be verified by a carefulinspection of the matrix-vector multiplications.

B. Pilot Interference

The total impulse response seen by the pilot channels fromthe transmit antennas to the output of the ns-th equalizationfilter is given by the NTLc length vector

g(ns)Tp = f (ns)TH = f (ns)T

(H + H∆

). (13)

Again, we can identify a deterministic interference

γ(ns)

p,H=

PCPICH

NT·

·NT∑

nt=1

Lc∑k=1k 6=τ

∣∣∣∣(f (ns)TH)

k+(nt−1)Lc

∣∣∣∣2 (14)

and a stochastic interference

γ(ns)p,H∆

=PCPICH

NT·

· EH∆

NT∑

nt=1

Lc∑k=1k 6=τ

∣∣∣∣(f (ns)TH∆

)k+(nt−1)Lc

∣∣∣∣2 ≈

≈ PCPICH

NT·MSE · (Lh − 1)

∥∥∥f (ns)∥∥∥2

2. (15)

Here, PCPICH denotes the total pilot channel power for alltransmit antennas.

C. Synchronization and Control Channels Interference

For the calculation of the interference emerging from thesynchronization and control channels we assume that thesechannels are transmitted on all antennas simultaneously, thatis, the power is equally distributed on all transmit antennas.Since the Synchronization Channel (SCH) and the controlchannel (PCCPCH) are transmitted time multiplexed, we as-sume that both channels have equal power; that is, PSCH =PCCPCH. The total impulse response of the synchronizationand control channels to the output of the ns-th equalizationfilter is given by

g(ns)TSCH =

1√NT

f (ns)T(H + H∆

)(1NT ⊗ ILc) . (16)

Here, 1NT denotes an NT × 1 dimensional vector withall entries equal to one. Note that the multiplication with(1NT ⊗ ILc) represents the summation of the individual trans-mit antenna impulse responses. This is required because weassumed that synchronization and control channels are equallydistributed on all transmit antennas. The deterministic part ofthe interference is calculated as

γ(ns)

SCH,H=

PSCH

NT·

Lc∑k=1k 6=τ

∣∣∣(f (ns)TH (1NT ⊗ ILc))

k

∣∣∣2 (17)

and the stochastic interference as

γ(ns)SCH,H∆

=PSCH

NT·

· EH∆

Lc∑

k=1k 6=τ

∣∣∣(f (ns)TH∆ (1NT ⊗ ILc))

k

∣∣∣2 ≈

≈ PSCH

NT·MSE · (Lh − 1)

∥∥∥f (ns)∥∥∥2

2. (18)

Additionally to the two interference terms above, interferenceat delay lag k = τ emerges from the SCH since it istransmitted without spreading and scrambling and is thus notorthogonal to the data channels

γ(ns)

SCH,H,τ= 0.1

PSCH

NT

∣∣∣(f (ns)TH (1NT ⊗ ILc))

τ

∣∣∣2 , (19)

γ(ns)SCH,H∆,τ =0.1

PSCH

NTEH∆

{∣∣∣(f (ns)TH∆

)τ

∣∣∣2} =

=0.1PSCH

NT·MSE·

·NR∑

nr=1

Lh∑k=1

∣∣∣∣(f (ns))

k+τ−Lh+(nr−1)Lf

∣∣∣∣2 . (20)

The constant factor 0.1 originates from the time-multiplexingof the SCH with the PCCPCH since the SCH occupies onlythe first 10% of all chips in every transmitted slot. Note thatthe PCCPCH does not contribute to the interference terms (19)and (20) since it is transmitted with spreading.

Copyright 2001 SS&C. Published in the Conference Record of the Fourtysecond Asilomar Conference on Signals, Systems and Computers, 2008, October 26-29, 2008, Pacific Grove, CA, USA.

TABLE IHSDPA SYSTEM PARAMETERS.

Parameter ValueCPICH Ec/Ior −10 dBSCH/PCCPCH Ec/Ior −12 dBUser equipment categories [1] 16Measured channel realizations 784Maximum transmit power 39 dBmCenter frequency 2.5 GHz

D. Post Equalization Noise

According to our system model (8), the variance of the postequalization noise at the output of the ns-th equalization filteris given by

σ(ns)2v′ = ‖f (ns)‖2

2σ2v . (21)

E. Post Equalization SINR

The total deterministic interference caused by the pilot,control, and synchronization channels is given by

γ(ns)

H= γ

(ns)

p,H+ γ

(ns)

SCH,H+ γ

(ns)

SCH,H,τ. (22)

Analogously, the stochastic interference is calculated by

γ(ns)H∆

= γ(ns)p,H∆

+ γ(ns)SCH,H∆

+ γ(ns)SCH,H∆,τ . (23)

Knowing all the above interference terms we can calculate thepost equalization SINR [8, 9] of the ns-th data stream at thedespreader output

SINRest(ns) =

SF∣∣∣(g(ns)

s

)τ

∣∣∣2 PHS−PDSCHNs

γ(ns)

s,H+ Cγ

(ns)

H+ γ

(ns)s,H∆

+ γ(ns)H∆

+ σ(ns)2v′

.

(24)

Here, SF is the spreading factor of the HS-PDSCH. The factorC is equal to 0 if the receiver uses interference cancelation ofthe synchronization, pilot, and control3 channels and equalto 1 if no interference cancelation is employed. If moreadvanced interference cancelation schemes like in [15] areemployed, also the term γ

(ns)

s,Hcan be reduced. The stochastic

interference terms can only be reduced by more advancedchannel estimation algorithms, and/or by iterative receivertechniques that use decision directed channel estimation.

IV. SIMULATIONS AND MEASUREMENTS

The accuracy of the SINR estimation was evaluated bylink level simulations and outdoor measurements. In the sim-ulations, we used an uncorrelated ITU Pedestrian B channelmodel [16]. The measurements were carried out in the alpinevalley Drautal in Carinthia, Austria and utilized the ViennaMIMO testbed described in detail in [7] and [17–19]. Thetransmitter and receiver were placed on the opposing sides ofthe valley4 in a distance of 5.7 km (Fig. 2). At the transmitter,

3Note that we assume here that the control channel can be detected errorfree due to the large spreading factor of 256 which provides a large spreadinggain and thus high SNR.

4Detailed transmitter and receiver positions can be downloadedfor Google Earth at http://www.nt.tuwien.ac.at/fileadmin/data/testbed/Carinthia-RX-TX-GPS.kmz

XYΦ table

radios

Sync

radiosPC

RA

M

C/Matlab

A D

RA

MA D

HDDsP2P

connection

TX (outdoor)

RX (indoor)

UTM coordinates:33T 353811mE 5178652mN

653 m.a.s.l.

33T 358943mE 5176167mNUTM coordinates:

991 m.a.s.l.

5.7 km

SyncPC

C/Matlab

HDDsP2Pconnection

Fig. 2. Measurement setup in the alpine valley Drautal in Carinthia, Austria.

we utilized a Kathrein 800 10543, 60◦ XX-pol basestationpanel antenna [20] with ±45◦ polarization. A center frequencyof 2.5 GHz and a maximum transmit power level of 39 dBmwas used. At the receiver, four Linksysr 5 dBi rod antennaspointing into different directions were mounted on a brasssheet in the size of the top of a Linksys WLAN router.The receive antennas were rotated and moved by an XYΦ-positioning table to generate 784 channel realizations withinan area of 3 λ×3 λ (36 cm×36 cm). At every position wemeasured twelve different transmit power levels.

The measurements were carried out using the followingprocedure: At every receive antenna position and transmitpower value, we first transmitted one HSDPA subframe whichwas evaluated in a “mini-receiver”. This mini-receiver onlyperformed synchronization, channel estimation and subse-quently the CQI and PCI calculation. The CQI and the PCIvalue was then fed back to the transmitter which transmittedan HSDPA subframe corresponding to the CQI and PCI value.This subframe was then stored at the receiver for later off-lineevaluation. More details about the measurement procedure canbe found in [21].

In the off-line data evaluation, we employed a low complex-ity LMMSE channel estimator [22] with three iterations at thereceiver. The iterations were required to alternately improvethe channel estimate and the channel autocorrelation estimate.The channel estimate was then used to calculate an estimate ofthe deterministic part of the receive signals (synchronizationand pilot signal) which were subtracted from the received sig-nal to perform interference cancelation. A subsequent LMMSEequalizer [21] was then applied to obtain an estimate ofthe transmitted data chip stream. Compared to the optimumML receiver [23], LMMSE equalization with interferencecancelation provides a good performance/complexity tradeoff.Log-likelihood values of the received bits were calculated ina soft max-log-MAP demapper and further processed in aTurbo decoder with eight iterations. If the CRC of the HSDPAsubframe was correct, the number of data bits was consideredin the throughput calculation. If the CRC failed, the datathroughput of this frame was set to zero.

Copyright 2001 SS&C. Published in the Conference Record of the Fourtysecond Asilomar Conference on Signals, Systems and Computers, 2008, October 26-29, 2008, Pacific Grove, CA, USA.

0

5

10

15

20

25

thro

ughp

ut [M

bit/

s]

-10 0-5

0

5

10

15

20

25

30

35

40

450

5

10

15

20

25

-5

0

5

10

15

20

25

30

35

40

45

Ior/Ioc [dB]

post

equ

aliz

atio

n SI

NR

[dB

]

thro

ughp

ut [M

bit/

s]po

st e

qual

izat

ion

SIN

R [dB

]

1x1 SISO

1x2 SIMO2x1 TxAA2x2 TxAA

2x2 D-TxAA

1x1 SISO

2x1 TxAA

2x2 D-TxAA

1x2 SIMO

2x2 TxAA

1x1 SISO

1x2 SIMO2x1 TxAA2x2 TxAA

2x2 D-TxAA

1x1 SISO

2x1 TxAA

1x1 SISO

1x2 SIMO

2x2 TxAA

1x2 SIMO

2x1 TxAA

2x2 TxAA

2x2 D-TxAAStream 1 & 2

-5 5 10 15 20 25 30

MEASUREMENTnon-urban outdoor-to-indoor

tap-wise LMMSE channel estimation

SIMULATIONuncorrelated ITU Pedestrian B channel

perfect channel knowledge

estimated SINRest

observed SINR

total transmit power [dBm]

5 10 15 20 25 30 35 40

observed SINR

estimated SINRest

1x1 SISO

1x2 SIMO

2x2 TxAA

1x2 SIMO

2x1 TxAA

2x2 TxAA

2x2 D-TxAAStream 1 & 2

2x2 D-TxAA

1x2 SIMO

2x2 TxAA

Fig. 3. Simulated (left side) and measured (right side) throughput and post equalization SINRs.

V. RESULTS

The results of our simulations and measurements are shownin Fig. 3 in form of throughput and SINR plots. The plots onthe left side in Fig. 3 are obtained by physical layer simulationswith perfect channel knowledge at the receiver. The plots onthe right side in Fig. 3 are obtained by measurements withchannel estimation at the receiver.

The “observed SINR” in Fig. 3 is the SINR observed at thedemapper input and is obtained as follows: Consider the trans-mitted data symbol vector d(ns) of the ns-th symbol streamand the corresponding received symbol vector at the demapperinput d(ns). The “observed” or “true” post equalization SINRis given by

SINR =‖d(ns)‖2

2

‖d(ns) − d(ns)‖22

. (25)

In the simulations, we observe a very good fit of the estimatedSINRs (calculated according to Equ. (24)) over the full Ior/Ioc

(ratio of the energy of the desired basestation to the energyof the interfering basestations) range. The SINR increaseslinearly with increasing Ior/Ioc. The estimated SINRs in themeasurements also show a good fit at all transmit powers. Incontrast to the simulations, the SINR saturates at about 30 dB

which is caused by residual interference due to the channelestimation error.

In the measured scenario, the 2×1 TxAA system almostachieves the performance of the 1×2 SIMO system, whereasin the simulations a larger distance of almost 5 dB is observed.The large distance in the simulation is caused by the ratherlarge delay spread of the Pedestrian B channel model. Atsuch a large delay spread, the precoding with a complexscalar for every transmit antenna is far from optimal. In ourmeasurements, however, the channel has a very small delayspread (about three chips) and thus the precoding works verywell.

In Fig. 4, the individual post equalization interference termsof the 2×2 TxAA system are shown over Ior/Ioc. BelowIor/Ioc ≈ 10 dB, the system performance is mainly dominatedby the post equalization noise. Above Ior/Ioc > 10 dB, thedeterministic interference of pilot and synchronization channelbecomes dominant and has to be canceled to achieve highperformance. If interference cancelation is performed, thesystem performance at high Ior/Ioc ≈ 10 is mainly dominatedby the channel estimation error.

Copyright 2001 SS&C. Published in the Conference Record of the Fourtysecond Asilomar Conference on Signals, Systems and Computers, 2008, October 26-29, 2008, Pacific Grove, CA, USA.

−10 0 10 20 3010-5

10-4

10-3

10-2

10-1

100

rela

tive

int

erfe

renc

e an

d no

ise

term

s

Ior/Ioc [dB]

- noise:

- deterministic self interference of data channels:- deterministic part:- stochastic part:interference of pilot, control, and synchronization channels:

- without interference cancelation:total interference and noise:

- with interference cancelation:

Fig. 4. Interference terms for a two receive antennas TxAA transmissionover an uncorrelated ITU Pedestrian B channel.

VI. CONCLUSIONS

In this work, we evaluated a throughput-maximizing methodthat jointly determines the CQI and the PCI value. The methodwas implemented in a physical layer link level simulator wherea very good agreement between the estimated and the truepost equalization SINR was achieved. Measurements withour testbed also showed a very good agreement between theestimated and the true SINR. A detailed analysis of the postequalization interference terms proofs that the performance atlarge receive SNRs is limited by interference caused by pilotand synchronization channels and by residual interference dueto channel estimation errors. Therefore, interference cancela-tion as well as high performance channel estimation techniqueshave to be implemented in a MIMO HSDPA receiver in orderto achieve high data throughput.

ACKNOWLEDGMENTS

The authors would like to thank the C12 project team fortheir work on the HSDPA simulator and fruitful discussions.We would also like to thank Walter Schuttengruber for histireless support on the Vienna MIMO testbed. This work hasbeen funded by the Christian Doppler Laboratory for DesignMethodology of Signal Processing Algorithms and the Insti-tute of Communications and Radio Frequency Engineering.

REFERENCES[1] 3GPP, “Technical specification group radio access network; physical layer

procedures (FDD) (Tech. Spec. 25.214 V7.7.0),” Nov. 2007. [Online]. Available:http://www.3gpp.org/ftp/Specs/html-info/25214.htm

[2] S. McBeath, M. Ahmed, and K. Rohani, “Impact of imperfect estimators onW-CDMA receiver performance with MIMO antenna systems,” vol. 2, pp.

1152–1156, Oct. 2003. [Online]. Available: http://ieeexplore.ieee.org/iel5/9004/28569/01285202.pdf?arnumber=1285202

[3] Motorola and Nokia, “Revised HSDPA CQI proposal,” 3GPP, Tech. Rep. TSG-RAN Working Group 4 Meeting #22, R4-020612, Apr. 2002. [Online]. Available:http://www.3gpp.org/ftp/tsg ran/WG4 Radio/TSGR4 22/Docs/R4-020612.zip

[4] S. Narayanaswamy and M. Rupp, “Antenna diversity in wireless communicationterminals,” US Patent 5,905,467, 1999. [Online]. Available: http://www.freepatentsonline.com/5905467.html

[5] E. Zacarias, S. Werner, and R. Wichman, “Partial update adaptive transmitbeamforming with limited feedback,” in Proc. IEEE International Conference onAcoustics, Speech, and Signal Processing (ICASSP 2006), Toulouse, France, May2006. [Online]. Available: http://ieeexplore.ieee.org/iel5/11024/34760/01661070.pdf?tp=&isnumber=&arnumber=1661070

[6] E. Zacarias, “Adaptive transmit beamforming with closed loop feedback in MIMOsystems,” Signal Processing Laboratory, Helsinki University of Technology, Tech.Rep., 2004. [Online]. Available: http://www.comlab.hut.fi/opetus/333/2004 2005slides/txbf closed loop text.pdf

[7] S. Caban, C. Mehlfuhrer, R. Langwieser, A. L. Scholtz, and M. Rupp, “ViennaMIMO testbed,” EURASIP Journal on Applied Signal Processing, Special Issueon Implementation Aspects and Testbeds for MIMO Systems, vol. 2006, Article ID54868, 2006. [Online]. Available: http://publik.tuwien.ac.at/files/pub-et 10929.pdf

[8] A. Szabo, N. Geng, A. Seeger, and W. Utschick, “Investigations on link tosystem level interface for MIMO systems,” in Proc. 3rd International Symposiumon Image and Signal Processing and Analysis (ISPA2003), pp. 365–369, Rome,Italy, Sept. 2003. [Online]. Available: http://ieeexplore.ieee.org/iel5/9084/28837/01296924.pdf?tp=&isnumber=&arnumber=1296924

[9] M. Wrulich, S. Eder, I. Viering, and M. Rupp, “Efficient link-to-system levelmodel for MIMO HSDPA,” in Proc. of the 4th IEEE Broadband Wireless AccessWorkshop, 2008.

[10] S. Geirhofer, C. Mehlfuhrer, and M. Rupp, “Design and real-time measurementof HSDPA equalizers,” in Proc. of the 6th IEEE Workshop on Signal ProcessingAdvances in Wireless Communications (SPAWC 2005), pp. 166–170, New YorkCity, USA, June 2005. [Online]. Available: http://publik.tuwien.ac.at/files/pub-et9722.pdf

[11] L. Mailaender, “Linear MIMO equalization for CDMA downlink signals withcode reuse,” IEEE Transactions on Wireless Communications, vol. 4, no. 5, pp.2423– 2434, Sept. 2005. [Online]. Available: http://ieeexplore.ieee.org/iel5/7693/32683/01532226.pdf

[12] M. Wrulich, C. Mehlfuhrer, and M. Rupp, “Interference aware MMSE equalizationfor MIMO TxAA,” in Proc. Third International Symposium on Communications,Control, and Signal Processing (ISCCSP 2008), pp. 1585–1589, St. Julians, Malta,Mar. 2008. [Online]. Available: http://publik.tuwien.ac.at/files/pub-et 13657.pdf

[13] C. Mehlfuhrer, M. Wrulich, and M. Rupp, “Intra-cell interference awareequalization for TxAA HSDPA,” in Proc. IEEE International Symposium onWireless Pervasive Computing (ISWPC 2008), pp. 406–409, Santorini, Greece,May 2008. [Online]. Available: http://publik.tuwien.ac.at/files/pub-et 13749.pdf

[14] B. Hassibi and B. Hochwald, “How much training is needed in multiple-antennawireless links?” Information Theory, IEEE Transactions on, vol. 49, no. 4, pp.951–963, Apr. 2003. [Online]. Available: http://ieeexplore.ieee.org/stamp/stamp.jsp?arnumber=1193803

[15] D. Bosanska, C. Mehlfuhrer, and M. Rupp, “Performance evaluation of intra-cellinterference cancelation in D-TxAA HSDPA,” in Proc. International ITGWorkshop on Smart Antennas (WSA 2008), Darmstadt, Germany, Feb. 2008.[Online]. Available: http://publik.tuwien.ac.at/files/pub-et 13677.pdf

[16] “Recommendation ITU-R M.1225: Guidelines for evaluation of radio transmissiontechnologies for IMT-2000,” Tech. Rep., 1997.

[17] C. Mehlfuhrer, S. Caban, and M. Rupp, “Experimental evaluation of adaptivemodulation and coding in MIMO WiMAX with limited feedback,” EURASIPJournal on Advances in Signal Processing, Special Issue on MIMO Systemswith Limited Feedback, vol. 2008, Article ID 837102, 2008. [Online]. Available:http://publik.tuwien.ac.at/files/pub-et 13762.pdf

[18] C. Mehlfuhrer, S. Geirhofer, S. Caban, and M. Rupp, “A flexible MIMOtestbed with remote access,” in Proc. of the 13th European Signal ProcessingConference (EUSIPCO 2005), Antalya, Turkey, Sept. 2005. [Online]. Available:http://publik.tuwien.ac.at/files/pub-et 9732.pdf

[19] M. Rupp, C. Mehlfuhrer, S. Caban, R. Langwieser, L. W. Mayer, andA. L. Scholtz, “Testbeds and rapid prototyping in wireless system design,”EURASIP Newsletter, vol. 17, no. 3, pp. 32–50, Sept. 2006. [Online]. Available:http://publik.tuwien.ac.at/files/pub-et 11232.pdf

[20] Kathrein, “Technical specification Kathrein antenna type no. 800 10543.” [Online].Available: http://www.kathrein-scala.com/catalog/80010543.pdf

[21] C. Mehlfuhrer, S. Caban, and M. Rupp, “Measurement based evaluation of lowcomplexity receivers for D-TxAA HSDPA,” in Proc. of the 16th European SignalProcessing Conference (EUSIPCO 2008), Lausanne, Switzerland, Aug. 2008.[Online]. Available: http://publik.tuwien.ac.at/files/PubDat 166132.pdf

[22] C. Mehlfuhrer and M. Rupp, “Novel tap-wise LMMSE channel estimation forMIMO W-CDMA,” in Proc. 51st Annual IEEE Globecom Conference, 2008, NewOrleans, LA, USA, Nov. 2008.

[23] M. Rupp, M. Guillaud, and S. Das, “On MIMO decoding algorithms forUMTS,” in Conference Record of the Thirty-Fifth Asilomar Conference onSignals, Systems and Computers, 2001, vol. 2, pp. 975–979, Nov. 2001.[Online]. Available: http://ieeexplore.ieee.org/iel5/7744/21273/00987640.pdf?tp=&isnumber=&arnumber=987640

Copyright 2001 SS&C. Published in the Conference Record of the Fourtysecond Asilomar Conference on Signals, Systems and Computers, 2008, October 26-29, 2008, Pacific Grove, CA, USA.