radio propagation characteristics of mimo systems in...
TRANSCRIPT
www.mobilevce.com
© 2007 Mobile VCE
NPL EM Day29 November 2007
Radio Propagation Characteristics of MIMO Systems in Practical Deployment Scenarios
Matthew Webb, Mark Beach and Mythri Hunukumbure
University of Bristol
www.mobilevce.com
© 2007 Mobile VCE
Introduction
MIMO now finding commercial interest802.11n – ‘pre-N’ already on the shelvesWiMAX3GPP LTEStandards in development: IEEE 802.20, 802.22
Some, e.g. spatial-multiplexing ‘Mobile WiMAX’use feedback to improve throughput / reliability
Broadens the modelling vs. experiment questionImposes additional constraints on the systemImposes additional computational load on the system
Need to have ways of quantifying how the radio environment affects these constraints and loads
www.mobilevce.com
© 2007 Mobile VCE
Introduction
Measurement campaign (2 GHz) to characterise outdoor urban environment with ‘prototype’user devices – PDA, laptop, reference antennas
MIMO throughput may increase significantly if channel feedback is available
Key is to exploit spatial eigenmodes efficientlyNeed to understand ‘dynamics of eigen modes’Variations in time and frequency increase feedback load
“overhead => reduction in throughput to users”
Study eigenmode characteristicsCoherence time/bandwidth of eigenvalues (key to MIMO)Investigate relationship of eigen coherences to classical parameters
www.mobilevce.com
© 2007 Mobile VCE
Measurement equipment
Medav RUSK sounder customised for 2GHz outdoor MIMO operation:
20MHz measurement bandwidth (128 fingers)4096 snapshots in 6.3 s
MIMO Ant.Interface
MIMO Ant.Interface
TxM
ux
LNA
RFfeed
Controlsignals
Initial parameter downloading / MIMO sync
MuxControl
RF i/p
PA 1
MEDAVRx.
MEDAVRx.
Tx 1
Tx 4
Rx 1
Rx 4
PA 4MEDAV
Tx.
MEDAVTx.
Rx
Mux
www.mobilevce.com
© 2007 Mobile VCE
Prototype devices
Laptop: 4 Printed Inverted F Antennas (PIFAs) in display lid. PDA : 4 linear slot antennas with V and H polarisations.Reference array: 4 dipoles fitted onto a cycle helmet (at 45˚
slant on two planes)No spacing and orientation constraintsNo user blocking
www.mobilevce.com
© 2007 Mobile VCE
Measurement area
58 locations (standing and walking measurements)Drive tests also conducted140GB of (raw) channel data
Area 1: Broadmead
Area 2: Victoria Street & Knights TemplarArea 3: Queens Square,
Waterfront & City Centre
Area 4: Eye Hospital & Bus Station
www.mobilevce.com
© 2007 Mobile VCE
MIMO campaign – measurements
6 m
1 m/s
Reference antenna worn on head to avoid user blockingMeasurements made in pairs
Ref. antenna with PDARef. antenna with laptop
Walking measurements spanning 6m, 2 routes per location
Standing measurements (6s) with 4 different orientations
Drive tests
www.mobilevce.com
© 2007 Mobile VCE
Example measurement
www.mobilevce.com
© 2007 Mobile VCE
Data analysis
For each location:RMS delay spreadRician K factorAverage SNRDynamic range of the SNR (max(SNR)dB- min(SNR)dB)Channel correlation coefficientsMIMO capacities at fixed 20dB SNRMIMO capacities at actual received SNRMIMO capacities for a 2x2 configuration selected on the highest SNR, for Laptop and PDA measurementsCoherence time for eigenmodesCoherence bandwidth for eigenmodes
www.mobilevce.com
© 2007 Mobile VCE
Correlation
www.mobilevce.com
© 2007 Mobile VCE
Received SNR variations
Received SNR
-5 0 5 10 15 20 25 30 350
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
SNR (dB)
Pro
babi
lity
(SN
R<a
bsci
ssa)
Ref. with PDAPDARef. with LaptopLaptop
0 5 10 15 20 25 30 35 400
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
Dynamic SNR range (dB)
Pro
babi
lity
(Dyn
amic
SN
R ra
nge
< ab
scis
sa)
Ref. with PDAPDARef. with LTLT
Dynamic range
www.mobilevce.com
© 2007 Mobile VCE
MIMO capacity
www.mobilevce.com
© 2007 Mobile VCE
Eigen coherence time and bandwidth
Eigenvalue coherence timeWhen is the eigenvalue’s temporal auto-correlation <0.7 ?Snapshot resolution 6.144 ms for interpolationAverage results across 128 frequencies per location
Eigenvalue coherence bandwidthWhen is the eigenvalue’s spectral auto-correlation <0.7 ?Frequency finger width is 156.25 kHz for interpolationAverage results across 1024 snapshots per location
Insight into how frequently feedback is needed over how many individual frequencies
www.mobilevce.com
© 2007 Mobile VCE
Eigen coherence time
0 50 100 150 200 250 300 350 4000
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
Coherence time for the maximum eigen value Tc (ms)
Pro
babi
lity
(Tc<
Abs
ciss
a) Standing DipolesStanding PDAStanding LaptopWalking DipolesWalking PDAWalking Laptop
Reference dipoles lowest Tc when walkingBroader azimuth view, no body shadowing, cross-polar response antenna ‘sees’ more multipath and associated dynamics
WalkingAv = 35ms Standing
Av = 120ms
www.mobilevce.com
© 2007 Mobile VCE
Eigen coherence bandwidth
0 200 400 600 800 1000 1200 1400 1600 1800 20000
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
Coherence bandwidth for the maximum eigen value Bc (kHz)
Pro
babi
lity
(Bc<
Abs
ciss
a)
Ref. DipolesPDA moduleLaptop module
Walking / standing indistinguishableReference dipoles have lowest Bc
Wider azimuth view more multipath components higher delay spread lower coherence bandwidth
⇒ ⇒⇒
www.mobilevce.com
© 2007 Mobile VCE
Eigen coherence bandwidth vs. delay spread
Inverse power law y = axb + c
0.69187.9-1.525.7Laptop
0.72203.5-2.289.72PDA
0.82211.6-2.486.55Reference dipoles
Corr. coeff.cbaAntenna
Good fitRMS delay easier metric for feedback requirements
www.mobilevce.com
© 2007 Mobile VCE
Eigen coherence time vs. Doppler spread
0.42129.4-0.36305Laptop
0.521.56-1.56380PDA
0.464.7-1.95567Reference dipoles
Corr. coeff.cbaAntenna
Inverse power law y = axb + c
Weak trend
www.mobilevce.com
© 2007 Mobile VCE
Conclusions
City-centre is highly-scattering, so most MIMO statistics are close to idealExplored temporal and spectral dynamics of MIMO eigenmodes in measured environmentsHead-worn dipoles (unobstructed by body) experienced more multipath activity => reduced coherence time/bandwidth RMS delay spread a good predictor of eigen coherence bandwidth, and easier to calculateFurther exploration of eigen coherence time vs. RMS Doppler spread is required