Trends in Wireless Communications

1
Trends in Wireless Communications

• Geert Leus
• Delft University of Technology
• g.leus_at_tudelft.nl
• Acknowledgements
• STW via VIDI-TVCOM and VICI-SPCOM
• TNO via UCAC
• University of Perugia
• Michigan Technological University
• Katholieke Universiteit Leuven

2
Outline

• Communications over time-varying channels
• Feedback in single-user and multi-user MIMO
  systems
• Ultra wideband communications
• Cognitive radio

3
Communications over time-varying channels
4
Problem Statement

• Many wireless communications standards assume the
  channel is time-invariant over a block (mainly
  OFDM) IEEE802.11, IEEE802.16d, DVB-T,
• When used in high mobility situations, problems
  occur and the orthogonality among subcarriers
  gets lost IEEE802.16e, DVB-H, underwater
  communications, ...
• Special transceiver signal processing techniques
  are required to solve this self-interference
  problem

5
OFDM

• Input-output relation

• Using a cyclic prefix, we get a circular
  convolution

6
OFDM

• How does this circular convolution look like?

7
OFDM

• We take IDFT and DFT at transmitter and receiver

8
OFDM

• We assume guard intervals are removed

9
OFDM Equalization
Non-banded equalizers
Banded equalizers
block MMSE banded
block MMSE non-banded
Block equalizers
Choi et al, TCOM 01
Rugini et al, COML 05
serial MMSE non-banded
serial MMSE banded
Serial equalizers
Cai-Giannakis, TCOM 03
Schniter, TSP 04
10
Improving Band Assumption

• Transmitter and receiver windows or pulse shapes
  have been developed to improve the subcarrier
  orthogonality and reduce the cyclic prefix length
• We relax these windows to improve the band
  approximation instead of the subcarrier
  orthogonality
• These methods are low-complexity in the sense
  that they have a complexity that is linear in the
  number of subcarriers
• Such schemes have also been labeled as
  generalized multicarrier systems

11
Channel Estimation

• There are too many unknowns to estimate
• We need a reduced model that exploits the
  correlation

Basis expansion model (BEM)
12
Channel Estimation

• Polynomial BEM
• Complex Exponential BEM

13
Channel Estimation

• Pilots are inserted in the frequency domain

selected received samples
pilots
14
Extensions

• Non-linear equalization
• Decision-feedback equalization
• Iterative equalization
• Improved channel estimation
• Semi-blind channel estimation
• Iterative channel estimation
• Extensions to MIMO
• Spatial multiplexing (with or without precoding)
• Space-time block coding

15
Simulation Results
N 128 subcarriers NA N active subcarriers L
5 channel taps fD / ?f 0.35 Doppler spread Q
3 bandwidth equalizer
Channel estimation parameters 8 pilots K2N256
resolution CE-BEM P4 basis functions LMMSE
estimator
16
Application

• UCAC project UUV - Covert Acoustic
  Communications

17
Application

• Different partners are testing different
  technologies to transfer a certain number of bits
  at the lowest SNR
• TNO and Delft University of Technology study OFDM
• Loss of orthogonality among subcarriers is a
  major problem when using OFDM in this set-up
• The proposed methods can be used to solve problem
• Multi-band OFDM is used to reduce complexity

18
Application
Time-invariant estimator
Time-varying estimator
19
Feedback for Single- and Multi-User MIMO Systems
20
Why feedback?

• Feedback of the channel state information (CSI)
  in a single-user multiple-input multiple-output
  (MIMO) system allows for improved capacity, SNR,
  BER,
• Example
• Feedback in a multi-user MIMO system allows for
  the exploitation of the so-called multi-user
  diversity by selecting the right set of users

Capacity MISO with CSIT
Capacity MISO without CSIT
21
Feedback for Single-User MIMO

• Both spatial multiplexing and space-time coding
  are incorporated in the above model
• The precoder adapts the transmitted signal to
  the current channel conditions

CSI est. and detection
Spat. mux. or ST code
precoder
Low-rate feedback link
22
Feedback for Single-User MIMO

• Many different feedback schemes have been
  proposed
• Statistical feedback of the CSI useful if the
  channel varies too rapidly to track accurately
• Quantized feedback of the CSI can exploit strong
  spatial modes if channel varies slowly
• We focus on quantized feedback

23
Quantized Feedback

• Quantized feedback is based on codebook of
  precoders

• Quantization steps are related to vector
  quantization

precoder construction (decoder)
index selection (encoder)

• The funtion can yield different
  measures
• Some distance between and
• Some performance measure of a chosen receiver
• Capacity loss due to quantization

24
Quantized Feedback

• Codebook design procedures
• Grassmannian sphere packing the precoders are
  optimally packed w.r.t. some subspace distance
• Generalized Lloyd algorithm the precoders are
  designed by iteratively minimizing the average
  distortion (done in 2 steps)
• Monte-Carlo algorithm randomly generate a large
  set of codebooks and select the one that
  minimizes the average distortion
• Last two approaches make use of a large training
  set of channels, generated according to some
  statistics

25
Quantized Feedback Extensions

• MIMO-OFDM systems
• Correlation between carriers can be exploited to
  reduce feedback and/or improve performance
• Entropy coding
• Clustering
• Finite-state vector quantization
• Time-varying MIMO systems
• Similar methods can be used to exploit the time
  correlation of the channel to reduce feedback
  and/or improve performance

26
Feedback for Multi-User MIMO

• In this case feedback is also used for user
  scheduling
• Let us consider the single-antenna users case

User 1
User 2
beamformer
User 3
Low-rate feedback links
27
Feedback for Multi-User MIMO

• Basic scheme opportunistic beamforming (OBF)
• The base station broadcasts a random beam
• Every user estimates its received SNR
• This received SNR is fed back to the base station
  
• Base station selects the user with the highest
  SNR
• Extensions
• OBF with beam selection (OBF-S)
• Opportunistic SDMA (OSDMA)
• OSDMA with beam selection (OSDMA-S)
• Fairness and delay play an important role here
• Difficult to exploit frequency- and
  time-correlation

28
Feedback for Multi-User MIMO

• Alternatively, feedback of channel information
  allows exploitation of frequency- and
  time-correlation, as well as specific spatial
  correlation patterns
• Every set of users is
  related to a quantized channel matrix
  and a beamformer
  (ZF or MMSE)
• Take the set that maximizes the sum rate
• Codebook design is the same as before

29
Feedback of Multi-User MIMO

• Frequency- and time-correlation can for instance
  be exploited by predictive vector quantization

2-antenna base station 2 users treated per slot 3
bits feedback per user
30
Ultra Wideband Communications
31
UWB Drivers

• Demand for short-range high-rate wireless
  capability
• Smaller semiconductor costs and power consumption
• Fragmented spectrum and discontinuous use of bands

32
Key Features of UWB

• High rate for short range
• Low-complexity and low-cost equipment
• Low transmit power and noise-like spectrum
• Multipath and interference immunity
• High penetration capability
• Accurate positioning
• Use of radio as a sensor (radar features)

33
IEEE Standardization

• IEEE 802.15.3a
• High-rate
• Not restricted to UWB but lends itself to it
• 100 Mbps within 10 m and 480 Mbps within 2 m
• Activities stopped in February 2006
• IEEE 802.15.4a
• Low-rate / low-complexity
• Operate in unlicensed bands
• Focus on WPAN, sensor networks, smart badges,
• Standard is being finalized

34
Generic Pulsed UWB Receiver
35
Subsampling UWB
PPM bits
n(t)
...
...
PPM
h
(t)
spr.
mod
c
t
0
t
PAM
...
...
mod
...
...
t
t
0
0
t
t
PAM bits
FFT
ADC
est t
PPM bits
equal.
DC
despr.
k
est c
PAM bits
k
Rx analog
Rx digital
36
Subsampling UWB

• PAM

PRR 20 MHz
Sample rate
78 MHz
156 MHz
313 MHz
625 MHz
1.25 GHz
37
Subsampling UWB

• PPM

PRR 20 MHz
Modulation index
? 2 ns
? 4 ns
? 8 ns
? 16 ns
Sample rate
156 MHz
38
Transmitted Reference UWB
n(t)
reference pulse
...
...
delay
spr.
h
(t)
c
t
0
t
PAM
...
...
mod
...
...
t
t
0
0
t
t
PAM bits
ADC
PAM bits
despr.
equal.
DC
delay
Rx analog
Rx digital
39
Transmitted Reference UWB
40
Transmitted Reference UWB
PRR 20 MHz
AWGN
CM1
Sample rate
20 MHz
41
UWB Testbed
AWG
RS232
PC
ADC EVALUATION BOARD
USB CABLE
FPGA BOARD
42
Cognitive Radio
43
Introduction

• Current wireless systems are characterized by
  wasteful static spectrum allocation
• Dynamic spectrum allocation (DSA) shows promises
  to alleviate the inefficient usage of the
  spectrum
• Frequency-agile cognitive radios (CRs) are key to
  this

44
Introduction

• The term cognitive radio was first coined by
  Mitola in 1999 and can be defined as in 2006 by
  IEEE A type of radio that can sense and
  autonomously reason about its environment and
  adapt accordingly. This radio could employ
  machine learning mechanisms in establishing,
  conducting or concluding communication and
  networking functions with other radios
• Two CR-related standards are under development
• IEEE 802.22 high rate access (1.5 Mb/s) in rural
  areas up to 100 km in coverage
• IEEE 802.11h WLANs with dynamic frequency
  selection transmit power control capabilities

45
Considered Set-Up

• A peer-to-peer CR network where each user
  corresponds to a single transmitter-receiver pair
• On top of that there is interference from primary
  users

46
How does it work?

• CRs dynamically decide the allocation of the
  available resources to improve the network-wide
  spectrum efficiency, a.k.a. dynamic resource
  allocation (DRA)
• The DRA task can be efficiently performed in a
  distributed fashion where every CR iteratively
  senses the available resources, and adjusts its
  own usage accordingly
• The resources can be represented by transmitter
  and receiver basis functions (carriers, pulses,
  codes, wavelets, etc.) which can be chosen to
  enable various agile platforms, such as
  frequency-, time-, or code-division multiplexing
  (FDM, TDM, CDM)

47
How does it work?

• Sensing part
• Sensing its own link is done by training
  techniques
• Sensing the interference is difficult due to the
  large number of possible resources, but since the
  actual number of used resources is small,
  compressive sampling mechanisms can be used
• Adapting part
• Given its own link and the interference, the CR
  optimizes its spectral efficiency under certain
  power and spectral mask constraints

48
Some Results

• Assume ideal case with carriers as waveforms

Large interference
Small interference
49
Discussion and Extensions

• Generally, DRA is done independently from
  waveform optimization, but this has a number of
  cons
• DRA has to run on a central level
• If distributed DRA is used, every CR requires the
  knowledge of the links to the other CRs and the
  decisions taken at the other CRs
• Sparsity constraints can be included in the
  optimization to limit the actual number of used
  resources
• Band-limited feedback is required from the
  receiver to the transmitter, which can be taken
  into account in the optimization procedure

50
Comments? Questions?