OFDM Systems for High Rate Wireless Communications

1
OFDM Systems for High Rate Wireless
Communications

• SUMIT ROY
• roy_at_ee.washington.edu
• University of Washington Intel
  Labs
• Seattle, WA
  Hillsboro, OR
• Feb. 11, 2003

2
Outline

• High Rate Wireless Communications
• System Design Challenges
• Why OFDM?
• Key OFDM Aspects
• - Peak-to-Average Power Ratio
• - Channel Estimation
• - Carrier Frequency Offset

3
Why high rate wireless commun.

• Current wireless cellular standards
• 2G voice rudimentary data services (low rate,
  10 Kbps)
• 3G voice high rate data services
• Stationary Users 2Mbps
• Pedestrian 384Kbps
• In Vehicle 144Kbps
• Current Wireless LAN standards
• .11b (2.4 GHz) - 11 Mbps
• .11a (5 GHz) - 54 Mbps (shared)
  
• Future Multimedia Services require still higher
  rates
• Audio/Video Streaming (multiple
  simultaneous flows)

4
Challenges in high rate wireless

• Multipath fading channels
• Time Dispersive a different transmitted symbols
  overlap
• Time varying (received signal amplitude varies)
• To combat dispersive channel
• ? Equalizer
• Complexity increases with rate for single-tone
  modulaton

5
Basic idea of OFDM

• Parallel transmission using N ( 64) carriers or
  sub-channels
• Longer symbol duration for sub carriers
• Robust to dispersive channel
• Sub carriers (SC)
• Multiples of base freq.
• Orthogonal in time
• a FFT based implementation
• Overlapped in freq.
• a spectral efficiency

spectrum
6
Orthogonal Freq. Division Multiplexing

• Disadvantages
• Channel and System imperfections that destroy the
  orthogonality of the sub-carriers
• 1. Dispersive Channel (OFDM uses Cyclic
  Prefix
• to compensate for this)
• 2. Carrier Frequency Offset (Needs estimator
  to compensate)
• High Peak-to-Average Power Ratio (due to
  simultaneous transmission over parallel channels)

7
Impact of Multipath

• Multi-path delay spread
• Time spread between the arrival of the first and
  last multipath signal, seen by the receiver.
• Received radio signal consisting of a direct
  signal, plus reflections from objects
• Multi-path delay spread effect
• Inter-Symbol Interference (ISI) when the delayed
  multipath signal overlaps with the symbols
  following it

8
Completely Eliminating ISI Cyclic Prefix

• Cyclic Prefix
• Add the last part of the packet to the beginning
  of the signal
• Duration of the CP larger than multipath delay
  spread
• Orthogonality of the carriers not affected.

9
Typical OFDM diagram and application

• Digital broadcasting DAB, DVB-T
• High speed WLAN IEEE 802.11a/ HiperLAN2
• High speed fixed wireless IEEE 802.16
• ADSL
• Considered promising candidate for 4G

10
Critical issues in OFDM for high rate

• Channel estimation
• - Preamble based
• Timing and Sub-Carrier (SC) Synchronization
• - Multi-carrier Systems more susceptible to
• frequency offset errors
• High Peak-to-Average Power Ratio (PAPR)
• - Due to simultaneous transmission of many
  carriers many techniques to reduce PAPR (e.g.
  coding, clipping, spectral shaping..)

11
Channel estimation (CE)

• Distortion over SC due to dispersive channels
• CE indispensable for
• Complex but spectral efficient signal
  constellations, QAM etc.
• Adaptive modulation over SCs
• Intermediate measure to obtain higher spectral
  efficiency

12
Cyclic Prefix (CP)

• Dispersive channel leads to ISI destroys SC
  orthogonality
• Solution CP insertion
• Constrain ISI in the samples corresponding to CP
• SC orthogonality preserves after CP removal
• Simple channel effect in freq. domain
• Can use per-tone estimation/equalization
• Reduces utilization (overhead)

CP insertion
13
CFO
No CFO
CFO in presence

• Due to base carrier frequency difference between
  Xmit and Rcvr
• Leading to serious inter-channel-interference
  (ICI)
• Must be estimated and compensated - .11 uses
  specific preambles to achieve this

14
PAPR problem

• OFDM symbols may have high peak-to-average power
  ratio
• May cause saturation in Power Amplifier in Xmit
• Counter measure
• High dynamic range PA
• Solutions
• Coding and/or Signal Processing Methods (clipping)

15
Training based CE (1/2)

• CE based on reception of special training symbols
  (a.k.a. pilots)
• Simple for systems with sufficient CP (CPgtchannel
  delay spread)
• CE problem estimate Gi, given yi(k) and si(k)
• LS method
• MMSE method better CE than LS by exploiting
  correlation of Gi ( )
• Lower complexity MMSE CE by approximating
  with its principle eigenvalues Edfors1998

16
Training based CE (2/2)

• For burst transmission (WLAN)
• Quasi-static channel
• One shot CE

• For continuous transmission (DAB, DVB-T)
• Time varying channel
• Sparse pilots distributed over frequency and time
• Interpolation uses channel correlation in both
  frequency and time

Pilot pattern for burst transmission
Pilot pattern for continuous transmission
17
Training based Vs. Blind

• Blind methods
• Problem formulation given channel outputs,
  estimate the channel
• WITHOUT ANY TRAINING SEQUENCES
• Made possible by exploiting redundant information
  
• Higher Complexity !!

18
Sources of redundancy

• CP
• Repetition in transmitted signal
• Virtual carriers (VCs)
• Ease implement of spectrum-limiting filter
• Provide extra degree of freedom than CP
• Example IEEE 802.11a, 12 VCs in 64 SCs
• Receiver diversity
• Multiple receive antennas or over-sampling

19
Exploiting redundancy in 802.11
20
Blind CE using CP (1/2)

• Previously reported blind CEs
• Time domain methods
• Applied on channel outputs before FFT
• Cyclic spectra method Heath etal1999
• CP a Repetition a cyclostationarity in
  transmitted and received signals
• Cyclic spectra imposes constraints on channel
  impulse response
• Eigen-structure method Cai2000
• CP a MIMO system (over-determined)
• Extracted noise subspace
• Special structure of filtering matrix A

a constraints on channel
21
Blind CE using CP (2/2)

• Rethinking about blind CE upon CP
• CP incurs notable channel utilization loss
• CP 25 of effective OFDM symbol duration
• Example in 802.11a, CP16, SC64
• Can we improve the channel utilization? And how?

22
Blind CE using other redundancy (1/2)

• Improve channel utilization by
• Reducing CP (CPltmax delay spread)
• Previous two CEs are applicable other method
  possible?
• After CE
• Channel shortening filter followed by
  conventional one-tap equalizer
• MMSE equalizer
• Eliminating CP
• High channel utilization
• New low complexity equalizer Trautmann2002
• W/o CP, previous CEs wont work

23
Blind CE using other redundancy (2/2)

• Two novel subspace based CEs
• By exploiting VCs and CP Li2001
• Includes Cai2000 as a special case
• By exploiting receiver diversity Roy Li2000
• Multiple receive antennas or over-sampling
• Features and theoretical significance
• Eigen-structure method
• Applicable to OFDM system w/o CP or w/
  insufficient CP catering for high rate
  communications
• Gives channel identifiability condition

24
Subspace based CE upon VCs
System diagram

• Q sub carriers D-length CP
• P data carriers, or, Q-P VCs

25
CFO estimation Training based (1/2)

• Simple idea
• Sending two identical symbols
• Due to CFO e, get
• CFO can be estimated by phase(y(2)/y(1))

26
CFO estimation Training based (2/2)

• Representative methods
• Moose (Moose1994) Frequency domain
• Two identical OFDM symbols
• CFO capture range lt ½ carrier spacing
• Schmidl (Schmidl1997) Time domain
• One symbol consisting of two identical halves (in
  time)
• Larger capture range
• All assume sufficient CP

27
Blind CFO estimation

• Higher channel utilization
• No training symbols needed
• Esp. efficient for CFO tracking
• Blind CFO estimators
• Sufficient CP assumed
• Exploiting CP introduced repetition
  (deBeek1998)
• MUSIC-like method by exploiting VC (Liu1998)

28
(No Transcript)