Unlicensed Broadband Wireless Systems: Adaptive Antenna Arrays for an Uncontrolled Interference Envi

1
Unlicensed Broadband Wireless Systems Adaptive
Antenna Arrays for an Uncontrolled Interference
Environment

• Van Sreng (vsreng_at_sce.carleton.ca)
• Fayyaz Siddiqui (fasiddiq_at_sce.carleton.ca)
• Prof. David Falconer Prof. Florence
  Danilo-Lemoine
• ddf_at_sce.carleton.ca fdanilo_at_sce.carleto
  n.ca

2
Outline

• Motivation
• Channel Model
• Simulation Scenarios Observations
• Interference Distribution Model.
• Depiction of Training Schemes.
• Interference Traffic Distribution Model.
• Channel and Interference Model for an OFDM
  system.
• Conclusions
• Future Direction

3
Motivation

• Problem Source
• Uncontrolled interference environment due to the
  absence of control coordination in unlicensed
  band systems.
• Unpredictable interference traffic pattern.
• Time and Frequency selectivity of the channel.
• High performance loss for any individual entity
  due to the above.
• Proposed Solution
• Using array processing to null out as many
  interferers as possible.
• Using OFDM along with array processing to combat
  multipath interference effects of the channel.

4
Composite Channel Model

• Long-term Effects
• Pathloss due to distance pathloss exponent 4.
• Lognormal Shadowing (both uncorrelated
  correlated) s 10 dB.
• Short-term Effects
• Flat Rayleigh fading.
• The channel coefficient hlk is a zero-mean
  complex gaussian random variable such that
• is a gaussian random variable with mean
• and standard deviation
• Frequency selective Rayleigh fading.
• Receiver Structure (see Fig. 1)

Fig. 1 Using smart antenna arrays at the receiver
(Rx) to combat against interference. The
receiver is equipped with multiple antenna
elements linearly uniformly spaced with
inter-element spacing of at least l/2 (the h
coefficients are a composite of loss due to
distance attenuation, lognormal shadowing, and
Rayleigh fading).
5
Simulation Scenarios

• Estimation Techniques based on Training

• Training Schemes
• Post-amble Training (Post)
• Pre-amble Training (Pre)
• Distributed Training (Dist)
• Tracking using Distributed Pre-amble
  (Dist-Pre)
• Tracking using Pre-amble Post-amble (Pre-Post)

• The channel auto-correlation matrix is given by,

..
The ith received symbol vector at the array
input.
Complex conjugate-transpose of r(i).
..
N Total number of training symbols used
14 symbols

• The estimate of desired channel propagation
  vector is given

Complex conjugate-transpose of ith known
training symbol.

• From the above, the estimated weights are
  obtained
• as follows

..
(3)

• In tracking, weights found with either of the
  training technique are used simultaneously on
  each symbol and the better one is selected.

..
..
Fig. 2 Various training schemes for channel
interference estimation.
6
Simulation Scenarios (Cont.)

• Continuous Interference Traffic
• Uniformly distributed interferers in a ring shape
  of radius between 1-0.1 km (Fig. 3).
• Uncorrelated Lognormal Shadowing.

• Interferers uniformly distributed in a sector
  confined to between A B (see Fig. 4).
• Correlated Lognormal Shadowing, where the
  correlation among users is a function of both the
  distance ratio and the angle of arrival (AOA)
  difference (all w.r.t Rx)
• (4)
•   (5)

Fig. 5 Correlation coefficient as a function of
AOA difference and distance ratio between desired
user and interferer the smaller the AOA
difference, the stronger the correlation (K in
(5) point of lowest distance-dependent
correlation value (20 dB), 60 in (4) point of
lowest AOA difference-dependent correlation
value).
Fig. 4 Users configuration when correlated
lognormal shadowing is considered (users are
confined to a small sector between A B, which
defines the max. AOA difference between users).
Fig. 3 Uniformly distributed interferers in a
ring shape of radius (B-E), B1 km, E0.1 km.
7
Performance (BER) Under Continuous Interference
Traffic

• Under Uncorrelated Lognormal Shadowing (Figs.
  6-7)
• Observations
• No great performance loss from using estimated
  weights compared to optimum weights.
• As the distance between the desired Tx Rx
  increases, the BER degrades as expected.
• BER performance is reasonably good provided the
  of antenna elements is greater than the of
  interferers.

Fig. 6 Uncorrelated lognormal interferers ( of
interferers4, of antenna elements 6).
Fig. 7 Correlated lognormal interferers ( of
interferers10, of antenna elements 6).
8
Performance (BER) Under Continuous Interference
Traffic (Cont.)

• Under Correlated Lognormal Shadowing (Figs. 8-9)
• Observations
• At a high correlation (above 0.9) in lognormal
  shadowing among users, the BER performance is
  worse than the uncorrelated case.
• While, at a low correlation among the
  interferers, but still some correlation between
  the desired user and the interferers, the
  performance is actually better than the
  uncorrelated case.

Fig. 8 Correlated lognormal interferers ( of
interferers10, of antenna elements 6).
Fig. 9 BER vs. Correlation among correlated
lognormal users based on optimal combining
technique ( of interferers 10, of antenna
elements6) .
9
Intermittent Interference Traffic

• Why intermittent Traffic?
• Most of the current and planned wireless systems
  are block based.
• Which kind of environment we are going to
  face.License Exempted !
• Interfering users get on and off the network at
  random times.
• Hence interferers arrive in an asynchronous
  fashion and will not be present for the whole
  duration of the desired users block, Part-Time
  interference.
• A Traffic Model required for simulating such
  environment.

10
Intermittent Interference Traffic (contd)

• Pictorial depiction..(yellow?desired user,
  red?an interferer)
• Why Pre-amble, Post-amble or mid-amble may fail?

11
Intermittent Interference Traffic (contd)

• Traffic Model
• Interference traffic follows a Batch Poisson
  Traffic with a certain , each interferers
  block/frame length is generated using the
  probability distribution given in table 1.
• Desired users block length is fixed (162
  Symbols)
• Traffic load at any instant can be calculated
  as Average duty cycle (defined below) multiplied
  by the maximum possible potential users at that
  specific instant. This will give the average
  number of actual users at any instant.
• Different performance measures are made against
  Offered traffic load in Erlangs, which can be
  defined as
• Erlangs Avg. Duty Cycle Maximum Number of
  interferers

table 1
12
Intermittent Interference Traffic (contd)

• Snap-Shot From Simulation

13
Performance (BER) Under Intermittent Interference
Traffic

• BER Performance under Equal Avg. received powers
• All training schemes are compared with same
  number of training symbols. (14 symbols).
• Observations
• As long as the number of antenna elements is
  greater than 2 to 3 times the traffic load in the
  network, one can achieve a BER performance of
  10-3 with distributed training scheme.
• The best choice for combating a traffic with
  part-time interferers would be the distributed
  training scheme.

14
Performance (BER) Under Intermittent Interference
Traffic (contd)

• Effect of Path Loss Log. Normal Shadowing
  (Users Spatially Uniformly Distributed)
• Distributed Training Only
• Observations
• As the distance between the desired Tx Rx
  increases, the BER degrades as expected.
• Difference is very small in case of 6 10
  antenna elements for larger distances. As in both
  cases same number of training symbols are used.

15
Performance (Outage prob.) Under Intermittent
Interference Traffic (contd)

• Outage Probability (another Performance measure)
• Probability that the instantaneous symbol error
  probability exceeds a specific threshold
• Observations
• Pre-Post Dist-Pre training Schemes are compared
  with Optimum case. The threshold is set for
  MMSEgt-10dB.
• Larger number of antenna elements results in a
  lower outage probability.
• With distributed training, no outage occurs below
  a traffic load of 2.8 Erlangs with 8 antenna
  elements.

16
Performance (Effect of Block Length) Under
Intermittent Interference Traffic (contd)

• Effect of Variable Block/Frame Length (training
  symbols fixed)
• Objective is to make the block length adaptive to
  achieve optimum performance in the presence of
  intermittent interference traffic.
• Factors to be considered efficiency, throughput,
  and acceptable output SINR.
• With certain number of training symbols
  distributed in the frame, information can be
  extracted about the tradeoff between efficiency
  and the overhead.

17
Performance of Arrays Under Multipath Channel and
Intermittent Interference Traffic

• In an unlicensed band, such as U-NII, high data
  rates up to 20 Mb/s are expected.
• Under a multipath frequency selective
  environment, ISI may span 50 to 100 symbols.
• Question arises How to deal with such Frequency
  Selective channels?
• Modulation Alternatives
• Single carrier modulation ---Rx equalization in
  Time domain.
• OFDM
• Single carrier modulation ---Rx equalization in
  Frequency domain.
• Every scheme has its own advantages
  disadvantages.
• OFDM has already been chosen as the foundation of
  wireless LAN.

18
Performance of arrays in multipath channel Under
Intermittent interference Traffic (contd)

• The Objective is to merge OFDM with existing
  Antenna Arrays Interference Cancellation
  Techniques, especially in an asynchronous or
  sparse interference environment.

19
Performance of arrays in multipath channel Under
Intermittent interference Traffic (contd)

• Multipath Model Used
• SUI-5 Channel..Power-Delay Profile

20
Performance of arrays in multipath channel Under
Intermittent interference Traffic (contd)

• Simplified Block Diagram for OFDM system

• Different Pilot (training symbols) placement
• Pilot carriers distributed in fixed fashion among
  info carriers (4/64).
• Preamble (One OFDM symbol with known training
  data).
• Distributed training symbols in a specific way
  known to the receiver (Hopping in a pseudo random
  pattern may be possible desirable in dealing
  with sparse interference).

21
Performance of arrays in multipath channel Under
Intermittent interference Traffic
(contd)

• Channel Estimation
• (Time Domain Interpolation)
• With Pilot Scheme 1
• Let Hp as the frequency response of the channel
  at the pilot sub carriers such that
• IFFT then Padding
• FFT

22
Performance of arrays in multipath channel Under
Intermittent interference Traffic (contd)

• Effect on MMSE with increase in antenna elements
  (using interpolation between the pilot carriers)
• By using more elements and interpolated weights
  on the carriers without pilot, improvement in
  output SINR is observed (synchronous CCI).
• Poor performance due to Pilot distribution scheme
  (not a good estimation)

23
Performance of arrays in multipath channel Under
Intermittent interference Traffic (contd)

• Asynchronous Interference and training symbols
  placement effect. (Preamble kind of training)
• Observation
• With asynchronous transmission as expected in
  unlicensed band, effect on the BER with CCI
  caught by the training is close to the optimum
  one.

24
Conclusions

• Continuous Interference Traffic
• The performance based on the estimated weights is
  near optimum.
• Under an interference-limited system (high SNRs),
  the BER is reasonably good (below 10-3) provided
  the number of antenna elements is greater than
  the number of interferers.
• The BER increases at a very high correlation in
  lognormal shadowing among users.
• Intermittent Interference Traffic
• Performance of the estimated weights deteriorates
  compared to the optimum weights. The more bursty
  the interference traffic is, the worse the BER
  becomes.
• Distributed training offers the best performance
  among the three basic training schemes. The
  reason behind is that to null out part time
  interferers by arrays, the training symbols (at
  least some of them) must lie in the region of the
  packets affected portion.
• As long as the number of antenna elements is
  greater than 2 to 3 times
• the traffic load in the network, one can
  achieve a BER performance of
• 10-3 with distributed training scheme.
• We can find an optimum length for the block
  depending upon current traffic load to get an
  optimum performance with best possible
  efficiency.
• In Non-adaptive OFDM systems, to compensate the
  severely attenuated sub-carriers, interleaving
  and coding are necessary.

25
Future Direction

• Using adaptive modulation for an OFDM system in
  order to overcome the deep fading effects in
  certain sub-carriers.
• An overall system analysis using other
  performance measures such as the overall channel
  capacity (defined in bits/s/Hz/m2) and
  throughput.
• Interference averaging technique such as use of
  spreading gain in direct-sequence CDMA or
  frequency hopping in OFDM to further mitigate the
  interference effects.
• When array processing alone is not adequate (this
  might be the case when the of interferers is
  one or more greater than the of antenna
  elements), we have to decide some etiquettes to
  deal with the situation (like in UPCS.LBT)

26
References

• Y. Bresler, et al., Optimum Beamforming for
  Coherent Signal and Interferences, IEEE
  Transactions on Acoustics, Speech, and Signal
  Processing, Vol. 36, pp. 833-842, June 1988.
• D. Falconer, Spatial-Temporal Signal Processing
  for Broadband Wireless Systems, Chapter in
  Wireless Comm. in the 21st Century, to be
  published by IEEE Press, 2002.
• L. Godara, Applications of Antenna Arrays to
  Mobile Communications, Part II Beam-Forming and
  Direction-of-Arrival Considerations, Proceedings
  of The IEEE, Vol. 85, pp. 1193-1239, Aug. 1997.
• T. Rappaport, Wireless Communications Principles
  and Practice, Prentice Hall PTR, 1996.
• T. Klingenbrunn, P. Mogensen, Modelling
  Cross-Correlated Shadowing in Network
  Simulations, IEEE VTC, pp. 1407-1411, 1999.
• Y. Bresler, et al., On the Number of Signals
  Resolvable by a Uniform Linear Array, IEEE
  Transactions on Acoustics, Speech, and Signal
  Processing, pp. 1361-1375, 1986.
• V. U. Reddy, et al., Performance Analysis of the
  Optimum Beamformer in the Presence of Correlated
  Sources and Its Behavior Under Spatial Smooting,
  IEEE Transactions on Acoustics, Speech, and
  Signal Processing, pp. 927-936, 1987.
• J. H. Winters, Optimum Combining in Digital
  Mobile Radio with Cochannel Interference, IEEE
  Journal on Selected Areas in Communications, pp.
  528-539, 1984.
• A. Haimovich and A. Shah, Reduced-Rank Array
  Processing for Wireless Communications with
  Applications to IS-54/IS-136, IEEE Transactions
  on Communications, pp. 743-747, 2002.
• A. Haimovich and A. Shah, The Performance of
  Space-Time Processing for Suppressing Narrowband
  Interference in CDMA Communications, Wireless
  Personal Communications, pp. 233-255, 1998.
• I. Reed, et al., Rapid Convergence Rate in
  Adaptive Arrays, IEEE Transactions on Aerospace
  and Electronic Systems, pp. 853-863, 1974.
• Document IEEE 802.14/96-08 3R2.
• Lettieri, P. Srivastava, Adaptive frame length
  control for improving wireless link throughput,
  range, and energy efficiency, M.B.INFOCOM '98.
  Seventeenth Annual Joint Conference of the IEEE
  Computer and Communications Societies.
  Proceedings. IEEE, pp. 564-571, vol.2, 1998.

27
QUESTIONS!