DIGITAL BEAMFORMING (DBF)
-demand for increased capacity is a major driving force for incorporating DBF
-marriage between antenna technology and digital technology
-3 major components: antenna array, digital transceiver, digital signal processor
-based on capturing the RF signals at each of the antenna elements and converting them into two streams of binary baseband signals (I & Q). Included in the digital baseband signal are the amplitude and phase of signals received at each of the elements of array. Beamforming is carried out by weighting these digital signals, thereby adjusting this amplitude and phases such that when added together they form the desired beam. This process can be carried out by special purpose digital signal processor
-Attractive features:
- A large number of independently steered high-gain beams can be formed without any resulting degradation in signal-to-noise ratio.
- All of the information arriving at the antenna array is accessible to the signal processors so that system performance can be optimized.
- Beams can be assigned to individual users, thereby assuring that all links operate with maximum gain
- Adaptive beamforming can be easily implemented to improve the system capacity by suppressing cochannel interference. Any algorithm that can be expressed in mathematical form can be implemeneted. As a byproduct, adaptive beamforming can be used to enhance the system immunity to multipath fading.
- DBF systems are capable of carrying out antenna system real-time calibration in the digital domain. Therefore, one can relax the requirements for a closematch of amplitude and phase between transceivers, because variation in these parameters can be corrected in real time.
- DBF has potential for providing a major advantage when used in satellite communications. If, after the launch of the satellite, it is found that the performance of the beamformer needs to be upgraded, a new suite of software can be telemetered up to the satellite. This means that the life of the satellite can be expanded by retrofits at various intervals, during which the satellite’s capabilities are upgraded.
Adaptive beamforming
-adaptive beamformer: device that is able to separate signals collocated in the frequency band but separated in the spatial domain, separating a desired signal from interfering signals
-algorithm based on maximization of SNR at the array output & least mean squares (LMS) errors
-Minimum-variance distortionless response (MVDR)
-Sample matrix inversion – fast adaptivity
Benefits of using adaptive antennas
- Coverage
-increase the cell coverage range substantially through antenna gain and interference rejection.
-Fewer sites required with adaptive antennas employed in base stations
-Larger coverage if antenna at greater height above average terrain. Can be eased by using the number of antenna elements
- Capacity
-It is possible to have multiple mobiles on the same RF channel but different spatial channels at a particular cell site
-allows a reuse factor of unity, that is a single frequency can be used in all cells.
-Can increase the number of available voice channels through directional communication links, depends on the propagation environment, the number of antenna elements and the amount of dynamic channel assignment allowed.
-Transmission bit rate can be increased due to the improved SIR at the output of the adaptive beamformer
-Allow RF channels to be adjusted through link power control to meet the requirements of user-selectable data transfer rates.
- Signal quality
-in noise-limited environment, minimum receiver thresholds are reduced by 10 logM dB on average.
-In interference-limited environments, the additional improvement in tolerable SIR at a single element results from interference rejection afforded against directional interferers.
-Can be considered as spatial equalizers
- Access technology
-In uplink, paths from different angles of arrival are separated by using a particular adaptive beamforming technique
-Downlink: energy can be focused at the mobile so that long delay multipath components can be reduced substantially
->combat ISI through spatial discrimination of “interfering” signals on both links
- Power control
-Eased thru the inclusion of adaptive antenna technology
- Handover
-antenna tech provides mobile unit location information that can be used by the system to substantially improve handoffs in both the low and high tiers. Accurate position estimates, prediction of velocities is possible.
- Base station transmit power
-the maximum peak EIRP required per user on a particular channel is decreased compared to without adaptive beamforming
- Portable terminal transmit power
-with adaptive antennas at cells, the transmit power levels from and to the mobile can be kept minimum to provide the requested service.
DIFFERENT TYPES OF ADAPTIVE BEAMFORMING
1. Adaptive beamforming for uplink
Reasons studies for uplink:
- traditionally used for radar, remote sensing and sonar reception system
- spatial channel information available on the uplink
Adaptive criteria
- optimum weights using different criteria are all given by the Wiener solution because it provides the upper limit on the theoretical adaptive beamforming steady-state performance
Adaptive Algorithms
- LMS algorithm: simple to implement , but limited in dynamic range over which it operates. Required power control or alternatively use normalized LMS algorithm
- SMI technique: fast convergence rate but increase computational complexity & numerical instability
- RLS algorithm: reduce computational complexity while maintaining similar performance, convergence rate faster than LMS provided that SNR is high, but has forgotten factor that is very dependent on fading rate of the channel
- Conjugate gradient method
- Eigenanalysis algorithm
- Rotational invariance based method
- Linear least squares error algorithm (LSSE)
- Hopfield neural network
Reference Signals
- If explicit reference signal available in communication it should be used as much as possible for less complexity, high accuracy, fast convergence
a)spatial reference:
- referred to as angle of arrival(AOA) information of desired signal and its multipath components
- AOA estimation techniques:
wavenumber estimation: based on decomposition of a covariance matrix whose terms consist of estimates of the correlation between the signals at the elements of an array antenna. Example: Multiple signal classification (MUSIC), modified forward-backward linear prediction (FBLP), Principal Eigenvector Gram-Schmidt (PEGS), Estimation of Signal Parameters by Rotational Invariance Techniques (ESPRIT)
parametric estimation: variety of maximum likelihood estimation (MLE) : particular likelihood function is formulated for the given radio signals, high computational complexity
- requirement for array calibration, extra processing load required for estimating AOA
b)temporal reference: may be a pilot signal that is correlated with the wanted signal, or known PN code in CDMA
Blind Adaptive Beamforming
- when explicit reference signal is not available
a)Constant Modulus Algorithm (CMA)
- For both compensating fading and canceling cochannel interference
- Applied to advanced mobile phone services ( AMPS), IS-54 signals, GMSK signals, 16-QAM signals
b)Decision-Directed Algorithm
- Low cost since not computationally intensive, and no array calibration required
- fast convergence, typically within 50 symbols
- locks on desired signal with probability of 99.9% at SIR levels as low as 1dB
- cochannel rejection is typically more than 20 dB
- implementation based on incoherent differential binary phase-shift keying (DBPSK) demodulation and LMS algorithm
- converge faster than CMA and SCORE
c)Cyclostationary Algorithm
- Developed and applied to AMPS
- AMPS exhibit cyclostationary properties due to presence of supervisory audio tone
- Show considerable improvement in MSE compared to the case of omnidirectional antennas
- Cylic beamforming can be applied to GMSK signals, only require that cochannel users have slightly different frequencies. Used in GSM and DECT. Shown capacity improvement
2. Adaptive Beamforming for Downlink
- Objective: to maximize the received signal strength at the desired mobile and to minimize the interference to other mobiles and adjacent base stations, thereby maximizing the downlink SINR
- If transfer function of the channel at the downlink is known, the downlink SINR can be maximized by multiplying the desired signal with a set of downlink weights.
- The weights are a scale version of the uplink weights, provided that the frequency of both links is same and the channel is relatively static during reception and transmission. Weight reuse can be applied to TDD systems ( CT2/CT2+, DECT, PHP, DCS1800).
- In FDD system (IS-54, IS-95, GSM), weight reuse cannot be used because far frequency separation
- Essence of the problem: to estimate the downlink transfer function
- Feedback technique was proposed. Using probing signal transmitted by BS. The mobile measure it own response to the probe signal and report them back to BS. The transfer function is estimated using the report. Simple but require complete redesign of protocols and signaling & applicable only for slowly change environments
- Other way: mobile directly transmit a narrowband testing signal at downlink frequency so that the BS can directly estimate the downlink channel transfer function from that. Not interrupt the normal uplink transmission but still require complete protocol redesign & require additional hardware in MS.
- Approached using AOA info. Downlink weights are derived by maximizing SIRN based on the same AOA.
- Use fixed multiple beams for both reception and transmission at the BS. On uplink, BS determine the direction of the path on which the strongest component of the desired signal arrived. On downlink, the BS points the beam in the corresponding direction. Not optimal, but SINR improved since narrowband signal is pointed, and can use high power beam to boost the SINR.
Ref:
John Litva, Titus Kwok-Yeung Lo, “ Digital Beamforming in Wireless Communications”, pg 157-184, Artech House Publisher, 1996
Ref: John Litva, Titus Kwok-Yeung Lo, “ Digital beamforming In Wireless Communications”, Artech House, London, 1996
Receive beamforming concept
-radiation pattern should match the energy profile in order to merge all the radiated power
-In wide angular spread case, pointing to a specific direction with narrow beam pattern is not optimal because some part of power spills over
Transmit beamforming
-It is suitable to transmit pointing towards the most significant reflector in order to minimize the interference between different users located at different angles.
-Suitable for narrowband transmission but not for MC-CDMA schemes where many carriers need to be considered.
Ref: Santiago Zazo, Ivana Raos, “Transmit Beamforming Design in Wide Angle Spread Scenarios for B3G MC-CDMA Systems”, IEEE Workshop on Signal Processing Advances in Wireless Communications, 2004
BLAST
- each antenna transmit an independently modulated signal simultaneously and on the same carrier frequency
-minimize redundancy between the various antenna signals in order to favor maximum data rate
Space-Time Coding
-introduce a lot of redundancy in an effort to maximize the diversity gain and achieve a minimum bit error rate
Space selectivity
-occurs when the received signal amplitude depends on the spatial location of the antenna, and is a function of the spread of angles of departure of the multipaths from the transmitter, and the spread of angles of arrival of multipaths at the receiver
General principle of LA is to:
-define a channel quality indicator, or so-called channel state information (CSI), that provides some knowledge on the channel. Metrics used as CSI : SNR & SINR(available from physical layer), PER & BER (Link layer)
-adjust a number of signal transmission parameters to the variations of that quality indicator over the signaling dimension explored (time, freq, space or combination thereof)
Adaptation based on Mean SNR
- Measure SNR at receiver (assessment of CSI)
- Convert the SNR info into BER info for each mode candidate (computation of adaptation thresholds, the minimum required SNR for a given mode to operate at a given target BER)
- Based on target BER, select for each SNR measurement the mode that yields the largest throughput while remaining within the BER target bounds (selection of the optimal mode)
- Feedback the selected mode to the transmitter
-this assuming ideal conditions (SNR can be measured instantaneously, ideal coherent detection, fading over time only, SNR measured in very short window so it is effectively nonfading)
-in practice, feedback delays and other implementation limitations will not allow instantaneous mode adaptation. Conversion of SNR to BER is not simple because the channel may exhibit some fading within the SNR window. > use of second and higher order statistics of SNR instead of Mean.
Adaptation based on Multiple Statistics of Received SNR
-If multicarrier modulation is used, a two dimensional time- frequency window may be used.
-The mapping between SNR and average BER is determined using pdf of the SNR over that window
-In physical channel this pdf cannot be obtained via simple analysis because it is a function of many parameters
-It can be simplify by estimating limited statistical info such as the k-order moment over the adaptation window, instead trying to estimate the full pdf
-Moment based CSI > simplicity and flexibility to LA algorithm, do not depends on any assumption made on the number or transmit and receive antenna
“How effective these methods can be in realistic traffic and bandwidth constraints is an open research problem. In particular, it is critical to measure the ability of the scheme to lend itself to a very fast adaptation scenario without significant bandwidth loss”
Pros and cons of CSI
-SNR based : offer flexibility to adapt the modes on a very fast basis; however, it relies on the computation and adaptation/switching thresholds that maybe inaccurate
-Error-based : captures accurate information of the modes, however this accuracy is reach only after a substantial amount of traffic observed.
“An important topic of current research is to combine all types of CSI together to yield both accuracy and robustness over a wide range of channels, adaptation rates, and traffic conditions”
In multiple antenna system the SNR varies not only over time and frequencies but also depends on:
-the way the transmitting signals are mapped and weighed onto the transmit antennas
-the processing techniques used at the receiver
-antenna polarization and propagation
Space-time-frequency adaptation> the adaptation algorithm desired to be able to select the best way of combining antennas at all time (choose between space-time coding approach, BLAST or beamforming approach)
LA algorithm design challenges:
-determination of adaptation thresholds: picking the least amount of statistical information to be computed while still describing the essence of channel behavior
-adaptation rate: fast adaptation consumes higher bandwidth, trade-off btw performance gain and amount of resource allocated to control messages
Ref:
Adaptive algorithms for weight calculations in adaptive antenna arrays > determine the convergence rate and hardware complexity
1. Time domain processing
- Applebaum algorithm: applicable only when DOA of the desired signal is known beforehead
- Least Mean Square (LMS): has been widely used for tap coefficient adaptations of an adaptive processor in antenna array, but it causes signal acquisition and tracking problems due to its slow convergence in multipath fading channel
- Constant Modulus Algorithm (CMA): useful when the constant envelope of modulated signal is maintained.
- Direct Matrix Inversion (DMI): fast convergence, but computationally too complex and may cause numerical instability
- Recursive Least Square (RLS): achieve faster convergence than LMS, less computational than DMI
2. Spatial domain processing : focused on DOA estimation by spectral analysis in the space domain
- Discrete Fourier Transform (DFT)
- Maximum Entropy Method (MEM)
- Multiple Signal Classification (MUSIC)
- Estimation of Signal Parameters via Rotation Invariance Technique (ESPRIT)
Ref:
Jin Young Kim, Jae Hong Lee, “Performance of a Multicarrier DS/CDMA System with Adaptive Antenna Array in Nakagami Fading Channel”, IEEE Conference paper, 1998
Steering vector
-contains the responses of all elements of the array to a narrow-band source of unit power
-associated with each directional source.
-For array of identical elements, each component of this vector has unit magnitude
-The phase of its ith component is equal to the phase difference between signals induced on the ith element and the reference element due to the source associated with the steering vector
-Also known as space vector and array response vector
W= array weight vector = weights of the beamformer
X= array signal vector= signals induced on all elements
R= array correlation matrix = its element denote the correlation between various elements of the array
Si= steering vector associated with ith source with direction (Øi, θi)
It is useful to express R in terms of its eigenvalue and their associated eigenvectors
-eigenvalues can be divided by two sets when the environment consists of uncorrelated direction sources and uncorrelated white noise
-noise eigenvalues: eigenvalues in one set are of equal values and it does not depend upon directional sources and is equal to the variance of the white noise
-2nd set: signal eigenvalues: a function of the parameters of directional sources, their number is equal to the number of these sources. Each eigenvalue of this set is associated with a directional source, its value changes with the change in the source power, and bigger than those associated with white noise
-R of an array of L elements immersed in M directional sources and the white noise has M signal eigenvalues and L-M noise eigenvalues
-R can be represented in the form of spectral decomposition of R, matrix with eigenvalues as diagonal matrix and multiplied by their corresponding unit-norm eigenvectors
-
BEAMFORMING METHODS
1. Conventional beamformer
-simple, sometimes known as delay-and-sum beam former, with all its weights of equal magnitudes
-the phases are selected to steer the array in particular direction, known as the look direction
-has unity response in the look direction, that is, the mean output power of the processor due to a source in the look direction is the same as the source power
-in environment consisting only uncorrelated noise and no directional interferences, this beam former provides maximum SNR