Detailed Description
The word "exemplary" is used herein to mean "serving as an example, instance, or illustration. Any embodiment or design described herein as "exemplary" is not necessarily to be construed as preferred or advantageous over other embodiments or designs.
Fig. 1 illustrates an exemplary subband structure 100 that may be used in an OFDM system. The overall system bandwidth of the OFDM system is BW MHz, which is divided into N orthogonal subbands using OFDM. The bandwidth of each sub-band is BW/NMHz. In a spectrally shaped OFDM system, only M of the total N subbands are used for data/pilot transmission, where M < N. The remaining N-M subbands are not used for data/pilot transmission, but rather are used as guard subbands to enable the OFDM system to meet spectral mask requirements. The M usable subbands include subband F through subband F + M-1, which are typically centered on all N subbands.
The N subbands of an OFDM system may experience different channel conditions (e.g., different fading and multipath effects) and may be associated with different complex channel gains. Accurate estimates of the channel response typically require processing (e.g., demodulation and decoding) of the data at the receiver.
Wireless channels in OFDM systems may be responded to by time domain channel impactshN×1Or corresponding frequency domain channel frequency responseHN×1And (5) characterizing. As used herein, the term "channel impulse response" is the time domain response of a channel, and the "channel frequency response" is the frequency domain response of the channel, consistent with conventional terminology. Channel frequency responseHN×1Is the channel impulse responsehN×1Discrete Fourier Transform (DFT). This relationship can be expressed in matrix form as follows:
HN×1＝WN×NhN×1equation (1)
WhereinhN×1Is an N x 1 vector for wireless channel impulse response between a transmitter and a receiver in an OFDM system;
HN×1is an N x 1 vector for the wireless channel frequency response; and
WN×Nis made byTo pairhN×1Performing DFT to obtainHN×1N × N DFT matrix.
DFT matrixWN×NIs defined as given the (n, m) th term w according to the following equationn，m：
N-1 … N, m-1 … N, equation (2)
Where n is the row index and m is the column index.
The impulse response of a wireless channel may be characterized by L taps (tap), where L is typically much smaller than the number of full subbands (i.e., L < N). That is, if the transmitter applies an impulse to the wireless channel, then L time domain samples (at BW MHz sampling rate) will be sufficient to characterize the wireless channel response based on the impulse excitation. The number of taps (L) of the channel impulse response depends on the delay spread (spread) of the system, which is the time difference between the earliest and latest arriving signal instances with sufficient energy at the receiver. Longer delay spreads correspond to larger values of L and vice versa. Vector quantityhN×1Each tap of the channel impulse response includes a non-zero term. Vector for delay spread of LhN×1The first L terms of (a) may include non-zero values, with the remaining N-L terms all being zero.
Since the channel impulse response only requires L taps, the channel frequency responseHN×1Spread out in a subspace of dimension L (not N). Thus, the frequency response of the wireless channel may be adequately characterized based on channel gain estimates for as few as L appropriately selected subbands rather than all N subbands. Even can obtainBy estimating channel gain for more than L subbands, an improved estimate of the wireless channel frequency response may also be obtained by suppressing noise components outside the subspace.
Fig. 2 illustrates a pilot transmission scheme 200 that may be used to derive a frequency response estimate for a wireless channel in an OFDM system. A pilot symbol is transmitted on each of the P pilot subbands, where L ≦ P ≦ M in general. The pilot subbands are distributed among M usable subbands and are indexed by s1To sP. Typically, the number of pilot subbands is much smaller than the number of usable subbands (i.e., P < M). The remaining M-P usable subbands may be used for transmission of user-specific data, overhead data, and so on.
The model of an OFDM system can be expressed as:
rN×1＝HN×1οxN×1+nN×1equation (3)
WhereinxN×1Is an N x 1 vector with N "transmit" symbols sent by the transmitter on N subbands, with zeros sent on inactive subbands;
rN×1is an N x 1 vector of N "received" symbols with N subbands derived by the receiver;
nN×1is an N × 1 noise vector for N subbands; and
"o" represents the Hadmard product, which is the product of the element-wise, whererN×1The ith element of (a) isxN×1AndHN×1the product of the ith element of (a).
Supposition of noisenN×1Is a mean of zero and a variance of σ2Additive White Gaussian Noise (AWGN).
The radio channel frequency can be obtained by the following equationInitial estimation of response
Equation (4)
WhereinxP×1PIs a P x 1 vector with P pilot symbols sent on P pilot subbands;
rP×1Pis a P x 1 vector of P received pilot symbols with P pilot subbands;
HP×1Pis a P x 1 vector of actual frequency responses for the P pilot subbands;
is a P x 1 vector used for initial frequency response estimation;
nP×1Pis the P × 1 noise vector for P pilot subbands; and
whereinAnd P(s)i) Respectively, a pilot subband siAnd transmit pilot symbols.
P x 1 vectorxP×1P、rP×1PAndnP×1Peach containing only Nx 1 vectors corresponding to P pilot subbandsxN×1、rN×1AndnN×1p entries of (a). As shown in equation (4), the receiver may derive an initial frequency response estimate based on the P-element ratios of the received pilot symbols to the transmitted pilot symbols for the P pilot subbandsThat is to say that the first and second electrodes,whereinIs a pair of sub-bands siThe channel gain estimate of (2). Vector quantityRepresenting the wireless channel frequency response for the P pilot subbands.
Estimation based on initial frequency response may be performed using various techniquesFrequency response estimates for all N subbands are obtained. For the direct least squares estimation technique, a least squares estimate of the wireless channel impulse response is first obtained based on the following optimization:
equation (5)
WhereinhL×1Is an lx 1 vector of assumed impulse responses for the wireless channel;
WP×Lis thatWN×NP × L sub-matrices of (a); and
is an lx 1 vector used for least squares channel impulse response estimation.
Matrix arrayWP×LComprising a matrixWN×NCorresponding to P rows of P pilot subbands.WP×LEach row of (a) includes L elements, which areWN×NThe first L elements of the corresponding line. The optimization in equation (5) covers all possible channel impulse responseshL×1. Least squares channel impulse response estimationEqualing the assumed channel impulse responsehL×1，hL×1Resulting in an estimation of the initial frequency responseAnd correspond tohL×1Produces a minimum mean square error between the frequency responses of (1), whereinhL×1ByWP×LhL×1And (6) calculating.
The solution to the optimization problem proposed by equation (5) can be expressed as:
equation (6)
A frequency response estimate for the wireless channel may then be derived from the least squares channel impulse response estimate, as follows:
equation (7)
WhereinWN×LIs provided withWN×NThe first L columns of the nxl matrix; and
is an N x 1 vector used for frequency response estimation for all N subbands.
The vectors may be calculated in a variety of waysFor example, first, a vector is calculated as shown in equation (6)Then using the vectorVector calculation as shown in equation (7)For equation (6), (b)WP×LHWP×L)-1WP×LHIs an L × P matrix that can be pre-computed. Therefore, the impulse response estimation can be obtained by L.P times of complex operation (or multiplication)For equation (7), the L1 vector is expanded by (1)(by zero padding) to obtain an Nx 1 vectorAnd (2) pairPerforming an N-point FFT more efficiently computes a frequency response estimateThis requires 0.5N · logN complex operations, therefore, the frequency response estimate can be obtained by a total of (L · P +0.5N · logN) complex operations of equations (6) and (7)
Alternatively, by combining equations (6) and (7), the vector may be formedDirectly calculating the vectorThe following were used:
equation (8)
WhereinWN×L(WP×LHWP×L)-1WP×LHIs an N × P matrix that can be pre-computed. Thus, the frequency response estimate can be obtained by a total of N.P complex operations
For both calculation methods as described above, the result is for one OFDM symbolThe minimum number of complex operations required is Nop＝min{(L·P+0.5N · logN), N · P }. If pilot symbols are transmitted in each OFDM symbol, then the calculated rate is Nop/TsymMillion operations per second (Mops), NopBW/N Mops, where TsymIs the duration of one OFDM symbol, which is equal to N/BW μ sec (described below) when there is no cyclic prefix. Number of complex operations NopIt is large for OFDM systems with a large number of subbands. For example, for an OFDM system with a total bandwidth of BW 6MHz, N4096 total subbands, P512 pilot subbands, and L512 taps, the calculations are calculated using equations (6) and (7)420Mops is required. Since equation (6) requires 384Mops and equation (7) requires 36Mops, the computation of the least squares channel impulse response estimate in equation (6) is much more burdensome than the N-point FFT computation in equation (7).
The pilot transmission scheme of fig. 2 does not limit the locations of the pilot subbands. Matrix arrayWP×LComprising a matrixWN×NCorresponding to P rows of P pilot subbands. This results in the need for a vectorEach of the L terms of (a) performs P complex operations.
FIG. 3 illustrates that the estimation of the impulse response of the channel for least squares can be simplifiedThe calculated uniform pilot transmission scheme 300. For scheme 300, the P pilot subbands are uniformly distributed across all N subbands, such that adjacent pilot subbands are spaced N/P subbands. Further, assume that the number of taps is equal to the number of pilot subbands (i.e., L ═ P). In this case, it is preferable that the air conditioner,WP×Pis a P x P DFT matrix and,whereinIIs a unit matrix, andand equation (6) can be simplified as:
equation (9)
Equation (9) shows that the initial frequency response can be estimated byPerforming P-point IFFT to obtain channel impulse response estimationVector quantityMay be zero padded to length N. The zero-filled vector can then be implemented with an N-point FFTTransforming to obtain a vectorThe following were used:
equation (10)
Can also be based on vectorsDeriving an S1 vector for frequency response estimation for S subbands of interestWherein N.gtoreq.S.gtoreq.P is usually present. If S is a power of 2, an S-point FFT may be performed to obtain
Using pilot transmission scheme 300, the result for one OFDM symbolThe number of complex operations required is NopThe calculation rate was 0.5 BW · N · logN)/N Mops (P · logP + N · logN). For the exemplary OFDM system described above, the pilot transmission scheme 300 may be utilized to calculate at 39.38MopsThis is much less than 420Mops required for pilot transmission scheme 200.
As described above, the reduced complexity least squares channel impulse response estimation in equations (9) and (10) relies on two key assumptions:
p pilot subbands are periodically distributed over all N subbands, an
2. The number of taps is equal to the number of pilot subbands (i.e., L ═ P).
These two assumptions impose importance on the actual OFDM systemIs limited/restricted. First, for some OFDM systems, it may not be possible to transmit pilot symbols on P subbands that are uniformly distributed across all N subbands. For example, in a spectrally shaped OFDM system, no symbols are transmitted on guard subbands in order to meet the requirements of the spectral mask. As another example, an OFDM system may not allow pilot/data transmission on certain subbands (e.g., zero or DC subbands). As yet another example, pilots may not be available on certain subbands due to implementation of receiver filters and/or other reasons. For these systems, it is generally not possible to have the P pilot subbands achieve strict periodicity across all N subbands. Second, the assumption of L-P (which is less important than the first assumption) degrades the final channel frequency response estimateThe quality of (c). It can be seen that if (1) L is assumed to be equal to P, (2) the pilot symbol energy is equal to the data symbol energy, and (3) not correctOrPerforming time-domain filtering to obtain additional energy, the quality of the channel estimate may be reduced by about 3dB from the optimal channel estimate. This reduction in channel estimation quality may be unacceptable for some systems.
Various techniques may be used to overcome the two limitations described above. First, channel gain estimates for the P uniformly spaced subbands may be derived based on received pilot symbols using extrapolation and/or interpolation, as desired. Thus, a P-point IFFT can be used to derive a channel impulse response estimateSecond, can be aligned withPerforms tap selection to obtain a higher quality channel estimate. The extrapolation/interpolation and tap selection are described in detail below.
Fig. 4 shows a uniform pilot transmission scheme 400 for a spectrum shaping OFDM system. For scheme 400, similar to scheme 300, the P pilot subbands are uniformly distributed across all N subbands such that adjacent pilot subbands are spaced N/P subbands. However, pilot symbols are sent only on pilot subbands that are located between the M usable subbands (or simply "active pilot subbands"). No pilot symbols are transmitted on the pilot subbands that lie between the N-M guard subbands (or simply "inactive pilot subbands"). Thus, the receiver obtains pilot symbols for the active pilot subbands and does not obtain pilot symbols for the inactive pilot subbands.
FIG. 5 illustrates a method for deriving a frequency response estimate for a wireless channel in a spectrally shaped OFDM systemProcess 500. Initial frequency response estimates for the first set of P uniformly spaced subbands are derived based on, for example, pilot symbols received on the second set of subbands used for pilot transmission (block 512). The first group includes at least one subband not included in the second group (e.g., pilot subbands between guard subbands). An impulse response estimate for the wireless channel is then derived based on the initial frequency response estimate (block 514). The channel impulse response estimates for the multiple OFDM symbols may be filtered to obtain higher quality channel estimates (block 516). A final frequency response estimate for the wireless channel is then derived based on the channel impulse response estimate (filtered or unfiltered) (block 518). Filtering may also be performed on the initial or final frequency response estimate (rather than the channel impulse response estimate) to obtain a higher quality channel estimate.
FIG. 6 illustrates a method for deriving frequency response estimates in a spectrally shaped OFDM systemThe specific process 600. First, P with pilot transmission is obtainedactReceived pilot symbols for the active pilot subbands (block 610.) then, pairs P may be derived based on the received pilot symbolsactChannel gain estimation for individual active pilot subbands(block 612). The output of block 612 is for PactP of initial frequency response estimation of individual active pilot subbandsactX 1 vectorAs described below, extrapolation and/or interpolation is performed as necessary to obtain pairs PextChannel gain estimates for the subbands not undergoing pilot transmission (block 614). The output of block 614 is P for non-pilot transmissionextP of initial frequency response estimation of individual sub-bandsextX 1 vectorThen, based on the vector fromAndforming a P x 1 vector for initial frequency response estimation for the P uniformly spaced subbandsFor example(block 616). Channel gain estimates for each of the P subbands may be derived based on received pilot symbols or extrapolation/interpolation.
Then, as shown in equation (9), the vectorPerforming a P-point IFFT to obtain a P1 vector for least squares channel impulse response estimation(block 618). Channel impulse response estimation for multiple OFDM symbolsTime domain filtering is performed to obtain a higher quality channel estimate (block 620). The time domain filtering may be omitted or performed on the frequency response estimate instead of the impulse response estimate. (filtered or unfiltered) vectorP terms for L taps are included, where L is typically less than P. The vectors are then processed as described belowTo select the "good" taps and discard or zero the remaining taps (block 622). Zero padding may also be performed to obtain an nx 1 vector for channel impulse response estimation(block 624). Then to vectorPerforming an N-point FFT to obtain a vector for final frequency response estimation for all N subbands(block 626).
Extrapolation/interpolation
For block 614 in fig. 6, extrapolation may be used to derive channel gain estimates for the inactive pilot subbands located between the guard subbands. For a function y ═ f (x) where a set of y values can be obtained for a set of x values within a known range, extrapolation can be used to estimate the y values for x values outside the known range. For channel estimation, x corresponds to the pilot subbands and y corresponds to the channel gain estimate. The extrapolation can be performed in different ways.
In one extrapolation scheme, the channel gain estimate for each inactive pilot subband is set equal to the channel gain estimate for the nearest active pilot subband as follows:
equation (11)
WhereinIs a pair of sub-bands siIs estimated by the channel gain ofbIs the first active pilot subband, seIs the last active pilot subband as shown in fig. 4.
In another extrapolation scheme, the channel gain estimate for each inactive pilot subband is derived based on a weighted sum of the channel gain estimates for the active pilot subbands. If the number of taps L is less than or equal to the number of active pilot subbands (i.e., L ≦ P)act) Then (in the absence of noise) may be entirely guided by the pair activationFor extrapolation, each inactive pilot subband is associated with a respective set of extrapolation coefficients, one for each active pilot subband, where each coefficient may be zero or non-zero, the extrapolation/interpolation for the inactive pilot subbands may be represented in the form of a matrix, as follows:
equation (12)
WhereinIs P of the extrapolated coefficientext×PactAnd (4) matrix.
The number of complex operations required for the extrapolation in equation (12) is Pext·Pact. The number of inactive pilot subbands isWhere G is the number of guard subbands, "[ x [ ]]"is the ceiling operator that assigns x the next larger integer. If the number of guard subbands is small, the number of inactive pilot subbands in the system is typically small. For example, if there are 80 guard subbands (i.e., G-80), an OFDM system as described above may have only 10 of the 512 pilot subbands (i.e., P-512) inactive pilot subbands (i.e., P-512)ext10). In this case, extrapolationThe required calculations do not add significantly to the computational complexity. Computational complexity can also be significantly reduced by limiting extrapolation to use a subset of the active pilots.
The extrapolation coefficients may be fixed and may be determined (i.e., pre-computed) offline based on criteria such as least squares, Minimum Mean Square Error (MMSE), and the like. For least squares extrapolation, the coefficient matrix can be alignedThe following definitions are made:
equation (13)
WhereinIs thatWN×NP ofactxL submatrix. In a real system, the matrixCan make it possible toIs "ill-conditioned", which means that the inversion operation on the matrix may face problems of numerical stability. In this case, the correction term can be used to overcome the ill-conditioned problem and the modified least squares extrapolation matrix can be appliedThe following definitions are made:
equation (14)
Where δ is a small correction factor.
For MMSE extrapolation, the coefficient matrix can be alignedThe following definitions are made:
equation (15)
Where γ is the signal-to-noise ratio (SNR) of the received pilot symbols; and
η is a factor used to derive an unbiased estimate.
Without SNR information, γ can be considered as a parameter that can be selected to optimize performance. The factor η is also a scalar that can be used to optimize performance. By usingThe obtained vectorIs the channel MMSE estimate assuming that the taps in the time domain are uncorrelated and have equal energy. Equation (15) assumes for PactNoise vector for individual active pilot subbandsThe autocovariance matrix of (2) is an identity matrix. The method can be modified if the receiver knows the autocovariance matrixAnd (15) to calculate the autocovariance matrix.
In yet another extrapolation scheme, the channel gain estimate for each inactive pilot subband is set equal to zero, i.e.si＜sbAnd si＞seFor example, functional extrapolation (functional extrapolation) techniques such as linear and quadratic extrapolation may be used. Non-linear extrapolation techniques can also be used, which will fall within the general framework of equation (12). The pilot transmission scheme may not evenly distribute the active pilot subbands over the M available subbands. In this case, interpolation may also be used to derive channel gain estimates for uniformly spaced subbands among the M usable subbands. Similar to the above description of extrapolation, interpolation can be performed in various ways. In summary, extrapolation and/or interpolation may be performed as needed based on available received pilot symbols to derive channel gain estimates for the P subbands, which are uniformly spaced across all N subbands.
Tap selection
For block 622 in FIG. 6, the vector is subtendedTap selection is performed to select good taps for channel impulse response estimation. Tap selection may be performed in different ways.
In a tap selection scheme, channel impulse response estimationTruncated into L values for L taps of the wireless channel. Vector quantityComprises P elements, wherein P is more than or equal to L. To pairAt the determined tap selection scheme, the tap selection scheme is to be performedThe first L elements of (a) are considered good taps and are retained, while the last P-L elements are replaced with zeros. When L < P, a least squares channel impulse response estimate with L taps (without loss of performance) can be obtained by assuming that the channel has P taps, performing a P-point IFFT, and truncating the P-L taps. This may be beneficial in some situations. For example, if L < P/2, then the least squares channel impulse response estimate can be derived using the computational advantage of the FFT, without computing the last P/2 taps.
In another tap selection scheme, the tap selection signal is processedThe low energy elements in (1) are replaced by zero.Corresponds to a tap having low energy, which may be due to noise rather than signal energy. The threshold is used to determine whether a given element/tap has sufficient energy and should be retained or zeroed out. This process is referred to as "thresholding".
The threshold may be calculated in various ways and based on various factors. The threshold may be a relative value (i.e., dependent on the measured channel response) or an absolute value (i.e., not dependent on the measured channel response). The relative threshold may be calculated based on the (e.g., total or average) energy of the channel impulse response estimate. The application of the relative thresholds ensures that: (1) the thresholding does not depend on the variation in received energy, and (2) does not zero out elements/taps that are present but have low signal energy. The absolute threshold may be calculated based on the noise variance/noise floor at the receiver, the lowest energy expected for the received pilot symbols, and so on. The use of absolute thresholds is such thatMust meet a certain minimum value to be preserved. The threshold may also be calculated based on a combination of factors for relative and absolute thresholds. For example, a threshold may be calculated based on the energy of the channel impulse response estimate and further limited to be equal to or greater than a predetermined minimum value.
Thresholding may be performed in various ways. In one thresholding scheme, thresholding is performed after truncation, which can be expressed as:
equation (16)
WhereinThe last P-L elements are replaced by zeros by truncation;
is the energy of the nth tap;
is the energy of the L taps of the channel impulse response estimate; and
is a threshold for zeroing out low energy elements/taps.
‖x‖2Is a vectorxIs not more than the vectorxThe sum of the squares of all elements in (c).
In equation (16), based on L binsThe average energy of the head defines a threshold. The coefficient alpha is selected based on a trade-off between noise suppression and signal cancellation. Higher values of alpha provide more noise suppression but also increase the likelihood that elements/taps of low signal energy are zeroed out. The coefficient α may be a value in the range of 0 to 1 (e.g., α ═ 0.1). Or based on channel impulse response estimationRather than the average energy, to define the threshold. The threshold may be fixed or may be adjusted based on (1) the particular coding and modulation scheme or rate of the data stream being demodulated, (2) Bit Error Rate (BER), Packet Error Rate (PER), block error rate (BLER) or other error rate performance requirements, and/or (3) other parameters and considerations.
In another thresholding scheme, a single threshold pair is utilized, similar to that shown in equation (16)Performs thresholding (i.e., no truncation). In another thresholding scheme, multiple threshold value pairs are utilizedPerforms thresholding. For example, a first threshold may be used forThe first L elements of (1), the second threshold value may be used forThe last P-L elements of (c). The second threshold may be set lower than the first threshold. In another thresholding scheme, only the pairsPerforms thresholding on the last P-L elements, without thresholding on the first L elementsA thresholding process is performed. The thresholding may be performed in other ways and will fall within the scope of the invention.
Thresholding is well suited for "sparse" wireless channels, such as those in a macro cellular broadcast system. Sparse wireless channels concentrate a large amount of channel energy in a few taps. Each tap corresponds to a resolvable signal path with a different time delay. Sparse channels include few signal paths, even though the delay spread (i.e., the time difference) between these signal paths may be of great importance. Taps corresponding to weak or non-existent signal paths may be zeroed out.
For block 518 in fig. 5 and block 620 in fig. 6, the channel impulse response estimate may be filtered in the time domain using a low pass filter such as a Finite Impulse Response (FIR) filter, an Infinite Impulse Response (IIR) filter, or other type of filter. The low-pass filter may be a causal filter (which performs filtering on past and current samples) or a non-causal filter (which performs filtering on past, current and future samples obtained by buffering). The characteristics (e.g., bandwidth) of the filter may be selected based on the characteristics of the wireless channel. Time-domain filtering may be performed separately for each tap of the channel impulse response estimate for multiple OFDM symbols. The same or different filters may be used for multiple taps of the channel impulse response estimate. The coefficients of each of the filters may be fixed or adjustable based on detected channel conditions. An advantage of performing filtering in the time domain is that the pilot subbands may be staggered in the frequency domain (i.e., different sets of pilot subbands may be used for different OFDM symbols). Interleaving of the pilot subbands is useful when the channel has excessive delay spread (i.e., the length of the channel impulse response is greater than P taps). With additional and different pilot subbands provided by interleaving, channel impulse response estimates with more than P taps may be obtained. Filtering may also be performed on the initial or final frequency response estimates.
OFDM system
Fig. 7 shows a block diagram of an access point 700 and a terminal 750 in a spectrally shaped OFDM system. On the downlink, at access point 700, a Transmit (TX) data processor 710 receives, formats, codes, interleaves, and modulates (i.e., symbol maps) traffic data and provides modulation symbols (or simply "data symbols"). An OFDM modulator 720 receives and processes the data symbols and pilot symbols and provides a stream of OFDM symbols. OFDM modulator 720 multiplexes data and pilot symbols onto the appropriate subbands, provides a signal value of zero for each inactive subband, and obtains a set of N transmit symbols for the N subbands in each OFDM symbol period. Each transmit symbol may be a data symbol, a pilot symbol, or a signal value of zero. As shown in fig. 4, pilot symbols may be sent on the active pilot subbands. The pilot symbols may be sent continuously in each OFDM symbol period. Alternatively, the pilot symbols may be Time Division Multiplexed (TDM) with the data symbols on the same subband.
OFDM modulator 720 also converts each set of N transmit symbols to the time domain using an N-point IFFT to obtain a "converted" symbol that includes N time-domain chips. Typically, OFDM modulator 720 copies a portion of each transformed symbol to obtain a corresponding OFDM symbol. The duplicated portion is called a cyclic prefix and is used to suppress delay spread in the wireless channel.
A transmitter unit (TMTR)722 receives and converts the stream of OFDM symbols into one or more analog signals and further conditions (e.g., amplifies, filters, and frequency upconverts) the analog signals to generate a downlink signal suitable for transmission over the wireless channel. The downlink signal is then transmitted via an antenna 724 to the terminals.
At terminal 750, an antenna 752 receives the downlink signal and provides a received signal to a receiver unit (RCVR) 754. Receiver unit 754 conditions (e.g., filters, amplifies, and frequency downconverts) the received signal and digitizes the conditioned signal to obtain samples. OFDM demodulator 756 removes the cyclic prefix appended to each OFDM symbol, converts each received converted symbol to the frequency domain using an N-point FFT to obtain N received symbols for N subbands in each OFDM symbol period, and converts the received symbols to N subbandsFrequency symbolIs provided to a processor 770 for channel estimation. OFDM demodulator 756 then receives a frequency response estimate for the downlink from processor 770Performs data demodulation on the received data symbols to obtain data symbol estimates, which are estimates of the transmitted data symbols, and provides the data symbol estimates to an RX data processor 758. An RX data processor 758 demodulates (i.e., symbol demaps), deinterleaves, and decodes the data symbol estimates to recover the transmitted traffic data. The processing by OFDM demodulator 756 and RX data processor 758 is complementary to the processing by OFDM modulator 720 and TX data processor 710, respectively, at access point 700.
Processor 770 obtains received pilot symbols for the active pilot subbands and performs channel estimation as described in fig. 5 and 6. Processor 770 performs extrapolation and/or interpolation as needed to obtain pairs PdnChannel gain estimation for uniformly spaced subbands (where P isdnIs the number of pilot subbands for the downlink), a least squares impulse response estimate for the downlink is derivedTo pairPerforms tap selection and derives final frequency response estimates for the N subbands of the downlink
On the uplink, a TX data processor 782 processes traffic data and provides data symbols. An OFDM modulator 784 receives and multiplexes the data symbols with pilot symbols, performs OFDM modulation, and provides a stream of OFDM symbols. Pilot symbolsP that may have been allocated to terminal 750 for pilot transmissionupTransmitting on a number of subbands, where the number of uplink pilot subbands (P)up) May be associated with the number (P) of downlink pilot subbandsdn) The same or different. The pilot symbols may also be multiplexed with the data symbols using TDM. A transmitter unit 786 then receives and processes the stream of OFDM symbols to generate an uplink signal, which is transmitted via an antenna 752 to the access point.
At access point 700, the uplink signal from terminal 150 is received by antenna 724 and processed by a receiver unit 742 to obtain samples, an OFDM demodulator 744 processes the samples and provides received pilot symbolsAnd data symbol estimation for the uplink. RX data processor 746 processes the data symbol estimates to recover the traffic data transmitted by terminal 750.
As shown in fig. 5 and 6, processor 730 performs channel estimation for each active terminal transmission on the uplink. Multiple terminals may transmit pilot concurrently on the uplink on their respective sets of pilot subbands, where the sets of pilot subbands may be interlaced. For each terminal m, processor 730 performs extrapolation and/or interpolation for the terminal as needed to obtain an initial frequency response estimate for the terminal uplinkBased onDeriving a least squares channel impulse response estimate for a terminalPerforming tap selection and further deriving a final frequency response estimate for the terminalFrequency response estimation for each terminalIs provided to an OFDM demodulator 744 and is used to demodulate the data for that terminal.
Processors 730 and 770 control operation at access point 700 and terminal 750, respectively. Memory units 732 and 772 store program codes and data used by processors 730 and 770, respectively. Processors 730 and 770 also perform computations as described above to derive frequency and impulse response estimates for the uplink and downlink, respectively.
For a multiple-access OFDM system (e.g., an Orthogonal Frequency Division Multiple Access (OFDMA) system), multiple terminals may transmit simultaneously on the uplink. For such a system, the pilot subbands may be shared among different terminals. The channel estimation techniques may be used in cases where the pilot subbands for each terminal span the entire operating band (possibly excluding the band edges). Such a pilot subband structure would be suitable to obtain frequency diversity for each terminal.
The channel estimation techniques described herein may be implemented in different ways. For example, these techniques may be implemented in hardware, software, or a combination thereof. For a hardware implementation, the processing units used for channel estimation may be implemented in hardware as one or more Application Specific Integrated Circuits (ASICs), Digital Signal Processors (DSPs), Digital Signal Processing Devices (DSPDs), Programmable Logic Devices (PLDs), Field Programmable Gate Arrays (FPGAs), processors, controllers, micro-controllers, microprocessors, other electronic components designed to perform the functions described herein, or a combination thereof.
For a software implementation, the channel estimation techniques may be implemented with modules (e.g., procedures, functions, and so on) that perform the functions described herein. The software codes may be stored in memory units (e.g., memory units 732 and 772 of fig. 7) and executed by processors (e.g., processors 730 and 770). The memory unit may be implemented within the processor or external to the processor, in which case it can be communicatively coupled to the processor via various means as is known in the art.
Headings are included herein for reference and to assist in locating particular sections. These headings are not intended to limit the scope of the concepts described therein under, which concepts may be employed in other sections throughout the entire specification.
The previous description of the disclosed embodiments is provided to enable any person skilled in the art to make or use the present invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the generic principles defined herein may be applied to other embodiments without departing from the spirit or scope of the invention. Thus, the present invention is not intended to be limited to the embodiments shown herein but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.