HK1120682B - An ofdm base station, and a method of inserting a pilot symbol into an ofdm frame at a base station - Google Patents

Description

OFDM base station and method for inserting pilot symbols into OFDM frames at base station

The invention relates to a divisional application, the invention name of the parent application is 'scattered pilot frequency pattern and channel estimation method for MIMO-OFDM system', the national application number of the parent application is: 02825159.8, respectively; the international application numbers of the parent applications are: PCT/CA02/01541, the filing date of which is 10/15/2002.

Technical Field

The present invention relates to OFDM communication systems and, more particularly, to more efficient use of pilot symbols in such systems.

Background

Multiple-input multiple-output-orthogonal frequency division multiplexing (MIMO-OFDM) is a new highly spectrally efficient technique that is used to transmit high-speed data over wireless channels with fast fading in frequency and time.

In a wireless communication system employing OFDM, a transmitter transmits data to a receiver by using many parallel subcarriers. The frequencies of the subcarriers are orthogonal. Transmitting the data in parallel allows the symbols containing the data to have a longer duration, which reduces the effects of multipath fading. The orthogonality of the frequencies allows the subcarriers to be closely arranged while minimizing inter-carrier interference. At the transmitter, the data is encoded, interleaved, and modulated to form data symbols. Overhead information including pilot symbols is added and the symbols (data plus overhead) are organized into OFDM symbols. Each OFDM symbol typically uses 2ⁿAnd (4) a frequency. Each symbol is assigned a component representing a different orthogonal frequency. An Inverse Fast Fourier Transform (IFFT) is applied to the OFDM symbols (and is therefore preferably 2)ⁿOne frequency) to generate time samples of the signal. Adding a cyclic extension to the signal and passing the signal to a digital-to-analog converter. Finally, the transmitter transmits the signal along a channel to the receiver.

When the receiver receives the signal, the inverse operation is performed. The received signal is passed through an analog-to-digital converter and timing information is then determined. The cyclic extension is removed from the signal. The receiver performs an FFT on the received signal to recover the frequency components of the signal, i.e., the data symbols. Error correction may be applied to the data symbols to compensate for phase and amplitude variations caused during propagation of the signal along the channel. The data symbols are then demodulated, deinterleaved, and decoded to produce the transmitted data.

In systems employing differential detection, the receiver compares the phase and/or amplitude of each received symbol to adjacent symbols. The adjacent symbols may be adjacent in the time direction or in the frequency direction. The receiver recovers the transmitted data by measuring the change in phase and/or amplitude between one symbol and the adjacent symbol. If differential detection is used, no channel compensation needs to be applied to compensate for the phase and amplitude variations caused during signal propagation. However, in systems employing coherent detection, the receiver must estimate the actual phase and amplitude of the channel response, and must apply channel compensation.

The changes in phase and amplitude caused by propagation along the channel are referred to as the channel response. The channel response is typically frequency and time dependent. If the receiver can determine the channel response, the received signal can be corrected to compensate for channel degradation. The determination of the channel response is a so-called channel estimation. The inclusion of pilot symbols in each OFDM symbol allows the receiver to perform channel estimation. The pilot symbols are transmitted with values known to the receiver. When the receiver receives the OFDM symbol, the receiver compares the received value of the pilot symbol with the known transmitted value of the pilot symbol to estimate the channel response.

The pilot symbols are overhead and should be as small as possible in order to maximize the transmission rate of the data symbols. Since the channel response may vary over time and over frequency, pilot symbols are scattered among data symbols to provide as complete a range of the channel response as possible over frequency and time. The set of frequencies and times at which pilot symbols are inserted is referred to as a pilot pattern. The optimal time interval between pilot symbols is typically specified by the maximum expected doppler frequency, and the optimal frequency interval between pilot symbols is typically specified by the expected delay spread of multipath fading.

Existing pilot-assisted OFDM channel estimation methods are designed for use with a conventional one-transmitter system. For scattered pilot placement, there are three types of algorithms:

1-D frequency or time interpolation

Frequency 1-D interpolation of transforms

Independent time and frequency 1-D interpolation

The first type of algorithm is based on pilot OFDM symbols (all subcarriers are used as pilots) or comb pilots. This method, shown in the flow chart of fig. 1A, is simple but only applicable to channels with high frequency selectivity or channels with high time fading. The method involves pilot extraction in the frequency domain (step 1A-1), followed by interpolation over time (step 1A-2), or interpolation over frequency (step 1A-3).

The second method, shown in the flow chart of fig. 1B, is for a channel with slow doppler fading and fast frequency fading. It improves the first approach by using FFT to reconstruct the channel response back to the time domain for noise reduction processing at the expense of separately computing FFT/IFFT for the channel estimate. The method starts with pilot extraction in the frequency domain (step 1B-1), possibly followed by interpolation in frequency (step 1B-2). Then, the inverse fast Fourier transform (step 1B-3), smoothing/denoising process (step 1B-4), and finally the fast Fourier transform (1B-5) step are performed.

The third method shown in the flow chart of fig. 1C may be used to estimate a channel for mobile applications where there is both fast time fading and frequency fading. However, it requires a relatively high pilot density and a complete interpolator. The method involves pilot extraction in the frequency domain (step 1C-1), followed by interpolation in time (step 1C-2) and in frequency (step 1C-3).

In a propagation environment with high frequency dispersion and time fading, channel estimation performance can be improved by increasing pilot symbol density at the expense of reducing the spectral efficiency of the data transmission. Interpolating and reconstructing the channel response function from the limited pilots to achieve reliable channel estimation with minimal additional overhead is a challenging task.

There are a variety of standard pilot patterns available. In an environment where the channel only varies slowly in time and frequency, pilot symbols may be inserted cyclically, at adjacent frequencies after each time interval. In an environment where the channel is of high frequency correlation, pilot symbols may be inserted periodically at all frequencies simultaneously. However, such pilot patterns are only suitable for channels that vary very slowly over time. In an environment where the channel is of high time correlation, pilot symbols may be inserted continuously at only one particular frequency in the comb arrangement to provide a constant measure of the channel response. However, such pilot patterns are only suitable for channels that vary slowly with frequency. In an environment where the channel is highly frequency-dependent and highly time-dependent (e.g., a mobile system with large multi-path fading), pilot symbols may be inserted periodically in time and frequency such that when the symbols are plotted on a time-frequency plot, the pilot symbols form a rectangular grid.

In an OFDM communication system employing coherent modulation and demodulation, the receiver must estimate the channel response at the frequency of all subcarriers and at all times. While this requires more processing than systems employing differential modulation and demodulation, a greater gain in signal-to-noise ratio can be obtained using coherent modulation and demodulation. The receiver determines the channel response at the time and frequency at which the pilot symbols are inserted into the OFDM symbols and performs interpolation to estimate the channel response at the time and frequency at which the data symbols are located within the OFDM symbols. Placing the pilot symbols more closely together within the pilot pattern (in frequency if a comb pattern is used; in time if a periodic pattern is used; or in frequency and time if a rectangular grid pattern is used) results in a more accurate interpolation. However, because pilot symbols are an additional overhead, a tighter pilot pattern is at the expense of the transmitted data rate.

Existing pilot patterns and interpolation techniques are generally sufficient if the channel changes slowly over time (e.g., for nomadic applications). However, if the channel changes very quickly over time (e.g., for mobile applications), the time interval between pilot symbols must be reduced to allow accurate estimation of the channel response by interpolation. This increases additional overhead in the signal.

The problem of minimizing the number of pilot symbols while maximizing the accuracy of the interpolation is also particularly troublesome in multiple-input multiple-output (MIMO) OFDM systems. In a MIMO OFDM system, a transmitter transmits data through more than one transmit antenna, and a receiver receives data through more than one receive antenna. Binary data is typically split between transmit antennas, although the same data may be transmitted through each transmit antenna if spatial diversity is desired. Each receive antenna receives data from all transmit antennas so that if there are M transmit antennas and N receive antennas, the signal will propagate through M x N channels, each with its own channel response. Each transmit antenna inserts pilot symbols into the same subcarrier locations of the OFDM symbol it is transmitting. To minimize interference between the pilot symbols of each transmit antenna at the receiver, each transmit antenna typically flashes its pilot pattern on and off. This increases the time separation of the pilot symbols for each transmitter, reducing the accuracy of the interpolation used to estimate the channel response. In a MIMO-OFDM system, a simple and fast channel estimation method is particularly critical because of the limitations of the computational power used to estimate the mxn channels, whereas in a SISO-OFDM system only one channel needs to be estimated.

Disclosure of Invention

A channel estimation method based on partial interpolation of scattered pilots by using true 2-D interpolation is provided; and additionally a simple 1-D interpolation is used to reconstruct the entire channel. This approach has reduced scattered pilot overhead and its computational complexity is at least an order of magnitude less than some existing approaches. In general, the proposed channel estimation method is more robust in channels with high doppler spread and provides better performance than some existing methods and requires less OFDM symbol buffering for coherent detection at the receiver than some methods.

The method allows fewer pilot symbols to be placed within each OFDM symbol while still allowing accurate interpolation of the channel response. The data rate of the MIMO-OFDM system is thereby increased.

A first broad aspect of the present invention provides a method of inserting pilot symbols at an OFDM transmitter having at least one transmit antenna into Orthogonal Frequency Division Multiplexing (OFDM) frames, the OFDM frames having a time domain and a frequency domain, each OFDM frame comprising a plurality of OFDM symbols. The method involves, for each antenna, inserting scattered pilot symbols in an identical scattered pattern in time-frequency.

In some embodiments, the equivalent dispersed pattern is a common diagonal grid.

In some embodiments, inserting pilot symbols into the equivalent diagonal grid comprises: for each point in the equivalent diagonal grid, a number of pilot symbols are inserted on a single subcarrier for N consecutive OFDM symbols, where N is the number of transmit antennas.

In some embodiments, the diagonal grid is a diamond grid.

In some embodiments, for each point in the diagonal grid, L uncoded pilot symbols are generated. Space-time block coding (STBC) is performed on a group of L uncoded pilot symbols to produce nxn STBC blocks, L and N determining the STBC coding rate. Then, one row or column of the STBC block is transmitted through each antenna on a specific subcarrier.

In some embodiments, transmitting pilot symbols is done with a power level higher than the power level of the data symbols, depending on the value reflecting the channel conditions.

In some embodiments, transmitting pilot symbols is done with a power level that is dynamically adjusted as a function of the type of modulation applied to the data-carrying subcarriers to ensure sufficiently accurate reception.

In certain embodiments, the diagonal grid pattern has a first plurality of equally spaced subcarrier locations and a second plurality of equally spaced subcarrier locations offset from the first plurality of subcarrier locations. The pilot symbols are alternately inserted in time using a first plurality of equally spaced subcarrier locations and a second plurality of equally spaced subcarrier locations.

In some embodiments, the second plurality of subcarriers are offset from the first plurality of equally spaced subcarrier locations by half the spacing between adjacent subcarriers of the first plurality of subcarrier locations, thereby forming a diamond-shaped grid pattern.

In some embodiments, the pilot pattern is cyclically shifted in both the time and frequency directions for at least one neighboring base station to form a reuse pattern.

Another broad aspect of the invention provides an OFDM transmitter. The OFDM transmitter has a plurality of transmit antennas and is adapted to insert pilot symbols into Orthogonal Frequency Division Multiplexing (OFDM) frames having a time domain and a frequency domain, each OFDM frame comprising a plurality of OFDM symbols by inserting the pilot symbols in an identical dispersion pattern in time-frequency for each antenna.

In some embodiments, the equivalent dispersed pattern is a diagonal grid.

In some embodiments, inserting pilot symbols into the equivalent scattered pattern involves: for each point in the equivalent scattered pattern, a number of pilot symbols are inserted on a single subcarrier for N consecutive OFDM symbols, where N is the number of transmit antennas, where N > -1.

In certain embodiments, the dispersed pattern is a diamond grid.

In some embodiments, for each point in the dispersion pattern, the OFDM transmitter is adapted to generate L uncoded pilot symbols, perform space-time block coding (STBC) on groups of L pilot symbols to produce nxn STBC blocks, and transmit one row or column of STBC blocks on each antenna.

In some embodiments, the OFDM transmitter is further adapted to transmit pilot symbols at a power level higher than the power level of the data symbols, depending on the value reflecting the channel conditions.

In some embodiments in which a diamond-shaped grid pattern is employed, the diamond-shaped grid pattern has a first plurality of equally spaced subcarrier locations and a second plurality of equally spaced subcarrier locations offset from the first plurality of subcarrier locations. The pilot symbols are alternately inserted in time using a first plurality of equally spaced subcarrier locations and a second plurality of equally spaced subcarrier locations.

Another broad aspect of the invention provides a method of estimating a plurality of channel responses at an Orthogonal Frequency Division Multiplexing (OFDM) receiver having at least one receive antenna. The method involves receiving at each receive antenna an OFDM frame transmitted by at least one transmit antenna, the OFDM frame having a time domain and a frequency domain, the OFDM frame transmitted by each antenna having pilot symbols inserted in an identical scattered pattern over time-frequency, each OFDM frame comprising a plurality of OFDM symbols. For each transmit antenna, the receive antenna combines: (a) the pilot symbols of the received OFDM frame are used to estimate the channel response at each point on the dispersion pattern; (b) estimating channel responses of a plurality of points not in the dispersion pattern by performing two-dimensional (time direction, frequency direction) interpolation of the channel responses determined for the points in the dispersion pattern; (c) interpolation is performed in the frequency direction to estimate the channel response corresponding to the remaining OFDM subcarriers in each OFDM symbol.

In some embodiments, a filtering function is performed on the channel responses before interpolation in the frequency direction is performed to estimate the channel responses corresponding to the remaining OFDM subcarriers within each OFDM symbol.

In some embodiments, the channel responses of a plurality of points not on the dispersion pattern are estimated by performing a two-dimensional (time direction, frequency direction) interpolation of the channel responses determined for the points in the dispersion pattern grid, which involves averaging, for each subcarrier to be estimated, the channel responses for a given channel estimation period for subcarriers preceding (when present) and following (when present) the subcarrier to be estimated in frequency and the channel responses for a previous (when present) and a subsequent (when present) estimation period.

In some embodiments, the method is applied to a single transmitter, single receiver system.

In other embodiments, the method is applied to a single transmitter system, where each point in the scattered pattern contains a single pilot symbol.

In some embodiments, the method is applied to a system where there are N > -2 antennas, and each point in the scattered pattern contains N consecutive coded pilot symbols sent on a subcarrier, a single channel estimate being determined for every N coded pilot symbols.

In some embodiments, the N coded pilot symbols comprise L pilot symbols that are STBC block coded, where N and L together determine the STBC coding rate.

Other aspects and features of the present invention will become apparent to those ordinarily skilled in the art upon review of the following description of specific embodiments of the invention in conjunction with the accompanying figures.

Drawings

The invention will now be described in more detail with reference to the accompanying drawings, in which:

fig. 1 shows a flow chart for three examples of conventional OFDM channel estimation;

fig. 2 is a block diagram of a multiple-input multiple-output Orthogonal Frequency Division Multiplexing (OFDM) transmitter provided by an embodiment of the present invention;

FIG. 3 is a block diagram of an OFDM receiver;

fig. 4 is a flow diagram of a method for an OFDM transmitter to insert pilot symbols into an OFDM frame in accordance with one embodiment of the present invention;

FIG. 5 is a diagram of generating a pilot pattern by using the method of FIG. 4;

FIG. 6 is a block diagram of a MIMO system showing channel transfer functions between two transmit antennas and two receive antennas;

FIG. 7 shows a time-frequency plot of channel estimation positions for pilot channel estimation;

FIG. 8 schematically shows the steps of filtering estimated and interpolated pilot channel estimates;

FIG. 9 schematically shows the steps of interpolating between previously determined channel estimates to provide channel estimates for all subcarriers and all times;

FIG. 10 is a flow chart summarizing the overall channel estimation methodology provided by embodiments of the present invention; and

FIG. 11 is an example of a set of performance results obtained by using the method of FIG. 10.

Detailed Description

The following sections describe a MIMO-OFDM transmitter/receiver and scattered pilot insertion. By way of introduction, an OFDM frame consists of a preamble OFDM symbol and a regular OFDM symbol. Each OFDM symbol uses a set of orthogonal subcarriers. When there are two transmit antennas, two OFDM symbols form an STTD block. For a conventional OFDM symbol, some of the subcarriers are used as pilot subcarriers carrying pilot symbols, while other subcarriers are used as data subcarriers carrying data symbols. The pilot subcarriers are modulated by pilot symbols generated by QPSK. The data subcarriers are modulated by complex data symbols generated by QAM mapping. STTD coding is applied to pilot subcarriers located on the same frequency within one STTD block.

Referring to fig. 2, there is shown a block diagram of a multiple-input multiple-output (MIMO) Orthogonal Frequency Division Multiplexing (OFDM) transmitter provided by an embodiment of the present invention. The OFDM transmitter shown in fig. 2 is a two-output OFDM transmitter, although more generally there may be multiple (M) transmit antennas. The OFDM transmitter 10 takes binary data as input, but may accommodate other forms of data. The binary data is passed to a coding/modulation primitive (private) 12 which is responsible for coding, interleaving, and modulating the binary data to generate data symbols, as is well known to those skilled in the art. The coding/modulation element 12 may comprise a plurality of processing blocks, not shown in fig. 2. The encoder 14 applies space time block coding (SBTC) to the data symbols. Encoder 14 also divides the data symbols into a first processing path 16 and a second processing path 18 by alternately transmitting the data symbols along each of the two processing paths. In the more general case where the OFDM transmitter 10 includes M transmit antennas, the encoder 14 separates the data symbols into M processing channels.

The data symbols transmitted along the first processing path 16 are sent to a first OFDM component 20. The data symbols are first passed to the demultiplexer 22 of the first OFDM part 20, after which the data symbols are treated as sub-carrier components. The data symbols are then sent to a pilot inserter 24, where pilot symbols are inserted between the data symbols. Collectively, the data symbols and pilot symbols are hereinafter referred to simply as symbols. The symbols are passed to an Inverse Fast Fourier Transform (IFFT) processor 26 and then to a multiplexer 28 where they are recombined into a serial stream. The guard inserter 30 adds a prefix to the symbol. Finally, the OFDM signal is passed through a hard limiter 32, a digital-to-analog converter 34, and a Radio Frequency (RF) transmitter 36, which transmits the OFDM symbols as signals through a first transmit antenna 37. In most embodiments, each unit in the first OFDM component 20 is a processor, a component of a larger processor, or a collection of processors, or any suitable combination of hardware, firmware, and software. These may include general purpose processors, ASICs, FPGAs, DSPs, to name a few.

The pilot inserter 24 is connected to receive space-time coded pilot symbols from a pilot STBC function 23, which pilot STBC function 23 performs STBC on the pilot symbols. The pilot STBC block 23 takes two pilot symbols at a time, e.g., P₁And P₂As shown in fig. 2, and generates an STBC block consisting of a 2 × 2 matrix having (P) in the first row₁，P₂) And the second row has (-P)₂ ^*，P₁ ^*). It is the first row of this STBC block that is inserted by the pilot inserter 24.

The data symbols transmitted along the second processing path 18 are sent to a second OFDM component 38 which includes a processor similar to that included in the first OFDM component 20. However, the pilot inserter 40 inserts the encoded pilot symbols of the second row of the STBC block generated by the pilot STBC function 23. The symbols transmitted along the second processing path 18 are ultimately transmitted as signals by a second transmit antenna 42.

Referring now to fig. 3, a block diagram of a MIMO-OFDM receiver is shown. The OFDM receiver 50 includes a first receive antenna 52 and a second receive antenna 54 (although more generally there will be one or more receive antennas). The first receiving antenna 52 receives a first received signal. This first received signal is a combination of the two signals transmitted by the two transmit antennas 37 and 42 of fig. 2, although each of the two signals will have been altered by the respective channel between the respective transmit antenna and the first receive antenna 52. The second receiving antenna 54 receives a second received signal. This second received signal is a combination of the two signals transmitted by the two transmit antennas 37 and 42 of fig. 2, although each of these two signals has been altered by the respective channel between the respective transmit antenna and the second receive antenna 54. The four channels (between each of the two transmit antennas and each of the two receive antennas) may vary over time and over frequency, and will typically be different from each other.

The OFDM receiver 50 includes a first OFDM component 56 and a second OFDM component 58 (although there will typically be N OFDM components, one for each receive antenna). The first OFDM component 56 comprises an RF receiver 59 and an analogue-to-digital converter 60 which converts the first received signal into digital signal samples. The signal samples are passed to a frequency synchronizer 62 and a frequency offset corrector 64. The signal samples are also fed to a frame/time synchronizer 66. Collectively, these three components produce synchronized signal samples.

The synchronized signal samples represent a time series of data. The synchronized signal samples are passed to a demultiplexer 68 and then in parallel to a Fast Fourier Transform (FFT) processor 70. An FFT processor 70 performs an FFT on the signal samples to generate estimated received symbols, which are multiplexed in a multiplexer 76 and sent as received symbols to a decoder 78. Ideally, the received symbols would be the same as the symbols fed to the IFFT processor 26 in the OFDM transmitter 10. However, since the received signal will likely be altered by various propagation channels, the first OFDM component 56 must correct the received symbols by taking into account the channel. The received symbols are passed to a channel estimator 72 which analyzes the received pilot symbols at known times and frequencies within the OFDM frame. The channel estimator 72 compares the received pilot symbols with the values of the pilot symbols known to the channel estimator 72 to be transmitted by the OFDM transmitter 10 and generates an estimated channel response for each frequency and time within the OFDM symbol. The estimated channel response is transmitted to the decoder 78. The channel estimator 72 is described in more detail below.

The second OFDM section 58 includes components similar to those included in the first OFDM section 56 and processes the second received signal in the same manner as the first OFDM section 56 processes the first received signal. Each OFDM component conveys OFDM symbols to a decoder 78.

Decoder 78 applies STBC decoding to the OFDM symbols and passes the symbols to a decoding/demodulation element 80 which is responsible for decoding, deinterleaving, and demodulating the symbols to generate output binary data, as is well known to those skilled in the art. The decoding/demodulation element 80 may include a number of additional processing blocks, not shown in fig. 2. Each unit in OFDM components 56 and 58 is a processor, a component of a larger processor, or a collection of processors.

Referring now to fig. 4, a method of inserting pilot symbols among data symbols by each of the pilot inserters 24 and 40 of fig. 2 is shown. The method will be described with reference to the pilot inserter 24 in the first OFDM part 20. The pilot inserter 24 receives the data symbols from the demultiplexer 22 at step 100. At step 102, the pilot STBC function 23 generates (or receives) two pilot symbols. At step 104, the pilot STBC function 23 applies STBC coding to the pilot symbols to generate STBC blocks of coded pilot symbols. The encoded pilot symbols generated for the first transmit antenna 37 will be one row of the STBC block and the number will be equal to the number of transmit antennas in the OFDM transmitter. Thus, for a two antenna system, a 2 × 2STBC block is generated.

The pilot inserter 24 inserts the coded pilot symbols into the OFDM symbols at step 106. The coded pilot symbols are inserted in a diamond grid pattern. The diamond grid pattern uses the same frequency as the other diamond grid patterns, but with a time offset from the other diamond grid patterns. Preferably, the time offset for each diamond-shaped grid pattern is one symbol (in the time direction) from another diamond-shaped grid pattern, such that the diamond-shaped grid pattern uses consecutive symbols in the time direction of the OFDM frame.

Wherein each coded pilot symbol is inserted into a diamond-shaped grid pattern, preferably a perfect diamond-shaped grid pattern, within the OFDM frame. To achieve this, coded pilot symbols are inserted on each frequency of the first subset of frequencies. The frequencies within the first subset of frequencies are equally spaced by one pilot spacing. The coded pilot symbols are inserted on each frequency of the first subset of frequencies for one STBC block (two OFDM symbols). At some later time, coded pilot symbols are inserted on each frequency of the second subset of frequencies. The frequencies in the second subset of frequencies are shifted from the frequencies in the first subset of frequencies by half the pilot spacing in the frequency direction. The pilot inserter 24 continues to insert the coded pilot symbols to alternate between the first subset of frequencies and the second subset of frequencies.

Alternatively, different pilot patterns may be used as long as the same pilot pattern is used for each of at least one coded pilot symbol unique to the transmission antenna 37 and as long as the pilot patterns used for the coded pilot symbols are offset from each other in the time direction of the OFDM frame. For example, a common diagonal grid pattern may be used, diamond grid being a special case of such a pattern.

The pilot inserter 40 inserts pilot symbols by using the same method, although the pilot symbols will be the other half of the STBC block 42. Coded pilot symbols unique to the second transmit antenna 42 are inserted into the OFDM frame at the same symbol position where the coded pilot symbols corresponding to the first transmit antenna 37 were inserted.

Referring to fig. 5, an example of a pilot pattern generated by using the method of fig. 4 is shown. The pilot and data symbols are spread over the OFDM frame in the time direction 120 and the frequency direction 122. Most of the symbols within the OFDM frame are data symbols 124. A first set of coded pilot symbols 126 corresponding to the first transmit antenna 37 are inserted in a diamond-shaped grid pattern. A second set of coded pilot symbols 128 corresponding to the first transmit antennas 37 is inserted in the diamond lattice structure at the same frequencies as the first set of coded pilot symbols, but offset by one OFDM symbol position in the time direction 120. In the illustrated example, two of every four OFDM symbols carry coded pilot symbols. Each of the other transmit antennas transmits by using the same pattern. Pairs of pilot symbols that are consecutive on a subcarrier are composed of two rows of STBC encoded pilot symbols. The same pattern is transmitted by the second antenna.

The power of the coded pilot symbols 126, 128 may be increased compared to the traffic data symbols 124. The power increase of the coded pilot may be dynamically adjusted relative to the transmitted data symbol power level or modulation type (QAM size), or as a function of channel quality. The position of the diamond grid pattern may also be optimized to allow fast extraction of scattered pilots without the use of calculations. This can be achieved if the pilot subcarriers are spaced 2^ n apart in the frequency direction. In a multiple base station transmission arrangement, the position of the diamond-shaped grid pattern may be cyclically shifted between adjacent base stations in the time and frequency directions to form a diamond-shaped grid reuse pattern.

Referring now to fig. 6 to 10, a channel estimation method based on the above-described pilot insertion method will be described. The present invention provides a simple two-dimensional channel inserter for a MIMO-OFDM system with low pilot density for fast fading channels in time and frequency. The goals of channel estimation are: for each possible transmit antenna, receive antenna combination, the channel characteristics are estimated for each subcarrier and at each time. Referring to fig. 6, for a two transmit antenna, two receive antenna example, two transmit antennas Tx 1140 and Tx 2142 and two receive antennas Rx 1144 and Rx 2146 are shown. The channel estimate estimates are denoted as per H for each subcarrier and between Tx 1140 and Rx 1144 at each time₁₁148 between Tx 1140 and Rx 2146 is represented as a transfer function H₁₂150, represented as transfer function H for transmitters Tx 2142 through Rx 1144₂₂152 ofThe channel estimate, and finally, for transmitter Tx 2142 to receiver Rx 2146, is represented as a transfer function H₂₁154.

Some advantages of the proposed method over some existing methods are: (1) robustness to high mobility speeds, (2) reduction of scattered pilot grid density, and thus reduced pilot overhead.

Let P₁And P₂Are two pilot symbols encoded in the STBC block, which are transmitted in consecutive OFDM symbols on one subcarrier by two antennas. Then, at the first receiving antenna, there is a relationship for each subcarrier on which a pilot symbol is transmitted, in which a channel response H is assumed_ijIs constant over two OFDM frames:

Y_1，1is received data on a first antenna on a subcarrier in a first one of two consecutive OFDM symbols, and Y_1，2Is the received data on the first antenna on a subcarrier in the second of two consecutive symbols. This may be for H_11，H₂₁Solving, yielding:

a similar process for the second antenna yields:

wherein Y is_2，1Is the received data on the second antenna on a subcarrier in the first of two consecutive OFDM symbols, and Y_2，2Is at two endsReceiving data on a second antenna on the subcarrier in a second one of the consecutive OFDM symbols.

Using this technique, channel estimates are made for each pilot subcarrier and for each pair of OFDM symbols used to transmit the STBC block.

For the example of fig. 6, the result is a channel estimate for each possible channel (these are for 4 channels in this example) for each pair of pilot symbols transmitted. This is shown in fig. 7, where only the sub-carriers used for transmitting the pilot are shown. Channel estimates 150 are generated for each pair of (successive in time) OFDM frames for each pilot subcarrier. This results in channel estimates 150, 152, 154 for the first and second frames and channel estimates 156, 158, 160 for the fifth and sixth frames, and so on.

The channel estimation is performed on a STBC block-by-STBC block basis such that the pattern of channel estimation shown in fig. 7 gradually appears over time. The next step in the process is to perform interpolation based on the channel estimates of fig. 7 to obtain channel estimates for locations on fig. 7 that do not represent pilot channel locations. The manner in which this is done will be described for a single example, namely the unknown channel estimate, denoted as 163 of fig. 7. The channel estimates are buffered on an ongoing basis and when the four channel estimates 152,156, 158 and 164 forming diamond 162 around the unknown channel estimate 163 have been calculated, it is time for interpolation to obtain the channel estimate for the unknown point 163. The channel transfer function on the sub-carrier located in the center of the diamond can be obtained from a simple 4-point two-dimensional interpolator. A three-point two-dimensional interpolator may be used to derive the channel estimates corresponding to the first or last useful subcarriers:

here, (k 2.... ang., N)_pilot-1)

Where k is the pilot subcarrier index, N is the channel estimate index (or STBC block number — one channel estimate per subcarrier for every two symbols), and N_pilotIs the number of pilot subcarriers (6 in the example of fig. 7). H_newIs the new interpolated channel estimate for the ith channel estimation period and the jth pilot subcarrier. H (i, j) is the channel estimate determined from the pilot symbols as previously described. The three-point interpolator will also be performed for the last STBC block (i.e., the last two OFDM symbols) of the OFDM frame.

These calculations are done for each transmit antenna, receive antenna combination. It should be noted that this is just one example of how the channel estimates are interpolated.

If the original distance between pilot subcarriers in the frequency direction is D_fThen, after the first interpolation step described above, the spacing of the pilot subcarriers becomes D_f/2。

In some embodiments, the channel estimates thus calculated are filtered at each channel estimation period in order to remove noise. This is illustrated in fig. 6, where channel estimates 170 over a channel estimation period are shown as being input to a filter 172 to produce filtered channel estimates. For example, a simple 3-point moving iterative smoothing algorithm may be applied to H':

wherein k is 3_pilot-2. It should be understood that other filtering algorithms may be employed.

After interpolation of the pilot channel estimates as outlined in fig. 7, there will be one channel estimate for each subcarrier over which pilot channel information is transmitted and for each two OFDM symbol period during which pilot channelization information is transmitted. Referring to fig. 5, this means: for each antenna, there will be one channel estimate for the time-frequency points shaded to indicate when the pilot channel information is transmitted. There will also be channel estimates for the time-frequency points in the center of the diamond-shaped lattice structure of fig. 7. However, for those points that are not the pilot symbol transmission time-frequency points, nor the points that are at the center of the diamond-shaped grid of such points, there will be no channel estimates computed. The next step is to perform another interpolation step to develop the channel estimates for these other points.

In some embodiments, Cubic Lagrange (Cubic Lagrange) interpolation and linear interpolation (for subcarriers near the first and last useful subcarriers) in the frequency direction are used to derive the channel transfer function (for each pair of OFDM symbols) at all subcarriers for each STBC block.

The coefficients of the cubic Lagrangian interpolator may be calculated as:

the channel transfer function on the data subcarriers is given by:

j is 2_pilot-2.

This is shown in fig. 9, where the estimated channel response is fed to the lagrangian cubic interpolator function 175, which outputs values for all the intermediate subcarriers. Other interpolators may alternatively be employed.

In some embodiments, each OFDM symbol contains some pilot insertion points, and as such, this completes the interpolation process. In other embodiments, there are some OFDM symbols that do not have any pilot insertion points. To derive the channel estimates for these OFDM symbols, an interpolation of the previously computed channel estimates over time is performed. In high mobility applications, pilots should be included in each OFDM symbol to avoid the need for this final interpolation in time steps.

Fig. 10 gives a general block diagram of the proposed interpolation method for two transmit antennas. An exemplary set of performance results for the proposed MTMO-OFDM channel estimation algorithm is shown in fig. 10. At very high doppler spreads, the performance of the 2-D channel estimation algorithm is close to that of the ideal channel (with only 0.5dB loss).

Referring now to fig. 10 and 3, a channel estimation method is carried out by the channel estimator 72 to estimate the channel response for each subcarrier and each OFDM symbol within an OFDM frame. The channel estimation method begins at step 500 by extracting pilot symbols in the frequency domain for each receive antenna. This is followed by a channel response matrix calculation step 502; thereby decoding the received signal received by the receiving antenna, which is in fact a time average of the pilot symbols that are encoded at each point of the pilot pattern. For example, assume that the receive antennas receive an OFDM frame with a pilot pattern as shown in fig. 5 (although symbol 126 will now be a linear combination of the coded pilot symbols sent by each transmit antenna at this location, and symbol 128 will be a linear combination of the coded pilot symbols sent by each transmit antenna at this location). After decoding, the pilot symbol at symbol position 126 will be the average of the pilot symbol received at symbol position 126 and the pilot symbol received at symbol position 128. During step 503, the time-averaging effect produced by the STBC decoding can be seen as a pre-processing step, as steps 500 and 502 can accomplish. The actual channel estimation method can be broadly described in four steps. Following step 503, during step 504, the channel estimator 72 estimates a channel response for each of the plurality of pilot symbols. For a diamond grid pattern, the plurality of pilot symbols would be four pilot symbols forming a single diamond pattern. The channel estimator 72 estimates a channel response of a center symbol having a time direction value and a frequency direction value limited by time direction value and frequency direction value constraints of a plurality of pilot symbols. The center symbol preferably has a frequency direction value equal to the frequency direction values of two of the plurality of pilot symbols and a time direction value intermediate between the time direction values of the two pilot symbols having the same frequency direction value as the center symbol. This can be generally described as four-point 2-D interpolation of the channel response between pilot symbols. Third, the channel estimator 72 smoothes the channel response in the frequency direction (corresponding to the coded pilot symbols and the center symbol), preferably by performing a three-point smoothing, as per step 505. Fourth, the channel estimator 72 performs interpolation in the frequency direction to estimate the channel response for the remaining symbols, as per step 506. The interpolation may be a linear interpolation for symbols having frequency direction values equal to the first or last useful sub-carrier within an OFDM symbol, and a cubic lagrange interpolation otherwise.

The method of interpolating pilot symbols (described above with reference to fig. 4) and the channel estimation method (described above with reference to fig. 10) need not be used together. Any channel estimation method may be used by the OFDM receiver to estimate the channel response for an OFDM frame containing encoded pilot symbols inserted using the method described above. However, the two-dimensional interpolation method is preferable to the one-dimensional interpolation method due to the sparse distribution of pilot symbols in the pilot patterns described above with reference to fig. 4 and 5. Likewise, the channel estimation method may be applied to OFDM frames containing any pattern of pilot symbols.

The present invention has been described in connection with a MIMO-OFDM communication system. The present invention can also be advantageously used in a single-input multiple-output OFDM communication system because the method of inserting pilot symbols (described above with reference to fig. 4) and the channel estimation method (described above with reference to fig. 10) do not depend on the number of receive antennas. Each receive antenna within OFDM receiver 50 performs channel estimation independently regardless of the number of receive antennas present.

The channel estimation method described with reference to fig. 10 would also be advantageous in an OFDM communication system having only one transmit antenna, as it provides improved interpolation of the channel response, regardless of the number of transmit antennas. The method of inserting pilot symbols described with reference to figure 11 may be used in an OFDM communication system having only one transmit antenna but would not be as advantageous as in an OFDM communication system having more than one transmit antenna because the additional overhead would not be reduced.

The method of inserting pilot symbols and the channel estimation method are preferably implemented in the form of software instructions readable by a digital signal processor at the OFDM transmitter and OFDM receiver, respectively. Alternatively, the methods may be implemented in logic circuitry within an integrated circuit. More generally, any computing device containing logic for performing the described functionality may implement the methods. The computing device implementing the methods, particularly the pilot inserter or channel estimator, may be a single processor, more than one processor, or a component of a larger processor. The logic may include external instructions stored on a computer-readable medium, or may include internal circuitry.

The description is only illustrative of the application of the principles of the invention. Other devices and methods may be implemented by those skilled in the art without departing from the spirit and scope of the present invention.

Claims (8)

1. A method of inserting pilot symbols into OFDM frames at an orthogonal frequency division multiplexing, OFDM, base station having at least two transmit antennas, the OFDM frames having a time domain and a frequency domain, each OFDM frame comprising a plurality of OFDM symbols, the method comprising the steps of:

inserting scattered pilot symbols for each antenna into a time-frequency identical scattering pattern, the OFDM base station being adjacent to a second OFDM base station having at least two transmit antennas, wherein the scattering pattern is offset from the scattering pattern of the adjacent OFDM base station;

wherein inserting scattered pilot symbols comprises: for each point in the equivalent dispersed pattern:

generating L uncoded pilot symbols, wherein L is more than or equal to 1;

performing space-time block coding, STBC, on the uncoded pilot symbols to produce an nxn STBC block, L and N determining a STBC coding rate, where N is the number of transmit antennas; and

one row or column of the STBC block is transmitted on a particular subcarrier on each antenna.

2. The method of claim 1, wherein for each OFDM base station in the plurality of OFDM base stations, transmitting comprises transmitting using an equivalent scattered pilot pattern in time-frequency.

3. The method of claim 1, wherein inserting pilot symbols into an equivalent diagonal grid for each antenna comprises: a plurality of pilot symbols are inserted for each point in the equivalent diagonal grid on a single subcarrier for N consecutive OFDM symbols, where N is the number of transmit antennas.

4. The method of claim 1, further comprising inserting scattered pilots in a frequency domain and at intervals having a frequency that is optimized to allow fast extraction of scattered pilot symbols without the need to compute a full FFT.

5. A method, comprising:

implementing the method of claim 1 in a first orthogonal frequency division multiplexing, OFDM, base station;

the method of claim 1 implemented in at least one other OFDM base station using a dispersion pattern offset from that used by the first OFDM base station.

6. An orthogonal frequency division multiplexing, OFDM, base station comprising:

a plurality of transmit antennas;

the OFDM base station is adapted to insert pilot symbols into OFDM frames located at the OFDM base station having a time domain and a frequency domain, each OFDM frame comprising a plurality of OFDM symbols, wherein the pilot symbols are inserted for each antenna into an identical dispersion pattern in time-frequency, the OFDM base station being adjacent to a second OFDM base station having at least two transmit antennas, wherein the dispersion pattern is offset from the dispersion pattern of the adjacent OFDM base station;

the OFDM base station is adapted to insert scattered pilot symbols for each point in an equivalent scattered pattern by:

generating L uncoded pilot symbols, wherein L is more than or equal to 1;

performing space-time block coding, STBC, on the uncoded pilot symbols to produce an nxn STBC block, L and N determining a STBC coding rate, where N is the number of transmit antennas; and

one row or column of the STBC block is transmitted on a particular subcarrier on each antenna.

7. The OFDM base station of claim 6, wherein the OFDM base station is adapted to insert pilots in an equivalent scattered pattern of time-frequency for each antenna by:

for each point in the equivalent scattered pattern, a row or column is inserted comprising a number N of pilot symbols, constituted by one pilot symbol per antenna.

8. The OFDM base station of claim 6, wherein the OFDM base station is adapted to insert scattered pilots in the frequency domain at intervals optimized to allow fast extraction of scattered pilot symbols without requiring full FFT contention.