TWI387234B - Method and apparatus for adaptive reduced overhead transmit beamforming for wireless communication system - Google Patents

Description

Method and device for adaptive transmission beamforming for wireless communication systems

The present invention relates to a wireless communication system, and more particularly to a method and apparatus for adaptive transmission beamforming for reducing the amount of transmitted data for a multiple input multiple output communication system.

The communication system can be divided into wired and wireless standards, and its applications range from home local wireless networks, domestic and international mobile phone networks, and the global Internet.

Each communication system conforms to at least one existing communication standard, and includes wireless IEEE standard IEEE 802.11, wireless local area network (WLAN), advanced mobile telephone system (AMPS), Bluetooth, global mobile communication system. (GSM), code division multiplex access (CDMA), regional multipoint distribution system (LMDS), multi-channel multipoint distribution system (MMDS), and other communication standards.

A wireless device, such as a notebook computer, personal digital assistant (PDA), video projector, or WLAN phone, can communicate directly or indirectly with other users or devices under the network. A communication system that communicates in a direct manner is also commonly referred to as a point-to-point communication system. The second device is assigned to at least one radio frequency (RF) channel and communicates directly through the assigned channel. In the indirect communication system, the second device communicates on the assigned channel through a media device, wherein the media device can be one of the base stations of the mobile phone service, an access point of the home or public WLAN network (Access Point, AP). To complete the communication connection, the access point or base station communicates directly with one system controller, the Public Switch Telephone Network (PSTN), the Internet, or other wide area network.

Please refer to FIG. 1. As shown, the conventional WLAN home network includes an access point (AP) or base station 200, a network interface hardware 204, and other WLAN home networks. Electronic devices 206-214. The electronic device includes a WiFi mobile phone 206, a PDA/WiFi camera 208, a notebook computer 210, a home audio system 212, and a High Definition Television (HDTV)/projector 214. The electronic devices 206-214 communicate with each other on the assigned channel through the access point 200. The access point 200 has a beamforming transceiver module 216 and a multiple input multiple output (MIMO) antenna 218. Each of the electronic devices 206-214 has a beamforming transceiver module and a MIMO antenna. The access point 200 connects to an Internet wide area network (WAN) or a local area network (LAN) through the network interface hardware 204.

To receive data on a network, the structure of a conventional digital device includes an antenna coupled to a radio frequency (RF) signal processing circuit. The antenna first receives an RF signal and transmits the RF signal to the RF circuit to filter unnecessary signals and adjacent channel noise, and then convert the RF signal to a fundamental frequency (center frequency is 0, direct current (DC)). Or Intermediate Frequency (IF). The analog RF output signal of the intermediate frequency or the fundamental frequency is then converted into a digital stream and processed by the baseband module. The baseband module demodulates and decodes the above-mentioned baseband signal, and finally obtains the original data.

Similarly, in order to transmit data to the network, the digital baseband encodes the bit stream and the modulated encoded stream, and converts the encoded stream to an IF signal. The digital signal is converted to a analog signal by a digital analog converter and then transmitted to the RF circuit. The RF circuit converts the analog fundamental frequency into an RF signal through a carrier frequency corresponding to one of the channels allocated by the specific UE, and then transmits the signal to the wireless channel by using the RF antenna.

A conventional device, especially a WLAN product that conforms to the 802.11 standard before 2005, which utilizes a physical RF antenna to transmit and receive data. The RF antenna is shared by Time-Domain Duplexing (TDD), that is, the transceiver transmits or receives data only at a specific time, and cannot receive and transmit data at the same time. Therefore, the antenna is shared by the receiver and the transmitter. When an antenna is used for the two ends of the communication link, the established channel can be called a single-input/single-output (SISO). Recent advanced systems based on the 802.11 standard use multiple antennas to simultaneously receive and transmit data. For example, a dual antenna device that utilizes antenna switching diversity automatically exchanges antennas based on received signal quality. If the dual antenna device uses two antennas to transmit data at the same time, and the receiving end of the communication link has only one antenna, the channel can be called a multi-input/single-output (MISO) channel; conversely, if the transmitting end uses a single antenna, and the receiving end With two antennas, the channel can be called a single-input/multiple-output (SIMO) channel.

At present, most of the products are multiple-input/multiple-output (MIMO) products, and the purpose is to increase the processing rate, increase the connection range and stability. However, today's MIMO product technology has surpassed current open standards, such as the Institute of Electrical and Electronics Engineers (IEEE), which only proposes the draft of the industry standard 802.11n 2.0 to develop the operation of a multi-antenna WLAN system. Standard 802.11n develops protocols and technologies for multi-antenna systems that enable systems to support each other and maximize beamforming. For example, a multi-antenna system can transmit and receive more than one data stream simultaneously by combining time or spatial coding functions. Therefore, depending on the number of parallel streams, a multi-antenna system can increase processing efficiency by a factor of two or three.

To improve signal reception quality, Draft 2.0 includes several beamforming modes of operation. Basically, beamforming is a technique of an antenna array that can be applied to a set target or source by adjusting the gain and phase of the array elements. By adjusting the gain and phase of the array elements, the antenna shape or beam can determine a preferred direction to receive or transmit data, or to attenuate other directions to reduce sources of interference. For beamforming methods, refer to the following: Spectrum Signal Processing Publishing, 1998, Introduction to Digital Beamforming by Toby Haynes, and Digital Publications, 1986, by Hans Steyskal Beamforming foundation. The above reference provides the basic mathematics for forming a beam pattern by concentrating the antenna array to improve the quality of receiving and transmitting RF signals.

In 802.11n draft 2.0, many beamforming methods involve Orthogonal Frequency Domain Multiplexing (OFDM) modulation. OFDM is modulated into multiple carriers, which converts N data symbols into a time domain signal transmission using an Inverse Fast Fourier Transform (IFFT) procedure. In the receiver, the time domain signal is a symbol, and after demodulation, it becomes N frequency domain symbols. The received signal y can be expressed as y=Hx+n on each sub-carrier, where x=Qs, s is the transmission symbol, Q is the premultiplier, and H is between the transmitter and the receiver. Channel, n is effective noise. Under the MIMO link, channel H is a matrix. For example, in a two-transmitter and three-receiver (2T3R) system, H is a 3x2 matrix, so the received signal formula can be expressed as:

The transmission signal formula can be expressed as:

In the process of decoding the received data, the channel matrix H is estimated by the Preamble of the data packet. To complete the specific format of beamforming, Channel State Information (CSI) is first processed by the receiver and then passed back to the transmitter. The transmitter sets a beamforming matrix Q according to the CSI, and uses the matrix Q to transmit the concatenated packet to the corresponding receiver. Such a beamforming method of returning a CSI to a transmitter may be referred to as Explicit beamforming.

Explicit beamforming is a method developed according to the 802.11n standard. The above CSI information has three feedback forms. The first form of feedback uses the actual size of the matrix H, which can also be referred to as full channel state information (Full CSI). This way of returning in the form of complete CSI will cause the channel to require a large dynamic range representation due to channel degradation, causing CSI data to be prohibited from being transmitted. In order to avoid channel degradation, CSI data cannot be fed back, and another feedback form is implemented by a steering matrix (Steering Matrix). In this form, the channel matrix is first decomposed using Singular Value Decomposition (SVD), which can be expressed as:

H = U Σ V * Eq. (1)

After matrix decomposition (Decomposition), the matrix V is passed back to the transmitter. The advantage is that the matrix V is a unit matrix (ie, the matrix V is a single norm (Norm)), so the channel can be represented by less data, and the amount of data transmission can be reduced by this method. However, the lack of this method is that when the channel condition is slightly mutated, the matrix V may be drastically changed. Therefore, this method cannot be applied to all channels and causes sudden failure in processing rate and connection stability.

Further, in order to reduce the amount of data for beamforming feedback, the third feedback form is a compression channel state data, which is parameterized by Givings rotation decomposition and Polar Coordinates. Matrix and pass back to the sender. As mentioned above, the lack of prior art includes the form of returning a complete CSI to the transmitter, resulting in beamforming processing, and the matrix of matrix decomposition and compressed channel state data being overly sensitive to channel variations, resulting in a larger matrix. The change.

Therefore, the present invention provides a method and apparatus for external explicit feedback transmission beamforming, which can reduce the amount of information and completely and correctly return CSI to the transmitter.

Briefly, the embodiment of the present invention includes a multiple input/multiple output transceiver having a channel estimation module for receiving samples to generate time domain beamforming parameters, wherein the receiving channel samples include a channel status information ( Channel State Information, CSI). The multi-input/multi-output transceiver further includes an adaptive beamforming parameter module for receiving the time domain beamforming parameters to generate time domain adaptive beamforming parameters and frequency domain adaptive beamforming parameters; a group for receiving the frequency domain adaptive beamforming parameter to generate a data bit; a channel parameter module for receiving the data bit to obtain the time domain adaptive beamforming parameter; And receiving the time domain adaptive beamforming parameter to generate a data packet, and encoding the data packet to generate a modulated data stream; and a beamforming matrix module for receiving the modulation Data streaming, and generating a beamforming data stream according to the frequency domain adaptive beamforming parameter; the MIMO transceiver is configured to process the beamforming data stream to generate an output signal and form a beam pattern. The output signal is transmitted.

Referring to FIG. 2, as shown in the figure, a multiple input multiple output (MIMO) transceiver 10 includes a transmitter receiver switch 20, a radio frequency (RF) module 22, and a signal. The filter module 24, an Inverse Fast Fourier Transform (FFT) module 26, a channel estimation module 28, an adaptive beamforming parameter module 44, and a decoding module 49. The multiple input multiple output transceiver 10 of the embodiment of the present invention further includes an encoding module 16, a beamforming matrix module 58, an inverse fast Fourier transform (IFFT) module 60, and a cyclic delay diversity. The (Cyclic Delay Diversity, CDD) module 62, and the one-channel filter module 64 are part of the transmission path. In addition, the MIMO transceiver 10 includes a channel parameter module 68, a fast Fourier transform module 70, a boot matrix module 72, and a memory module 74.

Further, the encoding module 16 includes a data shaping module 46, an encoder/puncture device 50, an interleaver 52, a modulator 54, and a stream parser 56. The decoding module 49 includes an equalizer 32, a deinterleaver 34, a de-puncturing device 36, and a decoder 40 as part of the receiving path.

The antennas 12 and 14 are coupled to the transmitter receiver switch 20, the transmitter receiver switch 20 is coupled to the radio frequency module 22, the radio frequency module 22 is coupled to the signal filtering module 24, and the signal filtering module 24 is coupled to the fast Fourier transform. The module 26, the fast Fourier transform module 26 is coupled to the channel estimation module 26, and the channel estimation module 26 is coupled to the adaptive beamforming parameter module 44. Next, the adaptive beamforming parameter module 44 is coupled to the data shaping module 46 and the equalizer 32, and the equalizer 32 is coupled to the deinterleaver 34, the deinterleaver 34 is coupled to the anti-puncture device 36, and the anti-puncture device The decoder 40 is coupled to the decoder 40.

The decoder 40 is coupled to the channel parameter module 68. The channel parameter module 68 is coupled to the fast Fourier transform module 70. The fast Fourier transform module 70 is coupled to the boot matrix module 72. The boot matrix module 72 is coupled to the memory module. 74, and the memory module 74 is coupled to the beamforming matrix module 58.

The data shaping module 46 is coupled to the encoder/puncture device 50. The encoder/puncture device 50 is coupled to the bit interleaver 52. The bit interleaver 52 is coupled to the modulator 54. The modulator 54 is coupled to the stream parser 56. The stream parser module 58 is coupled to the beamforming matrix module 58. The beamforming matrix module 58 is coupled to the inversion fast Fourier transform module 60. The inverting fast Fourier transform module 60 is coupled to the cyclic delay diversity module 62. The diversity module 62 is coupled to the channel filter module 64, and the channel filter module 64 is coupled to the transmitter receiver switch 20.

The two antennas 12 and 14 are coupled to the multiple input multiple output transceiver 10. The multi-input multi-output transceiver 10 transmits data to a dual antenna receiver (not shown in FIG. 2). After receiving the data packet, the dual antenna receiver estimates downlink channel state information (CSI), which is the former Part of the processing. In the explicit feedback beamforming system, the channel status information is transmitted back to the dual transmitter dual receiver (2T2R) multiple input multiple output transceiver 10 through the transmission module. Specifically, the channel status information of the backhaul process (as shown in FIG. 2) is included in the data shaping module 46 to form a data packet and transmitted via the transmission processing path. Similarly, the boot matrix module 72 is used in the beamforming matrix module 58 to return the data packet to the receiver at the other end of the communication link. In another embodiment of the invention, the multiple input multiple output transceiver 10 and the receiver each have at least one antenna.

Please continue to refer to FIG. 2, after the antenna receives the signal, it is transmitted to the transmitter receiver switch 20. The operation mode of the transmitter receiver switch 20 is preset to receive data. The radio frequency module 22 filters the signals of the noise and unnecessary adjacent channels, and then converts the signals from the radio frequency to the fundamental frequency or intermediate frequency (IF) to generate the fundamental frequency signal. The baseband signal is converted into a digital stream by the signal filtering module 24, and the digital stream is converted into a frequency domain receiving sample by the fast Fourier transform module 26.

Channel estimation module 28 processes the received samples described above to generate time domain beam parameters. The channel state information of the embodiments of the present invention is included in the received samples, and the estimation of the channel state information is performed in the time domain, which provides a method of improving the prior art. The channel estimation module 28 can cooperate with the channel state to transmit the necessary information indicating the channel on the communication link, thereby reducing excessive system information and increasing the system processing rate.

The adaptive beamforming parameter module 44 processes the time domain beamforming parameters to generate time domain adaptive beamforming parameters and frequency domain adaptive beamforming parameters. In the technique of the present invention, the channel state information is estimated to be applied to the channel estimation module 28 and the adaptive beamforming module 44 to reduce feedback data and provide near-complete channel status information. In addition, the present invention provides an automatic matching of the channel state, reducing the channel state information amount, thereby improving the accuracy of the feedback channel state information.

The time domain beamforming parameters are included in the feedback channel status information and transmitted to the encoding module 16 and then input to the packet of the data shaping module 46 to form a data packet. The data packets are encoded and punctured by the encoder/puncture device 50 to produce encoded data that is interleaved by a bit interleaver 52 to produce a data sample. The data samples are modulated by the modulator 54 to form a plurality of constellation points and parsed into a stream by the stream parser 56 to form a modulated data stream.

The adaptive beamforming parameter module 44 transmits the frequency domain adaptive beamforming parameters to the equalizer 32 to form an equalized sample. The equalized samples are deinterleaved by deinterleaver 34 to produce deinterleaved data, avoiding successive bits falling on an invalid subcarrier. The de-interlaced data is then reversed through the anti-puncture device 36 to produce anti-puncture data, wherein the data is transmitted at half rate. The anti-puncture data is decoded by decoder 40 to generate data bits.

The data bit is transmitted to the channel parameter module 68, and the channel status information is retrieved from the data bit. The channel state information includes time domain adaptive beamforming parameters and is converted to frequency domain adaptive beamforming parameters by a Fast Fourier Transform (FFT) module 70. As described in more detail below, the steering matrix module calculates the steering matrix based on the frequency domain adaptive beamforming parameters. Therefore, when the channel status information is updated, the boot matrix is updated and stored in the memory module 74. The steering matrix is then passed to a beamforming matrix module 58, where the beamforming matrix is then applied to the modulated data stream.

The beamforming matrix is combined with the modulated data stream to generate beamforming data through the beamforming matrix module 58. In a 2T2R system, assuming that the modulated data stream is represented as [s ₁ , s ₂ ], the 2X2 beamforming matrix is represented as Q, and the beamforming data stream is represented as [x ₁ , x ₂ ], the following formula is obtained :

The steering matrix combines with the beamformed data stream to produce an output data, which is converted to the time domain by an inverse fast Fourier transform (IFFT) module 60, and then transmitted to a cyclic delay diversity (CDD) module 62, wherein the CDD module 62 The output data stream is cyclically converted to avoid overlapping of invalid values between streams. The CDD module 62 outputs the data stream, which is then transmitted to the channel filtering module 64 to filter unnecessary signals and generate output signals. The output signal is transmitted to the transmitter receiver switch 20. The operation of the transmitter receiver switch 20 in the transmit mode transmits the output signal through the antennas 12 and 14 to the receiver.

In the general literature, since the dimensions have been reduced, most of the estimates in the time domain are proposed. On the other hand, estimates in the time domain increase its complexity. In the case of a single transmitter single receiver (1T1R), studies have shown that the better estimate is less unknown. For example, in a standard 802.11g single-input single-output (SISO) system, the channel estimate has 52 independent unknowns, whereas the channel in the time domain typically contains 16 significant numbers, exponential decay taps (which are based on standard 802.11a, Cyclic Prefix (CP) is not violated. Therefore, in a Tone basis, 52 measurements are used to determine 52 independent channel taps, and 52 measurements can be used to estimate the potential 16-point time domain sequence to increase the accuracy of the equalization steps. degree. In other words, the basic problem is that more than three coefficients are over-determined. Therefore, some functions in the current standard 802.11g and "smart" antenna (SA) systems cannot be performed when estimating through pure frequency domain channels. If a 16th-order channel module is used, the Mean-Square Error (MSE) estimate improves the pitch by approximately 5.1 dB. (However, not all tones can be improved, please refer to the instructions below, the edge tones near the guard band cannot be completely improved). Overdetermined systems are systems that are more stable and less sensitive to estimation errors. Furthermore, in any estimation problem, over-parameterization will reduce the estimated performance, so the preferred way is to assume that the shorter the channel length, the better.

In addition, another preferred way is to use prototype material. In the least square (Least-Squares) estimation, only the channel length is used; in terms of Minimum Mean Square Error (MMSE) estimation, power delay extension data is used in addition to the channel length. The above error estimation technique is further explained as follows.

Measurement model: single input and single output

The least-squares (LS) channel estimate can be easily applied to multiple input multiple output situations when the result of a single transmitter single receiver (1T1R) is obtained. The orthogonal frequency domain multiplexing (OFDM) fundamental frequency model can be expressed by the following formula:

y = XFg _zp + n . Eq. (3)

The variable g _{zp is} zero padded to the channel coefficient vector of length N (N=64 for standard 802.11a/g/n 20MHz systems); F is an FFT operation matrix; X is one of the transmission symbols (such as preamble) Diagonal matrix; n is the noise vector; and y is the output of the FFT. The channel in the frequency domain can be expressed as:

h = Fg _xp = Tg , Eq. (4)

Where T is the first L column of the FFT matrix and it is assumed that L (L < N) is the maximum length of the channel. This measurement equation simplifies the estimation of the result. Further, it is assumed that only columns of corresponding subcarrier frequencies are reserved in the F matrix to reduce the size of T to 56XL (standard 802.11n) or 52XL (standard 802.11a/g).

Minimum mean square error channel estimation (single input and single output case)

Suppose the linear estimate of h is:

The parameter M can be obtained from the above formula. To calculate M _mmse , the cost function needs to be minimized, ie:

J _mmse =< e , e 〉, Eq. (6)

Wherein, the brackets represent the expected value of the inner product, and the definition error is: To meet orthogonal conditions:

〈 e , y 〉=0, Eq. (7)

Reduce the cost function and replace the error, the orthogonal condition can get the following relationship:

among them

Therefore, the MMSE channel in the frequency domain is estimated to be:

By reducing the dimension (L-order), the time domain is estimated as:

This estimate is a more complicated calculation. It is necessary to assume the channel auto-covariance and noise size in advance to perform the inversion operation and store it in advance.

Least Squares (LS) channel estimation (single input and single output conditions)

A simplified estimate can be calculated by simple linear least squares, and the cost function described here is:

J _ls =( y - XTg )*( y - XTg ). Eq. (12)

Take out the partial cost function of the corresponding channel and set it to zero.

Solve the channel to get the least squares estimate:

Assume that the entire FFT matrix F is used instead of T (representing a zero-fill channel length estimate: , to simplify the problem into a single tone (frequency domain) estimate:

It is currently the standard 802.11g (1T1R) estimate, due to the over-parameterized channel, it is necessary to reduce the order estimate.

Least Squares Channel Estimation: Multiple Input Multiple Output Case, Draft 802.11N Preamble

The above formula can be applied to multiple input and multiple output situations, such as 2T2R configuration, without losing its generality. The formula described below can be applied to 1T1R, 1T2R, or 2T3R through relevant pre-information.

The extended application of MIMO first expands the measurement equation. Each subcarrier (k=1,...56) corresponds to a 2X2 matrix:

Under each subcarrier, the Walsh-Hadamard presymbol vector corresponds to the receive vector, which can be expressed as:

Among them, the coefficient k is used to simplify the representation formula; S is the double phase shift key (BPSK) pre-symbol under the k ^th subcarrier. A cyclic delay diversity (CDD) is included in the second transport stream, where the second transport stream is part of the channel. Since the data section contains delay diversity, the above information is valid. Since the equation system can solve four unknown channels at a time, a single tone estimate can be used:

The time domain channel matrix can be expressed as:

Assuming that the above-mentioned time domain channel coefficient l is limited by L (for example, 16 or 24), the output vector of the first receiving FFT module can be expressed as:

In the measurement equation (19), each FFT output vector ( r ₀₀ , r ₀₁ ) has 56 elements each, so there are 112 measurements. For example, if each channel vector ( g ₀₀ , g ₀₁ , etc.) has 16 elements each, there will be 64 unknowns. Therefore, the above equation is an overdetermined equation system. The matrix S is a diagonal matrix containing 11n long training field (LTF) symbols. Similarly, the output vector of the second receiving FFT module can be expressed as: In formula (20), the above A can also be expressed as:

Using the general least squares result for the above channel estimates, you can get:

Substituting the above matrix with MIMO measurement gives:

The last channel vector in the frequency domain is:

Since the channel parameters do not need to be transmitted back to the transmitter, the time domain results do not need to be explicitly calculated. All estimation processes can be combined into a matrix and stored in advance. The total estimate can be expressed as the following formula:

It is worth noting that conventional MIMO receivers using linear monophonic frequency domain estimation can be represented in a simplified format in conventional communication systems:

Through the matrix operation, the above channel estimation can be improved, and a block diagonal matrix of the format 56X56 element is obtained.

M _ls = T ( T * T ) ^-1 T * . Eq. (27)

To simplify the calculation, the only matrix operation that needs to be calculated is the inversion part: ( T * T ) ^-1 because T * and T can be calculated by the IFFT and FFT modules, respectively. The complete channel estimate includes both time and frequency domain estimates, as well as feedback on time domain channel parameters, see Figure 3.

As shown in FIG. 3, the channel estimation module 86 of the embodiment of the present invention is coupled to the adaptive beamforming parameter module 44, which includes front modules 90 and 92, junction points 94 and 96, and level module 98. 102, and IFFT modules 106 and 108. In addition, FIG. 3 further includes FFT modules 80 and 82.

The FFT modules 80 and 82 are respectively coupled to the front modules 90 and 92. The front modules 90 and 92 are coupled to the junctions 94 and 96, the junction 94 is coupled to the level module 98, and the level module 98 is coupled to the IFFT. The module 106 and the junction point 96 are coupled to the level module 102, and the level module 102 is coupled to the IFFT module 108.

The FFT modules 80 and 82 output the received samples to the pre-modules 90 and 92. Further, the first received sample (r ₀₀ r ₀₁ ) having the two elements is processed by the pre-modules 90 and 92 having a long pre-training sequence, wherein the long pre-training sequence includes the Walsh-Hadamard preamble yuan. The output of the front module 90 is subtracted from the output of the pre-module 92 at the junction 94 and multiplied by 1/2 at the level module 98 to produce a first element of the frequency domain beamforming parameters. The output of the front module 92 is added to the output of the pre-module 90 at the junction 96 and multiplied by 1/2 at the level module 102 to generate a second element of the frequency domain beamforming parameters according to equation (25). . Time domain beamforming parameters and It is calculated by applying T* to frequency domain beamforming parameters via IFFT modules 106 and 108. It is worth noting that and Also through the foregoing and The calculation is done to complete the 2x2 time domain channel matrix estimation (Equation (19)).

The adaptive beamforming parameter module 88 of the embodiment of the present invention includes window modules 112 and 114, matrix modules 116 and 118, multiplication modules 120 and 122, and FFT modules 124 and 126. The window modules 112 and 114 are coupled to the matrix modules 116 and 118, respectively, and coupled to the multiplication modules 120 and 122, respectively.

The time domain channel is estimated to be transmitted to the window modules 112 and 114. The window modules 112 and 114 have a window of threshold values and adaptive channel lengths such that the window can achieve a desired channel estimate of tap energy or other matrix. For example, the window can be expanded until the tap value outside the window is less than a set threshold. Taps not included in the window have a minor effect on the channel model and are not in the channel estimation procedure. Different techniques can be used to select the window center. In an embodiment of the invention, the tap with the highest tap energy (or magnitude) is the center of the window, and the window is open approximately at the center until the tap outside the window is below the threshold. In another embodiment of the present invention, a center of gravity algorithm is used to find the center of the window, and the window is expandable until the total tap energy outside the window is lower than the preset ratio of the tap energy, wherein the preset ratio is included in the window. Medium, and can be called window tap energy. In the following cases, the channel taps outside the window may not be included in the estimation procedure, which provides a fraction of the total channel energy. The operation of the window is further explained below. Please refer to Figure 4, which opens the channel tap until a fixed ratio of total energy is obtained. After the window is centered, channel tap data is transmitted from the window modules 112 and 114 to the matrix modules 116 and 118, respectively.

According to the window length, the matrix C = ( T * T ) ^-1 can be calculated or selected from the pre-calculated value and the stored value, and then the matrix modules 116 and 118 transmit it to the multiplication modules 120 and 122, wherein the time domain The adaptive beamforming parameters are calculated according to the following formula:

Time domain adaptive beamforming parameters (such as channel estimation, also known as channel status information (CSI)) for two different paths. The first path, the time domain adaptive beamforming parameters are transmitted to the transmission modules 128 and 130 (as shown in FIG. 3), and the transmission modules 128 and 130 convert the time domain adaptive beamforming parameters into data packets and transmit back To the other end of the communication link. The second path, the CSI is sent to the FFT modules 124 and 126 for processing, and after the Zero Padding time domain adaptive beamforming parameters are passed, the FFT modules 124 and 126 perform FFT operations on the time domain adaptive beamforming parameters to Generating frequency domain adaptive beamforming parameters. The above FFT operation is the same as equation (23), and both include multiplication with the T matrix. Zero padding is used to increase the length of the CSI parameters so that the vector has an appropriate length (such as 64 or 128) and then perform an FFT operation. The frequency domain adaptive beamforming parameters are then passed to equalizer modules 132 and 134 to adjust the Frequency Domain Equalizer (FEQ).

Please refer to FIG. 4, which is a block diagram of a window module and a matrix module according to an embodiment of the present invention. First, step 138 calculates the total tap energy of all channels, and then sets a k _cp value and a critical value (THRESH); in the case of RATIO < THRESH and w (window length value) < w_max, step 142 divides the tap energy in the window. The ratio (RATIO), which can be referred to as the optimal window ratio (RATIO), is calculated as the total tap energy. After each calculation of the ratio (RATIO), the value of w is incremented by one until the calculated ratio (RATIO) is equal to or greater than the set threshold.

The threshold (THRESH) and w_max need to be values that can be programmed. Increasing the energy threshold (THRESH) allows most of the channel taps to be included in the estimate and has a larger C matrix (such as C ₂₄ ) to choose from. Conversely, decreasing the threshold (THRESH) will exclude more taps from the estimate and treat them as negligible, resulting in the use of smaller C matrices for estimation. The w_max parameter defines the channel estimation window size. The above approach is preferably used for channel presets with negligible multipath effects and a short pulse response (e.g., with fewer important taps).

Based on the best window RATIO, step 144 determines a C matrix. For example, in practical applications, when the distance of the transmitted output signal is shorter than the path length of the wireless channel, the front is less tapped, such as w<4, containing most of the information, and the dimension 8x8 is selected from step 144. Matrix C ₈ .

Similarly, when the transmission distance is medium, such as w < 8, the matrix C _{16 of the} dimension 16x16 is selected, and when the transmission distance is long, the matrix C _{24 of the} dimension 24x24 is selected. Therefore, the C matrix can be adapted to the channel condition and output at step 146.

In the channel estimation procedure, the most important element is the matrix C = ( T * T ) ^-1 . The basic features of matrix C are explained below:

1. C is a real number (no complex number) and a symmetric matrix.

2. In addition, each diagonal (primary diagonal and secondary diagonal) is a symmetrical sequence, thus reducing its storage space and its complexity.

3. The C matrix is Shift Invariant, that is, different time domain windows are used for different substring estimation. and When the C matrix remains unchanged. Since cyclic shift diversity is applied to different preamble substreams, the system must comply with the 802.11n standard.

Through the simplified matrix, multiple matrices can be stored depending on the channel conditions. For example, the three different matrices are stored, respectively as a 8x8 matrix C of dimension _8, 16x16 matrix C of dimension _16, and dimension of the 24x24 matrix C _24. Intermediate vector Can be calculated by the window to get the width of the window, so that in the vector Next, the channel tap energy of the partial vector exceeds a preset ratio of the total tap energy. Refer to Figure 4 for the implementation of the channel window.

Preferably, the matrix C multiplication operation uses a singular value decomposition (SVD), that is, C = ( T * T ) ^-1 = V * Σ V , and the matrices V and Σ are stored in advance. This decomposition method allows all eigenvalues to be accurately stored in the diagonal matrix Σ, which has the same rank and row when the directional matrix is filled.

The above method can be extended to the MIMO system, and the detailed description thereof is as follows.

Extended application to 2T3R

In terms of completeness, to support 3 receivers, the additional subchannels are estimated using the same estimation method, ie:

Where r ₂ is the output of the third FFT module.

Extended application to 3 streams and 3T3R

The above described technique can be extended to the case of three streams. According to Eq. 20-27 on page 283 of the 802.11n standard, the polar structure of the Walsh-Hadamard preamble of the 802.11n standard will be expressed as:

When three streams are transmitted, the first three lines determine the polarity structure of the transmitted preamble. In this case, the four columns of the matrix represent the need to use an additional measurement to determine the 3X3 channel matrix. Using the above formula, the time domain channel coefficients are:

The above formula can obtain the first column of the channel matrix, and the above method can determine the second column and the third column of the channel matrix by substituting r ₁₀ for r ₀₀ and r ₁₁ for r ₀₁ .

Extended application to 4 streams and 4T4R

Since the preamble has four columns and only one extra transmission line, the application of Walsh-Hadamard extension to the four stream cases is based directly on the case of three streams. The fourth element of the channel estimate can be expressed as:

In order to show the improvement of the mean square error (MSE) of the embodiment of the present invention, the estimation formula of a standard 802.11g 1T1R system is as follows:

M _{ls ,1 T} = T ( T * T ) ^-1 T * S ₀ . Eq. (33)

In the standard 802.11g system, since two training symbols and 52 of the 64 subcarriers are used, the fundamental frequency domain (single tone) is estimated as:

Channel_Estimate_MSE=Channel_SNR+10*log(2)+10*log(64/52)~34dB.

In addition, if the actual channel has only 16 taps, the time domain estimation is expected to improve the mean square error (MSE): MSE_gain (16 taps) = 10 * log (52 / 16) = 5.1 dB, or MSE_gain (24 taps) = 10 * Log (52/24) = 3.4 dB.

As shown in Figure 5, except for a few sidetones and near DC or 0 frequency (DC), the gain effect is clearly seen. A similar curve as in Fig. 5 can also be obtained for the case of 8 taps.

Please refer to FIG. 5, which is a graph of the mean square error and subcarrier of the frequency domain estimation 152, the 24-tap time domain estimation 154, and the 16-tap time domain estimation 156 according to an embodiment of the present invention. The curve shown in Figure 5 is the result of an additive white Gaussian noise channel.

A decaying channel yields a similar curve result. Please refer to FIG. 6 , which is a mean square error estimation of a decay channel g=(1 0.1 1) according to an embodiment of the present invention. As shown, it has two invalid inner frequency bands. Figure 6 contains a plot of a standard mean square error and subcarriers for a frequency domain estimate 158, a 24-tap time domain estimate 160, and a 16 tap time domain estimate 162.

As shown in the figure, the time domain channel estimation can improve the mean square error of the tap estimation by about 5.4 decibels (dB), and the 1.25~2.0 dB improvement can be obtained by the RT2830C code base simulation estimation. Please refer to FIG. 7 , which is a packet-error rate (PER) time domain channel estimation analysis of the first, second, and third receiver transmission packets (single input and multiple output cases) according to the embodiment of the present invention. As shown, when smoothing, Maximum Ratio Combining (MRC) has an improvement of nearly 2 dB, such as applying a time domain channel estimate to a channel estimate configured with 3 receivers.

Please refer to Figure 8. In the case of multipath, the industry standard IEEE 802.11n Modulation Coding Scheme (MCS) 12 is a dual-stream six-point quadrature amplitude modulation (QAM). For 3T2R transceivers in standard IEEE channel B and E conditions, it can be seen from the figure that embodiments of the present invention can improve channel B conditions by about 1.8 dB and improve channel E (larger multipath) when using the smoothing algorithm described above. ) about 0.8dB.

In order to explain in detail the advantages of the present invention in reducing time consumption, the feedback method of the present invention is compared with other conventional techniques. The results are shown in Fig. 9, which lists the formula for calculating the zero-element load for different feedback methods. The purpose of the parameters in the formula: the number of channel rows (or the number of receivers) N _r , the number of channel columns (or the number of transmitters) N _c , the number of bits per coefficient N _b , the number of subcarriers N _sc full and the time domain channel Estimated number of taps N _L . Assume that each parameter has an 8-bit coefficient, a 3x3 channel ( N _r = N _c = 3), and 56 orthogonal frequency-division multiplex sub-carriers.

As can be seen from Fig. 9, according to the channel length, the embodiment (method 4) of the present invention reduces the data overload by 86%, 72% and 58% compared with the complete CSI (method 1). According to the standard 802.11n, in the case of poor channel conditions, if N _L = 24 must obtain the full length of the multipath, the method of the present invention reduces the data overload by 58% compared to the complete CSI method. It should be noted that in the case of a general channel, the method of the present invention is similar to the number of bits of the compression feedback (method 3), but the compression feedback method is to return an incomplete CSI. In contrast, the present invention can be returned. Grant almost complete CSI to the other end of the communication link.

The above are only the preferred embodiments of the present invention, and all changes and modifications made to the scope of the present invention should be within the scope of the present invention.

200. . . Access point, base station

204. . . Network interface hardware

206. . . WiFi mobile phone

208. . . Personal digital number / WiFi camera

210. . . Notebook computer

212. . . Home audio system

214. . . HDTV/projector

216. . . Beamforming transceiver module

218. . . Multiple input multiple output antenna

10. . . Multiple input multiple output transceiver

12, 14. . . antenna

16. . . Coding module

20. . . Transmitter receiver switch

twenty two. . . Wireless RF module

twenty four. . . Signal filter module

26, 80, 82, 124, 126. . . Fast Fourier Transform Module

28. . . Channel estimation module

46. . . Data shaping module

50. . . Coding/piercing device

52. . . Bit interleaver

54. . . Modulator

56. . . Stream parser

32. . . Equalizer

34. . . Deinterleaver

36. . . Detachment

40. . . decoder

44. . . Adaptive beamforming parameter module

49. . . Decoding module

58. . . Beamforming matrix module

60, 106, 108. . . Inverted fast Fourier transform module

62. . . Cyclic delay diversity module

64. . . Channel filter module

68. . . Channel parameter module

70. . . Fast Fourier Transform Module

72. . . Boot matrix module

74. . . Memory module

90, 92. . . Front module

94, 96. . . Junction point

98, 102. . . Level module

88. . . Parameter module

112, 114. . . Window module

116, 118. . . Matrix module

120, 122. . . Multiplication module

128, 130. . . Transfer module

132, 134. . . Equalizer module

138, 140, 142, 144, 146. . . step

152, 158. . . Frequency domain estimation

154, 160. . . 24-tap time domain estimation

156, 162. . . 16-tap time domain estimation

Figure 1 is a schematic diagram of a conventional wireless local area network.

2 is a schematic diagram of a multiple input multiple output transceiver according to an embodiment of the present invention.

FIG. 3 is a schematic diagram of a channel estimation module coupled to an adaptive beamforming parameter module according to an embodiment of the present invention.

4 is a block diagram of a window module and a matrix module according to an embodiment of the present invention.

FIG. 5 is a graph showing the mean square error of the frequency domain estimation value, a 24-tap time domain estimation, and a 16-tap time domain estimation according to an embodiment of the present invention.

Figure 6 is a graph showing the mean square error of the decay channel of g = (1 0.1 1) in the first embodiment of the present invention.

FIG. 7 is a graph showing a time domain channel estimation packet error rate of a single input multiple output transceiver according to an embodiment of the present invention.

FIG. 8 is a graph showing packet error rate of IEEE channels B and C according to an embodiment of the present invention.

FIG. 9 is a comparison diagram of a feedback method according to an embodiment of the present invention.

10. . . Multiple input multiple output transceiver

12, 14. . . antenna

16. . . Coding module

20. . . Transmitter receiver switch

twenty two. . . Wireless RF module

twenty four. . . Signal filter module

26. . . Fast Fourier Transform Module

28. . . Channel estimation module

46. . . Data shaping module

50. . . Coding/piercing device

52. . . Bit interleaver

54. . . Modulator

56. . . Stream parser

32. . . Equalizer

34. . . Deinterleaver

36. . . Detachment

40. . . decoder

44. . . Adaptive beamforming parameter module

49. . . Decoding module

58. . . Beamforming matrix module

60. . . Inverted fast Fourier transform module

62. . . Cyclic delay diversity module

64. . . Channel filter module

68. . . Channel parameter module

70. . . Fast Fourier Transform Module

72. . . Boot matrix module

74. . . Memory module

Claims (17)

A multi-input and multi-output transceiver includes a channel estimation module for processing a receive channel sample to generate a time domain beamforming parameter, wherein the receive channel sample includes channel status information (Channel State information (CSI); an adaptive beamforming parameter module for receiving the time domain beamforming parameters to generate time domain adaptive beamforming parameters and frequency domain adaptive beamforming parameters; a decoding module for Receiving the frequency domain adaptive beamforming parameter to generate a data bit; a channel parameter module for receiving the data bit to obtain the time domain adaptive beamforming parameter; and an encoding module for receiving The time domain adaptive beamforming parameter to generate a data packet and to encode the data packet to generate a modulated data stream; and a beamforming matrix module for receiving the modulated data stream to The frequency domain adaptive beamforming parameter generates a beamforming data stream; wherein the adaptive beamforming parameter module is used to estimate the channel shape a window of information, the received sample includes a plurality of taps, wherein each tap has a Tap Energy, and the window includes at least one tap and one window tap energy; the multi-input a multi-output transceiver for processing the beamformed data stream to generate Outputting a signal and transmitting the output signal by forming a beam; wherein the window has an adaptive length for placing a desired level of energy of the window tap, the center of the window being opened at a tap of the highest tap amount, the window Configuring a predetermined threshold, and the received sample has at least one tap outside the window, and the energy of the at least one tap is lower than the threshold; or the window finds a window center through a gravity algorithm, the received sample has At least one tap is located outside the window, and the tap has a tap energy that is less than a predetermined ratio of tap energy in the window. The multi-input multi-output transceiver of claim 1, wherein the window has a predetermined threshold for determining a length of the window, the threshold being increased to increase the length of the window to be placed in the tap. A large value that is reduced to reduce the length of the window to place a smaller value for the tap. The multi-input multi-output transceiver of claim 2, wherein the window has a w_max parameter, and the channel status information is preset as a short pulse response, and is placed in the window to limit the length of the window, wherein the window Short pulse responses have fewer important taps. The multiple input multiple output transceiver of claim 3, wherein the adaptive beamforming parameter module is configured to generate a ratio (RATIO) of the window to determine an optimal window ratio. The multi-input multi-output transceiver of claim 4, wherein the adaptive beamforming parameter module further comprises a matrix module for generating a C matrix according to the length of the optimal window ratio, the threshold is increased. To increase the size of the C matrix, the threshold is reduced to reduce the size of the C matrix. The multi-input multi-output transceiver as claimed in claim 5, further comprising a two-channel transmitter (Two Transmitter Two Receiver, 2T2R), the channel estimation module processing a matrix S, which includes a transmission 11n long training field (Long Training field, LTE) symbol, the channel estimation module processes the received symbol including (r ₀₀ , r ₀₁ ), and generates the frequency domain beamforming parameter according to the following formula and : The multiple input multiple output transceiver of claim 6, wherein the channel estimation module further comprises an Inverse Fast Fourier Transform (IFFT) module for processing the frequency domain beamforming parameters, Generating time domain beamforming parameters and . The multiple input multiple output transceiver of claim 7, wherein the adaptive beamforming parameter module further comprises a multiplication module for processing the C matrix, wherein the matrix C = ( T ^* T ) ^-1 can be calculated Deriving or selecting from pre-calculated values and stored values, and generating time domain adaptive beamforming parameters according to the following formula and : The multiple input multiple output transceiver according to claim 8, wherein the adaptive beamforming parameter module further comprises a Fast Fourier Transform (FFT) module for processing the time domain adaptive beamforming parameter. And by increasing the length of the time domain adaptive beamforming parameters, performing Zero Padding and generating an appropriate length vector through zero padding. The multiple input multiple output transceiver of claim 9, wherein the Fast Fourier Transform (FFT) module is configured to process the length vector to generate frequency domain adaptive beamforming parameters according to the following formula: and : The multi-input multi-output transceiver of claim 10, further comprising a Fast Fourier Transform (FFT) module coupled to the channel parameter module for receiving the time domain adaptive beamforming parameter, Zero padding is performed to process the time domain adaptive beamforming parameters to generate frequency domain adaptive beamforming parameters. The multiple input multiple output transceiver of claim 10, further comprising a steering matrix module for processing the frequency domain adaptive beamforming parameters and generating a guide A matrix is used to update the steering matrix when the frequency domain adaptive beamforming parameters are updated. The multiple input multiple output transceiver of claim 12, wherein the steering matrix module processes the modulated data stream [s1, s2], and generates a beamforming data stream [x1, x2] according to the following formula: : among them, A 2X2 beamforming matrix. The multiple input multiple output transceiver of claim 13, wherein the boot matrix module combines the steering matrix with the beamformed data stream to generate an output data. The multi-input multi-output transceiver of claim 1, wherein the encoding module further comprises: an encoding/piercing module for processing the data packet to generate encoded data; and a one-bit interleaver for Receiving the encoded data to generate a data sample; a modulator for receiving the data sample to generate a constellation point; and a stream analyzer for receiving the constellation point to generate the modulated data stream. The multiple input multiple output transceiver of claim 1, wherein the decoding module further comprises: an equalizer for processing the frequency domain adaptive beamforming parameters to generate equalization a sample; a deinterleaver for receiving the equalized samples to generate deinterleaved data; a depuncturing device for receiving the deinterleaved data to generate depuncturing data; and a decoder for receiving the solution Puncture data to generate data bits. A method for transmitting and receiving data, comprising: generating time domain beamforming parameters; generating time domain adaptive beamforming parameters; generating frequency domain adaptive beamforming parameters; generating data bits; from the data bits, Taking the time domain adaptive beamforming parameter; generating a window for estimating the channel state information, the channel state information being included in the received sample, the receiving sample comprising a plurality of taps, wherein each tap has a tap energy And the window includes at least one tap and one window tap energy; the center of the window is set to be opened at the highest tap amount tap, wherein the window has an adaptive length for placing the desired level of the window tap energy. The window is configured with a predetermined threshold, and the received sample has at least one tap outside the window, and the energy of the at least one tap is lower than the threshold; or the center of the window is found through a centroid algorithm, the receiving The sample has at least one tap located outside the window, and the tap has a tap energy less than the tap in the window One of the ratio of the amount of default; Generating a data packet; encoding the data packet to generate a modulated data stream; generating a beamformed data stream; and generating an output signal and transmitting the output signal by forming a beam.