RU2634915C1 - Device for estimation of frequency response characteristic of wireless channel - Google Patents

Abstract

FIELD: radio engineering, communication.

SUBSTANCE: consists of an LS-estimators and accurate estimators. The accurate estimator divides the band into sub-carrier groups of size S, and then performs stepwise linear polynomial regression independently on each of the groups. For this the samples received from the LS-estimator pass a serial matrix multiplier, and the result of its operation is recorded in a temporary memory device. Vector of values is fed to a parallel matrix multiplier from memory. Then the data goes to the decision device. Also, the decision is taken by the result of the operation of the digital filter, which is also connected to the output of the LS-estimator. The output values of the decision device are the result of the operation of the claimed device.

EFFECT: improving the accuracy and quality of the radio channel frequency response estimation while maintaining low computational complexity.

5 dwg

Description

Technical field

The invention relates to broadband communication systems that use orthogonal multi-subcarrier frequency division multiplexing (OFDM), and more particularly, to methods for estimating the frequency response of multipath channels in a preamble.

State of the art

For channel estimation in OFDM systems, the transmitter arranges pilot symbols on some subcarriers. These symbols and their location are known to the receiver. Using this information, the receiver performs channel estimation.

Channel estimation methods are divided into two classes according to the location of the pilot symbols. In the first case, an OFDM symbol is emitted that does not contain information symbols, but only pilot ones. This arrangement of pilot symbols is called blocky. In the second case, most of the subcarriers are used to transmit information, and only some transmit pilot symbols. This placement of pilot symbols is called combined. (S. Coleri, M. Ergen, A. Puri, A. Bahai, Channel Estimation Techniques Based on Pilot Arrangement in OFDM Systems, IEEE Transactions on Broadcasting, Vol. 48, Issue 3).

If the communication system operates in conditions of low signal-to-noise ratio (SNR), then in the block arrangement of pilot symbols it is advisable to use only every nth subcarrier, where n≥2. This allows you to increase SNR on pilot subcarriers. Hereinafter, for brevity, the preambles, where each n-th subcarrier is used, where n≥2, are called sparse.

Thus, in both methods of placing pilot symbols, the task of estimating a channel by pilot subcarriers is divided into three tasks: extracting a useful signal from noise, channel estimation on pilot subcarriers, and interpolating to subcarriers not containing pilot symbols.

Various methods for estimating the frequency response of a channel are given in the M.K. review. Ozdemir, N. Arslan, Channel Estimation For Wireless Ofdm Systems, IEEE Communication Surveys, Vol. 9, No. 2, 2007. A method for estimating by the least squares method (least squares, LS) is described. It is easy to implement, but has a low quality assessment, is highly susceptible to white noise. A device implementing this method may consist of one divider or one phase rotation device.

A known method of estimation by minimizing the mean square error (minimum mean square error, MMSE). Has a high quality channel estimation, can be used to simultaneously solve all three of the above problems. The disadvantage of this method is that it requires knowledge of the channel statistics (SNR and covariance matrix). When changing the channel statistics, it is required to perform matrix inversion, which leads to high computational complexity and the practical inapplicability of the method.

A device is known (K. E. Wazeer, M. Khairy, N. Fahmy, S. Habib, FPGA Implementation of an Improved Channel Estimation Algorithm for Mobile WiMAX, 2009 International Conference on Microelectronics) that uses pre-computed covariance matrices to reduce MMSE complexity.

The prototype of the proposed device is the patent US 7801230 B2, publ. 09/21/2010. In a known solution, a device is disclosed that can solve the problem of high computational complexity MMSE. It is assumed that on the receiving side the procedures for removing the cyclic prefix, analog-to-digital conversion, filtering, transfer to the video frequency, time synchronization, and fast Fourier transform (FFT) are performed.

A radio channel estimator consists of an LS estimator and an S-MMSE (split minimum mean square error) channel estimator. The device operates as follows. After passing through the FFT block, the signal in the frequency domain is fed to the LS-estimation block, where the frequency response value (FX) on each subcarrier is estimated independently by the LS-estimation method. Next, the signal is supplied to the S-MMSE estimation device, which groups the subcarriers into a predetermined number of groups and independently performs an S-MMSE evaluation on each group, which consists in filtering the signal. Filter coefficients are calculated by the formula

,

,

where R _S is the covariance matrix of the channel for subgroups of subcarriers of size S, N ₀ is the noise density, and I is the identity matrix.

The reduction in the number of computational operations is achieved due to the fact that the size of the group S can be chosen small. Sizes from 4 to 7 are suggested.

The disadvantage of this device is that it is necessary to know the correlation matrix of the channel and the SNR. Their approximate estimate in the device introduces an additional error and delay in the process of estimating the channel frequency response. This leads to a decrease in accuracy and quality of the channel frequency response estimate.

Disclosure of invention

The technical result of the claimed invention is to improve the accuracy and quality of the evaluation of the frequency response of the channel while maintaining low computational complexity. The specified technical result is achieved by replacing S-MMSE with stepwise linear polynomial regression (PLPR) and the use of sparse preambles. Using a phased linear polynomial regression allows you to abandon the measurement of SNR and the correlation matrix, as well as apply sparse preambles. In turn, the use of sparse preambles makes it possible to increase the SNR and, as a consequence, the accuracy and quality of the channel estimate. At the same time, in the inventive device for estimating a radio channel according to the preamble, as in the prototype, an LS-estimator is used, however, the S-MMSE estimator is replaced by an accurate estimator performing PLPR.

The inventive radio channel channel estimator according to the preamble includes an LS estimator and an accurate estimator consisting of a serial matrix multiplier, a temporary memory device, a parallel matrix multiplier, a resolver, and a digital filter. The input of the LS estimator is the input of the radio channel estimator according to the preamble. The output of the LS estimator is connected to the inputs of a serial matrix multiplier and a digital filter. The output of the serial matrix multiplier is connected to the input of the temporary memory device. The output of the temporary memory device is connected to the input of a parallel matrix multiplier, the output of which is connected to the second input of the resolving device. The output of the digital filter is connected to the first input of the resolver. The output of the decider is the output of the radio channel estimator.

The accurate estimator splits the band into groups of subcarriers of size S, after which each of the groups independently performs PLPR. For this, the samples received from the LS-estimation device pass through a sequential matrix multiplier, and the result of its operation is recorded in a temporary memory device. From the memory, the vector of values is fed to a parallel matrix multiplier. Next, the data goes to the decider. The decision device also receives the result of the digital filter, which is also connected to the output of the LS evaluation device. The output values of the decisive device are the result of the operation of the inventive device.

The consistent and pipelined nature of the calculations allows for low computational complexity and low computational latency. The use of linear polynomial regression allows one to obtain high accuracy and quality estimates of the channel frequency response.

Brief Description of the Drawings

FIG. 1 is a block diagram of a receive path in which the inventive channel estimation apparatus can be used.

FIG. 2 is a structural diagram of the inventive channel estimation apparatus.

FIG. 3 is a block diagram of a serial matrix multiplier.

FIG. 4 is a block diagram of a parallel matrix multiplier.

FIG. 5 is a graph comparing the quality of the assessment of the claimed device and prototype.

The implementation of the invention

An apparatus for estimating a radio communication channel is presented, which makes it possible to use preambles in a communication system in which pilot symbols occupy each n-th subcarrier, where n≥1, and intermediate subcarriers at n≥2 are not used for transmission. The device can be used in wireless data systems using OFDM technology, for example, but not limited to, Long Term Evolution (LTE) and Worldwide interoperability for microwave access (WiMAX).

The device is intended for use in the receiving path of the terminal. The details of the receive path performance may vary depending on the communication standard used, however, the main components of the receive path using OFDM technology and necessary for the application of the claimed invention are common to all systems. A part of the general functional diagram of the receiving path is shown in FIG. 1 and includes: a communication channel 1, a receiver 2, performing signal transfer to a video frequency and analog-to-digital conversion, a cyclic prefix removal device 3, a fast Fourier transform device 4 (FFT), a channel estimator 5, and a demodulation device 6. Receiver output 2, receiving a signal from communication channel 1, is connected to the input of the cyclic prefix removal device 3, the output of which is connected to the input of the FFT device 4, the output of which is connected to the input of the channel estimator 5 and to the second input of the demodulation device 6. B the output of the channel estimator 5 is connected to the first input of the demodulation device 6, the output of which is the output of the receive path.

For the operation of the proposed device channel estimation required regression matrix R1 and R2. They depend on the parameters of the preamble and are calculated on a computer as follows. The degree of the polynomial N and the number of subcarriers S are selected. The matrix X of linear polynomial regression is calculated (Draper N., Smith G., “Applied Regression Analysis”, Moscow: Finance and Statistics, 1986). Apply a singular decomposition to matrix X (Loginov N.V. Singular decomposition of matrices, Moscow: MGAPI, 1996), with the help of which matrices U, S and V are obtained such that X = USV *. The first regression matrix R1 is equal to the matrix V *, for which all rows are deleted, starting from the number N + 1. The second regression matrix R2 is equal to the matrix US, in which all rows are deleted, starting from the number N + 1. These calculations are performed once using a computer, after which the matrices R1 and R2 are recorded in the device memory. The use of sequential multiplication by low-ranking matrices allows such an implementation in a device in which significantly fewer logic elements are involved than in direct calculation.

A block diagram of a radio channel estimator is disclosed in FIG. 2. The device for evaluating the radio channel according to the preamble includes an LS estimator 7 and an accurate estimator consisting of a serial matrix multiplier 8, a temporary memory device 9, a parallel matrix multiplier 10, a resolver 11, and a digital filter 12.

The input of the device LS-assessment 7 receives the input signal of the device evaluating the radio channel. The output of the LS evaluation device 7 is connected to the inputs of the serial matrix multiplier 8 and the digital filter 12. The output of the serial matrix multiplier 8 is connected to the input of the temporary memory device 9. The output of the temporary memory device 9 is connected to the input of the parallel matrix multiplier 10, the output of which is connected to the second input solver 11. The output of the digital filter 12 is connected to the first input of solver 11. The output of solver 11 is the output of the radio channel estimator.

The serial matrix multiplier can be implemented as a block of scalar multipliers connected to a block of accumulating adders. The input of the multipliers are the rows of the matrix R1 and the input values. The block diagram of this implementation is shown in FIG. 3.

A parallel matrix multiplier can be implemented as a block of scalar multipliers, all the outputs of which are connected to the inputs of a multi-input adder. The input of the multipliers are the rows of the matrix R2 and the vector components stored in the temporary memory device. The block diagram of this implementation is shown in FIG. four.

These implementations of the serial and parallel matrix multiplier will be used as a non-limiting example of their implementation when describing the operation of the device.

The device operates as follows.

The input values, one per clock, go to the LS evaluation device 7. An LS assessment is independently performed on each value. One at a time from the LS estimate device, the rough estimate values are received.

The values obtained from the output of the LS-estimation device are sent to the serial matrix multiplier 8. There, the input value received at the ith step of the serial matrix multiplier is multiplied by the i-th row of the matrix R1 on the block of scalar multipliers, and the result is fed to the block of accumulative adders . This process continues all the time while the LS estimator is evaluating a group of pilot symbols. When the LS estimator completes the evaluation of this group of subcarriers, the accumulating adders block will contain the result of multiplying the vector of values of the LS estimates of the group and matrix R1.

After the LS evaluation device has processed all the pilot symbols of the group, the result of the accumulating adder is recorded in the temporary memory device 9.

The values are read from the temporary memory device and fed to a parallel matrix multiplier 10. There, the input vector at the ith step of the parallel matrix multiplier is multiplied elementwise by the ith row of the matrix R2. The result of the multiplication is fed to the multi-input adder. The adder output is a number representing an estimate of the frequency channel on the i-th subcarrier.

Simultaneously with the start of operation of the parallel matrix multiplier, LS estimates of the next group of subcarriers begin to come out of the LS estimator 7. The serial matrix multiplier 8 receives and processes them as described above. This process is repeated until all subcarrier estimates are obtained from the output of the LS estimator.

Simultaneously with the operation of the devices described above, a digital filter 12 is operating, which takes the values of the LS estimate and performs interpolation.

The values from the outputs of the parallel matrix multiplier 10 and the digital filter 12 are supplied to the inputs of the resolving device 11. This device, using a counter, tracks the number of the subcarrier to which the input estimate value corresponds. If the number indicates that the subcarrier is the last in the group, then the resolver outputs the value received from the digital filter. Otherwise, the value received from the parallel matrix multiplier is output. Moreover, if an unbroken preamble is used, the solver always selects a value from a parallel matrix multiplier.

From the description of the operation of the device it follows that it is a conveyor, that is, with a continuous supply of input values to the device after a certain initial delay, it continuously produces evaluation values.

When implementing the described device, the aforementioned low computational complexity is achieved by the fact that a relatively small number of arithmetic devices are used, which ensures the simplicity of the device and the possibility of its execution on a universal processor, digital signal processor (DSP), application-specific specialized integrated circuit (ASIC), field programmable gate array (FPGA) or other programmable logic device, on discrete logic elements or on transistor logic elements, on discrete hardware components, or any combination thereof, made to implement the functions considered.

The consistent and pipelined nature of the calculations allows for low computational complexity and low computational latency. The use of linear polynomial regression allows one to obtain high accuracy and quality estimates of the channel frequency response.

In addition, the matrix multipliers included in the inventive device are configured to be multiplied by fixed, pre-calculated matrices, which can significantly reduce the required amount of calculations in comparison with known methods where matrix equations with non-constant matrices and recursive least squares methods are required.

In FIG. 5 shows the simulation results of the prototype device and the claimed device. Channel characteristics were selected according to the Extended Vehicular A model (LTE; Evolved Universal Terrestrial Radio Access (E-UTRA); User Equipment (UE) radio transmission and reception (3GPP TS 36.101 version 10.1.1 Release 10)). In the prototype device, the number of subcarriers in the group was chosen equal to 4, in the inventive device - 64. The degree of the regression polynomial is 7. The estimation of the SNR and the correlation matrix in the case of the prototype device was assumed to be ideal. Despite this, the quality of the channel estimation by the claimed device coincides with the quality of the evaluation of the prototype at low SNR and exceeds with increasing SNR. In real conditions of operation of communication systems, one should expect a deterioration in the quality of the evaluation of the prototype due to errors in the estimation of channel parameters. In addition, modeling was performed for preambles with a continuous arrangement of pilot subcarriers, since the prototype does not allow the use of others. When using sparse preambles, the SNR increases, and the estimation error decreases, which increases the accuracy and quality of the estimate of the channel frequency characteristics.

Claims (1)

A device for estimating a radio channel according to the preamble, comprising a least squares (LS) estimator and an accurate estimator consisting of a serial matrix multiplier that multiplies the values from the LS estimator with the values of the matrix V *, a temporary memory device parallel to the matrix multiplier multiplying values from a temporary memory device with the values of the matrix US, where the matrices V *, U, and S are the matrices of the singular decomposition of the linear polynomial matrix regression, a solver and a digital filter, the input of the LS estimator being the input of the radio channel estimator according to the preamble, the output of the LS estimator is connected to the inputs of the serial matrix multiplier and digital filter, the output of the serial matrix multiplier is connected to the input of the temporary memory, the output of the temporary memory device is connected to the input of a parallel matrix multiplier, the output of which is connected to the second input of the deciding device, the output of the digital filter is connected Inen with the first input of the resolver, the output of the solver, which receives a signal from the output of a digital filter or from the output of a parallel matrix multiplier, is the output of the radio channel estimator.

Images