TWI269540B - Iterative fast fourier transform error correction - Google Patents

Abstract

Data is to be estimated from a received plurality of data signals in a code division multiple access communication system. The data signals are transmitted in a shared spectrum at substantially a same time. A combined signal of the transmitted data signals are received over the shared spectrum and sampled. A channel response for the transmitted data signals is estimated. Data of the data signals is estimated using the samples and the estimated channel response. The data estimation uses a Fourier transform based data estimating approach. An error in the data estimation introduced from a circulant approximation used in the Fourier transform based approach is iteratively reduced.

Description

</ RTI> </ RTI> <RTIgt; </ RTI> <RTIgt; </ RTI> <RTIgt; </ RTI> <RTIgt; </ RTI> <RTIgt; </ RTI> <RTIgt; The present invention generally relates to wireless communication systems. The invention particularly relates to data detection in a wireless communication system. The first figure illustrates a wireless communication system 10. The wireless communication system 10 has base stations 121 through 125 (12) in communication with user equipment (UEs M^ through 143 (14). Each base station 12 has an associated operational area in communication with UEs 14 in its operational area. • In some communication systems, such as code division multiple access (CDMA) and time division duplex (TDD/CDMA) using code division multiple access, multiple communications are transmitted over the same spectrum. The channel code is distinguished. To use this spectrum more efficiently, the TDD/CDMA communication system uses a frame that is divided into several time slots for communication. The communication transmitted in this system will have one. Or a plurality of codes and time slots assigned to themselves. The use of a code in a time slot is referred to as a source unit. Because multiple communications can be in the same spectrum at the same time. φ is transmitted, and the receiver in this system must distinguish the difference between multiple communications. One method for detecting such signals is multiuser detection (MUD). In MUD, with all UES User information of 14 At the same time, it is detected. Another method of detecting multi-code transmission from a single transmission is single user detection (SUD). In SUD, in order to restore the data transmitted from the receiver by multi-code, The received signal is despreaded by an equalization stage and using multiple codes. Methods for implementing levels such as MUD and SUD include using (::11〇81^ or Cholesky decomposition. These methods have a height Complexity. This high level of complexity leads to power consumption and reduces the battery life of 1269540 at the user end of UE 14. To reduce this complexity, MUD and SUD use Fast Fourier Transform (FFT) based methods. Has been developed. In some FFT methods, an approximation is used to facilitate the implementation of the FFT. This approximation produces small errors that are introduced into the estimate. Therefore, 'hope that there are other ways to detect the received Information. In the one-code multiple access system, the data will be received from the multiple

The signal is estimated. These data signals are transmitted over a common spectrum at substantially the same time. The signals combined with one of the transmitted data signals are received and sampled on the shared spectrum. A channel response is estimated for these transmitted data signals. Data signals are estimated using these sampled and estimated channel responses. This data estimate uses a Fourier transform-based data estimation method. The estimated error of the data generated from the cyclic approximation used in the method based on Fourier transform 1 is repeatedly reduced. BRIEF DESCRIPTION OF THE DRAWINGS The first figure is a wireless communication system. The second diagram is a simplified transmitter and a data detection receiver based on the correction of the error. The second picture is a flow chart of repeated error correction. The five maps selectively use a flowchart of the receiver that corrects the error correction. ^Six diagrams are flow diagrams of an example of an FFT-based SUD using repeated error correction. = Seven Diagrams is a flow chart of a FFT-based MUD using repeated error correction. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS 1269540 The second diagram illustrates a simplified transmitter 26 and receiver 28 using FFT-based data detection with inverse error correction in a TDD/CDMA communication system, although repeated error correction can be applied. For other systems, such as Frequency Division Duplex (FDD) CDMA. In a typical system, a transmitter is located in each UE 14 and multiple transmission circuits 26 that carry multiple communications are located in each base station 12. This reverse error correction receiver 28 can be in at least one base station 12, UE 14 or both. Transmitter 26 transmits data over radio frequency channel 30. The data generator 32 in the transmitter 26 produces the data to be transmitted to the receiver 28. A modulation/spreading/training device 34 expands the data and causes the expanded reference material and a midamble sequence to be multiplexed in a suitable time slot and code. A typical communication burst 16 has an intermediate walk 20, a guard section 18 and two data areas 22, 24, as shown in the third figure. The intermediate walk 20 separates the two data areas 22, 24, and the interval 18 separates the communication bursts to allow for differences in the arrival times of the bursts from the different transmitters 26. The two data areas 22, 24 contain information on the communication burst. This communication burst is modulated by the modulator 36 into radio frequency (RF). Antenna 38 transmits the RF signal to receiver 28 via radio frequency channel 30. The modulation pattern used to transmit communications can be any method known in the art, such as quadrature phase shift keying (QPSK) or quad matrix quad amplitude modulating (QAM). The antenna 40 of the 〇 receiver 28 receives a plurality of radio frequency signals. The received signal is demodulated by demodulator 42 to produce a fundamental frequency signal. The baseband signal is sampled by a sampling device 43, such as one or more analog to digital converters, to be sampled at a chip rate or a multiple chip 1269540 rate of the transmitted burst. The samples are processed, for example, by a channel estimation device 44 and an FFT-based data detector 46, in a time slot and assigned to the received bursty suitable code. Channel estimation device 44 uses the intermediate walk-through sequence components in the baseband samples to provide channel homing, such as channel impulse response. This channel impulse response can be viewed as a matrix H. The channel information and spreading code used by the transmitter are used by the data detecting device 46 to estimate the transmitted data of the received communication burst as a soft symbol. The error correction device 48 processes the estimate to correct errors resulting from the FFT-based detection.

Although the repeated error detection system uses the third generation cooperation project (3GPP) universal terrestrial radio access (UTRA) TDD system as the following communication system, it can be applied to other systems or others. FFT linear equation based application. This system is always a sequence of wideband CDMA (W-CDMA) systems in which uplink (UpHnk) and downlink (Shlink) transmissions are limited to mutually exclusive time slots. Data detection is generally based on a linear equation based on Equation 1. Program 1 is for usg and §, data detection is generally based on programs 2 and 3. r = n s + n , Cheng; 2 &quot; A program 3 Z: is used by the sampling device 43 to receive the extended data vector. Such as the expansion of the program 3, the fight Λ 应 should T T; Original transmission (four). The extended continuous vector light channel (3) is used to solve the minimum mean square error of the program 1 according to the programs 4 and 5 ο Η · -, 'program * 1269540 program 5 (· ) H represents the common transpose matrix function. Σ2 is the standard offset as determined by channel estimation device 44. I is the identity matrix 〇 Program 4 operates like a channel equalization level, while Program 5 acts like a diffusion. A cross-channel correlation matrix is defined by program 6. R^HNH+a2I Program 6 This linear program needs to be answered according to program 7. i? J = J The program is based on program 8. 1 = Uhl program 8

Although R is not a cycle of multiple chip rate sampling, a portion of R is cyclic. This loop is partially generated by eliminating the bottom and upper columns. W is the length of the channel impulse response. By approximating R as a circulant matrix Rcir, Rcir is decomposed via a Fourier transform, as in program 9.

R cir

P

DPADP program 9

Using a row of the R matrix to approximate the cyclic version of R can be determined by, for example, program 10. The extended

Hi) Program 10 (the first line of Rh.R, and (the first line of Hh.H, although any line can be used to replace the line. Preferably, the line from the left and right of the most W is Use, because these lines have more non-zero elements. For MUD, data detection is generally based on program 11. ϋ = Α ± + η program η 1269540 ^ should be matrix. The symbol response matrix is transmitted by the burst The frequency is multiplied by the mother's channel response. The MNSE method for answering private 11 is based on programs 12 and 13. λΗ program 12 _ refers to the interactive correlation matrix. %^13 The solution to the linear program is based on the program 14

The I program 14 is based on the program 15. l-AHr A program 15 If the elements of R are divided into K and K blocks (KbyK), the structure of R is 2 in the block cycle °K is the number of bursts arriving at the same time. &amp; bursts over each other during the period. For the 3GPP UTRATDD system, each data area of a time slot corresponds to an observation period. Loops using R Approximate Rbcir' Rbcir is decomposed via block Fourier transform, as in the program 16 〇

Rbcir ^D;]AD program 16 p^jD; ADp

Dp is based on program 17. The program 17 dp ^^N ^IKiP^K^Ns DN is an N-point Fourier matrix and the unit matrix of the 尺寸 system size κ. Ten represents the product of Kronecker (kr〇necker). Ns is the number of data symbols in a data area. A diagonal matrix of a block. Block a is iy?cz&gt;( :,1 : K). Therefore, this is program 18. The eight-tone program 18 uses FFrs, and the data vector is determined by the program 19. Disadvantages (4) Program 19 1269540 Therefore the FFT of '4 is determined. The data is estimated by taking the inverse FFT of J. To improve accuracy, the dimension of R can be extended to include the impulse response of the last symbol. The impulse response of the last symbol extends into the middle walking or interval interval. To capture the response of the last symbol, the block loop structure of R is another W-1 chip that is extended. w is the length of the impulse response. The same FFT method is executed by using an extended R matrix and by using an extension from the middle (after the middle walk is canceled) or the interval interval

If a prime factor algorithm (PFA) FFT is used, the R matrix can be extended so that the most efficient length P of P can be used. The R matrix is similarly extended using intermediate walk or interval information. While repeated error correction is explained in connection with the particular implementation of SUD and MUD, it can be applied to any FFT-based solution to a linear program that uses a cyclic approximation, substantially as described below. Program 1 is a general linear program. 2^y Program 1 is to determine X ’ use program 2 0. Ly program 2〇 The inverse matrix Z is a complex matrix. By approximating 2: to a cyclic matrix Zcir, Z is determined by the FFT decomposition of programs 21 and 22. ^dr =DplADp =~2)pADp P program 21

2-; =2) 71 ADP f Program 22 If Z is a block loop matrix, the block Fourier transform is used instead, using programs like programs 21 and 22. The cyclic approximation of &quot; z produces an approximation error. The difference between z and Zcir is based on program 23. Λ 11 1269540 Z = Zcir Program 23 Δ z is the difference matrix between z and zcir. Using this loop and the difference matrix, program 1 becomes program 24. (Zdr - Αζ) χ = ι program 24 The program 25 is generated by rearranging the program 24. ^Zcirl + ZciAzl program 25

A repetitive method can be used to resolve the program 25, depending on the programs 26, 27 and 28. χ (Ο=χΜ+^-0 Program 26 Program 27^-ι)=ζ2αζχ^ Program 28 Using programs 26, 27 and 28, Wang is solved as follows. The initial solution of the foot (()) is determined by the FFTs of the program 21 or 22 and the cyclic approximation, as shown in Equation 29. τ{0) one 7~ Wang Yizhen program 29

Afo) The initial error correction project is determined using program 30. 4}=^ΑζΙ1°) Program 3〇 According to the program 31, the initial error correction term L is added to the initial solution x(G). £(1)=^(〇)+4〇) The program 31 repeats N times according to the programs 32 and 33. Program 32 Wang (4) = /) + must, no = 1, 2, ·, · # a 1 program 33 12 1269540 忑 (9) is used to estimate 1. Fees are being processed to increase the force generated, the second winter can be based on the over-resolved error. This repetition continues the ancient repetitive f, N, can be not fixed, the difference is - the I value I until the left () is Equivalent values (solution convergence) or Bu's according to equations 34 and 35, respectively, 34 programs 35

&lt;T T is a critical value. The point is that it is optional. The initial peak, the additional accuracy of the friend's decision is the same estimate. Therefore, if you do not want to reverse the ambiguity, you will not be confronted or hope to have extra traits. However, if it is needed, it will be executed. Therefore, after % i corrects the appropriate balance between the additional complexity: 3⁄4 error and complexity noise addition and subtraction 喊 ' ' 反 反 反 反 反 反 反 反 反 反 反 反 反 反 反 反 反 反 反 反 反 反 反 反 反 反 反 反 反 反 反 , , , , , , , , SU SU The linear program of the group t is based on the program 36. The solution 36 2 and the system are based on the programs 37 and 13, respectively. External B program 37 ^-ίΓ^+σ2/ Program 13 In the connection between program 36 and program 1, R corresponds to Ζ, 济应, and 卫对庶. Using R's cyclic approximation Rdi, the implementation of the repeated error correction is determined as an initial extended data estimate β3 using FFTs, such as the program 38 households ^, Ming. ϋ 13 1269540

The program 38 is determined by an initial error correction project using the program 39. 4〇)=^&quot;γΔλ1{〇) Program 39 △ π is the difference between R and RCir. The initial error correction item is added to the initial solution ^(()), according to Equation 40. Meng (1) = Li (.) + △?) Program 40

Repeatedly, the programs 41 and 42 are repeated N times. Program 41 e+OqW+f,^1} 2...· #一1 Program 42 The data element is determined using Nth's repeated estimate of the extended data vector/% by using the channel that transmits the burst as shown in the program 43. Code solution extension. i = CHi(4) Program 43 Applying the Overlap Error Correction to MUD's MMSE Solution t, Resolved Program 44 Program 45 Program 46 In the relationship between Program 1 and Program 44, R corresponds to Z, 坌 corresponds to 而 and y corresponds to y. The Rb c i r is approximated using the block entropy of R, and the execution of the repeated error correction is as follows. An initial data estimate ί/(()) is determined using FFTs, as shown in program 47. Il〇)=^l The linear equation of program 47 is based on program 44. Rd^^y Wei and R are based on programs 45 and 46, respectively.

R = AHA + a2I The initial error correction item is determined using program 48. 14 1269540 △(aJ =dA^(0) Program 48 △ π is the difference between R and Rbcir. The initial error correction term is added to the initial solution i(G), according to program 49. Program 49 i(1) a) +4 ?) Repeatedly according to programs 50 and 51 are repeated N times. △(;)=〇〆(9) Program 50 々=1,2, ‘,· iv program 51

It is estimated that the data symbol is the Nth and the estimated data symbol/N). The continuous method is also used if the R array is extended to achieve the last symbol impulse response or extended to an efficient PFA length. 15 1269540 [Simple description of the diagram] The first diagram is a wireless communication system. The second diagram is a simplified transmitter and a Fourier-based data detection receiver using repetitive error correction. ^ The third figure illustrates a communication burst. The fourth chart is a flow chart for repetitive error correction. The fifth diagram is a flow diagram of a receiver that selectively uses a correction error correction. The sixth diagram is a flow chart of an example of an FFT-based SUD using repeated error correction. • Figure 7 is a flow chart of an example of an FFT-based MUD using repeated error correction. [Main component symbol description] 1〇: wireless communication system 12^125: base station 14^443: user equipment (UEs) 16: communication burst 18 • interval section (guardperiod) 20 · middle movement a 22, 24 · one Breeding area 26: Wheeler 28: Receiver 30: Radio frequency channel 32: Data generator 36: Modulator 38, 40: Antenna 42: Demodulator handle: Sampling device 44: Channel estimation device 46: Data detection Detector 48 ·•Reverse error correction device 16

Claims (1)

X. Patent application scope: 1. A base station for a code division multiple access communication system, the base station comprising: , an antenna, arranged to receive a combined signal of the transmitted signal, the transmitted data signal is The shared spectrum is transmitted at substantially the same time; a device is configured to estimate a channel response of the transmitted data signal; the bin estimating device is configured to use the sample, the estimated channel response, and the rate of determination The Fourier transform-based poor material estimation method estimates the data of the transmitted data message; and, for the difference correction device, is set to reduce a cycle approximation from the Fourier transform-based data estimation method resulting in the data estimation An error. 2. A base station as claimed in claim 1 wherein the error correction reduces the error in the estimate of the data. 3. A base station as claimed in claim 1 wherein the error correction means is arranged to be selectively used to reduce the error in the estimate of the data. 4. The base station of claim 1 wherein the error correction means is arranged to determine an error correction term by using a second non-approximation matrix and a cyclic approximation moment used by the data estimating means, And the mosquito-the data-estimating device outputs a -first-estimated output' and adds the error correction term to the first-estimate to reduce the error as a next estimate. 5. The patent of the claim 1 is the base station of item 1, wherein the error correction means is set to use the next error to repeatedly generate subsequent estimates. 7. The base station of claim 1 wherein the error correction means is set to reflect and subsequently estimate the last two of the subsequent estimates. In the scope of the patent, item 1, the floor is adjusted, and the device is adjusted to estimate that the difference between the last two estimates of the straight dragon and the like is less than a critical value.