WO2011140977A1

WO2011140977A1 - Method and apparatus for frequency offset compensation for an ofdma uplink

Info

Publication number: WO2011140977A1
Application number: PCT/CN2011/073903
Authority: WO
Inventors: James Peroulas; Shaoning Xiao
Original assignee: Zte Wistron Telecom Ab; Zte Corporation
Priority date: 2010-05-14
Filing date: 2011-05-10
Publication date: 2011-11-17

Abstract

An OFDM receiver includes a cyclic prefix removal module, a discrete fourier transform (DFT) module, a frequency offset estimation module, and a frequency offset compensation module. The cyclic prefix removal module is configured to receive a baseband signal and to remove a cyclic prefix of the baseband signal resulting in a modified baseband signal. The DFT module receives the modified baseband signal and produces a set of subcarrier symbols. The frequency offset estimation module produces an initial estimate of a product of a diagonal matrix and a column vector, wherein the diagonal matrix describes a channel response and the column vector represents the subcarrier symbols received by the DFT module. The frequency offset estimation module then produces a signal representing a refined estimate based on the initial estimate. The frequency offset compensation module receives this refined estimate and produces frequency offset compensated symbols.

Description

METHOD AND APPARATUS FOR FREQUENCY OFFSET

COMPENSATION FOR AN OFDMA UPLINK

PRIORITY

[0001] Priority is claimed to PCT Patent Application No. PCT/CN2010/072798 filed May 14, 2010, and PCT Patent Application No. PCT/CN2010/073971 filed June 13, 2010, the disclosures of which are incorporated herein by reference in their entirety

FIELD OF THE INVENTION

[0002] The field of the present invention is orthogonal frequency division multiplexing (OFDM) based communications systems, and in particular, to a process of performing frequency offset compensation on a received signal.

BACKGROUND

[0003] OFDM communications systems are sensitive to frequency offsets between the transmitter and the receiver. As such, mechanisms are in place to estimate the frequency offset and also to correct the frequency offset. Several frequency offset estimation methods are known to those skilled in the art.

[0004] An example of such an estimation method operates on a frequency domain signal R, which is received after an N Fast Fourier Transform (NFFT) point OFDM Discrete Fourier Transformation (DFT) operation. When a single mobile station is transmitting to a single base station, after passing through a frequency selective channel with a frequency offset, R can be expressed as:

where H is an Λ/FFT /VFFT diagonal matrix describing the channel response at each subcarrier, is an Λ/FFT point column vector containing the subcarrier symbols transmitted by the user equipment (with zeroes in the positions of the subcarriers where the transmitter is not transmitting any information), N is an Λ/FFT point Additive White Gaussian Noise (AWGN) column vector with variance

full matrix. This full matrix is defined as:

[0005] models the effect of the frequency offset on the received

signal. If the frequency offset is 0, then is simply an identity matrix I. The

frequency offset f is expressed as a multiple of the OFDM subcarrier spacing.

[0006] A base station receiving a signal from multiple users presents a different scenario. For example, let T represent a set containing indices of subcarriers on which the mobile station is transmitting, and let ST be a diagonal matrix that only contains ones in the diagonal elements indicated by the set T. If T = {10, 1 1 , 12}, then ST contains a "1 " at coordinates (10, 10), (1 1 , 1 1 ), and (12,12). All other entries contain the value "0".

[0007] If a mobile station is transmitting on subcarriers T1 , the signal received at the receiver's antenna, before receiver noise is added, will be:

where indicates keeping columns T₁ of and setting the

remaining columns to zero.

Note that:

The overall signal, including all mobile stations and noise, at the receiver is:

Because of Equation 6, one can simplify the above equations to:

where:

[0008] The

matrix can be thought of as a collection of sets of mutually orthogonal columns. The columns of the matrix associated with user n are mutually orthogonal. However, the columns associated with user m are only 'mostly' orthogonal to the columns of user n (for m not equal to n).

[0009] The base station receiver is interested in estimating

which represents the product of the channel response and the transmitted subcarrier symbols. One technique for estimating Z is by directly inverting and applying the result to Equation 9, which results in the following:

[0010] Empirical simulations have shown that

can be inverted if the frequency offsets of the user equipment are moderate (less than 0.4 times the subcarrier spacing), and if all the subcarriers are occupied. Empirical simulations have also shown that

is quite well behaved and has a very low condition number that is almost always less than 10. If a subcarrier is not actually occupied by a mobile transmitter, for the purposes of constructing an invertible , one can simply assume that the subcarrier is actually occupied by a mobile transmitter with a frequency offset of 0.

[0011] Although inverting

is numerically possible, in a modern communications system it may be impractical to invert such a matrix. For example, in an LTE uplink, there may be 2048 subcarriers and the scheduling assignments may change every 0.5 ms. Thus, inverting this matrix in an LTE uplink would require inverting a 2048x2048 matrix every 0.5 ms.

[0012] It would therefore be beneficial to perform accurate frequency offset compensation without employing a full matrix inversion.

SUMMARY OF THE INVENTION

[0013] The present invention is directed toward an apparatus and method for frequency offset compensation in an OFDM receiver.

[0014] In a first separate aspect of the present invention, to perform frequency offset compensation on a baseband signal, an OFDM receiver includes a cyclic prefix removal module, a discrete fourier transform (DFT) module, a frequency offset estimation module, and a frequency offset compensation module. The cyclic prefix removal module is configured to receive a baseband signal and to remove a cyclic prefix of the baseband signal resulting in a modified baseband signal. The DFT module receives the modified baseband signal and produces a set of subcarrier symbols from the modified baseband signal. The frequency offset estimation module produces an initial estimate of a product of a diagonal matrix and a column vector, wherein the diagonal matrix describes a channel response and the column vector represents the subcarrier symbols received by the DFT module. The frequency offset estimation module then produces a signal representing a refined estimate based on the initial estimate. The frequency offset compensation module receives this refined estimate and produces frequency offset compensated symbols.

[0015] In a second aspect of the present invention, the refined estimate is calculated by performing iterations of: subtracting from a product of the diagonal matrix and the column vector, a product of a corrective matrix and a previous refined estimate to obtain a current refined estimate; and setting the current refined estimate equal to the previous refined estimate. This iterative procedure can continue until a fixed number of iterations are performed or until a performance criterion is achieved.

[0016] Additional aspects and advantages of the improvements will appear from the description of the preferred embodiment. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Embodiments of the present invention are illustrated by way of the accompanying drawings, in which:

[0018] Figure 1 is a block diagram of a typical OFDM receiver.

[0019] Figure 2 is a block diagram illustrating an improved OFDM receiver according to an aspect of the present invention. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0020] Referring to Figure 1 , a block diagram of an OFDM receiver as known in the art comprises a cyclic prefix (CP) removal module 3, a Discrete Fourier Transform (DFT) module 6, an OFDM based frequency offset estimation module 8, a frequency offset compensation module 9, and a symbol processing module 10. The CP removal module 3 receives a baseband signal 1 , which is a baseband representation of a signal received from several users. The CP removal module 3 removes and discards the CP from the received signal 1 so that only the OFDM symbol remains. The received signal 1 , which now only consists of the OFDM symbol, is then sent through the DFT module 6, which produces a set of subcarrier symbols that are further processed by the OFDM based frequency offset estimation module 8.

[0021] As is known to those skilled in the art, the frequency offset estimation module 8 estimates the frequency offset of each received signal 1 and forwards these estimates to the frequency offset compensation module 9. The frequency offset compensation module 9 compensates for the frequency offset of each of the transmitters associated with the respective received signals, and forwards the frequency offset compensated signals to the symbol processing module 10.

[0022] The symbol processing module 10 processes the frequency offset compensated signals. This processing may involve channel estimation, channel compensation, demodulation, and block decoding before the final processed data is declared as having been properly received. [0023] Referring now to Figure 2, an improved OFDM receiver that may be used to implement aspects of the present invention is illustrated. This OFDM receiver includes a baseband signal 1 1 , a CP removal module 13, a DFT module 16, a frequency offset estimation module 18, an improved frequency offset compensation module 19, and a symbol processing module 20.

[0024] In a situation where a base station receives a signal from multiple users, the overall signal, including all mobile stations and noise experienced at the receiver, can be represented by Equation 9. As noted above, the base station receiver is interested in estimating Z, and one method of estimating Z is by directly inverting a

matrix, and applying the result of this inversion to Equation 9.

[0025] According to aspects of the invention, an initial estimate Z is first created, and then further refined. For example, let

represent the initial estimate which is calculated according to:

This initial estimate is refined to generate for n > 0 according to:

where:

where C is a corrective matrix.

[0026] Instead of having to first generate

and

and then multiplying, a Q matrix can be calculated directly. For example, if F(n) represents the frequency offset of the user that is transmitting on subcarrier n, then:

[0027] The iterative procedure described in Equation 15 can continue until a fixed number of iterations have been performed, or until some other performance criterion has been reached. For example, when a mobile station is transmitting a known reference signal, iterations can be performed until the known reference signal can be clearly decoded at the receiver.

[0028] According to numerical simulations, it was found that, by using only a small number of iterations (2 - 5), the difference between

(Z applying the iterative procedure) and

(Z using matrix inversion) was very small. Therefore, estimating Z by applying the iterative technique approaches the similar level of performance of the matrix inversion method, without requiring the matrix inversion. This iterative technique can be implemented as a series of matrix to vector multiplications which have an implementation cost significantly lower than the cost of inverting a matrix.

[0029] Taking into consideration the fact that the Q and

matrices have values that approach zero as one gets further from their respective main diagonals, these matrices can be approximated by keeping only some their respective diagonals.

[0030] Let the notation

represent keeping the main diagonal of matrix A, the d diagonals above the main diagonal of matrix A, and the d diagonals below the main diagonal of matrix A. All other entries are set to zero. In total, 2d + 1 diagonals of matrix A are retained. [0031] This aspect of the present invention can be described as:

[0032] By way of non-limiting example only, the di , d₂, and d₃ parameters are chosen based on simulation results and on the specifics of the OFDM receiver in question. For example, in an LTE system with 2048 subcarriers, satisfactory performance can be achieved if all of these parameters are set to the same value of 50. Different OFDM receivers will be able to use different values for these parameters. This aspect of the invention thus has an advantage that several of the matrix to vector multiplications have been simplified. Also, the implementation cost is reduced because many of the constants in several matrices have been set to 0.

[0033] Equation 14 and Equation 15 can be solved simultaneously to remove the recursion, yielding the following equations:

[0034] In other words, one can arrive at the result of the nth recursion step directly, without first performing steps 0 through n -1 . Only the first 3 iterations of the algorithm are shown in the above equations. However, one can continue expanding the equations to obtain equations for further iterations of the algorithm.

[0035] Although Equation 25 and Equation 26 contain powers of Q implying matrix to matrix multiplications, when it comes to implementing these equations, no matrix to matrix multiplications are needed. For example, to implement Equation 25, one can make several matrix to vector multiplications.

For example,

is calculated resulting in "T". QT can then be calculated resulting in "U". Next, QU can be calculated, resulting in "V". Lastly, V-3U+3T can be calculated to arrive at

without using any matrix to matrix multiplications.

[0036] As discussed above, aspects of the present invention recognize the fact that the Q and

matrices have values that approach zero as one gets further from their respective main diagonal. Thus, these matrices can also be approximated by keeping only some of their main diagonals.

[0037] This aspect of the present invention can be described as follows:

[0038] Similar to above, only the first 3 iterations of the algorithm are shown in the above equations. However, one can continue expanding the equations to obtain equations for further iterations of the algorithm.

[0039] While embodiments of this invention have been shown and described, it will be apparent to those skilled in the art that many more modifications are possible without departing from the inventive concepts herein. The invention, therefore, is not to be restricted except in the spirit of the following claims.

Claims

CLAIMS What is claimed is:

1 . An orthogonal frequency division multiplexing (OFDM) receiver for performing frequency offset compensation on a baseband signal, the OFDM receiver comprising:

a cyclic prefix removal module configured to:

receive the baseband signal; and

remove a cyclic prefix of the baseband signal to produce a modified baseband signal;

a discrete fourier transform (DFT) module configured to:

receive the modified baseband signal; and

produce a set of subcarrier symbols based on the modified baseband signal;

a frequency offset estimation module configured to:

receive the set of subcarrier symbols; and

provide an initial estimate of a product of a diagonal matrix and a column vector, wherein the diagonal matrix describes a channel response and the column vector is the set of subcarrier symbols received from the DFT module; and

produce a signal representing a refined estimate based on the initial estimate; and

a frequency offset compensation module configured to produce frequency offset compensated symbols based on the signal representing the refined estimate.

2. The OFDM receiver of claim 1 , wherein the refined estimate is calculated by performing at least one iteration of:

subtracting from the product of the diagonal matrix and the column vector, a product of a corrective matrix and a previous refined estimate to obtain a current refined estimate; and

setting the current refined estimate equal to the previous refined estimate.

3. The OFDM receiver of claim 2, wherein the frequency offset estimation module is configured to perform multiple iterations until a fixed number of iterations are performed or until a predetermined performance criterion is achieved.

4. The OFDM receiver of claim 3, wherein the predetermined

performance criterion is based upon a known test signal being clearly decoded at the OFDM receiver.

5. The OFDM receiver of claim 1 , wherein the diagonal matrix and column vector are transmitted in the baseband signal.

6. The OFDM receiver of claim 1 , wherein the baseband signal R is calculated as

^, herein H_c is the diagonal matrix, X_c is the column vector;

models an effect of the frequency offset, and N is a column vector with variance

7. The OFDM receiver of claim 6, wherein the initial estimate is calculated according to

, wherein represents the initial estimate and

represents a channel response.

8. The OFDM receiver of claim 7, wherein the initial estimate is refined to generate the refined estimate

for n>0 according to

wherein C=Q-I, where C is a corrective matrix, and Q is calculated directly according to

_where F(n) represents the frequency offset of user equipment that is transmitting on subcarrier n.

9. The OFDM receiver of claim 6, wherein the initial estimate is calculated according to

, wherein

represents the initial estimate and di is chosen based on the OFDM receiver.

10. The OFDM receiver of claim 9, wherein the initial estimate is refined to generate for n>0 according to:

wherein corrective matrix , where represents a

channel response, and d₂, and d₃ are chosen based on the OFDM receiver.

1 1 . The OFDM receiver of claim 6, wherein the initial estimate is calculated according to

, wherein represents a channel

response, and wherein represents the initial estimate, and the initial

estimate is simultaneously refined according to a predetermined equation associated with a predetermined number of iterations.

12. The OFDM receiver of claim 6, wherein the initial estimate is calculated according to , wherein represents the initial estimate,

represents the channel response, and the initial estimate is simultaneously refined according to a predetermined equation associated with a

predetermined number of iterations.

13. A method for performing frequency offset compensation on a baseband signal received at an orthogonal frequency division multiplexing (OFDM) receiver, the method comprising:

receiving, at a cyclic prefix removal module, the baseband signal; and removing, at the cyclic prefix removal module, a cyclic prefix of the baseband signal to produce a modified baseband signal;

producing a set of subcarrier symbols from the modified baseband signal;

transmitting a signal representing an initial estimate of a product of a diagonal matrix and a column vector, wherein the diagonal matrix describes a channel response and the column vector is the set of subcarrier symbols; transmitting a signal representing a refined estimate based on the initial estimate; and

producing frequency offset compensated symbols based on the signal representing the refined estimate.

14. The method of claim 13, wherein the refined estimate is calculated by performing at least one iteration of:

setting the current refined estimate equal to the previous refined estimate.

15. The method of claim 14, further comprising performing multiple iterations until a fixed number of iterations are performed or until a

predetermined performance criterion is achieved.

16. The method of claim 15, wherein the predetermined performance criterion is based upon a known test signal being clearly decoded at the OFDM receiver.

17. The method of claim 13, wherein the diagonal matrix and column vector are transmitted in the baseband signal.

18. The method of claim 13, wherein the baseband signal R is expressed _as

^, herein H_c is the diagonal matrix, X_c is the column vector; models an effect of the frequency offset, and Λ/ is a column vector with variance 19. The method of claim 18, wherein the initial estimate is calculated according to

, wherein represents the initial estimate, and

represents a channel response.

20. The method of claim 19, wherein the initial estimate is refined to generate for n>0 according to

wherein corrective matrix

C=Q-I, and the Q matrix is calculated directly according to

_wh_ere (_n) represents a frequency offset of user equipment that is transmitting on subcarrier n.

21 . The method of claim 18, wherein the initial estimate is calculated according to

wherein represents a channel response, and

wherein represents the initial estimate and d1 is chosen based on the OFDM receiver.

22. The method of claim 21 , wherein the initial estimate is refined to generate for n>0 according to: wherein

corrective matrix , and d₂, and d₃ are chosen based on

the OFDM receiver.

23. The method of claim 18, wherein the initial estimate is calculated according to represents a channel response, and

wherein represents the initial estimate, and the initial estimate is simultaneously refined according to a predetermined equation

associated with a predetermined number of iterations.

24. The method of claim 18, wherein the initial estimate is calculated according to , wherein represents a channel response, and

wherein represents the initial estimate, and the initial estimate is simultaneously refined according to a predetermined equation associated with a predetermined number of iterations.

25. An apparatus for performing frequency offset compensation on a baseband signal received at an orthogonal frequency division multiplexing (OFDM) receiver, the method comprising:

means for receiving, at a cyclic prefix removal module, the baseband signal;

means for removing, at the cyclic prefix removal module, a cyclic prefix of the baseband signal to produce a modified baseband signal;

means for producing a set of subcarrier symbols from the modified baseband signal;

means for transmitting a signal representing an initial estimate of a product of a diagonal matrix and a column vector, wherein the diagonal matrix describes a channel response and the column vector is the set of subcarrier symbols;

means for transmitting a signal representing a refined estimate based on the initial estimate; and

means for producing frequency offset compensated symbols based on the signal representing the refined estimate.

26. The apparatus of claim 25, wherein the refined estimate is calculated by performing at least one iteration of: subtracting from the product of the diagonal matrix and the column vector, a product of a corrective matrix and a previous refined estimate to obtain a current refined estimate; and

setting the current refined estimate equal to the previous refined estimate.

27. The apparatus of claim 26, further comprising performing multiple iterations until a fixed number of iterations are performed or until a

predetermined performance criterion is achieved.

28. The apparatus of claim 27, wherein the predetermined performance criterion is based upon a known test signal being clearly decoded at the

OFDM receiver.

29. The apparatus of claim 25, wherein the diagonal matrix and column vector are transmitted in the baseband signal.

30. The apparatus of claim 25, wherein the baseband signal R is expressed as

^, _w erein H_c is the diagonal matrix, X_c is the column vector; models an effect of the frequency offset, and Λ/ is a column vector with variance

31 . The apparatus of claim 30, wherein the initial estimate is calculated according to

, wherein

represents the initial estimate, and wherein represents a channel response.

32. The apparatus of claim 31 , wherein the initial estimate is refined to generate for n>0 according to represents a channel

response, wherein a corrective matrix C = Q - 1, and the Q matrix is calculated directly according to

_where F(n) represents a frequency offset of user equipment that is transmitting on subcarrier n.

33. The apparatus of claim 30, wherein the initial estimate is calculated according to

where represents a channel response,

34. The apparatus of claim 33, wherein the initial estimate is refined to generate for n>0 according to: , wherein

corrective matrix , and d₂, and d₃ are chosen based on

the OFDM receiver.

35. The apparatus of claim 30, wherein the initial estimate is

calculated according to where represents a channel

response, and wherein represents the initial estimate, and the initial

36. The apparatus of claim 30, wherein the initial estimate is

calculated according to , where represents a channel

response, and wherein represents the initial estimate, and the initial