WO2015154801A1

WO2015154801A1 - Device for estimating frequency offset in ofdm and method thereof

Info

Publication number: WO2015154801A1
Application number: PCT/EP2014/057098
Authority: WO
Inventors: Basuki Endah Priyanto; Fredrik RUSEK; Gengshi Wu; Sha HU
Original assignee: Huawei Technologies Co., Ltd.
Priority date: 2014-04-08
Filing date: 2014-04-08
Publication date: 2015-10-15
Also published as: EP3130122A1; CN105122752B; CN105122752A

Abstract

The present invention relates to a device for estimating frequency offset of a received. The device is adapted to: receive a signal comprising at least one Orthogonal Frequency Division Multiplexing, OFDM, symbol pair transmitted over a radio channel;extend a frequency capture range used for frequency offset estimation of said received signal by frequency shifting said received signal at least once in the frequency domain; and estimate a frequency offset Formula (I) of said received signal based on the extended frequency capture range. Furthermore, the present invention also relates to a corresponding method, a communication device comprising such a device, a computer program, and a computer program product.

Description

DEVICE FOR ESTIMATING FREQUENCY OFFSET IN OFDM

AND METHOD THEREOF

Technical Field

The present invention relates to a device for estimating frequency offset of a received multi carrier signal. Furthermore, the present invention also relates to a corresponding method, a communication device comprising such a estimation device, a computer program, and a computer program product.

Background

Orthogonal Frequency Division Multiplexing (OFDM) has been widely used in the recent wireless communication standards, including 4G Long Term Evolution (LTE), Wi-Fi, and Worldwide Interoperability for Microwave Access (WiMax). The OFDM technique enables high speed wireless data transmission by allowing spectrum overlap of multi-carrier transmissions and thus increases the spectral efficiency. On the other hand, the OFDM transmission systems become more sensitive to frequency offset than conventional single carrier transmission systems. OFDM requires the transmitted signal to be sampled at the centre frequency of each sub-carrier as shown in Fig. 1 . A frequency offset introduces signal degradation, phase rotation and Inter-Carrier Interference (ICI) which will lead to performance degradation. Therefore, frequency offset estimation and correction of frequency offset play an important role in an OFDM based receiver.

Frequency offset is known as a frequency drift between the carrier frequency of transmitter and receiver and is commonly known as Carrier Frequency Offset (CFO) £_CF0. CFO can be further divided into two parts, Integer Frequency Offset (IFO) and Fractional Frequency Offset (FFO), \.e:. &_CFO = &_IFO +&_FFO where &_IFO is an integer multiplied by the sub-carrier spacing and z_FFO is limited in magnitude to half the sub-carrier spacing. For example, LTE has a frequency spacing of 15 kHz. Thus, the IFO can be ± ΛΓ15 kHz, where N is an integer. FFO is limited within ± 7.5 kHz or ± 0.5 in normalized frequency offset. The user terminal, or user device, or also known as User Equipment (UE) in LTE, needs to estimate and correct the frequency offset when processing a received OFDM signal. The process is typically divided into two steps, namely acquisition stage and tracking stage.

The acquisition stage is targeting to estimate coarse frequency error and it is usually performed in every radio frame (e.g., 10 ms duration). On the other hand the tracking stage is designed to estimate fine frequency offset and is performed more frequently which can be every sub-frame e.g., 1 ms duration in LTE.

The typical operation in a receiver is shown in Fig. 2. The antenna 1 1 receives a transmitted signal and processes the received signal by amplifying the signal in the radio front-end and down converting the carrier frequency to a baseband signal. The digital baseband unit thereafter performs many operations including the frequency offset estimation described above. The frequency offset estimation output 12 is then used to adjust the receiver's local oscillator carrier frequency. A low cost crystal oscillator at the receiver can introduce large frequency deviations, especially if there is a temperature change in the receiver. Moreover, Discontinuous Reception (DRX) in LTE where the radio front-end is frequently switched on and off can also introduce a large frequency offset which reduces the receiver performance.

LTE UEs are designed to estimate and correct the frequency error to mitigate the effects of the same. Frequency offset estimation can be performed in the time domain or in the frequency domain. The time domain solution typically utilizes: the received dedicated training symbols/preamble sent by transmitter (e.g., a base station or an access point) to assist synchronization; and the received cyclic prefix part (or also known as guard interval) of the OFDM signal.

The time domain operation is performed by utilizing the signal characteristics of the received signal. Cyclic Prefix (CP) is essentially a copy of the last few samples of an OFDM symbol and placed in the beginning of the OFDM symbol. By using Maximum Likelihood (ML) estimation through correlation of the received CP part and the last few samples of OFDM symbols the frequency offset can be estimated. Time domain frequency offset estimation typically has quite wide frequency offset estimation range. According to prior art the normalized frequency offset estimation £_FOE can be estimated within | < 0.5 in the tracking stage. Time domain frequency offset estimation using either dedicated training symbols or cyclic prefix has however several practical problems, including sensitive to residual Direct Current (DC) offset, spurs signal, and narrow band interference. The presence of those imperfections can destroy the correlation output. The cyclic prefix can also be corrupted due to the multipath fading phenomenon, and thus the frequency offset estimate will no longer be accurate. On the other hand frequency domain solutions for frequency offset estimation typically use received pilot symbol at certain positions of the transmitted signal. The principle operation also relies on the correlation operation of the transmitted signal. The frequency offset captured range is also determined by the placement of the pilot symbols in the time domain.

Frequency domain frequency offset estimation has a fundamental issue that the frequency offset range is limited to the pilot structure of the received signal. According to some conventional methods the pilot symbols are restricted to be placed in two adjacent OFDM symbols. This will enable the frequency offset capture range within|s_p,_0i?| < 0.5 . However, e.g., LTE system does not provide that kind of pilot structure.

Another conventional method is to estimate Fractional Frequency Offset (FFO) by utilizing cell specific Common Reference Symbols (CRSs). This is quite well known method and can be named the baseline method.

The phase rotation measured by the correlation of CRS in two OFDM symbols can be used to estimate the frequency offset. The problem in LTE systems is that the reference symbol/pilot is not located at the same sub-carrier for two consecutive OFDM symbols. The receiver performs least square channel estimate: H_{l k} = R_{l k}S^* _k where R _i: is the received pilot, S_{l k} is the stored known pilot, / is the symbol index and k is the sub-carrier index.

Thereafter, a simple linear frequency domain interpolation of that least square channel estimate is performed to form a virtual channel estimate at sub-carrier without pilot as illustrated in Fig. 3. The correlation output is obtained as,

where K is a joint set of OFDM symbol indices / and sub-carrier indices^; where the CRS symbol is located. Furthermore, the phase rotation Θ_{Δ j} Can be expressed as, where Δ is the distance between two OFDM symbols carrying CRS symbol. The estimated frequency offset &_FOE is then given by,

1

'FOE

2π A(N_FFT + N_g )

The absolute maximum ran e for the normalized &_FOE is,

The CRS in LTE with normal CP configuration are located in OFDM symbol numbers 0, 4, 7, and 1 1 , respectively. The smallest distance between OFDM symbols of the CRS from the same antenna port, i.e., symbol 4 and 7 ( Δ=3). Thus, the frequency offset capture range of the baseline method is limited maximum up tols^J < 0.1553 , which is smaller than _FF0 . Another conventional solution can capture wider range within around|s^" _p0£ | < 0.45 . Thus, it cannot cover the whole fractional frequency offset. This method requires several numbers of correlations between OFDM symbols, including across sub-frame. Moreover, Look-Up Table (LUT) implementation is required. Thus, the result is heavily affected by the accuracy of the LUT implementation.

From the above description of conventional solutions there is a need in the art for an improved solution for frequency offset estimation.

Summary

An objective of the present invention is to provide a solution which mitigates or solves the drawbacks and problems of conventional solutions for frequency offset estimation.

Another objective of the present invention is to provide improved frequency offset estimations of received signals, such as multicarrier signals.

According to a first aspect of the invention, the above mentioned and other objectives are achieved with a device for estimating frequency offset of a received (multi carrier) signal, for example, in a wireless communication system. The device is adapted to: receive the signal comprising at least one Orthogonal Frequency Division Multiplexing, OFDM, symbol pair transmitted over a radio channel;

extend a frequency capture range used for frequency offset estimation of said received signal by frequency shifting said received signal at least once in the frequency domain; and estimate a frequency offset &_FOE of said received signal based on the extended frequency capture range.

According to a second aspect of the invention, the above mentioned and other objectives are achieved with a communication device adapted for communication in a wireless communication system, and comprising at least one device for estimating frequency offset according to embodiments of the present invention.

According to a third aspect of the invention, the above mentioned and other objectives are achieved by a method for estimating a frequency offset of a received (multi carrier) signal, for example in a wireless communication system, the method comprising the steps of:

receiving the signal comprising at least one Orthogonal Frequency Division Multiplexing, OFDM, symbol pair transmitted over a radio channel;

extending a frequency capture range used for frequency offset estimation of said received signal by frequency shifting said received signal at least once in the frequency domain; and

estimating a frequency offset £_FOE of said received signal based on the extended frequency capture range.

An OFDM symbol pair is two OFDM symbols carrying pilot symbols wherein the distance between the OFDM symbols in a pair will reflect the frequency offset estimation range. Frequency capture range is the range within which the frequency offset can be estimated and corrected by a receiver.

Embodiments of the present invention provide an estimation solution with a flexible frequency offset capture range which e.g., can cover the entire range of fractional frequency offset z_FFO or even more of an LTE system. Further, since the frequency estimation is performed in the frequency domain the present solution is robust against impairments, such as DC offset and narrow band interference. Moreover, the maximum complexity is expected to be approximately increased linearly with the multiplication factor of the range extension from the baseline frequency offset estimator which is an advantage. The proposed solution further offers the design option of trade-off between complexity and performance. Hence, the performance can be improved by utilizing more data/input whenever it is available.

According to an embodiment of the present invention, the device is further adapted to use a Finite Impulse Response, FIR, filter for frequency shifting said received signal so as to extend the frequency capture range. The FIR filter may have the filter coefficients C(m) calculable by

where m is the filter coefficient index, is a frequency shift, N_FFT is a number of Fast

Fourier Transform, FFT, points, N_gi is a length of guard interval or cyclic prefix of said received signal, and / is an OFDM symbol index within one sub-frame of said received signal. By frequency shifting in the frequency domain the use of fast multipliers (e.g. very high speed multipliers which are expensive) can be avoided which is the case if the frequency shift would be performed in the time domain. The computation complexity can be even lower as only sub-carriers carrying pilots will be passed through the FIR filter.

According to another embodiment of the present invention, the device is further adapted to frequency shift said received signal N times so as to obtain N + 1 frequency offset estimations e_k, {k e l,2,..., N+ l} extending the frequency capture range by N + 1 times a frequency capture range of an individual frequency offset estimation £_k so as to obtain adjacent frequency estimation regions. N + 1 frequency offset estimations are obtained since every frequency shift results in one frequency offset estimation each and additionally one frequency offset estimation is obtained without frequency shifting, i.e. for the region around the carrier frequency. This embodiment means that the total capture range will be maximal in relation to the number of frequency shifts as there is no overlap.

According to yet another embodiment of the present invention, the device is further adapted to frequency shift said received signal N times so as to obtain N + 1 frequency offset estimations e_k, {k e l,2,..., N+ l} extending the frequency capture range by less than N + 1 times a frequency capture range of an individual frequency offset estimation £_k so as to obtain overlapping frequency estimation regions. The complexity may be higher compared to the non-overlapping case but the performance is improved with overlapping ranges.

According to the embodiments with the N + 1 frequency offset estimations mentioned above frequency offset estimations are pair-wise symmetrically arranged around a transmit carrier frequency for said received signal according to another embodiment. This means optimal frequency capture ranges are achieved and also that the implementation of such an embodiment of the present invention is simplified.

According to yet another embodiment of the present invention, the present device is further adapted to obtain the N + 1 frequency offset estimations e_k, {k e l,2,..., N+ l} by correlating channel estimates determined from pilot symbols of the at least one OFDM symbol pair. According to this embodiment, the device may further be adapted to use a reduced number of sub-carriers of the at least one OFDM symbol pair for correlating channel estimates. This embodiment reduces computational load.

According to yet another am embodiment, the present device may further be adapted to use one OFDM symbol pair per each frequency offset estimation e_k, {k e l,2,..., N+ l} for correlating channel estimates, the OFDM symbol pairs used for each frequency offset estimation e_k, {k e l,2,..., N+ l} being the same OFDM symbol pair or having the same symbol distance. Therefore, since all frequency estimators will have the same offset capture range the final offset is easier obtained since the offset candidates are easily compared, e.g. by comparing correlation values. The device may yet further be adapted to use additional OFDM symbol pairs for correlating channel estimates for a frequency offset estimation £_k associated with a transmit carrier frequency for said received signal. This means that the frequency offset estimation is improved. According to yet another embodiment of the present invention, the device is further adapted to select a frequency offset estimation £_k having a highest absolute correlation value from the OFDM symbol pair as said frequency offset estimation &_FOE . This embodiment results in very low complexity in the selection process.

According to yet another embodiment of the present invention, the device is further adapted to use at least one Maximum Likelihood, ML, function for selecting said frequency offset estimation £_FOE . According to this embodiment a single OFDM symbol pair may be used for the frequency offset estimation, and the device further may be arranged to derive a ML function calculable by λ* = Re{_{a ε}χρ(- 2πε [Δ¾^ - /_Δ ]) +α₂ βχρ(-ι2π¾ [Δ¾^])

for each frequency offset estimation e_k,{k e l,2,...,N+l} , and to select a frequency offset estimation £_k which fulfils condition€_FOE - έ_ρ as said frequency offset estimation £_FOE , the in dex p being obtained as p = argmaxi , A is a symbol distance between two OFDM symbols, f_A is a constant, and a_k are coefficients of the ML function. This means that the frequency offset estimation will be more accurate by reducing false detection regions of frequency offset.

According to this embodiment L > 1 OFDM symbol pairs may be used for the frequency offset estimation, and the device further being adapted to derive a ML function calculable by λ* = Re{_{a ε}χρ(- 2πε [Δ¾^ - /_Δ ]) +α₂ exp(- 2^ [Δ¾^])

for each OFDM symbol pair v |v e l,2,...,Z] , and to linearly combine the ML functions

L

^_F ^k'^v =∑ K_' ^k'^v f°^{r eacn} frequency offset estimation £_k,{k e l,2,...,N+l] , and to select a v'=l

frequency offset estimation &_FOE - &_p as said frequency offset estimation &_FOE , the index p

being obtained as p =

According to yet another embodiment of the present device, two OFDM symbol pairs may be used for frequency offset estimation, and the device is further adapted to compute a minimum distance for two sets of frequency estimations, and to select a frequency offset ε~ , + ε ₂

estimation which fulfils £_FOE = ^p' ^r' , where indices p, r are obtained by p, r = argmin ι έ_{χ 1}— έ J) . The averaging according to this embodiment will improve the estimation results.

According to yet another embodiment of the present device, the radio channel is a Multiple Input Multiple Output, MIMO, channel, and the device is further adapted to compute correlation values of channel estimates for the at least one OFDM symbol pair for each MIMO stream; linearly combine the computed correlation values for each MIMO stream; and using the combined correlation values for estimating said frequency offset £_FOE . By using more samples from different MIMO paths the frequency offset estimation result can be improved.

The present invention also relates to a computer program, characterized in code means, which when run by processing means causes said processing means to execute any method according to the present invention. Further, the invention also relates to a computer program product comprising a computer readable medium and said mentioned computer program, wherein said computer program is included in the computer readable medium, and comprises of one or more from the group: ROM (Read-Only Memory), PROM (Programmable ROM), EPROM (Erasable PROM), Flash memory, EEPROM (Electrically EPROM) and hard disk drive. Further applications and advantages of the invention will be apparent from the following detailed description.

Brief Description of the Drawings

The appended drawings are intended to clarify and explain different embodiments of the present invention, in which:

- Fig. 1 illustrates sub-carriers representation of OFDM signal and frequency offset in LTE; - Fig. 2 illustrates frequency offset estimation and frequency offset estimation correction in a prior art receiver;

- Fig. 3 illustrates linear interpolation operations of least square channel estimates for sub-carrier without pilots and correlation between OFDM symbols carrying pilots; - Fig. 4 illustrates a block diagram of a frequency offset estimator with frequency shift;

- Fig. 5 illustrates non-overlapping baseline estimation with two frequency shifts;

- Fig. 6 illustrates a receiver structure according to an embodiment of the present invention;

- Fig. 7 illustrates non-overlapping N frequency shifts (4 re-sampling filters which gives 5 zones);

- Fig. 8 illustrates receiver block diagram for estimating frequency offset according to an embodiment of the present invention;

- Fig. 9 illustrates how multiple antennas generate more data samples in a MIMO scenario;

- Fig. 10 shows a flow chart for frequency offset estimation with different frequency offset ranges according to an embodiment of the present invention;

- Fig. 1 1 illustrates performance evaluation for various schemes with frequency offset ± 7 kHz;

- Fig. 12 illustrates performance evaluation for various schemes with frequency offset ± 10 kHz;

- Fig. 13 illustrates a method according to an embodiment of the present invention;

- Fig. 14 illustrates an estimation device according to an embodiment of the present invention; and

- Fig. 15 schematically illustrates downlink transmission of a multicarrier signal from a base station to a communication device according to an embodiment of the present invention.

Detailed Description

With the estimation device 10 according to an embodiment of the present invention, e.g. shown in Fig. 4, the frequency capture range used for frequency offset estimation of a received OFDM signal is extended by frequency shifting the received multicarrier signal at least once in the frequency domain. Thereafter a frequency offset estimation of the received signal is performed based on the extended frequency capture range so as to obtain the frequency offset estimation £_FOE . Therefore, a re-sampling (frequency shifting) method is proposed to perform frequency shift in the frequency domain so that the captured range for frequency offset can be extended.

Considering that the received OFDM signal has a frequency offset of ε and the receiver is equipped with a frequency offset estimator up to &_FFO the problem is the case in which or in other words in which the frequency offset is larger than the frequency

offset estimation range of the frequency estimator of the receiver. Therefore, not all frequency offsets can be estimated. Typically ε is a residual frequency offset within

Ι . In order to facilitate a wider range of the frequency offset estimator the range is extended by introducing a frequency shift so that + ε₅ | < |ε | in the estimation device 10. This

operation can be illustrated as in the block diagram of Fig. 4. Hence, embodiments of the present invention relate to a device 10 for estimating frequency offset of a received multi carrier signal in a wireless communication system 20, the device comprises optionally at least one processor 30 adapted to: receive a signal comprising at least one Orthogonal Frequency Division Multiplexing, OFDM, symbol pair transmitted over a radio channel; extend a frequency capture range used for frequency offset estimation of said received signal by frequency shifting said received signal at least once in the frequency domain to; and estimate a frequency offset 8_TO£ of said received signal based on the extended frequency capture range. Typically the capture range is far less than the actual frequency offset error. However, the embodiments of the present invention solve this problem of the conventional approaches by extending the capture range. Fig. 4 shows a block diagram in which both frequency shifting and frequency offset estimation is performed in the frequency domain. A DL signal is received by an antenna unit 1 1 and is converted to a baseband signal in the front end 12. Therafter, the time domain OFDM signal is converted to the frequency domain by an FFT unit 13. The frequency of the signal is frequency shifted by passing the frequency domain signal to an FIR filter which is performed in the block "re-sampling filter" 14. Finally, a frequency offset estimation is performed in block "FOE" 15 which will be further described in the following disclosure. According to an embodiment of the present invention a frequency shift can be implemented in the frequency domain by a FIR Filter. The filter coefficients C(m) can according to another embodiment be written in the form of,

where ε^ is a frequency shift, N_FFT is a number of Fast Fourier Transform, FFT, points, N_gi is a length of guard interval or cyclic prefix of said received signal, and / is an OFDM symbol index within one sub-frame of said received signal.

According to yet another embodiment of the present invention both re-sampling filters and baseline frequency offset estimators are used for obtaining multiple frequency offset estimations e_k, {k e l,2,..., N+ l} . Therefore, the present device according to this embodiment is adapted to frequency shift the received signal N times (N being an positive integer) so as to obtain N + 1 frequency offset estimations e_k, {k e l,2,..., N+ l} including one frequency offset without a frequency shift, i.e. around the carrier frequency for the transmitted signal. Concatenating all frequency offset estimations together extends the total frequency capture range by N + 1 times a frequency capture range of an individual frequency offset estimation £_k . In this way adjacent frequency estimation regions are obtained. The baseline frequency estimator has a certain range of frequency offset and the re-sampling filter frequency is carefully designed so that the baseline estimator can be constructed to achieve wider frequency offset capture range. In this embodiment there is no frequency estimate range overlap and thus the total range can be maximized. An illustration of the frequency shifts performed by this embodiment is given in the Fig. 5. Firstly, the anticipated frequency offset range is defined which is the maximum offset that can be estimated, e.g. within +/- 0.5 of the sub-carrier spacing. In this example, the target of anticipated frequency offset range is ± 7 kHz or (± 0.466 in normalized scale). Two frequency hypothesis denoted as -ε^ and ε^ , respectively, are defined in this example assuming that OFDM symbols [4, 7] are used in a LTE system. This will form a set of frequency shifts β_έ as

-ε, , Ο,ε, .

The receiver device according to this embodiment may have the structure as illustrated in Fig. 6. The digital OFDM baseband time domain signal is converted to the frequency domain by an FFT unit 16. The three Frequency Offset Estimation (FOE), blocks 18 in Fig. 6, generate three corresponding correlation values denoted as μ₁₅ μ₂, μ₃ . Three tentative FOE hypotheses 8_{L 5}8₂ , 8₃ are computed ase_k = K arg^) where κ is a constant that depends on the LTE setting. The selector 19 of the present device will select one out of three frequency offset hypotheses έ_ι,έ₂,έ₃ by first finding the index p from arg max | JL ^ | j- operation to determine the selected frequency offset ε . Then, the final frequency offset £_FOE will be ε_ρ - β_ρ , where β_ρ is the predefined frequency shift used for the re-sampling filter coefficient generation.

According to another embodiment of the present invention the frequency offset estimation is obtained by increasing the number of frequency hypothesis by having overlapping frequency estimation regions. For example, the number may be increased from N=2 to N=4. In this respect the anticipated capture range must be defined and being divided into Λ/+1 zones of equal frequency spacing. The frequency hypotheses are thereafter placed at the centre of each zone as shown in Fig. 7. It can be observed from Fig. 7 that the capture range of each zone is -0.093, 0.093 which is smaller than the individual estimator capture range -0.155, 0.155. Therefore, according to this embodiment the present device 10 is arranged to frequency shift the received signal N times so as to obtain N + 1 frequency offset estimations e_k, {k e l,2,..., N+ l} together extending the frequency capture range by less than N + l times a frequency capture range of an individual frequency offset estimation £_k .

The advantage of this embodiment is by allowing overlap from neighbouring estimators the receiver device 50 will only trust the estimator which provides the frequency estimates within its zone. Otherwise, the results are discarded. Finally, the best estimator is obtained by finding the highest absolute correlation magnitude from those trusted estimator outputs. In Fig. 7 this procedure is shown. In region 1 , a FFO estimate, the circle sign in Fig. 7, is obtained that falls within the same region, therefore this value is trusted. For region 2, the FFO estimate, i.e. the plus sign in Fig. 7, falls within the zone of the first estimator, therefore this estimate is not trusted as it is too far away from the centre of region 2. Also, the two regions to the right, i.e. regions 4 and 5, are producing FFO estimates that are too far away from the centres of the regions, and are therefore discarded. Region 3 in the middle produces a valid FFO estimate, i.e. the triangle sign. Hence, the final output will be either the FFO estimate shown in the circle sign or the triangle sign depending on which one that has the largest correlation magnitude according to an embodiment. According to yet another embodiment of the present invention at least one Maximum

Likelihood (ML) function is used for selecting the frequency offset estimation £_FOE of the received signal. Preferably, the method described in Fredrik usek, Basuki E. Priyanto, "Karhunen Loeve based Maximum Likelihood estimation of frequency offsets in OFDM systems using pilots", Patent Application EP13198573.1, is used but with a slightly different approach. Rather than using correlation of CRS symbols the present idea is based on constructing the likelihood function. Maximizing this function would qualify as optimal

frequency offset estimation, ;ε j . By using the method in EP13198573.1 only a single FFT needs to be executed per OFDM symbol in order to apply the results of

EP13198573.1 ; and only three function evaluations of λ ^|/- ,™ j ;s j are needed. Still the results coincide with those from EP13198573.1.

From EP13198573.1 it can further be observed that only three values of λ ^|/- ,™ j ;s j are needed in order to evaluate the entire function. The frequency shift ε^ can be denoted as β . The three values μ^~β , μ°, μ^β can be arranged to be the samples of λ |τ _/™| ;ε ] :

It follows from EP13198573.1 that the likelihood function λ ^|/- ,™ | ;ε | can be expressed as, λ ({¾} ;ε ) = φ₁ (ε) α₁+ φ₂(ε)α₂+ φ₃ (ε)α₃

wherea = Αμ , and A is a known 3x3 matrix which is pre-computed offline. The function φ_έ(ε) are complex exponential functions φ_έ(ε) = εχρ(ί(¾ε + β_έ)) where (¾, β_έ) are known in advance. The overall operation using the ML method of EP13198573.1 is illustrated in the block diagram of Fig. 8. The step by step operation of the frequency offset estimation can be described as follows with reference to Fig. 8:

1 . Use the C S symbols at OFDM symbols 4 and 7, take the FFT of the time signals y₄ and y₇ .

2. Apply re-sampling to both signals by a normalized frequency amount of {-β ,Ο, β } .

Thus, three branches of output are obtained, in this example β = 0.311 and 0 can be interpreted as no re-sampling. Then, perform least square channel estimate and interpolation for each branch and denote these by H₄ and H₇. We then have a total of

6 signals Η^ ,Η₄,Η₄ ^β ,Η ^β ,Η₇,Η₇ ^β .

3. Find three total correlation values as,

Arrangements have to be made as there are not pilots at all positions, and that they do not appear at the same sub-carriers in OFDM symbols 4 and 7 which can be done by interpolation between channel estimates in the frequency domain as illustrated in Fig. 3. Therefore, the above correlations are "conceptual".

Note that A Ms a fixed matrix. As described in

EP13198573.1 for OFDM symbols 4, 7 combination we have,

-0.5692 + .2977/ 1.1344 -0.5692 - .2977/ ^"

A^" = 1.1343 + 0.0077/ -1.2687 1.1343 - 0.0077/

-0.5651 - 0.3054/ 1.1344 -0.5651 + 0.3054/ where β = 0.311, /_Δ = 0.5.

5. Find the index p by finding the maximal correlation over the different frequency intervals: p = argmax J .

6. Find a tentative frequency offset as,

1 N

ε = - FFT

arg { μ_ρ } + (p ~ 2) β ,

2π A (_NpPT + N_gi ) where arg{ } is a function to provide the angle of complex numbe^ , and the last term is a re-positioning by the frequency shift into the frequency interval that gave the largest correlation value in step 6.

7. Construct a list of possible frequency offsets as Ξ = {ε -ε_Γ,ε,ε + ε_Γ} , where z_r is the range of baseline frequency offset. The combination of OFDM symbols

4 and 7 provides the range of 0.311. Denote these 3 values by £_k, 1 < k≤ 3. All of these possible frequency offsets should be within the anticipated frequency offset range plus a small margin to anticipate additional small frequency error due to noise. We consider a margin of 0.023 or 350 Hz. Otherwise it will be truncated to that maximum value. If the range is extended then the number of candidates is increased as a multiple of z_r since the range of the frequency offset is linearly related to a multiple of z_r .

8. Compute, λ* = Re{_{ai ε}χρ(- 2πε [_Δ¾½ - /_Δ]) +α₂ exp(-/2 [Δ¾^])

+ α₃ exp (-/2πέ, [Δ¾ ½ + f_A ])} for k=1 , 2, 3.

9. Take the final estimate as z_FF0 = £ , p = argmaxi where p is an index for the k

selected estimate.

The aforementioned embodiments have been exemplified in the single antenna system case for the transmitter and the receiver. In LTE, multiple antennas are usually employed at both the transmitter and receiver sides and this fact can be exploited by the present invention. Multiple antennas system can be illustrated as shown in Fig. 9 in which the radio channel is the Multiple Input Multiple Output (MIMO) radio channel.

In the MIMO radio channel, the receiver can produce more channel estimates based on multiple pairs of transmit and receive antennas. The LTE system has been designed so that each antenna port can transmit CRS symbols. Hence, more correlation outputs can be generated according to,

After combining the correlation outputs to the form μ^<0 , μ^ω ₂, μ^₃] the remaining operations is the same as described in the previous embodiments. This can also be extended in the case when time domain averaging is performed. In LTE, the averaging can be performed over more than 1 subframe. So, according to this embodiment with transmissions over the MIMO radio channel the present receiver device 10 is further arranged to compute correlation values of channel estimates for at least one OFDM symbol pair for each MIMO stream; thereafter linearly combine the computed correlation values for each MIMO stream; and using the combined correlation values for estimating the frequency offset &_FOE of the received signal.

According to further embodiments of the present invention more than one OFDM symbol pair can be used for estimating the frequency offset. According to these embodiments more than one OFDM symbol pairs are used as the input to the frequency offset estimator. For example, in LTE systems with normal CP configuration the reference symbols are located at OFDM symbol locations 0, 4, 7 and 1 1 . Thus, many combinations of OFDM symbols pairs can be formed, i.e., [0, 4], [4, 7], [4, 1 1], [0, 7], etc. It is noted that different OFDM symbol pairs result in different frequency offset ranges and thus it must be uniquely treated in order to combine the result.

The general methodology is described with reference to the flow chart in Fig. 10:

1 The device 10 needs to predefine the number of L OFDM symbol pairs to be used for estimation.

2 Perform L parallel operations to calculate the correlation values and tentative frequency offsets by re-sampling; and from each operation, find the maximum correlation value at with/without re-sampling outputs to indicate selected tentative frequency offset candidates.

3 Construct a list of possible frequency offset candidates. Considering two frequency estimates (εΌ ₄ , ε_{4 7} ) are available from the combination of symbols [0,4] and [4,7] using re-sampling, then a set of frequency hypotheses can be configured as, ^fcsei4,7

^{1 V ' fc}.R4,7 ' · · ·' V^fc4,7 ^fc.R4,7 ' ^fc4,7 ' V^fc4,7 ^{+ h} R4,l ^■■ -' V^fc4,7 ^{+ 1 V ' fc}.R4,7 J ^setO,A ^~ {(^0,4 ^_ M^■ £^o,4 )' " ·' (βθ,4 ^{~ £}Λ0,4 )' ^0,4 ' (^4,7 ^£Λ0,4 )' " ·' (^0,4 + ^ ^' ^R0,4 ^\ '

where f_{Rm n} is frequency range of the estimation using symbol m and n, f_{R4 7} equals to 0.31 1 and f_{R0 4} equals to 0.233, and N and M are integer values. Those frequency hypotheses should also be within a predefined anticipated frequency offset range (e.g., ±

0.466 or ± 7 kHz in LTE systems) plus a small margin (for compensation of noise errors). We consider a margin of 0.023 or 350 Hz. Otherwise, those frequency hypotheses are discarded. At this stage there are two proposals after step 3 above, namely use combined approximate ML or Minimum Distance (MD). Assuming L=2 we have two sets e.g., combination of symbols [0, 4] and [4, 7], each with 2Λ/+1 , 2Λ/+1 number of frequency hypotheses/candidates. If using the MD method, subtract the frequency estimate candidates from those two sets and find indices p, r by,

p, r = argmm \ \ε_χ— ε

xel,2,..,2M+\,ye\,2,..,2N+V'

where ε_χ is element of vector &_{setA 1} and ε is element of vector s_jei0;4 . The final FOE will be έ + έ

obtained as έ =— - . It should be noted that the MD method must ensure the whole

2

anticipated frequency offset range is covered by sufficient number of regions/re-sampling branches. The advantage with using the MD method is the low complexity involved.

If using the combined approximate ML method described above, combine those two sets in step 4 above into one set &_{set[A 1] [0 4]} with (2M+1) + (2N+1) candidates. Use these candidates as the input of L (L=2) approximate ML method to produce two sets of λ assuming L=2. ₁ ,

A_{K 2} are computed using the ML function above. Combine the results by simply adding them together i.e., _tot = , + _>2. Find the two largest λ and denote them as A_{max toi} , _ax__Uoi . If

(Th is a threshold value which should be kept small) then o m. ax max— 1

ε FOE . It is likely two of the selected candidates (from different sets) are almost

2

the same or within the Th value. Otherwise e_FOE = e_max . It is likely that two of the selected candidates (from different sets) are almost the same or within the threshold value Th which is typically set to a small value. Thus, we can average them to minimize the frequency offset estimation error.

Further embodiments of the present invention also relate to devices and methods for complexity reduction of the present frequency offset estimation solutions. The objective of complexity reduction is to reduce the complexity with relatively low performance degradation. The performance degradation can be contributed from higher probability of false detection and/or increasing RMS error due to less number correlation for noise averaging purpose. The main focus is then to reduce the number of correlations and avoiding an increase in the probability of false region detection.

In the following examples two re-sampling filters, combination of symbols [0, 4] and [4, 7] for LTE system with normal CP is assumed which requires three sets of correlations as shown below. Each set consists of two correlations, OFDM symbols # 0 and # 4, and OFDM symbols # 4 and # 7, respectively. It should however be noted that the method is not limited to the above mentioned combination or to LTE.

A first complexity reduction method is almost identical to the above methods and devices for frequency offset estimation except that the number of correlations in all sets is reduced by using a reduced number of sub-carriers of the at least one OFDM symbol pair for correlating channel estimates. Each branch after FFT outputs perform correlation for both 2 pair of symbols; symbol # 0 and symbol # 4 and symbol # 4 and symbol # 7. The phase rotation measured from two OFDM symbol with the distance of Δ symbols can be expressed as,

The correlations output can also be simply written as,

where K is the number of frequency domain least square channel estimate (i.e., K=2x2x50=200 for LTE 10 MHz), and χ is a reduction factor. After both phase rotations and correlations are obtained the remaining steps are the previously mentioned ML method which is identical to the method without complexity reduction.

In the second complexity reduction method only one OFDM symbol pair (OFDM symbols # 4 and # 7) is used at the branches with re-sampling filter (N branches). The OFDM symbol pair used for each frequency offset estimation e_k , {k e l, 2, ..., N+ l} (branch) is the same OFDM symbol pair or has the same symbol distance. This embodiment can be further improved by using additional OFDM symbol pairs for correlating channel estimates for a frequency offset estimation £_k associated with the transmit carrier frequency for said received signal. In this case the estimation branch without re-sampling uses two OFDM symbol pair.

The correlation outputs together with outputs from same OFDM symbol pair at branches without re-sampling are used as the input to the approximated ML method described above. This will produce coarse frequency estimate. Finally, fine frequency estimate is obtained by combining the coarse frequency offset and one measurement frequency estimate output from symbol # 0 and # 4. This combination is using the MD method. The detailed operations are:

1 . Use C S symbols at OFDM symbols 4 and 7, and take the FFT of the time signals y₀ , y₄ and y₇ .

2. Use combined re-sampling and the ML method for the pair of OFDM symbols 4 and 7 to produce &_FOEl .

3. Compute the frequency offset from the pair of OFDM symbols 0 and 4,

4. Construct a list of possible frequency offsets as,

Ξ = {ε^ - 0.466,8₀₄ - 0.233, 8₀₄ , ε^ + 0.233, έ₀₄ + 0.466} .

5. Find the nearest frequency to the coarse frequency offset estimate using minimum distance method,

(£_F0E2 idx) = min ( Ξ - &_F0El ) .

6. Select the final frequency offset estimate as,

ε 'FOEl ' ε ^FOEl

&FOE Link level simulation was carried out in order to evaluate the present frequency offset estimation device and method. The frequency offset estimation was in principle performed in 1 sub-frame. However, it can also be extended to more than 1 sub-frame by performing time domain averaging. The anticipated frequency offset range was within ± 7 kHz and ± 10 kHz. For benchmarking purpose, we also provide the result where the frequency offset range was limited to ± 2.33 kHz. This is the range of baseline frequency offset estimator with OFDM symbols [4, 7] in LTE with normal CP configuration. The common simulation parameters that were used are given in Table 1.

Table 1 : Simulation parameters

Fig. 1 1 shows the performance for various schemes assuming that the frequency offset is within ± 7 kHz. The single channel case performs worst. The performance can be improved by using multiple channels (MIMO) and also using more symbol combinations, both [0, 4] and [4, 7]. MIMO with minimum distance (MIMO [0, 4], [4, 7]) achieves the same performance as MIMO - ML [0, 4], [4, 7]. However, the ML method has better results for low SNR values. The best performance was achieved by performing time domain averaging. The result of "MIMO-ML [0, 4], [4, 7] - 4 Sub Frames" setup indicated it is quite close to the results with restricted frequency offset (± 2.33 kHz).

Fig. 12 shows the performance for various schemes assuming that the frequency offset is within ± 10 kHz. It can be seen the present device and method is still performing well even when the frequency offset has been extended to ± 10 kHz. There is only a small penalty observed in most of the cases. The performance of using the MD method is completely offset. It is mainly because the MD method requires proper placement of frequency shift which is for re-sampling filter. The MD method used here is designed for three zones which can theoretically cover up to 7 kHz frequency offset.

Furthermore, as understood by the person skilled in the art, the present invention also relates to methods for frequency offset estimation and complexity reduction. Any method according to the present invention may be implemented in a computer program, having code means, which when run by processing means causes the processing means to execute the steps of the method. The computer program is included in a computer readable medium of a computer program product. The computer readable medium may comprises of essentially any memory, such as a ROM (Read-Only Memory), a PROM (Programmable Read-Only Memory), an EPROM (Erasable PROM), a Flash memory, an EEPROM (Electrically Erasable PROM), or a hard disk drive.

Fig. 13 illustrates a method according to an embodiment of the present invention. The method for estimating a frequency offset of a received multi carrier signal in a wireless communication system 20, comprises the steps of: receiving 100 a signal comprising at least one Orthogonal Frequency Division Multiplexing, OFDM, symbol pair transmitted over a radio channel; frequency shifting 200 said received signal at least once in the frequency domain so as to extending a frequency capture range used for frequency offset estimation of said received signal; and estimating 300 a frequency offset £_FOE of said received signal based on the extended frequency capture range. It should be noted that the above method can be modified, mutatis mutandis, according to the different embodiments of the present device 10.

Moreover, it is realized by the skilled person that the present estimation device 10 and communication device 50 each comprises the necessary communication capabilities in the form of e.g., functions, means, units, elements, etc., for executing the present method. Examples of other such means, units, elements and functions are: processors, memory, encoders, decoders, mapping units, multipliers, interleavers, de-interleavers, modulators, demodulators, inputs, outputs, antennas, amplifiers, Rx unit, Tx unit, DSPs, MSDs, TCM encoder, TCM decoder, interfaces, communication protocols, etc. which are suitably arranged together.

Especially, the processors of the present user device or access node device may comprise, e.g., one or more instances of a Central Processing Unit (CPU), a processing unit, a processing circuit, a processor, an Application Specific Integrated Circuit (ASIC), a microprocessor, or other processing logic that may interpret and execute instructions. The expression "processor" may thus represent a processing circuitry comprising a plurality of processing circuits, such as, e.g., any, some or all of the ones mentioned above. The processing circuitry may further perform data processing functions for inputting, outputting, and processing of data comprising data buffering and device control functions, such as call processing control, user interface control, or the like.

Fig. 14 shows a device 10 according to the present invention. The device 10 comprises in this case a processor unit 30 coupled to an input unit and an output unit. The processor 30 is arranged to receive a signal (or a representation of a signal) comprising at least one OFDM symbol pair. The processor 30 is further arranged to process the signal as described in this application in conjunction with the various embodiments to obtain a frequency offset estimation which can be outputted for further processing such as correcting the signal from the frequency offset. In this example the device 10 also comprises a memory coupled to the processor for storing data. The memory may also include program instructions to be executed in the processor. The device 10 may be a standalone device or be integrated in communication device 50.

Fig. 15 shows a communication device 50 according to an embodiment of the present invention which comprises at least one frequency offset estimation device 10 according to an embodiment of the present invention. The communication device 50 in Fig. 15 receives a downlink multicarrier signal from a base station in this case. The communication system 20 may be a cellular multicarrier system such as LTE but the present invention is not limited to an LTE system. In mentioned LTE system the communication device 50 is a UE but can be any communication device arranged to receive radio communication signals transmitted in a multicarrier system using OFDM symbols, such as terminals or general receiver devices.

Finally, it should be understood that the present invention is not limited to the embodiments described above, but also relates to and incorporates all embodiments within the scope of the appended independent claims.

Claims

1 . Device (10) for estimating frequency offset of a received signal, the device (10) being adapted to:

receive the signal comprising at least one Orthogonal Frequency Division Multiplexing, OFDM, symbol pair transmitted over a radio channel;

2. Device (10) according to claim 1 , further adapted to use a Finite Impulse Response, FIR, filter for frequency shifting said received signal.

3. Device (10) according to claim 2, wherein the FIR filter has filter coefficients C(m) calculable by

Fourier Transform, FFT, points, N_gi is a length of guard interval or cyclic prefix of said received signal, and / is an OFDM symbol index within one sub-frame of said received signal.

4. Device (10) according to any of claims 1-3, further adapted to frequency shift said received signal N times so as to obtain N + 1 frequency offset estimations e_k, {k e l,2,..., N+ l} extending the frequency capture range by N + 1 times a frequency capture range of an individual frequency offset estimation £_k so as to obtain adjacent frequency estimation regions.

5. Device (10) according to any of claims 1-3, further adapted to frequency shift said received signal N times so as to obtain N + 1 frequency offset estimations e_k, {k e l,2,..., N+ l} extending the frequency capture range by less than N + 1 times a frequency capture range of an individual frequency offset estimation £_k so as to obtain overlapping frequency estimation regions.

6. Device (10) according to any of claims 4-5, wherein N of the N + 1 frequency offset estimations are pair-wise symmetrically arranged around a transmit carrier frequency for said received signal.

7. Device (10) according to any of claims 4-6, further adapted to obtain the N + 1 frequency offset estimations e_k, {k e l,2,..., N+ l} by correlating channel estimates determined from pilot symbols of the at least one OFDM symbol pair.

8. Device (10) according to claim 7, further adapted to use a reduced number of sub- carriers of the at least one OFDM symbol pair for correlating channel estimates.

9. Device (10) according to any of claims 7-8, further adapted to use one OFDM symbol pair per each frequency offset estimation e_k, {k e l,2,..., N+ l} for correlating channel estimates, the OFDM symbol pairs used for each frequency offset estimation £_k, {k e l,2,..., N+ l] being same OFDM symbol pair or having same symbol distance.

10. Device (10) according to claim 9, further adapted to use additional OFDM symbol pairs for correlating channel estimates for a frequency offset estimation £_k associated with a transmit carrier frequency for said received signal.

1 1 . Device (10) according to any of claims 4-10, further adapted to select a frequency offset estimation £_k having a highest absolute correlation value from the OFDM symbol pair as said frequency offset estimation &_FOE .

12. Device (10) according to any of claims 4-10, further adapted to use at least one Maximum Likelihood, ML, function for selecting said frequency offset estimation £_FOE .

13. Device (10) according to claim 12, wherein a single OFDM symbol pair is used for the frequency offset estimation, and the device (10) further being arranged to derive a ML function calculable by λ* = Re{_{ai ε}χρ(- 2πε [_Δ¾½ - /_Δ]) +α₂ exp(-/2 [Δ¾^]) + α₃ exp (-/2πέ, [Δ¾ ½ + f_A ])} for each frequency offset estimation e_k,{k e l,2,...,N+l} , and to select a frequency offset estimation £_k which fulfils condition€_FOE - έ_ρ as said frequency offset estimation £_FOE , the index p being obtained as p = argmax^ *) , A is a symbol distance between two OFDM symbols, f_A is a constant, and a_k are coefficients of the ML function.

14. Device (10) according to claim 12, wherein L > 1 OFDM symbol pairs are used for the frequency offset estimation, and the device (10) is further adapted to derive a ML function calculable by λ* = Re{_{ai ε}χρ(- 2πε [_Δ¾½ - /_Δ]) +α₂ exp(- 2^ [Δ¾^]) + α₃ exp (- 2πε, [Δ¾ ½ + f_A ])} for each OFDM symbol pair v |v e l,2,...,Z] , and to linearly combine the ML functions

L

^_F ^k'^v =∑ K_' ^k'^v f°^{r eacn} frequency offset estimation e_k,{k e l,2,...,N+l , and to select a v'=l

being obtained as p = 2LXg 2& k_F ^{k v}

k,v

15. Device (10) according to any of claims 4-10, wherein two OFDM symbol pairs are used for frequency offset estimation, and the device (10) further being adapted to compute a minimum distance for two sets of frequency estimations, and to select a frequency offset estimation which fulfils έ\ ' ρ,ι where indices p, r are obtained by p,r = argmm _ε - _ε . ..

xel,2,..,P,yel,2,..,R

16. Device (10) according to any of the preceding claims, wherein the radio channel is a Multiple Input Multiple Output, MIMO, channel, and the device (10) further being adapted to compute correlation values of channel estimates for the at least one OFDM symbol pair for each MIMO stream;

linearly combine the computed correlation values for each MIMO stream; and use the combined correlation values for estimating said frequency offset £_FOE .

17. Method for estimating a frequency offset of a received signal, comprising the steps of: receiving (100) the signal comprising at least one Orthogonal Frequency Division

Multiplexing, OFDM, symbol pair transmitted over a radio channel;

extending (200) a frequency capture range used for frequency offset estimation of said received signal by frequency shifting said received signal at least once in the frequency domain; and

estimating (300) a frequency offset £_FOE of said received signal based on the extended frequency capture range.

18. Computer program with a program code for performing the method according to claim 17, when the computer program runs on a computer.