WO2015144203A1

WO2015144203A1 - Interferene aware radio receiver with adapated covariance

Info

Publication number: WO2015144203A1
Application number: PCT/EP2014/055865
Authority: WO
Inventors: Thomas Wagner; Liping Zhang
Original assignee: Telefonaktiebolaget L M Ericsson (Publ)
Priority date: 2014-03-24
Filing date: 2014-03-24
Publication date: 2015-10-01

Abstract

The invention refers to method in a wireless receiver comprising a plurality of receive antennas (10) to receive a plurality of signals being transmitted from a wireless transmitter, wherein the plurality of signals form a receive signal vector, comprising the steps of determining a covariance matrix of the receive signal vector, determining a ratio value, SINR, indicative of noise and interference power with respect to a receive signal power, determining an adapted covariance matrix by changing values of certain covariance matrix elements in dependency of the SINR, and using the adapted covariance matrix to recover a transmitted symbol from the receive signal vector. The invention further refers to a corresponding wireless receiver and to a computer program.

Description

Title

INTERFERENE AWARE RADIO RECEIVER WITH ADAPATED COVARIANCE

Technical Field

The present invention generally relates to mobile network wireless receivers, and more specifically relates to 3GPP LTE MIMO receivers.

Background

3GPP LTE specifications prescribe so-called multiple-input, multiple-output, MIMO, transmission schemes to enhancing cellular wireless systems. Current LTE receiver thus may have a plurality of receive antennas, e.g. two, four or eight antennas, to receive a plurality of signals transmitted from a further plurality of transmit antennas, e.g. two, four or eight antennas.

On challenge of MIMO receivers concerns the estimation disturbances with respect to the received signal in order to detect the transmitted bits. LTE receivers are known that assume the signal noise being essentially white noise. For such receivers, in the following also being referred to as first generation (MIMO) receivers, the noise covariance matrix elements are assumed to be zero for all elements that are not part of the main diagonal. Such assumption leads to a rather simple algorithm structure. However, in case of signal interference at different receive antennas, i.e. in case of spatially colored noise, the performance of the first generation LTE receive may degrade significantly.

More recently, interference-aware LTE receivers, in the following also being referred to as second generation receivers, have been introduced to exploit the full noise covariance. These receivers provide significant gains in case of spatially colored interference.

The noise covariance estimation in the receiver is usually determined based on a pilot signal that is periodically transmitted according to a certain time and frequency grid (according to current 3GPP specifications) cell specific reference signal is present on OFDM symbols 0,1 ,4,7,8,1 1 in time, and every 6^th resource element in frequency per TX antenna). However, due to the dynamic character of the transmission channel, the noise covariance estimation is being based on a limited number of pilot symbols. Therefore, the noise covariance matrix estimation may not always be sufficiently accurate. Such lack of estimation accuracy may lead to significant detection losses. Losses of second generation receivers may even be bigger compared to first generation receivers in certain interference conditions, in particular in the presence of spatially white noise.

Second generation receivers may be non-linear so-called sphere decoders or linear so-called interference rejection combining, IRC, receivers. Sphere decoders perform noise whitening, which depends on the noise covariance matrix, in front of the sphere decoder. IRC receivers, e.g. as discussed in the 3GPP document R4-1 13528, titled "Performance of Interference Rejection Combining Receiver for LTE" and discussed on a TSG RAN WG4 meeting in Bucharest, June 27 -July 1 , 201 1 , use the noise covariance to determine the combining weights for the combination of the receive signal components. Both receivers show good performances in case of spatially non-white noise. However, in case of (dominating) spatially white noise, no improvement of IRC receiver compared to standard MMSE receiver is to be expected, as the off- diagonal elements of the noise covariance should be zero (or almost zero). To the contrary, both the IRC receiver and sphere decoder receiver may provide some degradation, in some cases of noise covariance estimation errors.

Summary

It is an object of the present invention to provide an improved wireless receiver. This object is achieved by the independent claims. Advantageous embodiments are described in the dependent claims and by the following description.

Distinguishing accurately between white noise and colored noise is difficult. If the noise covariance indicates colored noise, such indication may be correctly due to actual colored noise; however it may also be due to estimation errors. As a consequence, it is difficult to decide which kind of receiver as discussed above provides better results. It is an insight of the invention that an impact of a noise covariance estimation error may not be equal in all situations. Link-level simulations have been performed showing that owing to full usage of the estimated noise covariance matrix, degradation is significant at high signal-to-noise ratios, SNR, in case of spatially white interference or noise. This influences some higher order modulation performance (e.g. 64QAM), which is more sensitive to estimation errors and is used at high signal-to-noise ratios. Such behavior may be even more pronounced in situations with a high signal to interference plus noise ratio, SINR; if the operating point is above a certain high SINR, the receiver loss due to an "erroneous" noise whitening or application of IRC has been found being more pronounced than at low SINR.

Thus embodiments of the invention propose to combine the strengths of both first generation and second generation receiver concepts in that an estimated noise covariance matrix is adapted as a function of the SINR.

In an embodiment, a full (noise plus interference) covariance of the received signal (i.e. comprising the variance values in the main diagonal and the cross variances in the off-diagonal positions of the covariance matrix) is determined, wherein the off-diagonal values are adapted as a function of the SINR. Such adapted covariance matrix may then being used by the receiver, e.g. by an aforementioned IRC receiver to perform a recombination of the received signal in order to detect the transmitted symbols or by a sphere receiver to determine the noise whitening. Therewith an improved performance may be achieved for white and colored noise/interference compared to existing solution.

In an embodiment thereto, at least one of the following operations is performed to an estimated noise covariance matrix to get the adapted noise covariance matrix:

• setting the off-diagonal values to a certain value (e.g. zero) above a certain SINR threshold, and

• scaling the off-diagonal values according to a scaling factor being dependent on the SINR.

In an embodiment thereto, a smooth transition between using the full estimated noise covariance and using the main diagonal covariance matrix (i.e. the matrix wherein only the off-diagonal elements are set to zero). Thereto a scaling factor is determined to scale the off-diagonal matrix elements as a function of the SINR such that the off-diagonal values remain unchanged below a first SINR threshold, are set to zero (or to any small value) above a second SINR threshold, and are monotonically scaled in dependency of the SINR in between the first SINR threshold and the second SINR threshold.

The wireless receiver may be a receiver to be implemented in a user equipment, UE according to 3GPP LTE specifications. Alternatively the wireless receiver may a wireless LAN, WLAN receiver.

The present invention also concerns computer programs comprising portions of software codes in order to implement the methods as described above when operated by a respective wireless receiver, e.g. an LTE user equipment. The computer programs can be stored on a computer readable medium. The computer-readable medium can be a permanent or rewritable memory within the user device or the recipient device or located externally. The respective computer program can be also transferred to the receiver, for example via a cable or a wireless link as a sequence of signals.

In the following, detailed embodiments of the present invention shall be described in order to give the skilled person a full and complete understanding. However, these embodiments are illustrative and not intended to be limiting.

Brief Description of the Figures

Fig. 1 shows a block diagram illustrating an exemplary wireless receiver employing a covariance matrix,

Fig. 2 shows a block diagram illustrating the covariance matrix

determination of the wireless receiver of Fig. 1 ,

Fig. 3 shows an exemplary diagram of a scaling factor as a function of an

SINR to determine an adapted covariance matrix, and

Fig. 4 shows exemplary method steps of for detecting a transmitted

symbol in a multi-antenna wireless receiver. Detailed Description

Fig. 1 schematically illustrates a wireless receiver 1 comprising an antenna unit 10 with plurality (e.g. two) receive antennas, a Fourier transformer 12, a channel estimator 14, a (noise and interference) covariance determination unit 16 and an IRC equalizer 18.

The wireless receiver 1 , by way of example further comprises a processor P and a memory M. The memory M may store program code to be executed in the processor P to perform receiving functions described in the following. The processor P may be realized as a base band processor or a wireless receiver.

The Fourier transformation unit 12 transforms the antenna signal provided from the antenna unit 10 into the frequency domain, being referred to as receive signal y in the following. The receive signal y can be regarded as receive vector having as many vector components as receive antennas are available. In an exemplary case of two receive antennas, the receive signal y can be written as {y-i , y₂}. Further, a pilot signal comprising of periodically transmitted resource elements transmitted by the radio transmitter (e.g. by a so-called eNodeB as specified in 3GPP LTE documents) are detected, (e.g. cell specific reference signal, CRS, and/or a user specific reference signal URS).

The cannel estimator 14 performs a channel estimation based on the pilot signal resulting in an estimated channel matrix H which elements H represent the actual transmission channels between each the number of i transmit antennas of the wireless transmitter and the number of j receive antennas of the wireless receiver.

The covariance determination unit 16 determines a disturbances covariance matrix based on the estimated channel H and the pilot signal, being discussed in more details in the following Fig. 2. The disturbances may comprise signal noise and interference. It may additionally comprise estimation or measurement errors.

The equalizer 18 performs an estimation of the transmitted symbols (or soft bits if it is assumed that the equalizer additionally performs a de-mapping of symbols as a function of the covariance matrix P'). By way of example the equalizer 18 is (or comprises) an interference rejection combining, IRC, equalizer that determines the combining weights as a function of the covariance matrix and the channel estimate. Further details for IRC equalization can be drawn e.g. from above-cited 3GPP document R4-1 13528, Annex A, providing amongst others different approaches to obtain an IRC weight matrix as a function of the channel estimate and the covariance matrix.

It is to be understood that the structures as illustrated in Fig. 1 are merely schematic and that the node may actually include further components which, for the sake of clarity, have not been illustrated, e.g. further signal algorithm units , further interfaces or further processors. Also, it is to be understood that the memory M may include further types of program code modules, which have not been illustrated, e.g. program code modules for implementing known functionalities of the wireless receiver.

Fig. 2 shows a block diagram of an exemplary covariance determination unit 16 in more details. The covariance determination unit 16 comprises a disturbances covariance estimator 162, a scaling factor determination unit 164, a SINR determination unit 166 and a covariance adaptation unit 168. The covariance estimator 162 determines a covariance matrix R comprising the main diagonal matrix elements r denoting a noise power plus interference power at each antenna i and the off-diagonal matrix elements r denoting the cross power and interference terms. The disturbances may dominantly comprise noise and interference, but also some estimation, calculation and/or processing errors. In the following, the disturbances covariance (matrix) will also be denoted as noise plus interference covariance (matrix), as noise covariance (matrix) or simply as covariance (matrix).

For a two-antenna receiver, the estimated corresponding 2x2 noise covariance matrix can be written as:

* Λ

R = ^r21

r2\ ^r22 j wherein r -π and r ₂2 are the main diagonal covariance matrix elements, and r ₂i and r ₂ (r -i ₂ = r^* ₂ ) are the off-diagonal covariance matrix elements The scaling factor determination unit 164 calculates a scaling factor S for covariance adaptation. The scaling factor S is determined as a function of the SINR being determined by the SINR determination unit 166.

Above-described scaling factor S may be used in the covariance adaptation unit 168 to scale the off-diagonal elements (i.e. the matrix entries different from the main diagonal entries) of the estimated noise covariance matrix R. The resulting noise covariance may be denoted adapted covariance matrix R'.

For a two-antenna receiver, the adapted noise covariance matrix R' may be written as:

In an embodiment, the scaling factor is determined such to provide on one hand no loss for spatially white noise receive signals in an IRC receiver compared to a first generation (MMSE) receiver. On the other hand the scaling shall provide as much improvement as possible for colored noise.

By way of example, the scaling factor S may be defined as follows:

S = S2 - f{SINR)- (S2 - SI) V SINR e [Thrl, Thrl]

S = S2 V SINR < Thr\

S = Sl V SINR > Thrl wherein:

• S is a scaling factor used to adapt the covariance matrix,

• f(x) is an arbitrary function with the boundary conditions f(Thr1 ) = 0 and f(Thr2) = 1 . In an embodiment, the function f(x) is monotonically increasing within the input interval between the threshold values Thrl and Thr2,

• Thrl and Thr2 represent the lower and the upper SINR threshold respectively (as an example, Thrl may equal a value between 15dB and 20dB (e.g. 18 or 20dB), and Thr 2 may equal a value between 25dB and 30dB (e.g. 26, 28 or 30 dB),

• S1 and S2 represent the lower and the upper scaling factor limit, and

• SINR represents a current signal-to-interference and noise ratio value.

As discussed above, the adapted noise covariance matrix R' may be used for the weight computation of the linear IRC receiver or for the spatial whitening filter computation of the non-linear sphere receives as discussed above.

The SINR determination unit 166 may determine the ratio between the signal power and the disturbances power. The disturbances power may dominantly comprise noise power and interference power, thus representing a signal-to- noise-plus-interference ratio and in the following being also referred to as SINR.

The SINR may be determined based on the channel estimation and noise estimates of the noise covariance estimation unit 1 62. In an example, the SINR value may be determined with respect a selected antenna i of the plurality of receive antennas:

SiNR_; = !— ^

Alternatively, an overall or combined SINR value may be determined. For a two- antenna receiver the overall SINR may be written as:

SINR =

The scaling factor S can be calculated based on the estimated SINR on each received signal samples (e.g. RE in LTE).

Alternatively, in order to get an accurate SINR estimation, the scaling factor S may be left stable for a certain bandwidth or timing interval. Further alternatively, the scaling factor S may be filtered for a certain bandwidth or timing interval. Fig. 3 shows an exemplary diagram to illustrate an example implementation of determining the scaling factor S in dependency of the SINR. The diagram depicts the scaling factor S (y-coordinate) as a function of the SINR value (x-coordinate). By way of example, the scaling factor S shows a first constant value (upper scaling factor limit S2) in the SINR interval between 0 and the lower SINR threshold Thr1 . Between the lower SINR threshold Thr1 and the upper SINR threshold Thr2, the scaling factor S exemplarily drops down linearly from the first constant value (upper scaling factor limit S2) to a second constant value (lower scaling factor S1 ). Above the upper SINR threshold Thr 2, the scaling factor S shows a second constant value (lower scaling factor limit S1 ). In an exemplary practical implementation S2 may be selected to be equal to 1 , while S1 may be selected to be equal to 0.

Fig. 4 illustrates an exemplary sequence of steps to be performed in the wireless receiver to recover the transmit symbols (or soft bits) from the receive signal vector:

In a first step S1 1 , a covariance matrix of the receive signal vector is estimated.

In a second step S12, an SINR value indicative of a ratio of a noise and interference power with respect to a receive signal power is determined, based on elements of the channel estimate and on elements of the covariance matrix,

In a third step S13, an adapted covariance matrix is determined by adapting off- diagonal elements of the covariance matrix elements in dependency of the SINR value.

In a fourth step S14, the adapted covariance matrix is used to recover the transmit symbols (or soft bits), e.g. to determine an IRC weight matrix.

Claims

1 . A method in a wireless receiver (1 ) comprising a plurality of receive antennas (10) to receive a plurality of signals being transmitted from a wireless transmitter, wherein the plurality of signals form a receive signal vector, comprising the following steps:

• determining (S1 1 ) a covariance matrix of the receive signal vector,

• determining (S12) a ratio value, SINR, indicative of disturbances

power with respect to a receive signal power,

• determining (S13) an adapted covariance matrix by adapting certain covariance matrix elements in dependency of the SINR, and

• using (S14) the adapted covariance matrix to recover a transmitted symbol from the receive signal vector.

2. The method of the preceding claim, wherein the SINR is indicative of a ratio between the signal power of the transmit signal and a sum of noise power and a power of signals received from an interfering cell

3. The method of the preceding claim, wherein the SINR is determined as a function of estimated channel matrix values and covariance matrix values.

4. The method of anyone of the preceding claims, wherein the certain covariance matrix elements are set to a certain value, if the SINR is being determined above a certain threshold.

5. The method of the preceding claim, wherein the certain covariance matrix elements are set to zero, if the SINR is being determined above a certain threshold.

6. The method of anyone of the preceding claims 1 -3, wherein the certain covariance matrix elements are multiplied with a scaling factor within a certain SINR interval, wherein the scaling factor is determined as a function of the SINR.

7. The method of anyone of the preceding claims, wherein the certain matrix elements being adapted comprise the off-diagonal matrix elements denoting the matrix elements different from the main diagonal of the covariance matrix.

8. The method of the preceding claim, wherein the scaling factor is determined to scale the off-diagonal matrix elements as a function of the SINR such that the off -diagonal values are

• let unchanged below a first SINR threshold,

• set to zero above a second SINR threshold, and

• monotonically scaled in dependency of the SINR in between the first SINR threshold and the second SINR threshold.

9. The method of the preceding claim, wherein a scaling factor is a result of a filter operation of the SINR.

10. A wireless receiver (1 ) comprising a plurality of receive antennas (10) to receive a plurality of signals being transmitted from a wireless transmitter, wherein the plurality of signals form a receive signal vector, comprising a processor (P) adapted for:

• determining a covariance matrix of the receive signal vector,

• determining a ratio value, SINR, indicative of noise and interference power with respect to a receive signal power,

• determining an adapted covariance matrix by adapting certain

covariance matrix elements in dependency of the SINR, and

• using the adapted covariance matrix to recover a transmitted symbol from the receive signal vector.

1 1 . The wireless receiver (1 ) of the preceding claim, wherein the receiver comprises a combining weight matrix unit for combining the receive signal vector elements being adapted to use the adapted covariance matrix.

12. The wireless receiver (1 ) of claim 10, wherein the receiver comprises a pre- whitening unit being adapted to use the adapted covariance matrix.

13. The wireless receiver of anyone of claims 10-12, being adapted to implemented in a user equipment, UE, e.g. according to LTE specifications.

14. A computer program loadable into a memory (M) of a wireless receiver (1 ) the computer program comprising instructions adapted to execute the method of any of the preceding method claims.