阅读说明：本技术 Mimo系统中的信道估计 (Channel estimation in MIMO systems ) 是由 戴凌龙 谭竞搏 王碧钗 R·麦肯齐 郝墨 于 2020-01-23 设计创作，主要内容包括：一种在多输入多输出(MIMO)通信系统中执行信道估计的方法,该MIMO通信系统包括被设置为通过通信信道进行通信的下行链路装置和上行链路装置,该方法包括：在上行链路装置处：在角域中分析表示下行链路装置与上行链路装置的第一天线之间的信道增益的第一信道向量,以识别角域支持点集合,各角域支持点指示第一信道向量在角域中的各元素；根据角域支持点集合生成值向量,该值向量包含第一信道向量在角域中的表示最大信道增益的元素；从上行链路装置向下行链路装置反馈值向量的指示；在下行链路装置处：在角域中分析表示下行链路装置与上行链路装置的第二天线之间的信道增益的第二信道向量,以识别相同的角域支持点集合；根据角域支持点集合生成发送导向矩阵；以及根据从上行链路装置反馈的值向量的指示和所生成的发送导向矩阵生成第一信道向量的估计。(A method of performing channel estimation in a multiple-input multiple-output (MIMO) communication system comprising a downlink apparatus and an uplink apparatus arranged to communicate over a communication channel, the method comprising: at an uplink device: analyzing a first channel vector representing a channel gain between a first antenna of a downlink device and an uplink device in an angular domain to identify a set of angular domain support points, each angular domain support point indicating each element of the first channel vector in the angular domain; generating a value vector from the set of angular domain support points, the value vector containing an element of the first channel vector in the angular domain representing a maximum channel gain; feeding back an indication of a vector of values from the uplink device to the downlink device; at a downlink device: analyzing a second channel vector representing channel gain between a second antenna of the downlink device and the uplink device in the angular domain to identify a same set of angular domain support points; generating a transmission steering matrix according to the angular domain support point set; and generating an estimate of the first channel vector from the indication of the value vector fed back from the uplink device and the generated transmit steering matrix.)

1. A method of performing channel estimation in a multiple-input multiple-output, MIMO, communication system comprising a downlink apparatus and an uplink apparatus arranged to communicate over a communication channel, the method comprising the steps of:

at the uplink device:

analyzing a first channel vector representing a channel gain between a first antenna of the downlink device and the uplink device in an angular domain to identify a set of angular domain support points, each angular domain support point indicating each element of the first channel vector in the angular domain;

generating a vector of values from the set of angular domain support points, the vector of values containing an element of the first channel vector in the angular domain representing a maximum channel gain;

feeding back an indication of the vector of values from the uplink device to the downlink device;

at the downlink device:

analyzing a second channel vector representing channel gain between a second antenna of the downlink apparatus and the uplink apparatus in the angular domain to identify a same set of angular domain support points;

generating a transmission steering matrix according to the angular domain support point set; and

generating an estimate of the first channel vector from the indication of the vector of values fed back from the uplink device and the generated transmit steering matrix.

2. The method of claim 1, wherein the method further comprises the steps of: using the estimate of the first channel vector to generate an estimate of a channel matrix that models a channel between the downlink device and an uplink device.

3. The method of claim 2, wherein the method further comprises the steps of: deriving a digital precoding matrix using the estimated channel matrix for precoding data to be transmitted from the downlink apparatus.

4. A method according to any preceding claim, wherein the uplink device is arranged to use the first antenna only for receiving communications from the downlink device and to use the second antenna for receiving and transmitting communications from and to the downlink device.

5. The method according to any of the preceding claims, wherein the step of analyzing the first channel vector in the angular domain comprises: transforming the first channel vector to the angular domain using a Discrete Fourier Transform (DFT) matrix.

6. The method of any preceding claim, wherein the set of angular domain support points is identified from among non-zero elements of the first channel matrix in the angular domain.

7. The method according to any of the preceding claims, wherein the vector of values is generated by selecting the P largest elements of the first channel vector in the angular domain, where P is the number of resolvable paths between the downlink device and the uplink device.

8. The method according to any one of the preceding claims, wherein the method comprises the steps of: quantizing the vector of values at the uplink device using a codebook, the indication of the vector of values comprising the selected index of the codebook.

9. The method of claim 8, wherein the indication of the vector of values further comprises a magnitude of the vector of values.

10. The method of claim 9, wherein only the selected index and the magnitude of the vector of values are fed back from the uplink device to the downlink device.

11. The method of any of claims 8 to 10, wherein the selected index I_nIs based onTo select, among others,is the value vectorThe first-order conjugate of the first-order conjugate,is the codebook, c_iIs the quantized vector of the codebook C, and B is the number of bits to access the codebook.

12. The method of any preceding claim, wherein each angular support point of the identified set of angular support points is dependent on a beam transmission angle for each path between the downlink and uplink devices.

13. The method of claim 12, wherein the transmit steering matrix is generated using one or more beam emission angles calculated from the set of angular domain support points.

14. The method according to any of the preceding claims, wherein the set of angular domain support points Φ is given by:wherein d is the spacing between adjacent antenna elements of the downlink apparatus; λ is the wavelength of the signal transmitted from the downlink apparatus; n is the number of antenna elements at the downlink device, and phi_pIs the transmission angle of a signal beam transmitted by the downlink apparatus along the p-th path between the downlink apparatus and the uplink apparatus.

15. The method of claim 5, wherein analyzing the second channel vector in the angular domain comprises: transforming the second channel vector to the angular domain using the Discrete Fourier Transform (DFT) matrix.

16. The method of any preceding claim, wherein the MIMO communication system is a time division duplex, TDD, MIMO communication system.

17. A multiple-input multiple-output, MIMO, communication system comprising a downlink apparatus and an uplink apparatus configured to communicate over a communication channel;

the uplink apparatus includes:

a plurality of antennas; and

a processing unit configured to:

analyzing a first channel vector representing a channel gain between a first antenna of the downlink device and the uplink device in an angular domain to identify a set of angular domain support points, each angular domain support point indicating each element of the first channel vector in the angular domain;

generating a vector of values from the set of angular domain support points, the vector of values containing an element of the first channel vector in the angular domain representing a maximum channel gain; and is

Feeding back an indication of the vector of values to the downlink apparatus;

the downlink apparatus includes:

a plurality of antennas; and

a processing unit configured to:

analyzing a second channel vector representing channel gain between a second antenna of the downlink apparatus and the uplink apparatus in the angular domain to identify a same set of angular domain support points;

generating a transmission steering matrix according to the angular domain support point set; and is

Generating an estimate of the first channel vector from the indication of the vector of values fed back from the uplink device and the generated transmit steering matrix.

Technical Field

The present invention relates to performing channel estimation in a MIMO communication system.

Background

Millimeter wave (mmWave) wireless communication systems are receiving increasing attention as they ensure that the increasing bandwidth requirements of wireless devices are met. Millimeter wave systems typically operate in the 30GHz to 300GHz frequency band. This is a larger frequency band than the sub 6GHz band currently used for Long Term Evolution (LTE) networks, and therefore may support a larger bandwidth than is supported by existing systems currently operating in the sub 6GHz band.

One problem associated with millimeter wave communications is the high free space path loss that may be experienced. This high path loss may cause severe attenuation of signals that experience blocking or communication across long distances.

One way to overcome this problem is to implement millimeter wave communication within a multiple-input multiple-output (MIMO) system. The shorter wavelengths of millimeter wave communication enable reduced antenna spacing for MIMO systems, and thus enable packaging of larger antenna arrays (containing, for example, 256 to 1024 antenna elements) in smaller physical sizes than is achievable when operating in the sub-6 GHz band. These large antenna arrays can effectively compensate for the high path loss caused by the higher frequency communication of millimeter waves.

MIMO systems may be implemented with all-digital precoding. An example MIMO Base Station (BS) with all-digital precoding is shown in fig. 1.

Base station 100 includes a digital precoder 102 and a plurality of Radio Frequency (RF) chains, generally designated 104, coupled to digital precoder 102. Each RF chain is coupled to an antenna (generally designated 108). In the example shown here, each RF chain is coupled to each antenna by an amplifier. The amplifier is generally indicated at 106.

The digital precoder 102 receives a plurality of data streams, generally designated 110, and operates to control the amplitude and phase of each received data stream to achieve a transmit beam from the antenna 108 having a desired direction and gain. The data stream, once precoded, is communicated via the RF chain 104. Each RF chain may support a single data stream. The RF chain operates to convert the digitally precoded data streams into analog signals for transmission by the antennas 108. Each RF chain typically includes transceiver circuitry for generating analog signals from a received digitally precoded data stream. The transceiver circuitry may include, for example, digital-to-analog (DAC) converters, mixers, and frequency converters. The output signals generated by the RF chains are then amplified by respective amplifiers and transmitted from respective antennas.

Alternatively, MIMO systems may be implemented using hybrid precoding, which separates the precoding operation into the analog and digital domains. Digital precoding may be implemented using a digital precoder to control the weights associated with each RF chain. Analog precoding may be achieved by controlling the phase of the signals transmitted by the antennas using phase shifters.

It has been appreciated that accurate Channel State Information (CSI) estimation for a communication channel between a base station and a user device of a MIMO system is important for the performance of the MIMO system. For example, the downlink transmission rate (i.e., the transmission rate from the BS to the user device) may depend on precoding to mitigate interference between different data streams. Thus, accurate knowledge of CSI may be used to set parameters for precoding to mitigate interference and thus improve performance.

It has been found that the difficulty of obtaining accurate downlink channel estimates may increase as the number of antennas at the BS increases. In systems where channel reciprocity between uplink and downlink channels is preserved, these problems can be avoided by obtaining CSI for estimating precoding from the uplink channel. However, there are certain wireless systems where full channel reciprocity is not maintained. For example, some user devices are configured to employ fewer antennas for transmitting in the uplink direction than antennas employed for receiving in the downlink direction. This may be done on the basis that the transmit RF chain is more power and hardware intensive than the receive RF chain, to reduce the power and hardware requirements of the device. In these cases, only partial channel reciprocity remains, which makes some form of channel feedback from the user equipment to the BS desirable to obtain a full CSI estimate.

Disclosure of Invention

According to the present invention, there is provided a method of performing channel estimation in a multiple-input multiple-output (MIMO) communication system including a downlink apparatus and an uplink apparatus arranged to communicate over a communication channel, the method comprising the steps of:

at an uplink device:

analyzing a first channel vector representing a channel gain between a first antenna of a downlink device and an uplink device in an angular domain to identify a set of angular domain support points, each angular domain support point indicating each element of the first channel vector in the angular domain;

generating a value vector from the set of angular domain support points, the value vector containing an element of the first channel vector in the angular domain representing a maximum channel gain;

feeding back an indication of a vector of values from the uplink device to the downlink device;

at a downlink device:

analyzing a second channel vector representing channel gain between a second antenna of the downlink device and the uplink device in the angular domain to identify a same set of angular domain support points;

generating a transmit steering matrix (transmit steering matrix) from the set of angular domain support points; and

an estimate of the first channel vector is generated based on the indication of the value vector fed back from the uplink device and the generated transmit steering matrix.

The method may further comprise the steps of: the estimate of the first channel vector is used to generate an estimate of a channel matrix that models a channel between the downlink device and the uplink device.

The method may further comprise the steps of: a digital precoding matrix is derived using the estimated channel matrix for precoding data to be transmitted from the downlink apparatus.

The uplink device may be arranged to use the first antenna only for receiving communications from the downlink device and to use the second antenna for receiving and transmitting communications from and to the downlink device.

The step of analyzing the first channel vector in the angular domain may comprise: the first channel vector is transformed to the angular domain using a Discrete Fourier Transform (DFT) matrix.

The set of angular domain support points may be identified from non-zero elements of the first channel matrix in the angular domain.

The value vector may be generated by selecting the P largest elements of the first channel vector in the angular domain, where P is the number of resolvable paths between the downlink device and the uplink device.

The method can comprise the following steps: a vector of values is quantized at the uplink device using a codebook, the indication of the vector of values comprising a selected index of the codebook.

The indication of the vector of values may also include a magnitude of the vector of values.

Only the magnitude of the selected index and value vectors may be fed back from the uplink device to the downlink device.

Selected index I_nCan be based onTo select, among them,is a vector of valuesHermitian (Hermitian) conjugation,is a codebook, c_iIs the quantized vector of codebook c and B is the number of bits to access the codebook.

Each angular support point of the identified set of angular support points may depend on a beam transmission angle for each path between the downlink device and the uplink device.

The transmit steering matrix may be generated using one or more beam emission angles computed from a set of angular domain support points.

The set of angular support points φ can be given byWherein d is the spacing between adjacent antenna elements of the downlink apparatus; λ is the wavelength of the signal transmitted from the downlink device; n is the number of antenna elements at the downlink device, and phi_pIs the transmission angle (angle of the future) of the signal beam transmitted by the downlink device along the p-th path between the downlink device and the uplink device.

The step of analyzing the second channel vector in the angular domain may comprise transforming the second channel vector to the angular domain using a Discrete Fourier Transform (DFT) matrix.

The MIMO communication system may be a Time Division Duplex (TDD) MIMO communication system.

According to a second aspect of the present disclosure, there is provided a multiple-input multiple-output (MIMO) communication system comprising a downlink apparatus and an uplink apparatus configured to communicate over a communication channel;

the uplink device includes:

a plurality of antennas; and

a processing unit configured to:

analyzing a first channel vector representing a channel gain between a first antenna of a downlink device and an uplink device in an angular domain to identify a set of angular domain support points, each angular domain support point indicating each element of the first channel vector in the angular domain;

generating a value vector from the set of angular domain support points, the value vector containing an element of the first channel vector in the angular domain representing a maximum channel gain; and is

Feeding back an indication of a vector of values to a downlink apparatus;

the downlink apparatus includes:

a plurality of antennas; and

a processing unit configured to:

analyzing a second channel vector representing channel gain between a second antenna of the downlink device and the uplink device in the angular domain to identify a same set of angular domain support points;

generating a transmission steering matrix according to the angular domain support point set; and is

An estimate of the first channel vector is generated based on the indication of the value vector fed back from the uplink device and the generated transmit steering matrix.

Drawings

The invention will now be described by way of example with reference to the accompanying drawings. In the drawings:

fig. 1 shows an example of an all-digital precoded MIMO transmitter.

Fig. 2 shows an example of a MIMO communication system.

Fig. 3 shows a flow chart of steps for performing channel estimation according to the present disclosure.

Fig. 4A and 4B show examples of functions defined by channel vectors in the angular domain.

Fig. 5 shows simulation results comparing data rates achieved with the channel estimation techniques described herein and with conventional channel estimation techniques.

Detailed Description

One approach to solving the problem of obtaining full CSI estimates with only partial channel reciprocity is to feed back to the BS channel estimates (i.e., channel vectors) for the channel between the BS and the receive-only antennas of the user devices. This may be done by performing a limited feedback by use of a quantization codebook of vectors known to the BS and the user equipment. At the user equipment, the channel vector is quantized using the codebook, and an index of a selected vector of the codebook for the quantized channel vector is fed back to the BS. The feedback overhead of this approach is B bits, where B is the number of bits required to index the codebook (i.e., the codebook contains up to 2^BOne quantized vector). In this method, the following is generally the case: a full-dimensional channel vector is quantized (e.g., for a BS containing N antennas, an N-element channel vector is quantized and returned to the BS).

The present disclosure is directed to a method of performing channel estimation in a MIMO system, in which a channel vector is analyzed in an angular domain. The channel vectors represented in the angular domain may be referred to herein as angular domain channel vectors. A MIMO system includes a downlink device (e.g., BS) and an uplink device (e.g., user device). At an uplink device, a set of one or more elements of a channel vector expressed in an angular domain (i.e., an angular domain channel vector) is identified. The downlink device is not aware of the channel vector. The value of each identification element may represent a channel gain for each physical path between a downlink apparatus and an uplink apparatus of the MIMO system. In some examples, the set of identified elements of the angular domain channel vector is a set of non-zero elements of the vector. The set of identified elements is referred to herein as angular domain support points because these points identify the non-zero components of the channel vector in the angular domain. From the set of angular domain support points, a vector formed by elements of the angular domain vector representing the largest channel gain is identified. This vector may be referred to herein as a value vector. The value vector typically has a dimension (size) less than N, where N is the number of antennas at the downlink device, and in some examples, the value vector may have a dimension L, where L is the number of resolvable physical paths between the downlink device and the uplink device. For massive MIMO systems, the following is typically the case: l is significantly less than N. An indication of the identified value vector is then fed back from the uplink device to the downlink device, e.g., using a quantization codebook. At the downlink device, additional channel vectors known to the downlink device are analyzed in the angular domain to identify the same set of angular domain support points. These angular domain support points are used with the quantized vectors fed back to generate an estimate of the unknown channel vector at the downlink device. The method enables channel estimates to be formed at the downlink device without the need for feedback of quantized N-dimensional channel vectors. It also avoids the need for feedback that has been identified as a corner-area support point common to both uplink and downlink devices. This may reduce channel feedback overhead.

Various aspects of the disclosure are described in more detail below.

Fig. 2 illustrates a MIMO communication system 200. System 200 includes a downlink device 202 and an uplink device 204. The downlink apparatus 202 may be, for example, a Base Station (BS), or more generally a MIMO transmitter. The uplink device 204 may be a user device or a MIMO receiver. The downlink device 202 is operative to transmit wireless communication signals in a downlink direction over a communication channel 206 to the device 204. The number of physical communication paths defined by a channel is denoted P. In this example, for illustration, P is 3. P represents the number of valid or resolvable physical paths. A resolvable or valid path may be a path where the signal power exceeds a certain threshold. Scattering objects 208 and 210 operate to scatter signals received from downlink device 202 that partially define a resolvable path.

MIMO system 200 is a Time Division Duplex (TDD) system (TDD MIMO system).

Downlink device 202 includes a digital precoder 212, a set of RF chains, generally indicated as 214, an antenna array 216, and a processing unit 218. The antenna array 216 includes N antennas. The antenna array 216 is a Uniform Linear Array (ULA) in this example.

The digital precoder 212 receives a plurality (N)_sData streams 220 and performs digital precoding on the streams. The pre-encoded stream is then passed via the RF chain 214. Each RF chain may support a single data stream. The number of RF chains may be equal to the number of antennas N. The RF chains are connected to an antenna array 216.

Uplink device 204 includes a lens array 222, which in this example is also a ULA. Lens array 222 includes M antennas. In the example shown in fig. 2, M is 2. The lens array 222 is coupled to a set of RF chains 224. The set of RF chains is coupled to a digital combiner 226. The uplink device 204 further comprises a processing unit 228.

In operation, precoder 212 receives a plurality (N)_sOne) data stream. The precoder performs digital precoding on the data streams. Precoding may include adjusting the weight and/or phase of each data stream. Precoding may be performed to reduce interference between different data streams. The precoding may be performed by applying a digital precoding matrix F to the data stream. The precoded data streams are passed through the set of RF chains 214 to generate analog signals. A single data stream is conveyed by means of each RF chain; that is, each RF chain supports a single data stream. The analog signal generated by the RF chain indicates or represents the data stream communicated via the RF chain. The signals generated by the RF chains are passed to an antenna array 216 for transmission over channel 206 to uplink device 204. The signals are transmitted to the antennas of the antenna array 216 such that the signals transmitted by the array 216 form a plurality of discrete beams. A beam refers to a set of one or more signals focused along a particular direction or travel angle. Signals are transmitted from the antenna array 216 such that the beams travel along various communication paths; i.e. each beam travels along a respective physical path. In this illustrated example, the transmitted signal forms three beams: a first beam traveling on a first path, a second beam traveling on a second path, and a third beam traveling on a third path.

Each beam has an angle of emission (AoD) from the antenna array 216. The emission angle may be measured relative to a reference direction defined with respect to the array. The first pathThe emission angle of the beam in the path is denoted as phi₁The emission angle of the beam on the second path is denoted as phi₂And the emission angle of the beam on the third path is denoted as phi₃。

The transmitted beam is transmitted through a channel and received at device 204. Each beam is incident on antenna array 222 at an angle of arrival (AoA). The angle of arrival of each beam is measured relative to a reference direction defined with respect to array 222. The angle of arrival of the beam on the first path is denoted θ₁The angle of arrival of the beam on the second path is denoted as θ₂And the angle of arrival of the beam on the third path is denoted as θ₃。

The beam is received at an antenna array 222. The antennas of the array 222 are coupled to the RF chains of the set of chains 224. Each antenna of array 222 may be coupled to a respective RF chain. The RF chain 224 performs the inverse of the RF chain 214 of the device 202. That is, while RF chain 214 performs processing operations on the received precoded data stream to generate an analog signal, RF chain 224 operates to generate a digital signal that represents the precoded data stream from the received analog signal. The digital signals generated by the RF chains are then transmitted to the combiner 226. The combiner 226 operates to undo (reduce) or remove the precoding applied to the data stream by the precoder 212 of the device 202 to generate a decoded data stream 230.

The signal received at the device 204 may be modeled as:

y＝HFs+n (1)

in equation (1), y is the signal vector received at the downlink device 204; h is a spatial channel matrix that models the channel between device 202 and device 204; f is the digital precoder matrix applied by the digital precoder 212, s is the signal vector transmitted from the device 202, and n is the noise vector.

The vector y has the dimension M × 1, i.e., it is an M-element vector. The vector s has the dimension N × 1, i.e., it is an N-element vector. The channel matrix H is an mxn matrix, i.e., a matrix having M rows and N columns. The channel matrix H is defined as:

H＝[h₁，h₂，…，h_M]^T (2)

wherein，h_iRepresenting the channel vector between the antenna of device 202 and the i-th antenna of device 204. Thus, each channel vector h_iIs an N x 1 vector (i.e., an N element vector), the jth element [ h ] of the channel vector_i]_jRepresenting the channel between the jth antenna of the downlink apparatus 202 and the ith antenna of the uplink apparatus 204.

The digital precoder matrix F is an N × N matrix. The noise vector n is an additive white Gaussian (Gaussian) noise (AWGN) vector that satisfies:

n～CN(0，σ²I_M) (3)

wherein σ²Represents a change in noise power, and I_MIs an M × M identity matrix.

The precoder matrix F satisfies the following power constraint:

where ρ is the average received signal power at the receiver.

The transmission signal vector s satisfies:

E(ss^H)＝I_M (5)

the antenna array 216 is a Uniform Linear Array (ULA), and thus, the channel matrix that models the channel can be expressed as:

wherein, g_pIs the path gain for the p-th physical path, a (φ, N) and a (θ, M) are the array response vectors at the downlink device 202 and uplink device 204, respectively, φ_pIs the emission angle of the p-th path, and θ_pIs the angle of arrival of the p-th path. g_pSatisfy CN (0, 1) and

the array response vector a (φ, N) may be written as:

where λ is the wavelength of the signal and d is the spacing between adjacent antennas at the downlink apparatus 202. The values of λ and d are expressed in the same units (e.g., meters).

The matrix a may be defined as:

A＝[a(φ₁，N)，a(φ₂，N)…a(φ_P，N)] (8)

and vector g_mCan be defined as:

g_m＝[g₁a(θ₁，M)(m)，g₂a(φ₂，M)(m)，…，g_La(φ_P，M)(m)]^T (9)

by virtue of these definitions, there are (using equations (2) and (6)):

h_m＝Ag_m (10)

matrix a is a transmit steering matrix; i.e., the steering matrix of the array 216 of downlink devices 202.

Equation (10) summarizes the channel vector h_mCan be derived from a steering matrix that depends on the emission angle phi and can be viewed as a receiver gain vector g_mAnd (4) determining.

In the example shown in fig. 2, uplink device 204 includes two antenna elements, whereby H ═ b₁，h₂]. According to equation 10:

h₁＝Ag₁ (11)

h₂＝Ag₂ (12)

uplink device 204 is configured such that a first one of its antennas (referred to as antenna 1) is set to receive downlink transmissions from device 202 and transmit uplink transmissions to device 202, and a second one of its antennas (referred to as antenna 2) is set to receive downlink transmissions only from device 202. That is, device 204 cannot transmit uplink transmissions from antenna 2, i.e., it cannot transmit uplink transmissions from antenna 2The signal is transmitted in the uplink direction. Thus, there is only partial reciprocity of channel 206, meaning that downlink device 202 is able to determine, e.g., with an uplink channel estimation technique, a channel vector h representing the channel vector between the antenna of device 202 and antenna 1 of device 204₁But h cannot be estimated accurately using similar techniques₂(representing the channel vector between the antenna of device 202 and antenna 2 of device 204).

The method for estimating the channel vector h at the downlink device 202 will now be described with reference to the flowchart in fig. 3₂The method of (1). In the following description, it is assumed that downlink apparatus 202 is capable of calculating channel vector h₁E.g., by means of uplink channel estimation. It is further assumed that the uplink means 204 are able to calculate two channel vectors h, e.g. by means of downlink channel estimation₁And h₂Is estimated.

At step 302, the uplink device 204 analyzes the channel vector h in the angular domain₂To identify the vector h₂The angular domain of (2) supports a set of points. Step 302 may be performed by processing unit 208 of apparatus 204. The device 204 analyzes the channel vector h₂Is represented as (is represented as)) To analyze the channel vector h in the angular domain₂. In other words,is h₂The angular domain vector of (a), or in other words,is a channel vector h in the angular domain₂. Indicating or identifying angular domain vectors by each angular domain support pointEach element of (1).

Angular domain vectorCan be obtained by applying a channel vector h₂Performing a Discrete Fourier Transform (DFT) generation. This can be achieved using a DFT matrix. Channel vector h₂May be multiplied by the DFT matrix D to generate an angular domain vectorMathematically, means 204 generates an angular domain vector according to:

the DFT matrix D can be expressed as:

thus, the element of D is given by:

wherein p ', q' is 0, 1, …, N-1.

To understand how the analysis of angular domain vectors can be used to identify angular domain support points, consider the following example: the uplink device 204 contains only a single antenna (i.e., M ═ i), and there is only one physical path (i.e., P ═ 1). In this case, a (θ, M) ═ 1 and equation (10) are reduced to:

H＝h＝g₁a(φ₁，N) (16)

further comprising:

from equation (15), therefore, line pth of D is given by:

using equations (7), (16), (17) and (18),the pth element of (a) can be written as:

wherein the content of the first and second substances,

taking the modulus of equation (19) gives:

wherein the content of the first and second substances,

equation (20) showsIs at a sample point η_pFunction f (η) of (C)_p) The sample of (1). In other words, angular domain vectorIs determined by the value of the element at the sample point η_pFunction f (η) of (C)_p) It is given. Function f (η)_p) The sine (sine) -like distribution with center point f (0) is given.

According to equation 20, therefore, at η_pWhen the value is 0:

wherein p is₀Is the AND function f (η)_p) Corresponding to a central point ofAngular domain vectorOf (2) is used. Thus, p₀The values of (d) are given as:

wherein the content of the first and second substances,andrepresents p₀Is given the nearest integer (because p₀Identify elements of the angular domain vector).

At phi₁Angular domain vector satisfying equation (21)Having only one non-zero element p₀. If phi is₁Equation (21) is not satisfied, thenThere may be multiple non-zero elements. However, since the value of N (the number of antennas at the downlink device 202) is large in a typical system, it is not necessary to set the number of antennas at the downlink device 202 to be largeWill still surround the point p₀Is distributed and is only at point p₀Nearby vector elements will take on non-negligible values. In other words, even if equation (21) is not satisfied, the angular domain vectorWill still focus on the vector element p₀And (4) surrounding.

Fig. 4A and 4B illustrate an example f (η [)_p) AndFIG. 4A shows phi₁The case when equation (21) is satisfied, and FIG. 4B shows φ₁The case when equation (21) is not satisfied. As can be seen in FIG. 4A, only the non-zero elements of the angular domain vector are p₀. FIG. 4B shows that the angular domain vector contains a number of non-zero elements, but only at p₀Those within the vicinity (shown at 404 and 406) have non-negligible values. In both figures, the circle symbols represent the values of the elements of the angular domain vector h. The solid line represents the function f (η) defined by the channel vector in the angular domain_p)。

The above description illustrates the element p of a physical path forming and angular domain vector_oA single center point of the channel vector in the corresponding angular domain. In other words, the channel vector defines a function in the angular domain having a single center point, the center point corresponding to the element p of the angular domain vector_o. The center point may be associated with the element p_oExactly aligned (if phi)₁Satisfies equation (21)), or element p_oMay be the element located closest to the center point of the function (if phi)₁Equation (21) is not satisfied, e.g., as shown in fig. 4B). The center point (and thus the element p)_o) Dependent on the emission angle phi of the beam for the physical path₁。

Angular domain vectors when there are multiple physical paths between devices 202 and 204Defining each having an and vectorA linear combination of functions corresponding to a single center point for each element of (a). In other words, for the P physical paths, the angular domain vectorDefining a linear combination of P functions, wherein each of the P functions has elements related to an angular domain vectorCorresponding single centerAnd (4) point. Each elementDependent on the beam emission angle, i.e. element, for each physical pathDepending on the beam emission angle of the physical path P, where P is 1, …, P. This set of elements is referred to herein as a set of angular domain support points.

Referring back to the example system shown in fig. 2, the uplink device 204 analyzes the angular domain vectorTo calculate the set of angular support points phi for the vector given by₂：

Thus, each angular domain support point identifies each element of the angular domain vector. Angular domain vectors at various angular domain support pointsThe values of (b) represent the gains for each physical path.

At step 304, the uplink device 204 generates a vector of values from the set of angular domain support points. This step may also be performed by the processing unit 228.

The value vector is composed of angular domain vectors representing maximum channel gainThe element (c) is formed. The value vector may be generated by selecting the L largest elements of the angular domain vector. In some examples, L ═ P. In other examples, L is less than P.

Mathematically, the value vector may be generated by first vectoring the channel vector h₂The selected row of the DFT matrix D is applied for calculation. The rows of the DFT matrix to be applied are determined by the corner-domain control points. In particular, the line numberA row equal to the angular domain support point will be applied to the channel vector. This is mathematically expressed as:

or in other words:

wherein the content of the first and second substances,is supported by P angular domains₂Identified angular domain vectorP elements of (a). Vector quantityMay be referred to as an intermediate value vector.

(Vector)Then sorted by their norm in descending order and then the L elements representing the largest channel gain are selected. This may be done by selecting a vectorIs performed with the L largest elements.

Mathematically:

wherein the content of the first and second substances,is equal to a vector of intermediate values ordered in decreasing order by the magnitude of the element, andis formed by angular domain vectorIs used to form an L element vector. In other words, a vectorFormed by the L elements of the angular domain vector representing the L maximum channel gains.

The value vector is in this example a vectorNote that the subscript "2" indicates that a vector of values is formed for channel vector h₂And (4) forming.

As noted above, the value of L may be chosen according to the implementation, and in some examples is equal to the number P of resolvable physical paths. If the value of L is P, then no selection vector need be performedL maximum elements. Thus, in some arrangements, the vector of values is a vectorOr it may be a vector on the basis ofIf all P elements identified by the corner domain support points are maintained, then the vector elements do not need to be sorted.

At step 306, the uplink device 204 feeds back an indication of the generated vector of values to the downlink device 202.

Processing unit 228 may transmit indications of the vectors of values to RF chains 224, which then generate analog signals representing indications of the vectors of values that are passed to antenna 1 of antenna array 222 (which is the only antenna capable of transmitting in the uplink direction). Antenna 1 then sends an indication representing the vector of values to apparatus 202 in the uplink direction to feed back an indication of the vector of values to apparatus 202.

The indication of the vector of values may be generated using a quantization codebook C known to both uplink device 204 and downlink device 202. The quantization codebook is formed from a set of quantization vectors. The uplink device 204 may quantize the generated vector of values using a quantization codebook C. The index of the quantization codebook indexed to the quantization vector selected as the quantization value vector may then be fed back to the downlink device 202. In other words, the indication of the vector of values fed back to the downlink apparatus comprises the selected index of the quantization codebook. The quantized vector of the codebook may be a unit norm vector. In this case, the processing unit 228 may also calculate a magnitude of the value vector and feed back the magnitude together with an index of the quantization codebook. Thus, in implementations when the codebook contains unit norm vectors, the indication of the vector of values comprises a selected index of codebook C and a magnitude of the vector of values.

Codebook C can be mathematically represented as:

therein, 2^BIs the number of quantized vectors in the codebook and B is the number of bits required to index the codebook, i.e., each codebook index is B bits.

The processing unit 228 may derive the closest match value vector fromCodebook C selects a quantization vector C_i. In particular, processing unit 228 may select index I of codebook C_nTo quantize the value vector according to:

the value vector is then fed back to the downlink device 202Index I of_nAnd an amplitude value.

Value vectorIs received by the antenna array 216 and the RF chain 214 at the downlink device 202 and passed to the processing unit 218.

At step 308, the downlink device 202 analyzes the channel vector h known to the device₁To identify the same set of angular domain support points identified by the uplink device at step 302. This may be performed by the processing unit 218.

The downlink device 202 analyzes the channel vector h in the angular domain₁To identify angular domain support points. Channel vector h represented in the angular domain₁Is shown asThat is to say that the first and second electrodes,is an angular domain vector. Angular domain vectorChannel vector h can be mapped by using Discrete Fourier Transform (DFT) matrix D₁DFT generation is performed. Channel vector h₁May be multiplied by the DFT matrix D to generate an angular domain vectorMathematically, means 202 generates an angular domain vector according to:

downlink device 20How 2 the uplink device 204 analyzes the angular domain vectorIn a similar manner as described above with respect to step 302, to analyze the angular domain vectorTo generate angular domain support points. In doing so, downlink apparatus 202 generates a vector for use inIs represented as phi given by₁：

Thus, for vectorsIs equivalent to the support point for the vector given by equation (23)The angular domain of (1) supports points. To see this, note that for vectorsDepends only on the emission angle phi_pIndependent of angle of arrival θ_p. It has also been observed from equations (11) and (12): h is₁＝Ag₁And h is₂＝Ag₂And it has also been observed from equation (8): the steering matrix a depends on the same emission angle as the angular domain support points. Thus, vector h₁And h₂Have common angular domain support, i.e., they share identical angular domain support points. This is consistent with the physical interpretation of MIMO system 200: typically, the distance between adjacent antennas of uplink device 204 is much smaller than the scattering objects (e.g., 208 and 210) and the uplink deviceThe distance between them. This means that the physical propagation paths for the two antennas are via the same scattering object, and therefore the channel vectors for the two antennas depend on the same emission angle.

At step 310, the downlink device 202 supports the set of points Φ according to the calculated angular domain₁A transmit steering matrix a is generated. As will be explained in more detail below, the downlink device 202 supports the point Φ according to the slave angular domain₁The calculated one or more emission angles generate a transmit steering matrix. In general, downlink device 202 generates a transmit steering matrix from L transmit angles calculated from each L elements of an angular domain vector representing L maximum channel gains. That is, the transmission steering matrix a is generated from a set of L angular domain support points representing L maximum channel gains.

To support a point phi from the angular domain₁Identifying a set of L angular domain support points representing a maximum channel gain, from which the downlink device 202 depends₁A vector of values is generated. The value vector is composed of angular domain vectors representing maximum channel gainThe element (c) is formed. Value vector by selecting angular domain vectorIs generated.

Mathematically, the value vector is generated by first vectoring the channel vector h₁Applying the selected rows of the DFT matrix D to match the value vectorIn a similar manner. The row of the DFT matrix to be applied is controlled by the corner-Domain control Point phi₁To be determined. Specifically, a row with a row number equal to the corner domain support point will be applied to the channel vector. This is mathematically expressed as:

or in other words:

wherein the content of the first and second substances,is supported by P angular domains₁Identified angular domain vectorP elements of (a). Vector quantityMay be referred to as an intermediate value vector.

(Vector)Then sorted by their norm in descending order and then the L elements representing the largest channel gain are selected. This may be done by selecting a vectorIs performed with the L largest elements.

Mathematically:

wherein the content of the first and second substances,is equal to a vector of intermediate values ordered in decreasing order by the magnitude of the element, andis formed by angular domain vectorIs used to form an L element vector. In other words, a vectorBy an angular domain vector representing the L maximum channel gainsL elements of (a). Vector quantityIs a vector of values calculated by the downlink apparatus 202. The subscript "1" indicates that a vector of values is formed for the channel vector h₁And (4) forming.

Downlink device also computes descriptionIs arranged in a ranking dimensionIndex vector I of₁. Thus, vector I₁Is to indicateThe rearranged P element vector of (1). Thus, vector I₁Identifies the vector representing the maximum channel gainOr indexing it.

Thus, the set of points Φ can be supported from the angular domain according to₁Identifying a set of L angular domain support points representing a maximum channel gain:

thus, in summary, the downlinkLink means 202 is based on a channel vector h known to the means₁Calculating P angular domain support point sets phi₁. Each of these angular domain support points identifies an angular domain vectorEach element of (1). A set of L angular domain support points representing a maximum channel gain is then identifiedWherein L is less than or equal to P. In the above example, the set of points is supported by P angular domains by computing the set of points Φ₁Identified angular domain vectorP elements of (a) form an intermediate value vectorTo identify. The intermediate value vectors are then sorted in descending order by their norm, then by the index vector I₁L elements representing the maximum channel gain are identified. The index vector is then used to support a set of points Φ in P angular domains₁The corresponding L angular domains in (1) support point indexing.

Consistent with equations (31) and (36), the emission angle φ for 1, … L_lFrom the set of identified L corner region support points, it can be calculated as follows:

consistent with equation (8), the steering matrix a for the L identified launch angles is calculated by the downlink apparatus 202 as:

the matrix A represents the set of angular domain support points Φ calculated by the downlink device 202₁Come to countThe calculated transmit steering matrix. It is an N × L matrix.

At step 312, the device 202 performs a vector of values according to the feedback from the device 204And the steering matrix a calculated at the means 202 calculates the channel vector h₂Is estimated (denoted as uplink estimation of)。

Specifically, downlink apparatus 202 may calculate the channel vector estimate according to

Wherein, c_FIs from the index I according to feedback from the slave 204_nThe quantized vector of the selected codebook C, andis a vector of values fed back from the device 204The amplitude of (c).

(Vector)Is a channel vector h that cannot be calculated by means of a downlink estimation technique₂Is calculated.

The above-described method of computing a channel vector may provide several advantages over conventional techniques of channel estimation. The vector quantized and fed back to the downlink device 202 is a reduced-dimension L vector, where L is less than or equal to the number P of resolvable physical paths. The number of physical paths is typically much smaller than the number N of antennas at the downlink device, and thus, the method is advantageously compared with the conventional method of limited feedback where N-dimensional vectors are quantized and fed back. By reducing the size of the vector to be quantized, a more accurate quantization can be performed for a codebook of a given dimension. Furthermore, it has been understood that there is correlation between channels in the angular domain, whereby the channels of different receivers of the downlink device have common angular domain support (i.e. they share a common set of angular domain points). This enables omission of feedback for angular domain support points, which reduces channel feedback overhead.

Figure 5 shows simulation results obtained by the inventors showing that the above approach enables high summing rate performance with low channel feedback overhead.

In the simulation, a vector of valuesQuantized using a 6-bit Random Vector Quantization (RVQ) codebook. Summing rate performance, which can be achieved with ideal Channel State Information (CSI), is shown at 502. At 504 it is shown that the best (i.e. non-quantized) value vector may be usedThe achieved summing rate performance. A vector of values that can be quantized with the present feedback scheme and using a 6-bit RVQ codebook is shown at 506The achieved sum rate performance and the conventional limited feedback using the RVQ codebook is shown at 508. It is observed that the feedback scheme described herein outperforms the conventional RVQ codebook scheme that quantizes N-dimensional vectors. Furthermore, it can be observed that the performance loss due to the vector of quantized values is small, which means that a channel feedback scheme with good performance can be achieved without increasing the channel feedback overhead compared to the conventional scheme.

If the channel vector h has already been calculated₂The downlink device 202 may then use the estimated channel vectorAnd the calculated channel vector estimate h₁An estimate of the channel matrix H is computed (e.g., obtained from the uplink channel estimates) to obtain full CSI. If the channel matrix H has been calculated, the downlink device 202 may calculate an updated precoding matrix F. To this end, the downlink apparatus may decompose the channel matrix using Singular Value Decomposition (SVD) into:

H＝U∑V^H (40)

where U is the left singular matrix, Σ is a diagonal matrix composed of singular values in descending order, and V is the right singular matrix.

The precoding matrix F can then be calculated from the first M columns of V as:

F＝V[I_M 0]^T (41)

wherein, I_MIs an M × M identity matrix.

The downlink device may then transmit a signal through the antenna array 216 using the updated or optimized precoding matrix F.

In the example described herein, the uplink device 204 contains two antennas, one of which cannot transmit in the uplink direction. It will be appreciated that this is for illustrative purposes only, and that the techniques described herein are applicable to downlink devices having any suitable number of antennas. In summary, the techniques described herein enable the calculation, at a downlink device, of an estimate of a channel vector h between antennas of the downlink device and an uplink device using other channel vectors available to the downlink device (e.g., obtained using uplink estimation techniques) and an indication of a value vector containing elements of the channel vector h in the angular domain fed back from the uplink device. In some embodiments, for example, an uplink device may contain 8 antennas, four of which are limited to receiving downlink transmissions (i.e., only four antennas are capable of transmitting in the uplink direction).

In the above example, a set of L angular domain support points is identified and used to compute the channel vector estimate. In some examples, L is equal to P. In these cases, the pair vector may not be neededSorting and selecting the L elements to generate a vector of values at the uplink device: vector quantityCan be treated as a vector of values that are then quantized and fed back to the downlink device. Further, at the downlink device, it may not be necessary to compute the vector byIdentifying a set of L corner region support points: instead, the set of support points Φ can be supported directly from P, consistent with equation (37)₁P emission angles for calculating the matrix a are calculated.

The applicant hereby discloses in isolation each individual feature described herein and any combination of two or more such features, to the following extent: such features or combinations are capable of being carried out based on the present specification in its entirety in view of the common general knowledge of a person skilled in the art, irrespective of whether such features or combinations of features solve any problems disclosed herein, and without limiting the scope of the claims. The applicant indicates that aspects of the present invention may consist of any such individual feature or combination of features. In view of the foregoing description it will be evident to a person skilled in the art that various modifications may be made within the scope of the invention.