FIELD OF THE INVENTION
[0001] The present invention generally relates to a technique first developed for "802.11ac" version of the standard "802.11" and referred to as MU-MIMO (Multi-User Multiple-Input Multiple-Output) where transmitting communication device and receiving communication devices of a wireless local area network (WLAN) are equipped with multiple antennas.
[0002] The invention relates more specifically to a method for obtaining a steering matrix to be applied for MU-MIMO transmission. Note that, in literature, said steering matrix may also be called pre-coding matrix.
[0003] Applications can be found notably in a WiFi Access Point (AP) modem transmitter for DL(DownLink)-MU-MIMO transmission of data.
BACKGROUND OF THE INVENTION
[0004] Figure 1 highlights the main components of a SU-MIMO transmission model of digital data. Said digital data may be encoded by using Orthogonal Frequency-Division Multiplexing (OFDM) method.
[0005] The illustrated SU-MIMO transmission model is more particularly adapted to the transmission of digital data, encoded with OFDM method, from an AP, as a transmitter 200, to a user station, as a receiver 230.
[0006] Let define the following notations used notably onto Figure 1:
- NTX is the number of transmit antennas 201, 202 of the transmitter 200;
- NSS is the number of spatial streams; and
- NRX is the number of receive antennas 231, 232 of the receiver 230.
[0007] The number of spatial streams is defined as the number of data streams which are simultaneously broadcasted via the transmit antennas 201, 202. Obviously, it is mandatory that
NSS ≤
NTX.
[0008] Let call
Si(
m,k) a modulated symbol, for instance a QAM (Quadrature Amplitude Modulation) symbol, sent on a
kth subcarrier of a
mth OFDM symbol on the i
th spatial stream.
[0009] Let further define:
- S(m,k) = [S1(m,k) ...SNSS(m,k)]T, the NSS dimension multi-stream modulated symbol to be transmitted and
- T(m,k) = [T1(m,k) ... TNTX(m,k)]T, the NTXdimension multi-antenna transmit signal,
where exponent T characterizes the transpose operator.
[0010] Let have
T(
m,k) =
Q(
k)
S(
m,k), where
Q(
k) is a matrix used to convert the
NSS symbols to be transmitted into
NTX complex signals.
[0011] This conversion is done linearly by multiplying vectors
S(
m,k) of length
NSS by a
Q(k) matrix of dimensions
NTX ×
NSS. Each vector of length
NSS is intended to be transmitted during one OFDM symbol (In Wi-Fi, the symbol duration is often equal to 4 µs, but not always) and over a given subcarrier
k. The matrix
Q(k), which can be called spatial mapping matrix, may vary from one subcarrier to the other, but not from one OFDM symbol to the other, during a whole packet transmission.
[0012] The spatial mapping matrix
Q(k) is thus used for SU-MIMO beamforming. This technique is implemented by the transmitter 200 with multiple antennas 201, 202 to steer signals using knowledge of the physical channel between the AP and the user station in order to improve throughput or the covering range. In such a case, the transmitter 200 may be called a beamformer and the matrix
Q(
k) is merely called a steering matrix. It can be determined from a beamforming feedback matrix
V(
k) that is sent back to the beamformer 200 by the receiver 230 which may then be called beamformee, according to standard beamforming calibration processes, also called sounding processes (an example of which is detailed hereafter).
[0013] Before being emitted the
NTX dimension multi-antenna transmit signal
T(
m,k), and more particularly each component
Ti(
m,
k) of said signal, with
i = {1
, ...,
NTX}, is transformed in a conventional way according to Inverse Fast Fourier Transform (or IFFT) for converting each component of said signal from its original frequency domain to a representation in the time domain.
[0014] Now, let define
R(
m,k) = [
R1(
m,k) ...
RNRX(
m,k)]
, the signal received on the
NRX antennas 231, 232 of the receiver 230.
[0015] More particularly, in a conventional way, said received signal
R(
m,k)
, and more particularly each component
Ri(
m,k) of said received signal, with
i = {1,...,
NRX}
, is obtained by transforming the radio-frequency signal received by each
ith receive antenna according to Fast Fourier Transform (or FFT) for converting each component of said signal from the time domain to its original frequency domain.
[0016] Assuming an ISI (InterSymbol Interference) free transmission, and assuming that all offsets have been perfectly compensated, the frequency domain model is given by the following equation:

where
WGN ∼
N(0
,σ2) is a White Gaussian Noise, further with the assumption that all the received paths are affected by the same noise level, and where
Hφ(
k) is a
NRX ×
NTX matrix characterizing the physical channel between the antennas 201, 202 of the transmitter 200 and the ones 231, 232 of the receiver 230.
H(
k) is a
NRX ×
NSS matrix characterizing the effective channel between the transmitted symbol
S(
m,k) and the received signal
R(
m,k)
, with:
H(
k) =
Hφ(
k)
Q(
k)
.
[0017] In the SU-MIMO beamforming presented here above, all space-time streams of the signal to be transmitted are intended for reception at a single station 230. With DL-MU-MIMO beamforming, disjoint subsets of the space-time streams are intended for reception at different user stations.
[0018] From document
US 9,319,122 B1, it is known a wireless network device, such as an AP, of a WLAN which transmits simultaneously independent data streams to at least two user stations according to a multi-user (MU) mode via an antenna array. In order to reduce, or even in order to cancel out, interference at a receiving station due to simultaneous transmissions from several antennas of the AP to one or more user stations, the AP develops respective transmit beamsteering vectors for downlink transmissions towards the user stations.
[0019] Figure 2 highlights the main components of a DL-MU-MIMO-OFDM transmission model. The applicant's admitted prior art as discussed below in connection with Figure 2 may be considered as an adaptation of the disclosure of document
US 9,319,122 B1 to the transmission of digital data encoded with OFDM method. Hereafter, for sake of simplicity, subcarrier index
k has been omitted in comparison with the here above equations.
[0020] The transmitter (or beamformer) 200 should calculate a DL-MU-MIMO steering or pre-coding matrix
Q in order to cancel out crosstalk between participating user stations 230, 240. Said DL-MU-MIMO steering matrix
Q can be determined from the beamforming feedback matrices
V(d) of the participating user stations 230, 240.
[0021] Let add some additional notations:
- K, the number of user stations 230, 240 participating to the considered MU-MIMO transmitted frame, with K ≤ 4 according to the standard specifications;
- NSStot, the total number of spatial streams comprised in the MU-MIMO frame to be transmitted, with NSStot being constrained by NSStot ≤ NTX;
- NSS(d), the number of spatial streams intended for the dth user station, constrained by NSS(d) < NSStot since ∑dNSS(d) = NSStot; and
- NRX(d), the number of receive antennas (231, 232 or 241, 242) of the dth user station (230 or 240, respectively), with NRX(d) ≥ NSS(d) since each stream is intended to be received by one receive antenna.
[0022] Let further call

the
NSS(d)-dimension QAM-symbol to be transmitted to the
dth user.
[0023] Then, let

the
NRX(d)-dimension signal received by the
dth user station, verifies:

[0024] By splitting Q between the user stations 230, 240 as
Q = [
Q(1) ...
Q(K)]
, this equation can be decomposed into two parts :

[0025] The first part of the equation comprises the signal of interest, whereas the second part characterizes the interference of the signal(s) transmitted to the other user station(s) on the signal received by the
dth user station. As mentioned earlier, the transmitter 200 shall define the matrix
Q in order to minimize, or even cancel out, the interference part for each of the
K receivers 230, 240 or in order to optimize the SINR (signal-to-interference-plus-noise ratio) received by each user station.
[0026] More generally, the problem of all existing algorithms, like the one disclosed into document
US 9,319,122 B1, is that they assume that the number of receive antennas (231, 232 or 241, 242) of each user station (230 or 240, respectively) is equal to the number of spatial streams intended for this user station, e.g.
NRX(d) =
NSS(d). Consequently, they assume that ∑
dNRX(d) ≤
NTX.
[0028] In this context, i.e. in communication systems of the type briefly described above and notably of the type referred to as (DL-)MU-MIMO, it is an object of the invention to provide a method for obtaining a steering matrix Q to be applied for (DL-)MU-MIMO transmission that supports extended configurations of a (DL-)MU-MIMO communication system.
[0029] Another object of the invention is to provide a method for obtaining a steering matrix
Q to be applied for (DL-)MU-MIMO transmission that improves significantly the quality of a (DL-)MU-MIMO transmission when the number of antennas of the receivers is larger than the number of streams intended for these receivers, that is when ∑
dNRX(d) ≥
NTX.
[0030] A further object of the invention is to provide a method for obtaining a steering matrix to be applied for (DL-)MU-MIMO transmission that offers much better performance than existing algorithms when ∑
dNRX(
d) >
NTX.
SUMMARY OF THE INVENTION
[0031] To at least one of these ends, it is provided, according to a first aspect of the invention, a method for obtaining a steering (or pre-coding) matrix
Q = [
Q(1),.., Q(K)] to be applied for MU-MIMO transmission of data. The method is implemented by a MU-MIMO communication system which comprises at least one transmitter comprising at least
NTX = 2 transmit antennas and
K = 2 receivers, with each
dth receiver comprising at least
NRX(d) receive antennas, wherein ∑
dNRX(d) ≥
NTX. The method comprising, at the level of the transmitter, the following sequence of steps for each value of
d = {1,2} :
- Computing a channel inversion matrix C(d) of size NTX × NRX(d) wherein:

wherein F(d) is a matrix of size NTX × NRX(d) whose components depend on those of a matrix

of size NTX × NRX(d) which characterizes physical channel between the transmitter and the dth receiver, and more particularly wherein

with U(d) being a unitary matrix, i.e. U(d)HU(d) = INTX, and wherein exponent H refers to the Hermitian operator; - Computing a beamforming component B(d) by singular value decomposition of matrix product F(d)C(d); and
- Computing the steering matrix Q according to Q(d) = C(d)B(d).
[0032] This algorithm does not tend to determine a steering matrix which allows of cancelling out the inter-user interference at the input of each receive antenna.
[0033] This algorithm proposes a way to compute a steering matrix Q to be applied for (DL-)MU-MIMO transmission that supports extended configurations of a (DL-)MU-MIMO communication system, at least since it allows to use more receive antennas than transmit antennas. Indeed, this algorithm allows to avoid the need to use an antenna reduction protocol or an interference cancellation protocol.
[0034] This algorithm further proposes a way to compute a steering matrix Q which allows of improving significantly the quality of subsequent DL-MU-MIMO transmissions when the number of antennas of the receivers is larger than the number of antennas of the transmitter, e.g. when ∑
dNRX(d) ≥
NTX. This improvement is particularly more relevant when at least one among the number of antenna(s)
NRX(1) of the first receiver and the number of antenna(s)
NRX(2) of the second receiver is strictly inferior to the number of spatial streams intended for said receivers, i.e. ∑
dNSS(d). Since more hardware resources are put to contribution owing to the present algorithm, it comes that said algorithm offers much better performance than existing algorithms when ∑
dNRX(
d) >
NTX and in particular when at least one among the number of antenna(s)
NRX(1) of the first receiver and the number of antenna(s)
NRX(2) of the second receiver is further strictly inferior to the number of spatial streams intended for said receivers, i.e. ∑
dNSS(d).
[0035] Optionally, the invention may have any of the following facultative features that can be used separately or in combination.
[0036] According to an optional embodiment of the first aspect of the invention, the beamforming component
B(d) is more particularly computed as
NSS(d) singular vectors corresponding to the
NSS(d) strongest singular values of matrix product
F(d)C(d), with the steering matrix component
Q(d) being of size
NTX ×
NSS(d), where
NSS(d) is a number of spatial streams intended for the
dth receiver.
[0037] According to another optional embodiment of the first aspect of the invention, the MU-MIMO communication system is a WiFi communication system, the at least one transmitter is an access point (AP) of said WiFi communication system, the
K = 2 receivers are user stations of said WiFi communication system; the method according to the first aspect of the invention being more particularly a method for obtaining a steering matrix
Q to be applied for DL-MU-MIMO transmission of data.
[0038] Thus the method can be implemented in a WiFi AP modem transmitter for DL-MU-MIMO transmission.
[0039] According to another optional embodiment of the first aspect of the invention, the method further comprises, before said sequence of steps, a sounding step comprising at the level of the transmitter the following substeps:
- Emitting a Null Data Packet (NDP) frame comprising NTX streams from transmit antennas, then
- From a first receiver among the K = 2 receivers, receiving, as a response to the NDP frame, a first set of data on which depends the matrix

of size NTX × NRX(1) which characterizes physical channel between the transmitter and the first receiver, - From a second receiver among the K = 2 receivers, receiving, as a response to the NDP frame, a second set of data on which depends the matrix

of size NTX × NRX(2) which characterizes physical channel between the transmitter and the second receiver.
[0040] Thus the method is compatible with standard sounding processes as already performed to acquire knowledge about the physical channels between the transmitter and the receivers. More particularly, no change in the way to operate the sounding process is required at least at the level of the receivers with respect to standard sounding processes.
[0041] According to another optional embodiment of the first aspect of the invention, each of said first set of data and said second set of data comprises matrices
V(d) of size
NTX ×
NRX(d) and ∑
(d) of size
NRX(d) ×
NRX(d), with
d = 1 and
d = 2 respectively, said matrices
V(d) and ∑
(d) being determined as a singular value decomposition of

wherein
V(d) is an orthonormal matrix comprising the
NRX(d) left singular vectors of

and ∑
(d) is a diagonal matrix comprising the
NRX(d) strongest singular values of

At this stage, each of matrices

and

might be evaluated in function of the corresponding matrices
V(d) and ∑
(d) by applying

providing
U(d) is known for
d = 1 and
d = 2.
[0042] Thus, on the contrary of standard pre-coding processes according to which the transmitter uses only the values of the first column of each matrix
V(d) (the ones corresponding to the strongest singular vector), the transmitter implemented according to the method for MU-MIMO transmission of data as introduced below in connection with the second aspect of the invention uses the full
NTX x NRX(d) values of each matrix
V(d) as a measure from which the physical channel between the transmitter and the
dth receiver can be determined.
[0043] The method according to the first aspect of the invention is a computer implemented method. At least some steps of the method, in particular the steps of computing a channel inversion matrix
C(d), computing a beamforming component
B(d) and computing the steering matrix component
Q(d) are executed by a data processing device such as a processor or a microprocessor.
[0044] According to a second aspect of the invention, it is provided a method for MU-MIMO transmission of data. As the method according to the first aspect of the invention, the method according to its second aspect is implemented by a MU-MIMO communication system which comprises at least a transmitter comprising at least
NTX = 2 transmit antennas and
K = 2 receivers, each
dth receiver comprising
NRX(d) receive antennas, wherein ∑
dNRX(d) >
NTX. The method according to the second aspect of the invention comprises, with a steering (or pre-coding) matrix
Q = [
Q(1),..,
Q(K)] being obtained according to the method according to the first aspect of the invention, the following steps:
- At the level of the transmitter, pre-coding a modulated data symbol S(d)(m) of an mth symbol to be transmitted to the dth receiver among said K = 2 receivers according to Q(d)S(d)(m), then emitting simultaneously at least two signals, with each of said at least two signals being emitted from said at least NTX = 2 transmit antennas, in order for the method according to the second aspect of the invention to further comprises the following steps :
- At the level of the d = 1 receiver among the K = 2 receivers, receiving said at least two signals by the NRX(1) receive antennas, with:

where each R(1)(m) is an NRX(1)-dimension signal received by the d = 1 receiver, and
where

is a matrix of size NTX × NRX(1) characterizing physical channels and WGN(1)(m) are terms which characterize noise levels, between the transmitter and the d = 1 receiver; - At the level of the d = 2 receiver among the K = 2 receivers, receiving said at least two signals by the NRX(2) receive antennas, with:

where each R(2)(m) is an NRX(2)-dimension signal received by the d = 2 receiver, and
where

is a matrix of size NTX × NRX(2) characterizing physical channels and WGN(2)(m) are terms which characterize noise levels, between the transmitter and the d = 2 receiver.
[0045] According to an optional embodiment of the second aspect of the invention, the method further comprises, with each
dth receiver comprising an equalizer, for instance a maximal-ratio combining (MRC) equalizer, a step comprising, with
d = {1,2}:
- Equalizing the NRX(d)-dimension signal R(d)(m) by using coefficients W(d) according to Z(d)(m) = W(d)R(d)(m), said coefficients W(d) being pre-computed at the level of each dth receiver from an estimation of the effective channel matrix

and - Outputting a signal Z(d)(m) as an estimation of the modulated data symbol S(d)(m).
[0046] Thus it appears that the method allows of cancelling out the inter-user interference at the output of each receiver equalizer or of optimizing the SINR (Signal-to-Interference-plus Noise Ratio) received by each receiver in order to retrieve each symbol of transmitted encoded data.
[0047] According to another optional embodiment of the second aspect of the invention, said symbol to be transmitted are encoded by using one among OFDM encoding and COFDM (for Coded Orthogonal Frequency-Division Multiplexing) encoding.
[0048] Thus the method works naturally with different encoding methods and notably those which are of most use or of most recent development.
[0049] According to the above recited optional embodiment of the second aspect of the invention, said symbol can further be data modulated into data symbol by using one among PSK modulation and QAM modulation.
[0050] Thus the method works naturally with different modulation methods and notably those which are of most use.
[0051] The method according to the second aspect of the invention is a computer implemented method. At least some steps of the method, in particular the steps of pre-coding said modulated data symbol
S(d)(
m), equalizing said
NRX(d)-dimension signal
R(d)(
m), and outputting
Z(d)(
m) are executed by data processing devices such as a processor or a microprocessor.
[0052] According to a third aspect of the present invention, it is provided a computer program product comprising instructions which, when implemented by at least one digital processing device, performs at least the steps of at least one method among the method for obtaining a steering matrix
Q to be applied for MU-MIMO transmission of data according to the first aspect of the invention and the method for MU-MIMO transmission of data according to the second aspect of the invention.
[0053] According to a fourth aspect of the present invention, it is provided a non-transitory computer readable medium storing instructions which, when implemented by at least one digital processing device, performs at least the steps of at least one method among the method for obtaining a steering matrix
Q to be applied for MU-MIMO transmission of data according to the first aspect of the invention and the method for MU-MIMO transmission of data according to the second aspect of the invention.
[0054] According to a fifth aspect of the present invention, it is provided a transmitter which comprises at least
NTX = 2 transmit antennas and which is intended to be used into a MU-MIMO communication system comprising further
K = 2 receivers, with each
dth receiver comprising
NRX(d) receive antennas, wherein ∑
dNRX(d) ≥
NTX. Said transmitter is configured to implement the method for obtaining a steering matrix
Q to be applied for MU-MIMO transmission of data according to the first aspect of the present invention.
[0055] According to a sixth aspect of the present invention, it is provided a MU-MIMO communication system which comprises at least a transmitter which comprises at least
NTX = 2 transmit antennas and
K = 2 receivers, with each
dth receiver comprising
NRX(d) receive antennas, wherein ∑
dNRX(d) > NTX. Said MU-MIMO communication system is configured to implement the method for MU-MIMO transmission of data according to the second aspect of the present invention.
[0056] According to an optional embodiment of the sixth aspect of the invention, at least one among the number of antenna(s)
NRX(1) of the first receiver and the number of antenna(s)
NRX(2) of the second receiver is strictly inferior to the number of spatial streams intended for said receivers, i.e. ∑
dNSS(d).
[0057] Further objects, features and advantages of the present invention will become apparent to the ones skilled in the art upon examination of the following description in reference to the accompanying drawings. It is intended that any additional advantages may be incorporated herein.
BRIEF DESCRIPTION OF THE DRAWINGS
[0058]
Figure 1 illustrates schematically a SU-MIMO wireless communication system;
Figure 2 illustrates schematically a MU-MIMO wireless communication system by which the invention can be implemented;
Figure 3 illustrates schematically, in a different manner than that of figure 2, a MU-MIMO wireless communication system by which the invention can be implemented.
Figure 4 is a flowchart of the method according to an embodiment of the first aspect of the invention.
Figure 5 is a flowchart of a sounding process according to an embodiment of the first aspect of the invention.
Figure 6 is a flowchart of the method according to an embodiment of the second aspect of the invention.
Figure 7 shows some graphs presenting the performance achievable by the present invention compared to some reference solutions.
Figure 8 shows a timing diagram illustrating a detailed embodiment of the sounding process illustrated on figure 6.
[0059] The figures are given as examples and are not restrictive to the invention. They are principle schematic representations intended to facilitate the understanding of the invention.
DETAILED DESCRIPTION
[0060] Herein sounding or calibration is a term used to denote the process performed by the transmitter to acquire channel state information (CSI) from each of the different receivers by sending training symbols (or streams) and waiting for the receivers to provide explicit feedback comprising a measure from which the physical channels between the transmitter and the receivers can be determined.
[0061] The term "stream" or "data stream" denotes a flow of data circulating in a communication system from point of origin to destination.
[0062] In a transmission system comprising several transmit antennas, the bitstream to be transmitted can be split (spatial multiplexing) in several spatial streams. All spatial streams are transmitted within the same spectral channel, the spatial dimension being used to avoid/limit interference between the spatial streams.
[0063] The term "frame" denotes a data transmission unit which may consist of a set of successive OFDM symbols.
[0064] The term "Null Data Packet" denotes a frame used to sound the physical channel between the transmitter and the receiver. In WiFi, this frame is only composed of the preamble containing the training symbols used to estimate the CSI and it does not contain any symbol carrying data information.
[0065] The following detailed description of the invention refers to the accompanying drawings. While the description includes exemplary embodiments, other embodiments are possible, and changes may be made to the embodiments described without departing from the scope of the invention. The various steps illustrated in figures 4 and 6 which are enclosed by dashed lines are optional and are not restrictive to the illustrated method(s).
[0066] Figure 2 gives an example of the context to which the invention particularly applies, i.e. a multi-user MIMO communication system
20 where multiple receivers
230, 240 are intended to be addressed simultaneously taking advantage of the spatial dimension of the communication channels between the transmitter
200 and the receivers
230, 240 through the use of multiple transmit antennas
201, 202.
[0067] As illustrated, to achieve this, the transmitter or transmitting device or, according to a preferred embodiment, an AP
200, i.e. an access point (or base station) such as defined for instance in IEEE 802.11 specifications, is made able through its set of transmit antennas
201, 202 to create simultaneously constructive interferences, each one being intended for a given receiver or, according to the preferred embodiment, for a given user station
230 or
240. In 802.11 terminology, a user station is any device that has the capability to use the corresponding communication protocol, and according to the preferred embodiment the capability to use a Wi-Fi communication protocol. For example, a user station
230 or
240 is a laptop or a desktop personal computer (PC). For example, the AP
200 is a modem, such as a wireless mobile broadband modem operating at microwave frequencies.
[0068] Figure 2 more particularly illustrates a part of the computational background and mathematical formalism under which pre-coding of the transmitted signals is actually performed by an AP
200 in order to obtain constructive (and destructive) interferences adapted to the location of multiple user stations
230 and
240.
[0069] In the case of a single-user or SU, all space-time streams in the transmitted signal are intended for reception by a single user station. With MU-MIMO beamforming and, according to the preferred embodiment, with DL-MU-MIMO beamforming, disjoint subsets of the space-time streams are intended for multiple user stations
230 and
240. Then, the transmitter
200 must include a beamformer component
300 capable of calculating a steering matrix Q according to which pre-coding of the symbol to be transmitted can be performed to define a composite RF signal to be broadcasted by the transmit antennas
201, 202. The pre-coding can thus be performed so that crosstalk, i.e., inter-user interference at user stations
230 and
240 can be controlled.
[0070] In order to describe the pre-coding of the symbol to be transmitted in details, let define the following parameters with references to Figure 2:
- NTX is the number of transmit antennas 201, 202 of the AP 200;
- K is the number of user stations participating to the transmission of a considered frame;
- NRX(d) is the number of receive antennas of the dth user station, with d = {1,..,K};
- NSS(d) is the number of spatial streams allocated to the dth user station for transmitting the considered frame; and
- NSStot is the total number of spatial streams, NSStot = ∑dNSS(d).
[0071] Let also define, always with references to Figure 2:

the NSS(d)dimension modulated data symbol of an mth symbol to be transmitted to the dth user station, and

the NRX(d)-dimension signal received by the dth user station that verifies:

wherein

is a matrix of size
NTX ×
NRX(d) characterizing physical channel and wherein
WGN(d)(
m) are terms which characterize noise levels, between the AP
200 and the
dth user station
230 or
240.
[0072] According to well-known background of the invention, said symbol to be transmitted may indeed be encoded, for instance by using one among OFDM encoding and COFDM (for Coded Orthogonal Frequency-Division Multiplexing) encoding. OFDM or COFDM symbol may further be modulated by using for instance one among PSK modulation and QAM modulation.
[0073] By splitting the steering matrix
Q between the user stations as
Q = [
Q(1)...
Q(K)], the above equation can be decomposed into two parts:

[0074] The first part of the equation contains the signal of interest for the
dth user station, whereas the second part characterizes the interference of the other user stations on the signal received by the
dth user station. As mentioned earlier, the AP
200 shall define the steering matrix
Q in order to control the interference part for each of the
K user stations
230, 240.
[0075] Figure 3 illustrates a case of a MU-MIMO communication system
20 in which the AP
200 has only two transmit antennas:
201 and
202, communicating with
K = 2 user stations:
230 and
240, with each user station having two receive antennas:
231, 232 and
241, 242 respectively. While the AP
200 broadcasts the composite RF signal towards the two user stations, the objective is that modulated data symbol
S(1) be transmitted mainly, if not exclusively, to the first user station
230 and modulated data symbol
S(2) to the second one
240. To this end, as mentioned above, the broadcasted RF signal is pre-coded in the AP
200 with the help of a steering matrix
Q adapted, according to the present invention, to take care of the characteristics of the communication channels or CSI. These characteristics, which are assumed to be known, even indirectly, beforehand by the AP
200, may be described under the form of the two illustrated
Hφ(1) and
Hφ(2) matrices
321 and
322. Each matrix
Hφ(d) among these matrices characterizing the physical channel between the AP
200 and one
dth receiver among the
K = 2 user stations
230 and
240.
[0076] Hence, a set of two equations gives the signals
R(1)(
m) and
R(2)(
m) respectively received by the user stations
230 and
240: 
and

[0077] Thus, in order to calculate the steering matrix
Q, a beam-former component
300 needs to have, even indirectly, an accurate estimate of the physical channels over which the AP
200 is intended to transmit data to each
dth user station, with
d = {1,2}. The physical channels may be expressed directly in the form of matrices
Hφ(d), or indirectly, for instance through the knowledge of matrices
V(d) and
∑(d), as detailed below.
[0078] An explicit mechanism called sounding or calibration is put in place to provide the required information from the user stations
230, 240, also called beamformee, to the AP
200, also called beamformer, prior to the broadcasting of data from the AP
200. For WiFi communication system, this mechanism is described in 802.11.ac standard. This mechanism is described hereafter with references to
Figures 4, 5 and
8.
[0079] During sounding, each user station
230, 240 performs the following actions:
- measuring the physical channel Hφ(d) between itself and the AP 200;
- computing a standard singular value decomposition or SVD of Hφ(d) so that:

- sending feedback matrices V(d) and ∑(d) to the beamformer into a feedback frame in the form of a data set.
[0080] From the point of view of the AP
200, a sounding step
9 is performed which can comprise the following substeps:
- Emitting 91 a Null Data Packet (NDP) frame comprising NTX streams from transmit antennas 201, 202, then
- From a first user station 230, receiving 92, as a response to the NDP frame, a first set of data on which depends the matrix

of size NTX × NRX(1) which characterizes physical channel between the AP 200 and the first user station 230, - From a second user station 240, receiving 93, as a response to the NDP frame, a second set of data on which depends the matrix

of size NTX × NRX(2) which characterizes physical channel between the AP 200 and the second receiver 240,
wherein each of said first set of data and said second set of data comprises the above mentioned feedback matrices
V(d) and ∑
(d), with
d = 1 and
d = 2 respectively.
[0081] Generally, each stream among said
NTX streams of the Null Data Packet (NDP) frame is more particularly emitted by a single transmit antenna of the transmitter
200, but this is not compulsory.
[0082] In a WiFi communication system, said
NTX streams of the Null Data Packet (NDP) frame are usually emitted simultaneously, but on several successive symbols in order to obtain a view of all dimensions of each physical channel. Nonetheless, said
NTX streams could also be emitted successively, transmit antenna by transmit antenna.
[0083] Note that it does not matter if the first and second sets of data sent to the transmitter
200 are emitted by one and/or the other of the receiver antennas
231, 232 and
241, 242.
[0084] At this stage, each of matrices

and

might be evaluated in function of the corresponding matrices
V(d) and ∑
(d) by applying
U(d)∑
(d)V(d)H, providing
U(d) is known for
d = 1 and
d = 2. Actually, the AP 200 will use the feedback matrices
V(d) and ∑
(d), rather than matrices

in order to compute 11 a channel inversion matrix
C(d), as described above.
[0085] With references to
Figure 8, the sounding process
9 is described with more details hereafter. Firstly, the beamformer
200 initiates the calibration sequence by transmitting a VHT (Very High Throughput) NDP Announcement frame followed by a VHT NDP after a SIFS (Short InterFrame Space). The number of streams contained in the NDP,
NSTSNDP, shall be equal to the number of transmit antennas of the AP
200, NSTSNDP =
NTX. Then, the signal received by the
dth user station during the VHT-LTF (Long Training Field) training period of the NDP verifies:

where
SNDP(
n) is a
NTX dimension vector and where
NLTF is the number of LTF. In a WiFi communication system,
NLTF ≥
NSTSNDP (for instance: if
NSTSNDP = 2, then
NLTF = 2; if
NSTSNDP = 3
or 4,
NLTF = 4). From the observation of the NLTF VHT-LTF fields, each
dth beamformee
230, 240 can estimate the corresponding matrix

Each matrix

of size
NRX(d) x
NTX measured by one beamformee
230 or
240 upon the receipt of this NDP is a characterization or a mathematical representation of the physical channel between the beamformer
200 and the considered beamformee
230 or
240.
[0086] Then, each beamformee
230 or
240 determines the beamforming feedback matrices
V(d) and ∑
(d) by SVD of

wherein exponent
H refers to the Hermitian operator.
[0087] A difference between the sounding step
9 according to the present invention and the standard algorithms consists in that said matrices
V(d) and ∑
(d) are hereby determined as a singular value decomposition of
U(d)∑
(d)V(d)H, wherein
V(d) is an orthonormal matrix of size
NTX ×
NRX(d) comprising the
NRX(d) left singular vectors of

and wherein ∑
(d) is a diagonal matrix of size
NRX(d) ×
NRX(d) comprising the
NRX(d) strongest singular values of

each of matrices
V(d) and ∑
(d) are transmitted as such to the AP
200 in order to be subsequently used without approximation. On the contrary, according to the standard algorithms, the AP
200 uses only the first
NSS(d) column of each
V(d), i.e. the one corresponding to the strongest singular vector, and not said orthonormal matrix
V(d) which comprises the
NRX(d) left singular vectors of

[0088] Once the characteristics of the physical channels may be expressed in the form of matrices, such as

and

even indirectly through the knowledge of corresponding (
V(1),
∑(1)) and (
V(2),∑(2)) sets of matrices respectively, there exist several standard algorithms to compute the DL-MU-MIMO steering matrix
Q. The simpler is called "Channel Inversion". Another well-known algorithm is called "Block Diagonalization". "Channel Inversion" algorithm simply consists in applying the inverse of each physical channel for pre-coding the broadcasted data symbol. The steering matrix
Q computed according to said standard algorithms generally aims at cancelling out the inter-user interference; each user station would receive only the signal intended to it and no part of the signal intended to the other station(s).
[0089] Nonetheless, such standard solutions work only if the number of transmit antennas
NTX of the AP
200 is larger or equal to the sum of the receive antennas
NRX(d) of all the user stations
230 and
240, that is if
NTX ≥
∑dNRX(d). Indeed, the number of antennas of the AP
200 must be by theory larger than the sum of the number of spatial streams,
NTX ≥
NSStot; a problem of all the standard algorithms is that they assume that the number of receive antennas
NRX(d) of each
dth user station is equal to the number of spatial streams intended to this
dth user station (
NRX(d) =
NSS(d)) and consequently they assume that
NTX ≥ ∑
dNRX(d).
[0090] Now, let consider a scenario as the one illustrated on
Figure 3 where the AP
200 has two transmit antennas
201 and
202 and each of the two considered user stations
230 and
240 has two receive antennas
(231, 232 and
241, 242, respectively). The AP
200 may want to transmit a MU-MIMO frame to these two user stations. Nonetheless, because it has only two transmit antennas
201 and
202, the transmitter
200 can transmit only two streams, thus one single stream to each user station. This means that each user station will receive one stream while each user station has two receive antennas. Translated into mathematical formalism, this means that
NTX ≥
NSStot (as previously), but
NTX < ∑
dNRX(d).
[0091] In this case, the above mentioned standard algorithms can use two different strategies:
- 1. Antenna Reduction: each user station 230, 240 uses only one of its two receive antennas and feedbacks to the AP 200 data characterizing the physical channel between the AP 200 and the selected receive antenna.
- 2. Interference Cancellation: each receiver 230, 240 measures the full 2x2 matrix characterizing the physical channel between the AP 200 and its two receive antenna and computes a V(d) matrix of size 2 × 2. Nonetheless, the AP 200 uses only the first column of each V(d), that is the one corresponding to the strongest singular vector. In this case, the signal received on the receive antennas of the user stations includes a mix of both signals; it is not possible for an AP 200 having only two transmit antennas to cancel out the interference on two sets of two receive antennas. Nevertheless, if each user station includes an equalizer with interference measurement and cancellation, the signal can be properly demodulates.
[0092] The here disclosed algorithm does not attempt to have the inter-user interference terms that cancels out at inputs of the user stations, i.e., on their receive antennas (this is impossible in the considered configuration), but it rather consists in cancelling out them at the output of the equalizer of each receiver
230, 240.
[0093] Figure 4 shows the steps of the algorithm performed by the AP
200 to obtain
10 the steering matrix
Q according to the invention. It is described hereafter on the basis of the communication system
20 illustrated on Figure 3, i.e. with an AP
200 equipped with two transmit antennas
201, 202 while the
K = 2 user stations
230, 240 have each two receive antennas, i.e. in a communication configuration not efficiently supported by the above mentioned standard algorithms. However, it is important to note that the methods
10 and
100 according to the first and second aspects of the present invention are also implementable in communication systems, where the AP
200 has more than two transmit antennas. Moreover, the methods
10 and
100 according to the first and second aspects of the present invention is further implementable in communication systems where at least one among
NRX(1) and
NRX(2) is strictly inferior to the number of spatial streams intended for said receivers, i.e. ∑
dNSS(d).
[0094] According to the first aspect of the present invention, the first step of the method
10 for obtaining a steering matrix
Q = [
Q(1),..,
Q(K)] to be applied for MU-MIMO transmission of data consists in computing
11 a matrix
C(d), which can be called channel inversion matrix, of size
NTX ×
NRX(d) wherein:
C(d) = (
F(1)HF(1) +
F(2)HF(2))
-1F(d)H.
[0095] F(d) is a matrix of size
NTX ×
NRX(d) whose components depend on those of a matrix

of size
NTX ×
NRX(d) which as already established above characterizes physical channel between the AP
200 and the
dth user station
230 or
240. More particularly,

with
U(d) being a unitary matrix, i.e.
U(d)HU(d) =
INTX, wherein exponent
H refers to the Hermitian operator. This computing step
11 of the matrix product
F(d) thus depends on the two feedback matrices,
V(d) and ∑
(d), previously received by AP
200 during the sounding step
9 detailed above. As already discussed, these feedback matrices come from the SVD operation performed at each
dth user station on each

matrix measuring the physical channel between the AP
200 and
dth user station
230 or
240. Actually, according to the here above detailed sounding process
9, F(d) = ∑
(d)V(d)H, but the present invention is not limited to said sounding process and is implementable with various sounding processes, providing these latter allow evaluating the physical channels between the AP
200 and each
dth user station
230 or
240.
[0096] Then, in a second step of the method
10, the AP
200 computes
12 a component
B(d), which can be called beamforming component, by SVD of matrix product
F(d)C(d). More particularly, the beamformer component
300 of the AP
200 computes, for each user station
230 and
240, the matrix product
F(d)C(d).
[0097] Each beamforming component
B(d) is more particularly computed
12 as
NSS(d) singular vectors corresponding to the
NSS(d) strongest singular values of matrix product
F(d)C(d), with the steering matrix component
Q(d) being intended to be of size
NTX ×
NSS(d), where
NSS(d) is a number of spatial streams intended for the
dth user station.
[0098] Then, in a third step, the AP
200 computes
13 the steering matrix component
Q(d) according to
Q(d) =
C(d)B(d). More particularly, the beamformer component
300 of the AP
200 computes, for each user station
230 and
240, the matrix product
C(d)B(d) in order to obtain
10 the steering matrix components
Q(d) with
d = {1,2}.
[0099] Advantageously, the method
10 according to the first aspect of the invention thus allows of obtaining an optimized steering matrix
Q = [
Q(1) Q(2)] that really considers all various communication configurations, such as the one illustrated on Figures 2 and 3, where more receive antennas
231, 232, 241 and
242 than spatial streams
NSStot are involved.
[0100] The steering matrix
Q obtained by implementing the method
10 according to the first aspect of the invention is intended to be used for implementing a method
100 for (DL-)MU-MIMO transmission of data according to the second aspect of the invention. As the method
10, the method
100 is implementable by a (DL-)MU-MIMO communication system
20 as illustrated on Figures 2 and 3.
[0101] With references to
Figure 6, the method
100 according to the second aspect of the invention comprises, with a steering matrix
Q obtained according to the method
10 according to the first aspect of the invention as described above, the following steps:
- At the level of the transmitter 200, pre-coding 101 a modulated data symbol S(d)(m) of an mth symbol to be transmitted to the dth receiver among said K = 2 receivers 230, 240 according to Q(d)S(d)(m), then
- emitting 102 simultaneously at least two signals, with each of said at least two signals being emitted from said at least NTX = 2 transmit antennas 201, 202.
[0102] The preceding steps of the method
100 are much particularly such that the method
100 further comprises the following steps:
- At the level of the d = 1 receiver 230 among the K = 2 receivers 230, 240, receiving 103a said at least two signals by the NRX(1) receive antennas 231, 232, with:

where each R(1)(m) is an NRX(1)-dimension signal received by the d = 1 receiver, and
where

is a matrix of size NTX × NRX(1) characterizing physical channels and WGN(1)(m) are terms which characterize noise levels, between the transmitter 200 and the d = 1 receiver 230; - At the level of the d = 2 receiver 240 among the K = 2 receivers 230, 240, receiving 103b said at least two signals by the NRX(2) receive antennas 241, 242, with:

where each
R(2)(
m) is an
NRX(2)-dimension signal received by the
d = 2 receiver, and
where

is a matrix of size
NTX ×
NRX(2) characterizing physical channels and
WGN(2)(
m) are terms which characterize noise levels, between the transmitter
200 and the
d = 2 receiver
240.
[0103] Receiving steps
103a and
103b are performed independently from each other.
[0104] As already mentioned, the optimized steering matrix
Q obtained by implementing the method
10 according to the first aspect of the invention does not tend to cancel out the inter-user interference (characterized by terms

and

at the input of each receive antenna
231, 232, 241 and
242. The signals
R(1)(
m) and
R(2)(
m) received on the receive antennas of the user stations
230, 240 include a mix of both modulated data symbols
S(1)(
m) and
S(2)(
m).
[0105] In order to cope with said inter-user interference, each user station
230, 240 may imbed an equalizer
235, 245 capable of measuring signal interferences so that they become able to carry out themselves cancellation of said inter-user interference. For instance, a maximal-ratio combining (MRC) equalizer imbedded into each user station may be used.
[0106] The method
100 may then comprise the following steps performed independently at the level of each
dth receiver among the
K = 2 receivers
230, 240, with
d = {1,2}:
- Equalizing 104a, 104b the NRX(d)-dimension signal R(d)(m) by using coefficients W(d) according to Z(d)(m) = W(d)R(d)(m), said coefficients being pre-computed at the level of each dth user station from an estimation of the effective channel matrix

and - Outputting 105a, 105b a signal Z(d)(m) as estimation of the modulated data symbol S(d)(m).
[0107] More particularly, let call
W(d) the equalizer coefficients used by the
dth receiver upon the reception of the (DL-)MU-MIMO transmission described above.
[0108] The equalizer output
Z(d)(
m) =
W(d)R(d)(
m) of each
dth user station verifies:

[0109] There exists different formulas, such as minimum mean square error (MMSE) formula and zero forcing (ZF) formula, to compute the coefficients
W(d) of each linear equalization from an estimation of the effective channel

For most classical formulas, and at least for the above mentioned classical formulas, it can be observed that:

and

[0110] The method
100 according to the second aspect of the present invention thus tends to cancel out the inter-user interference at the output of the equalizer
235, 245 of each receiver
230, 240.
[0111] Thus the invention discloses a way of computing an improved steering matrix Q which retains the advantage of the standard solution, and more particularly the use of an equalizer within the receivers. This is achieved without limiting the total number of used receive antennas to the number of transmit antennas. The invention allows of implementing simple equalizers, typically maximum-ratio combining (MRC) or Zero Forcing (ZF) equalizers, which are already imbedded in most common user stations, even if more sophisticated equalizers (that carry out interference cancellation) allow of improving the performance of the method
100, as shown hereafter.
[0112] Figure 6 is a comparison of performance simulations obtained by implementing the methods
10 and
100 according to the first and second aspects of the invention versus standard implementations of a DL-MU-MIMO communication system.
[0113] Graphs
610, 620 show the obtained packet error rate (PER) of transmitted data symbol (extending on 1000 bytes) versus the signal-to-noise ratio (SNR) (in decibel, dB) of the communication channel. Those results are here obtained through simulations and more particularly through Monte Carlo simulations. Graph
610 shows the simulation results obtained for the first user station
230 and Graph
610 shows the simulation results obtained for the second user station
240. Graphs
610 and
620 should become exactly the same than each other, when the number of iterations of said Monte Carlo simulations tends to infinity. Four cases that have been discussed in the above description are considered:
- Curves 611, 621 show simulation results in the case where the user stations 230, 240 carry out an "Antenna Reduction" scheme according to which reception is performed by only one antenna per user station and where the AP 200 computes the DL-MU-MIMO steering matrix Q according to the standard "Channel Inversion" algorithm;
- Curves 612, 622 show simulation results in the case where the user stations 230, 240 measure the full 2 × 2 physical channels during the calibration phase. All of the four user station antennas are used for reception of the MU-MIMO frame. Moreover, in all user stations, a sophisticated equalizer that carries out interference cancellation is implemented. In this case also, the AP 200 computes the MU-MIMO steering matrix Q with the standard "Channel Inversion" algorithm.
- Curves 613, 623 show simulation results in the case where the user stations 230, 240 also measure the full 2 × 2 physical channels and use all of the four user station antennas for the reception of the MU-MIMO frames as above. However, user stations only implement in this case a basic equalizer, e.g. an MRC equalizer, that does not perform any interference cancellation. The AP 200 computes the MU-MIMO steering matrix Q with the method 10 according to the first aspect of the invention. Thus, these curves are representative of what can be obtained when pre-coding is performed with the Q(d) matrices computed as described with reference to Figure 4.
- Curves 614, 624 are the same as above and thus correspond to the use of the invention, except that user stations 230, 240 implement equalizers capable of interference cancellation. A further improvement in performance is observed with respect to the simulation results shown by curves 613 and 623.
[0114] Hence, curves
612 and
614 (and curves
622 and
624) can be directly compared to see the benefit of using the methods
10 and
100 according to the present invention since in both cases the user stations
230, 240 use a same equalizer capable of interference cancellation which is not however mandatory for implementing the invention (see curves
613 and
623). A gain improvement of more than 5 dB can thus be observed compared to the standard "Channel Inversion" method of computing the steering matrix
Q.
[0115] As mentioned above, the method
100 for MU-MIMO transmission of data as described above is particularly relevant when at least one among
NRX(1) and
NRX(2) is strictly inferior to the number of spatial streams intended for said receivers, i.e. ∑
dNSS(d). For example, the method
100 is particularly relevant when:
- NTX = 3, NSS(1) = 1, NRX(1) = 2, NSS(2) = 2, NRX(2) = 3; or
- NTX = 4, NSS(1) = 2, NRX(1) = 3, NSS(2) = 2, NRX(2) = 4.
[0116] Thus the method
10 according to the first aspect of the invention proposes a way to compute a steering matrix
Q to be applied for (DL-)MU-MIMO transmission that supports extended configurations of a (DL-)MU-MIMO communication system
20, at least since it allows to use more receive antennas
231, 232, 241 and
242 than available transmit antennas
201 and
202. Indeed, this method
10 allows of avoiding the need to use either an "Antenna Reduction" strategy or an "Interference Cancellation" strategy, as mentioned above.
[0117] The method
10 according to the first aspect of the invention further proposes a way to compute a steering matrix
Q which allows of improving significantly the quality of subsequent (DL-)MU-MIMO transmissions when the number of receive antennas
231, 232, 241 and
242 of the receivers
230 and
240 is larger than the number of transmit antennas
201 and
202 of the transmitter
200, e.g. when
∑dNRX(d) ≥
NTX, since it allows to use more receive antennas (belonging to two receivers
230 and
240) than the number of transmit antennas that the transmitter
200 has.
[0118] Since more hardware resources are efficiently put to contribution owing to the present methods
10 and
100 according to the first and second aspects of the present invention, it comes that said methods offer much better performance than existing algorithms when ∑
dNRX(
d) ≥
NTX.
[0119] Moreover, the communication system
20 can switch from or towards a processing mode implementing the methods
10 and
100 according to the present invention, when the conditions of implementation of these methods (in particular when
∑dNRX(
d) ≥
NTX) are fulfilled, other already known methods for obtaining and using other steering matrices Q being usable when these conditions are not fulfilled.
[0120] The particular implementations shown and described herein are illustrative examples of the invention and are not intended to otherwise limit the scope of the invention in any way.
1. A method (10) for obtaining a steering matrix
Q = [
Q(1),..,
Q(K)] to be applied for MU-MIMO transmission of data, the method being implemented by a MU-MIMO communication system (20) which comprises at least one transmitter (200) comprising at least
NTX = 2 transmit antennas (201, 202) and
K = 2 receivers (230, 240), with each
dth receiver (230 or 240) comprising
NRX(d) receive antennas (231, 232 or 241, 242, respectively), wherein ∑
dNRX(d) ≥
NTX, the method comprising, at the level of the transmitter (200), the following sequence of steps for each value of
d = {1,2} :
- Computing (11) a channel inversion matrix C(d) of size NTX × NRX(d) wherein: C(d) = (F(1)HF(1) + F(2)HF(2))-1F(d)H,
wherein F(d) is a matrix of size NTX × NRX(d) whose components depend on those of a matrix

of size NTX × NRX(d) which characterizes physical channel between the transmitter (200) and the dth receiver (230 or 240) and wherein exponent H refers to the Hermitian operator;
- Computing (12) a beamforming component B(d) by singular value decomposition of matrix product F(d)C(d); and
- Computing (13) the steering matrix Q according to Q(d) = C(d)B(d).
2. Method (10) according to claim 1, wherein the beamforming component B(d) is more particularly computed (12) as NSS(d) singular vectors corresponding to the NSS(d) strongest singular values of matrix product F(d)C(d), with the steering matrix component Q(d) being of size NTX × NSS(d), where NSS(d) is a number of spatial streams intended for the dth receiver.
3. Method (10) according to any one of the previous claims, wherein at least one among the number of antenna(s) NRX(1) of the first receiver (230) and the number of antenna(s) NRX(2) of the second receiver (240) is strictly inferior to the number of spatial streams intended for said receivers, i.e. ∑dNSS(d).
4. Method (10) according to any one of the previous claims, wherein the MU-MIMO communication system (20) is a WiFi communication system, the at least one transmitter (200) is an access point (AP) of said WiFi communication system, the K = 2 receivers (230, 240) are user stations of said WiFi communication system, with the method (10) being a method for obtaining a steering matrix Q to be applied for DL-MU-MIMO transmission of data.
5. Method (10) according to any one of the previous claims, further comprising, before said sequence of steps, a sounding step (9) comprising at the level of the transmitter (200) the following substeps:
- Emitting (91) a Null Data Packet (NDP) frame comprising NTX streams from transmit antennas (201, 202), then
- From a first receiver (230) among the K = 2 receivers (230, 240), receiving (92) a first set of data on which depends the matrix

of size NTX × NRX(1) which characterizes physical channel between the transmitter (200) and the first receiver (230),
- From a second receiver (240) among the K = 2 receivers (230, 240), receiving (93) a second set of data on which depends the matrix

of size NTX × NRX(2) which characterizes physical channel between the transmitter (200) and the second receiver (240).
6. Method (10) according to claim 5, wherein each of said first set of data and said second set of data comprises matrices
V(d) of size
NTX ×
NRX(d) and ∑
(d) of size
NRX(d) ×
NRX(d), with
d = 1 and
d = 2 respectively, said matrices
V(d) and ∑
(d) being determined as a singular value decomposition of
U(d)∑(d)V(d)H, wherein
V(d) is an orthonormal matrix comprising the
NRX(d) left singular vectors of

and ∑
(d) is a diagonal matrix comprising the
NRX(d) strongest singular values of
7. Method (100) for MU-MIMO transmission of data, the method being implemented by a MU-MIMO communication system (20) which comprises at least a transmitter (200) comprising at least
NTX = 2 transmit antennas (201, 202) and
K = 2 receivers (230, 240), with each
dth receiver (230 or 240) comprising
NRX(d) receive antennas (231, 232 or 241, 242, respectively), wherein ∑
dNRX(d) ≥
NTX, the method comprising, with a steering matrix
Q = [
Q(1),..,
Q(K)] being obtained according to the method of any one of claims 1 to 6, the following steps :
- At the level of the transmitter (200), pre-coding (101) a modulated data symbol S(d)(m) of an mth symbol to be transmitted to the dth receiver among said K = 2 receivers (230, 240) according to Q(d)S(d)(m), then emitting (102) simultaneously at least two signals from said at least NTX = 2 transmit antennas (201, 202), in order for the method (100) to further comprises the following steps :
- At the level of the d = 1 receiver (230) among the K = 2 receivers (230, 240), receiving (103a) said at least two signals by the NRX(1) receive antennas (231, 232), with:

where each R(1)(m) is an NRX(1)-dimension signal received by the d = 1 receiver, and
where

is a matrix of size NTX × NRX(1) characterizing physical channels and WGN(1)(m) are terms which characterize noise levels, between the transmitter (200) and the d = 1 receiver (230);
- At the level of the d = 2 receiver (240) among the K = 2 receivers (230, 240), receiving (103b) said at least two signals by the NRX(2) receive antennas (241, 242), with:

where each R(2)(m) is an NRX(2)-dimension signal received by the d = 2 receiver, and
where

is a matrix of size NTX × NRX(2) characterizing physical channels and WGN(2)(m) are terms which characterize noise levels, between the transmitter (200) and the d = 2 receiver (240).
8. Method (100) according to claim 7, further comprising, with each
dth receiver (230, 240) comprising an equalizer (235, 245), for instance a maximal-ratio combining (MRC) equalizer, the following steps, with
d = {1,2}:
- Equalizing (104a, 104b) the NRX(d)-dimension signal R(d)(m) by using coefficients W(d) according to Z(d)(m) ≡ W(d)R(d)(m), said coefficients being pre-computed from an estimation of the effective channel matrix

and
- Outputting (105a, 105b) a signal Z(d)(m) as an estimation of the modulated data symbol S(d)(m).
9. Method (100) according to any one of claims 7 and 8, wherein said symbol to be transmitted are encoded by using one among OFDM encoding and COFDM (for Coded Orthogonal Frequency-Division Multiplexing) encoding.
10. Method (100) according to claim 9, wherein said symbol are modulated by using one among PSK modulation and QAM modulation.
11. A computer program product comprising instructions which, when implemented by at least one digital processing device, performs at least the steps of the method (10) for obtaining a steering matrix Q = [Q(1),..,Q(K)] to be applied for MU-MIMO transmission of data according to any one of claims 1 to 6.
12. A computer program product comprising instructions which, when implemented by at least one digital processing device, performs at least the steps of the method (100) for MU-MIMO transmission of data according to any one of claims 7 to 10.
13. A transmitter (200) comprising at least NTX = 2 transmit antennas (201, 202) intended to be used into a MU-MIMO communication system (20) which further comprises K = 2 receivers (230, 240), with each dth receiver (230 or 240) comprising NRX(d) receive antennas (231, 232 or 241, 242, respectively), wherein ∑dNRX(d) ≥ NTX,
wherein said transmitter (200) is configured to implement the method (10) for obtaining a steering matrix Q = [Q(1),..,Q(K)] to be applied for MU-MIMO transmission of data according to any one of claims 1 to 6.
14. A MU-MIMO communication system (20) which comprises at least one transmitter (200) comprising at least NTX = 2 transmit antennas (201, 202) and K = 2 receivers (230, 240), with each dth receiver (230 or 240) comprising NRX(d) receive antennas (231, 232 or 241, 242, respectively), wherein ∑dNRX(d) ≥ NTX,
wherein said MU-MIMO communication system (20) is configured to implement the method (100) for MU-MIMO transmission of data according to any one of claims 7 to 10.
15. A MU-MIMO communication system (20) according to the previous claim, wherein at least one among the number of antenna(s) NRX(1) of the first receiver (230) and the number of antenna(s) NRX(2) of the second receiver (240) is strictly inferior to NSStot = ∑dNSS(d), where NSS(d) is a number of spatial streams intended for the dth receiver.
1. Verfahren (10) zum Erhalten einer Steuermatrix
Q =
[Q(1), .. , Q(K)], um für die MU-MIMO-Übertragung von Daten angewendet zu werden, wobei das Verfahren durch ein MU-MIMO-Kommunikationssystem (20) implementiert wird, das mindestens einen Transmitter (200) umfasst, umfassend mindestens
NTX = 2 Übertragungsantennen (201, 202) und
K = 2 Empfänger (230, 240), wobei jeder
d-te Empfänger (230 oder 240)
NRX(d) Empfangsantennen (231, 232 bzw. 241, 242) umfasst, wobei ∑
d NRX(d) ≥
NTX, wobei das Verfahren auf der Ebene des Transmitters (200) die folgende Sequenz von Schritten für jeden Wert von
d = {1,2} umfasst:
- Berechnen (11) einer Kanalinversionsmatrix C(d) der Größe NTXx NRX(d),
wobei: C(d) = (F(1)HF(1) + F(2)HF(2))-1F(d)H,
wobei F(d) eine Matrix der Größe NTX x NRX(d) ist, deren Komponenten von denjenigen einer Matrix

der Größe NTX x NRX(d) abhängen, die den physikalischen Kanal zwischen dem Transmitter (200) und dem d-ten Empfänger (230 oder 240) kennzeichnet, und wobei sich der Exponent H auf den hermiteschen Operator bezieht;
- Berechnen (12) einer Strahlbildungskomponente B(d) durch eine Singularwertzerlegung des Matrixprodukts F(d)C(d); und
- Berechnen (13) der Steuermatrix Q gemäß Q(d) = C(d)B(d).
2. Verfahren (10) nach Anspruch 1, wobei die Strahlbildungskomponente B(d) insbesondere als NSS(d) Singularvektoren berechnet wird, die den stärksten Singularwerten von NSS(d) des Matrixprodukts F(d)C(d) entsprechen, wobei die Steuermatrixkomponente Q(d) der Größe NTX x NSS(d) ist, wobei NSS(d) eine Anzahl von spatialen Strömen ist, die für den d-ten Empfänger ausgelegt sind.
3. Verfahren (10) nach einem der vorhergehenden Ansprüche, wobei mindesten eine aus der Anzahl von Antennen NRX(1) des ersten Empfängers (230) und der Anzahl von Antennen NRX(2) des zweiten Empfängers (240) strikt geringer als die Anzahl von spatialen Strömen ist, die für die Empfänger ausgelegt sind, d. h. ∑d NSS(d).
4. Verfahren (10) nach einem der vorhergehenden Ansprüche, wobei das MU-MIMO-Kommunikationssystem (20) ein WiFi-Kommunikationssystem ist, der mindestens eine Transmitter (200) ein Zugriffspunkt (AP) des WiFi-Kommunikationssystems ist, die K = 2 Empfänger (230, 240) Benutzerstationen des WiFi-Kommunikationssystems sind, wobei das Verfahren (10) ein Verfahren ist, um eine Steuermatrix Q zu erhalten, um für die MU-MIMO-Übertragung von Daten angewendet zu werden.
5. Verfahren nach einem der vorhergehenden Ansprüche, weiter umfassend, vor der Sequenz von Schritten, einen Sondierungsschritt (9), umfassend auf der Ebene des Transmitters (200) die folgenden Unterschritte:
- Senden (91) eines Nulldatenpaket (NDP)-Rahmens, umfassend NTX Ströme von Übertragungsantennen (201, 202), dann
- von einem ersten Empfänger (230) aus den K = 2 Empfängern (230, 240) Empfangen (92) eines ersten Satzes von Daten, von dem die Matrix

der Größe NTX x NRX(1) abhängt, die den physikalischen Kanal zwischen dem Transmitter (200) und dem ersten Empfänger (230) kennzeichnet,
- von einem zweiten Empfänger (240) aus den K = 2 Empfängern (230, 240), Empfangen (93) eines zweiten Satzes von Daten, von dem die Matrix

der Größe NTXx NRX(2) abhängt, die den physikalischen Kanal zwischen dem Transmitter (200) und dem zweiten Empfänger (240) kennzeichnet.
6. Verfahren (10) nach Anspruch 5, wobei jeder des ersten Satzes von Daten und des zweiten Satzes von Daten Matrizen
V(d) der Größe
NTX x NRX(d) und ∑
(d) der Größe
NRX(d) x NRX(d) umfasst, wobei
d = 1 bzw.
d = 2, wobei die Matrizen V
(d) und ∑
(d) als eine Singularwertzerlegung von

bestimmt wird, wobei
V(d) eine orthonormale Matrix ist, umfassend die
NRX(d) linken Singularvektoren von

und ∑
(d) eine diagonale Matrix ist, umfassend die
NRX(d) stärksten Singularwerte von
7. Verfahren (100) zur MU-MIMO-Übertragung von Daten, wobei das Verfahren durch ein MU-MIMO-Kommunikationssystem (20) implementiert wird, das mindestens einen Transmitter (200) umfasst, umfassend mindestens
NTX = 2 Übertragungsantennen (201, 202) und
K = 2 Empfänger (230, 240), wobei jeder d-te Empfänger (230 oder 240)
NRX(d) Empfangsantennen (231, 232 bzw. 241, 242) umfasst, wobei ∑
d NRX(d) ≥ NTX, wobei das Verfahren, mit einer Steuermatrix
Q =
[Q(1),..,
Q(K)], die gemäß dem Verfahren nach einem der Ansprüche 1 bis 6 erhalten wird, die folgenden Schritte umfasst:
- auf der Ebene des Transmitters (200) Vorcodieren (101) eines modulierten Datensymbols S(d)(m) eines m-ten Symbols, das an den d-ten Empfänger aus den K = 2 Empfängern (230, 240) gemäß Q(d)S(d)m übertragen werden soll, dann gleichzeitiges Senden (102) von mindestens zwei Signalen von den mindestens NTX = 2 Übertragungsantennen (201, 202), damit das Verfahren (100) weiter die folgenden Schritte umfasst:
- auf der Ebene des d = 1 Empfängers (230) aus den K = 2 Empfängern (230, 240) Empfangen (103a) der mindestens zwei Signale durch die NRX(1) Empfangsantennen (231, 232), wobei:

wobei jedes R(1)(m) ein Signal der NRX(1)-Dimension ist, empfangen durch den d = 1 Empfänger, und
wobei

eine Matrix der Größe NTX x NRX(1) ist, die physikalische Kanäle kennzeichnet und WGN(1)(m) Ausdrücke sind, die Geräuschniveaus zwischen dem Transmitter (200) und dem d = 1 Empfänger (230) kennzeichnen;
- auf der Ebene des d = 2 Empfängers (240) aus den K = 2 Empfängern (230, 240) Empfangen (103b) der mindestens zwei Signale durch die NRX(2) Empfangsantennen (241, 242), wobei:

wobei jedes R(2)(m) ein Signal der NRX(2)-Dimension ist, empfangen durch den d = 2 Empfänger, und
wobei

eine Matrix der Größe NTX x NRX(2) ist, die physikalische Kanäle kennzeichnet, und WGA(2)(m) Ausdrücke sind, die Geräuschniveaus zwischen dem Transmitter (200) und dem d = 2 Empfänger (240) kennzeichnen.
8. Verfahren (100) nach Anspruch 7, weiter umfassend mit jedem
d-ten Empfänger (230, 240), der einen Entzerrer (235, 245) umfasst, z. B. einen Maximalverhältniskombinations (MRC)-Entzerrer, die folgenden Schritte, wobei
d = {1,2}:
- Entzerren (104a, 104b) der Signale der NRX(d)-Dimension R(d)(m) durch Verwendung der Koeffizienten W(d) gemäß Z(d)(m) ≡ W(d)R(d)(m), wobei diese Koeffizienten aus einer Schätzung der effektiven Kanalmatrix Hφ(d) Q(d) vorberechnet werden, und
- Ausgeben (105a, 105b) eines Signals Z(d)(m) als eine Schätzung des modulierten Datensymbols S(d)(m).
9. Verfahren (100) nach einem der Ansprüche 7 und 8, wobei das Symbol, das übertragen werden soll, durch Verwendung eines aus der OFDM-Codierung und COFDM (für Coded Orthogonal Frequency-Division Multiplexing)-Codierung codiert wird.
10. Verfahren (100) nach Anspruch 9, wobei das Symbol unter Verwendung eines aus der PSK-Modulation und der QAM-Modulation moduliert wird.
11. Computerprogrammprodukt, umfassend Anweisungen, das, wenn es durch mindestens eine digitale Verarbeitungsvorrichtung implementiert ist, mindestens die Schritte des Verfahrens (10) zum Erhalt einer Steuermatrix Q = [Q(1),..,Q(K)] durchführt, um auf die MU-MIMO-Übertragung von Daten nach einem der Ansprüche 1 bis 6 angewendet zu werden.
12. Computerprogrammprodukt, umfassend Anweisungen, das, wenn es durch mindestens eine digitale Verarbeitungsvorrichtung implementiert ist, mindestens die Schritte des Verfahrens (100) für die MU-MIMO-Übertragung von Daten nach einem der Ansprüche 7 bis 10 durchführt.
13. Transmitter (200), umfassend mindestens NTX = 2 Übertragungsantennen (201, 202), die ausgelegt sind, um in einem MU-MIMO-Kommunikationssystem (20) verwendet zu werden, das weiter K = 2 Empfänger (230, 240) umfasst, wobei jeder d-te Empfänger (230 oder 240) NRX(d) Empfangsantennen (231, 232 bzw. 241, 242) umfasst, wobei ∑d NRX(d) ≥ NTX,
wobei der Transmitter (200) konfiguriert ist, um das Verfahren (10) zum Erhalt einer Steuermatrix Q = [Q(1), .. , Q(K)] zu implementieren, um auf die MU-MIMO-Übertragung von Daten nach einem der Ansprüche 1 bis 6 angewendet zu werden.
14. MU-MIMO-Kommunikationssystem (20), das mindestens einen Transmitter (200) umfasst, umfassend mindestens NTX = 2 Übertragungsantennen (201, 202) und K = 2 Empfänger (230, 240), wobei jeder d-te Empfänger (230 oder 240) NRX(d) Empfangsantennen (231, 232 bzw. 241, 242) umfasst, wobei ∑d NRX(d) ≥ NTX,
wobei das MU-MIMO-Kommunikationssystem (20) konfiguriert ist, um das Verfahren (100) zur MU-MIMO-Übertragung von Daten nach einem der Ansprüche 7 bis 10 zu implementieren.
15. MU-MIMO-Kommunikationssystem (20) nach dem vorhergehenden Anspruch, wobei mindestens eine aus der Anzahl von Antennen NRX(1) des ersten Empfängers (230) und der Anzahl von Antennen NRX(2) des zweiten Empfängers (240) streng geringer als NSStot = ∑dNSS(d) ist, wobei NSS(d) eine Anzahl von spatialen Strömen ist, die für den d-ten Empfänger ausgelegt sind.
1. Procédé (10) pour obtenir une matrice de direction
Q = [
Q(1), ..., Q
(K)] à appliquer pour une transmission MU-MIMO de données, le procédé étant implémenté par un système de communication MU-MIMO (20) qui comprend au moins un transmetteur (200) comprenant au moins
NTX = 2 antennes de transmission (201, 202) et
K = 2 récepteurs (230, 240), chaque
dième récepteur (230 ou 240) comprenant
NRX(d) antennes de transmission (231, 232 ou 241, 242, respectivement), dans lequel ∑
dNRX(d) ≥
NTX, le procédé comprenant, au niveau du transmetteur (200), la séquence d'étapes suivante pour chaque valeur de
d = {1, 2} :
- le calcul (11) d'une matrice d'inversion de canal C(d) de taille NTXx NRX(d)
dans lequel : C(d) = (F(1)HF(1) + F(2)HF(2))-1F(d)H,
dans lequel F(d) est une matrice de taille NTX x NRX(d) dont les composantes dépendent de celles d'une matrice

de taille NTX x NRX(d) qui caractérise un canal physique entre le transmetteur (200) et le dième récepteur (230 ou 240) et dans lequel l'exposant H désigne l'opérateur hermitien ;
- le calcul (12) d'une composante de formation de faisceau B(d) par une décomposition de valeur singulière de produit de matrice F(d) C(d) ; et
- le calcul (13) de la matrice de direction Q selon Q(d) = C(d)B(d).
2. Procédé (10) selon la revendication 1, dans lequel la composante de formation de faisceau B(d) est calculée (12) plus particulièrement sous forme de NSS(d) vecteurs singuliers correspondant aux NSS(d) valeurs singulières les plus fortes de produit de matrice F(d) C(d), la composante de matrice de direction Q(d) étant de taille NTX x NSS(d), où NSS(d) est un nombre de flux spatiaux destinés au dième récepteur.
3. Procédé (10) selon l'une quelconque des revendications précédentes, dans lequel au moins l'un parmi le nombre d'antennes NRX(1) du premier récepteur (230) et le nombre d'antennes NRX(2) du second récepteur (240) est strictement inférieur au nombre de flux spatiaux destinés auxdits récepteurs, c'est-à-dire ∑d NSS(d).
4. Procédé (10) selon l'une quelconque des revendications précédentes, dans lequel le système de communication MU-MIMO (20) est un système de communication WiFi, l'au moins un transmetteur (200) est un point d'accès (AP) dudit système de communication WiFi, les K = 2 récepteurs (230, 240) sont des stations utilisateur dudit système de communication WiFi, le procédé (10) étant un procédé pour obtenir une matrice de direction Q à appliquer pour une transmission DL-MU-MIMO de données.
5. Procédé (10) selon l'une quelconque des revendications précédentes, comprenant en outre, avant ladite séquence d'étapes, une étape de sondage (9) comprenant au niveau du transmetteur (200) les sous-étapes suivantes :
- l'émission (91) d'une trame de paquet de données nul (NDP) comprenant NTX flux en provenance d'antennes de transmission (201, 202), puis
- en provenance d'un premier récepteur (230) parmi les K = 2 récepteurs (230, 240), la réception (92) d'un premier ensemble de données dont dépend la matrice

de taille NTX x NRX(1) qui caractérise un canal physique entre le transmetteur (200) et le premier récepteur (230),
- en provenance d'un second récepteur (240) parmi les K = 2 récepteurs (230, 240), la réception (93) d'un second ensemble de données dont dépend la matrice

de taille NTXx NRX(2) qui caractérise un canal physique entre le transmetteur (200) et le second récepteur (240).
6. Procédé (10) selon la revendication 5, dans lequel chacun dudit premier ensemble de données et dudit second ensemble de données comprend des matrices
V(d) de taille
NTX x
NRX(d) et ∑
(d) de taille
NRX(d) x
NRX(d), d = 1 et
d = 2 respectivement, lesdites matrices
V(d) et ∑
(d) étant déterminées sous forme de décomposition de valeur singulière de

dans lequel
V(d) est une matrice orthonormale comprenant les
NRX(d) vecteurs singuliers restants de

et ∑
(d) est une matrice diagonale comprenant les
NRX(d) valeurs singulières les plus fortes de
7. Procédé (100) de transmission MU-MIMO de données, le procédé étant implémenté par un système de communication MU-MIMO (20) qui comprend au moins un transmetteur (200) comprenant au moins
NTX = 2 antennes de transmission (201, 202) et
K = 2 récepteurs (230, 240), chaque
dième récepteur (230 ou 240) comprenant
NRX(d) antennes de réception (231, 232 ou 241, 242, respectivement), dans lequel ∑
dNRX(d) ≥
NTX, le procédé comprenant, avec une matrice de direction
Q = [
Q(1), ..., Q
(K)] obtenue selon le procédé de l'une quelconque des revendications 1 à 6, les étapes suivantes :
- au niveau du transmetteur (200), le précodage (101) d'un symbole de données modulé S(d)(m) d'un mième symbole à transmettre au dième récepteur parmi lesdits K = 2 récepteurs (230, 240) selon Q(d) S(d)(m), puis l'émission (102) simultanée d'au moins deux signaux à partir desdites au moins NTX = 2 antennes de transmission (201, 202), afin que le procédé (100) comprenne en outre les étapes suivantes :
- au niveau du d = 1 récepteur (230) parmi les K = 2 récepteurs (230, 240), la réception (103a) desdits au moins deux signaux par les NRX(1) antennes de réception (231, 232), avec :

où chaque R(1)(m) est un signal de dimension NRX(1) reçu par le d = 1 récepteur, et
où

est une matrice de taille NTX x NRX(1) caractérisant des canaux physiques et WGN(1) (m) sont des termes qui caractérisent des niveaux de bruit, entre le transmetteur (200) et le d = 1 récepteur (230) ;
- au niveau du d = 2 récepteur (240) parmi les K = 2 récepteurs (230, 240), la réception (103b) desdits au moins deux signaux par les NRX(2) antennes de réception (241, 242), avec :

où chaque R(2)(m) est un signal de dimension NRX(2) reçu par le d = 2 récepteur, et
où

est une matrice de taille NTXx NRX(2) caractérisant des canaux physiques et WGN(2) (m) sont des termes qui caractérisent des niveaux de bruit, entre le transmetteur (200) et le d = 2 récepteur (240).
8. Procédé (100) selon la revendication 7, comprenant en outre, avec chaque
dième récepteur (230, 240) comprenant un égaliseur (235, 245), par exemple un égaliseur de combinaison à rapport maximal (MRC), les étapes suivantes, avec
d = {1,2} :
- l'égalisation (104a, 104b) du signal de dimension NRX(d) R(d)(m) en utilisant des coefficients W(d) selon Z(d)(m) ≡ W(d)R(d)(m), lesdits coefficients étant pré-calculés à partir d'une estimation de la matrice de canal effective

Q(d), et
- la fourniture en sortie (105a, 105b) d'un signal Z(d)(m) sous forme d'une estimation du symbole de données modulé S(d)(m).
9. Procédé (100) selon l'une quelconque des revendications 7 et 8, dans lequel ledit symbole à transmettre est encodé en utilisant au moins l'un parmi l'encodage OFDM et l'encodage COFDM (pour multiplexage par répartition orthogonale de la fréquence codé).
10. Procédé (100) selon la revendication 9, dans lequel ledit symbole est modulé en utilisant au moins l'une parmi la modulation PSK et la modulation QAM.
11. Produit programme d'ordinateur comprenant des instructions qui, lorsqu'elles sont implémentées par au moins un dispositif de traitement numérique, réalise au moins les étapes du procédé (10) pour obtenir une matrice de direction Q = [Q(1), ..., Q(K)] à appliquer pour une transmission MU-MIMO de données selon l'une quelconque des revendications 1 à 6.
12. Produit programme d'ordinateur comprenant des instructions qui, lorsqu'elles sont implémentées par au moins un dispositif de traitement numérique, réalise au moins les étapes du procédé (100) pour une transmission MU-MIMO de données selon l'une quelconque des revendications 7 à 10.
13. Transmetteur (200) comprenant au moins NTX = 2 antennes de transmission (201, 202) destiné à être utilisé dans un système de communication MU-MIMO (20) qui comprend en outre K = 2 récepteurs (230, 240), chaque dième récepteur (230 ou 240) comprenant NRX(d) antennes de réception (231, 232 ou 241, 242, respectivement), dans lequel ∑dNRX(d) ≥ NTX,
dans lequel ledit transmetteur (200) est configuré pour implémenter le procédé (10) pour obtenir une matrice de direction Q = [Q(1), ..., Q(K)] à appliquer pour une transmission MU-MIMO de données selon l'une quelconque des revendications 1 à 6.
14. Système de communication MU-MIMO (20) qui comprend au moins un transmetteur (200) comprenant au moins NTX = 2 antennes de transmission (201, 202) et K = 2 récepteurs (230, 240), chaque dième récepteur (230 ou 240) comprenant NRX(d) antennes de réception (231, 232 ou 241, 242, respectivement), dans lequel ∑dNRX(d) ≥ NTX,
dans lequel ledit système de communication MU-MIMO (20) est configuré pour implémenter le procédé (100) pour une transmission MU-MIMO de données selon l'une quelconque des revendications 7 à 10.
15. Système de communication MU-MIMO (20) selon la revendication précédente, dans lequel au moins l'un parmi le nombre d'antennes NRX(1) du premier récepteur (230) et le nombre d'antennes NRX(2) du second récepteur (240) est strictement inférieur à NSStot = ΣdNSS(d), où NSS(d) est un nombre de flux spatiaux destinés au dième récepteur.