US 7,917,107 B2Grant

Antenna selection with RF imbalance

Issue Date:2011-03-29

•9 Claims

•5 Drawing Sheets

Abstract

A method for selecting antennas in a multiple-input, multiple-output wireless communications network that includes multiple stations is presented. Each station includes a set of transmit RF chains and a set of receive RF chains, and each RF chain is connectable to a set of antennas. A first station transmits a training frame for each possible subset of the set of antennas, and the set of transmit RF chains are connected to the subset of antennas according to a connection mapping rule while transmitting each training frame. A subchannel matrix is estimated in a second station for each frame. The subchannel matrices are combined to obtain a complete channel matrix. A particular subset of the antennas is selected according to the complete channel matrix. The set of transmit RF chains is connected to the selected particular subset of antennas to transmit data from the first station to the second station.

Metadata

Assignee

Mitsubishi Electric Research Laboratories, Inc.

Inventors

Daqing Gu
Andreas F. Molisch
Jinyun Zhang
Hongyuan Zhang
Dong Wang

Application Information

Application Number:US 11/387,252

Filing Date:2006-03-23

Priority Date:2006-03-23

Art Unit:2618

Classifications

IPC:

H04B7/08

Patent Drawings (5 sheets)

Description

FIELD OF THE INVENTION

[0001] This invention relates generally to multiple-input, multiple-output (MIMO), wireless communication systems, and more particularly to selecting antennas in MIMO systems.

BACKGROUND OF THE INVENTION

[0002] Multiple-input, multiple-output (MIMO) techniques can significantly increase system capacity in a scattering environment of a wireless network. However, multiple antennas increases hardware complexity and cost because, in a typical system, each transmit/receive antenna requires a separate radio frequency (RF) chain including a modulator/demodulator, an AD/DA converter, an up/down converter, and a power amplifier. In addition, the processing complexity at the baseband also increases with the number of antennas.

[0003] Antenna selection (AS) can reduce the number of RF chains while still taking advantage of the capacity/diversity increase provided by multiple antennas. The idea is to select a submatrix from a complete channel matrix according to some predetermined criterion. To perform antenna selection, the complete channel matrix is estimated by sending training frames to measure the complete channel state information (CSI). The training frames can be sent in the same physical layer packet or by multiple packets.

[0004] Conventionally, on transmitting or receiving these AS training frames, the device conducting antenna selection switches different antenna subsets to the RF chains and estimates the corresponding subchannels matrices. The selection is based on the complete channel matrix composed of the estimated subchannels matrices.

[0005] However, conventional antenna selection schemes ignore the fact that each possible connection of a RF chain to an antenna introduces one unique RF response containing the effects of both amplitude gain and phase shift. As a result, in some circumstances, distortions are inevitable for antenna selection, because the selected antennas used for data transmission and reception may be connected to RF chains different from those used during training.

[0006] In the data transmission phase, the actual channel associated with the selected antennas may not be identical to that used in the training phase. This phenomenon is known as a RF imbalance problem.

[0007] It is desired to correct the RF imbalance problem in MIMO systems.

SUMMARY OF THE INVENTION

[0008] The embodiments of the invention provide solutions for antenna selection that reduce an RF imbalance problem. A first solution trains with all possible antenna subsets, instead of the disjointed subsets as in conventional schemes. To avoid large overhead, a second solution defines calibration procedures to correct the distortion.

BRIEF DESCRIPTION OF THE DRAWINGS

[0009] FIG. 1 is a block diagram of a MIMO system according to an embodiment of the invention;

[0010] FIG. 2 is a block diagram of an antenna selection cycle according to an embodiment of the invention;

[0011] FIG. 3 is a block diagram of training and calibration initiated by a transmit station; and

[0012] FIG. 4 is a block diagram of training and calibration initiated by a receive station.

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENT

System Overview

[0013] FIG. 1 shows a MIMO communication system according to an embodiment of the invention. The system includes a first station (STA A) 110, a second station (STA B) 120, and wireless channels 130. Either station can operate in receive or transmit mode. Generally, the station that transmits data is called the transmit station or transmitter, and the station that receives the data is called the receive station or receiver. The data are transmitted and received using a baseband signal.

[0014] Hereinafter, a ‘set’ is defined as including one or more elements; the number of elements in a ‘subset’ is equal to or less than the number of elements in the corresponding set.

[0015] Each station includes a set of receive (Rx) RF chains 111 and a set of transmit (Tx) RF chains 112, both connected to a set of antennas 113 by switches 114. Generally, in MIMO devices, the number of antennas is larger than the number of RF chains. Therefore, a subset of antennas is selected 115 from the set of total available antennas by a method according to an embodiment of the invention during an antenna selection (AS) training phase as described herein. The selection method can be initiated by either the transmitter or the receiver, and the selection can be conducted at the transmitter and/or at the receiver.

[0016] Association Phase

[0017] During an initial association phase, not shown, the stations exchange information about the number of RF chains, the number of antennas, and the type of antenna selection. In particular, the type of information contained in a feedback packet, e.g., whether information is indices of the antennas to be used, and/or a full (instantaneous) channel state information (CSI), and/or an average channel state information is transmitted during that time or, alternatively, as part of the feedback packet.

[0018] Training and Data Transmission Phases

[0019] As shown generally in FIG. 2, each AS training cycle is composed of an antenna selection training phase 201 and a data transmission phase 202. The cycle can be repeated periodically, or as the channel environment changes. Several AS training frames 211 are transmitted in each AS training phase 201. Each of the frames is transmitted from and/or received by one subset of antennas to be selected. The antenna selection 210 is based on a complete channel matrix. The complete channel matrix is estimated from a combination of the subchannel matrices estimated from the AS training frames 211. The selected antennas are connected to the RF chains by the switches 114 before the data frames 212 are transmitted during the data transmission phase 202. Another antenna selection cycle can be initiated when the current selected antenna subset is outdated, or when a predetermined time interval has elapsed.

[0020] System Model for MIMO Systems with Antenna Selection

[0021] In a MIMO system with conventional antenna selections, a transmit station A has a set of N_Aantennas with N_A_—_SStransmit RF chains, and a receive station B has a set of N_Bantennas with N_B_—_SSreceive RF chains. In the data transmission phase 202, a relationship between a transmitted signal and a received signal in a flat-fading channel can be expressed as
r_B=F_B^H({tilde over (H)}_A→BF_As_A+n)
where r_Bis a N_B_—_SS×1 received signal vector, s_Ais a N_A_—_SS×1 transmitted signal vector, and {tilde over (H)}_A→Bis a N_B×N_Aequivalent channel matrix containing physical channel responses and an effect of transmit and receive RF responses.

[0022] A noise vector n has N_B×1 entries that are independent and identically distributed (i.i.d.) zero-mean circular complex Gaussian random variables with variance N₀. F_Ais a N_A×N_A_—_SStransmit antenna selection matrix, and F_Bis a N_B×N_B_—SSreceive antenna selection matrix. Both F_Aand F_Bare submatrices of an identity matrix. The equivalent channel matrix after antenna selection is a N_B_—_SS×N_A_—_SSmatrix H_eq=F_B^H{tilde over (H)}_A→BF_A, which is a submatrix of the channel matrix {tilde over (H)}_A→B. The superscript ‘H’ means the conjugate transpose, which is used here for selection by the receiver.

[0023] The determination of F_Aand F_Boptimizes the capacity of the channel and the signal-to-noise ratio (SNR). If only one side antenna selection is considered, one of F_Aand F_Bequals the identity matrix, and the corresponding number of RF chains equals the number of antennas.

[0024] Antenna selection is performed by switching each output signal of a transmit RF chain to a selected transmit antenna, or each input signal of a selected receive antenna to a receive RF chain. The required number of RF chains to modulate/demodulate the transmitted/received signals is less than the total number of available transmit/receive antennas. Therefore, the cost of the system is reduced.

[0025] Conventional antenna selection schemes ignore the fact that the equivalent channel {tilde over (H)}_A→Bin the data transmission phase contains the impact caused by RF responses. Specifically,
{tilde over (H)}_A→B=C_B,Rx(F_B)H_A→BC_A,Tx(F_A),
where H_A→Bis the actual propagation channel between the transmit antennas of STA A and the receive antennas of STA B. C_A,Tx(F_A) is a N_A×N_Adiagonal matrix whose i^thdiagonal element, [C_A,Tx(F_A)]_ii, collects the RF response corresponding to the i^thtransmit antenna, which is a function of the antenna selection matrix F_A.

[0026] If the i^throw in F_Acontains all zeros, then the i^thantenna is not selected, and [C_A,Tx(F_A)]_ii=0. If an element at the i^throw and l^thcolumn of F_Ais one, then the i^thantenna is selected and is connected to the l^thtransmit RF chain during the data transmission phase. [C_A,Tx(F_A)]_ii=α_li^(Tx)is a complex number characterizing both the amplitude and phase shift of the RF response corresponding to the connection of transmit RF chain l and antenna i.

[0027] Therefore, in C_A,Tx(F_A) there are N_A_—_SSnon-zero diagonal elements. C_B,Rx(F_B) is similarly defined as [C_B,Rx(F_B)]_jj=β_li^(Rx)if the element at the j^throw and l^thcolumn of F_Bis one.

[0028] In the m^thconventional antenna selection training phase, a relationship between a transmitted signal and a received signal can be expressed as:
r_B,t(m)=T_B^H(m)({tilde over (H)}_A→BT_A(m)s_A,t+n),
where S_A,tis a N_A_—_SS×1 training vector known at both stations; r_B,tis the N_B_—_SS×1 received training vector at STA B; and T_A(m) and T_B(m) are the switching matrices in the m^thAS training frame, whose values are predetermined, indicating the connections of the RF chains to the m^thantenna subset. All these antenna subsets are exclusive with each other. For example, if
N_A=4×N_A_—_SS=2,N_B=2,N_A_—_SS=2
and

[0029]

T_{A} (1) = [\begin{matrix} 1 & 0 \\ 0 & 1 \\ 0 & 0 \\ 0 & 0 \end{matrix}] and T_{A} (2) = [\begin{matrix} 0 & 0 \\ 0 & 0 \\ 1 & 0 \\ 0 & 1 \end{matrix}],

then there are totally

[0030]

M = ⌈ N_{A} / N_{A_SS} ⌉ ⌈ N_{B} / N_{B_SS} ⌉

training frames.

[0031] Then, STA B can estimate the complete channel matrix, which is used for AS computations, by combining the M subchannel matrices. Consequently, by ignoring channel estimation errors from training frame m, the estimated subchannel matrix is
{tilde over (H)}′_A→B(m)=T_B^H(m)C_B,Rx(T_B(m))H_A→BC_A,Tx(T_A(m))T_A(m),
and the AS is based on the following estimated complete channel matrix:
{tilde over (H)}′_A→B=C′_B,RxH_A→BC′_A,Tx,
where the diagonal matrix C′_A,Txcontains all non-zero diagonal elements, and [C′_A,Tx]_ii=[C_A,Tx(T_A(m))]_iiif the i^thantenna is trained by the m^thtraining frame. Each antenna is trained only once, and C′_B,Rxis similarly defined.

[0032] RF Imbalance

[0033] The antenna selection distortion caused by the RF imbalance can be described as follows. Antenna selection is based on the estimated complete matrix {tilde over (H)}′_A→B, i.e., using a certain AS criteria X, the selection can be expressed as:

[0034]

{F_{A, opt}, F_{B, opt}} = \underset{F_{A}, F_{B}}{\arg \max} X (F_{B}^{H} {\tilde{H}}_{A \to B}^{'} F_{A}) .

There are

[0035]

N = (\begin{matrix} N_{A} \\ N_{A_SS} \end{matrix}) (\begin{matrix} N_{B} \\ N_{B_SS} \end{matrix})

possible antenna subsets to be selected according to the above expression. However, if F_A,opt, F_B,optare selected in the data transmission phase, the equivalent channel becomes
H_eq=F_B,opt^HC_B,Rx(F_B,opt)H_A→BC_A,Tx(F_A,opt)F_A,opt,
and X(H_eq) may not be optimal due to the differences between the RF responses in training phase and in data transmission phase. Therefore, the selection {F_A,opt, F_B,opt} may, in fact, be suboptimal.

[0036] In the above example of N_A=4, N_A_—_SS=2, N_B=2, N_A_—SS=2, i.e., only STA A conducts AS, F_B=T_B=I so C_B,Rxis always fixed. Then, the selection is based on

[0037]

{\tilde{H}}_{A \to B}^{'} = C_{B}^{(Rx)} H_{A \to B} [\begin{matrix} α_{11}^{(Tx)} \\ α_{22}^{(Tx)} \\ α_{13}^{(Tx)} \\ α_{24}^{(Tx)} \end{matrix}] .

[0038] If antennas 1 and 3 are selected at STA A during data transmission phase, then

[0039]

{\tilde{H}}_{A \to B}^{'} = C_{B}^{(Rx)} H_{A \to B} [\begin{matrix} α_{11}^{(Tx)} \\ 0 \\ α_{23}^{(Tx)} \\ 0 \end{matrix}] .

Obviously, there is a distortion caused by α₁₃^(Tx)≠α₂₃^(Tx), and transmit antennas 1 and 3 may not be the optimal selection.

[0040] RF Chain to Antenna Connection Mapping Rule

[0041] To improve the conventional selection process, we use the following connection mapping rule. The RF chains have corresponding RF chain indices and the antennas have corresponding antenna indices. For any selected antenna subset, the connection of RF chains to antennas is as follows. Without loss of generality, a RF chain with smaller RF chain index always connects to an antenna with smaller antenna index. For example, in the above example, if antennas 1 and 3 are selected at STA A, then transmit RF chain 1 is connected to antenna 1, and transmit RF chain 2 is connected to antenna 3.

[0042] According to the connection mapping rule, in the AS training phase 201 and the data transmission phase 202, there are a total of N_A_—_SS×(N_A−N_A_—_SS+1) connections of RF chains and antenna subsets at STA A, and all the possible RF responses can be expressed as:

[0043]

\begin{matrix} \begin{matrix} α_{11}^{(Tx)} & α_{22}^{(Tx)} & \dots & α_{N_{A_SS} N_{A_SS}}^{(Tx)} \\ α_{12}^{(Tx)} & α_{23}^{(Tx)} & \dots & α_{N_{A_SS} (N_{A_SS} + 1)}^{(Tx)} \\ ⋮ & ⋮ & ⋱ & ⋮ \\ α_{1 (N_{A} - N_{a_SS} + 1)}^{(Tx)} & α_{2 (N_{A} - N_{A_SS} + 2)}^{(Tx)} & \dots & α_{N_{A_SS} N_{A}}^{(Tx)} \end{matrix} . & (1) \end{matrix}

[0044] For example, in the above N_A=4, N_A_—_SS=2 case, all possible RF responses include:

[0045] α₁₁^(Tx)α₂₂^(Tx)

[0046] α₁₂^(Tx)α₂₃^(Tx)

[0047] α₁₃^(Tx)α₁₂a₂₃

[0048] α₁₃^(Tx)α₂₄^(Tx).

[0049] There are totally 2×3=6 possible values of α_li^(Tx). Note that the above connection mapping rule does not introduce any performance loss compared with conventional antenna selection schemes, which do not take RF responses into considerations.

[0050] Antenna Selection Considering RF Imbalance

[0051] One solution for the RF imbalance problem is to train all chains in combination with all possible subsets of antennas, instead of only disjoint subsets of antennas, as in the prior art. Consequently, in each AS training phase 201, there are N training frames, one for each possible subset of antennas connected to the transmit RF chains according to the above connection mapping rule. If the m^thpossible subset is selected, i.e., F_{A, opt}=T_A(m), F_{B, opt}=T(m), then, in both the training frame m and the data transmission phase, the estimated subchannel matrix is

[0052]

\begin{matrix} {\tilde{H}}_{A \to B}^{'} (m) = T_{B}^{H} (m) C_{B, Rx} (T_{B} (m)) H_{A \to B} C_{A, Tx} (T_{A} (m)) T_{A} (m) \\ = F_{B, opt}^{H} {\tilde{H}}_{A \to B} F_{A, opt} . \end{matrix}

[0053] Thus, there is no distortion between the AS training phase 201 and the data transmission phase 202. The drawback of this scheme is the increased training overhead because N>M.

[0054] Calibration for Antenna Selection

[0055] In addition, calibration can be performed to reduce the RF imbalance problem. The RF responses cannot always be determined because the RF responses vary with the environment over time, e.g., changes in frequency, temperature, etc. Therefore, an over-the-air calibration processes can be used. The overhead for calibration is small because the calibration processes can be conducted infrequently, e.g., only when the environment varies.

[0056] Calibration for Transmit Antenna Selection

[0057] FIG. 3 shows the calibration procedure when the transmitter (STA A) conducts antenna selection. STA 110 sends a training frame 301 with the following transmit RF chain and antenna connections:
RF1→Ant1,RF2−Ant2, . . . ,RFN_A_—_SS→AntN_A_—_SS,
i.e., the transmit RF responses equal to the first row of expression (1). On receiving this training frame, STA B 120 estimates 310 the corresponding subchannel matrix {tilde over (H)}′_A→B(1).

[0058] STA A sends the other (N_A−N_A_—_SS) training frames 302, in each of which the connections of RF chains to antennas follows the corresponding row in expression (1). On receiving these training frames, STA B 120 estimates 310 the corresponding subchannel matrices {tilde over (H)}′_A→B(2)˜{tilde over (H)}′_A→B(N_A−N_A_—_SS+1).

[0059] After receiving all the (N_A−N_A_—_SS+1) training frames 301-302, STA B feeds back 320 all the estimated subchannel matrices to STA A. STA A determines 330 RF imbalance TX correction coefficients based on all the estimated subchannel matrices fed back from STA B.

[0060] If N_B>N_B_—_SSand STA B also conducts receive antenna selection, on receiving all the (N_A−N_A_—_SS+1) training frames, STA B uses a predetermined subset of receive antennas, each connected to a predetermined receive RF chain.

[0061] The correction coefficients are determined as follows, by ignoring channel estimation errors,

[0062]

{\tilde{H}}_{A \to B}^{'} (1) = [\begin{matrix} {\tilde{h}}_{A \to B, 11}^{(11)} & {\tilde{h}}_{A \to B, 12}^{(22)} & \dots & {\tilde{h}}_{A \to B, 1 N_{A_SS}}^{(N_{A_SS} N_{A_SS})} \\ {\tilde{h}}_{A \to B, 21}^{(11)} & {\tilde{h}}_{A \to B, 22}^{(22)} & \dots & {\tilde{h}}_{A \to B, 2 N_{A_SS}}^{(N_{A_SS} N_{A_SS})} \\ ⋮ & ⋮ & ⋱ & ⋮ \\ {\tilde{h}}_{A \to B, N_{B_SS} 1}^{(11)} & {\tilde{h}}_{A \to B, N_{B_SS} 2}^{(22)} & \dots & {\tilde{h}}_{A \to B, N_{B_SS} N_{A_SS}}^{(N_{A_SS} N_{A_SS})} \end{matrix}],

where {tilde over (h)}_A→B,n_B_i^(li)=β_l_B_n_B^(Rx)h_A→B,n_B_iα_li^(Tx)stands for the equivalent channel involving all the RF responses, and h_A→B,n_B_iis the actual physical channel coefficient from transmit antenna i to receive antenna n_Bconnected to receive RF chain l_B. Therefore, β_l_B_n_B^(Rx)is the corresponding receive RF response at STA B, and similarly α_li^(Tx)is the transmit RF response at STA A with the connection of transmit antenna i and transmit RF chain l. {tilde over (H)}′_A→B(2)˜{tilde over (H)}′_A→B(N_A−N_A_—_SS+1) can be expressed similarly based on different transmit RF chain and antenna connections following the corresponding rows of the expression (1).

[0063] As in expression (1), in any case antenna 2 at STA A can only be connected to RF chain 1 or 2. Then, for a predetermined n_B, the following calculation can be conducted:

[0064]

κ_{l 2} = \frac{{\tilde{h}}_{A \to B, n_{B} 2}^{(12)}}{{\tilde{h}}_{A \to B, n_{B} 2}^{(l 2)}} = \frac{β_{l_{B} n_{B}}^{(Rx)} h_{A \to B, n_{B} 2} α_{12}^{(Tx)}}{β_{l_{B} n_{B}}^{(Rx)} h_{A \to B, n_{B} 2} α_{l 2}^{(Tx)}} = \frac{α_{12}^{(Tx)}}{α_{l 2}^{(Tx)}},

for l=1 and 2, and k₁₂=1. Then, k₁₂is used as the correcting coefficient, which is multiplied on the baseband stream transmitted from the l^thRF chain whenever the l^thRF chain is connected to antenna 2. Therefore, any transmission from antenna 2 leads to a transmit RF response of α₁₂^(TX)at STA A in the AS training phase 201 and in the data transmission phase 202.

[0065] Similarly, for the i^thtransmit antenna:

[0066]

κ_{li} = \frac{{\tilde{h}}_{A -> B, n_{B} i}^{(\min {L_{i}} i)}}{{\tilde{h}}_{A -> B, n_{B} i}^{(li)}} = \frac{α_{\min {L_{i}} i}^{(Tx)}}{α_{li}^{(Tx)}}, for l \in L_{i}

where L_iis the set of RF chain indices that are possible to be connected to antenna i as defined in expression (1). Then, k_liis applied whenever RF chain l is connected to antenna i, and any transmission from antenna i leads to a corresponding transmit RF response of α_min^(Tx)_{L_i_}i. As special cases, transmit antennas 1 and N_Aare always connected to RF chain 1 and N_A_—_SS, respectively, so no correction is needed for antennas 1 and N_A. By doing the same calculations up to I=N_A−1 and by applying the results, at any time the equivalent complete channel matrix can be expressed as:

[0067]

{\tilde{H}}_{A \to B} = \begin{matrix} C_{B, Rx} H_{A \to B} \cdot \\ diag {α_{11}^{(Tx)}, α_{12}^{(Tx)}, \dots, α_{\min {L_{i}} i}^{(Tx)}, \dots, α_{N_{A_SS} N_{A}}^{(Tx)}} \end{matrix},

and there is no distortion between the AS training phase and the data transmission phase for STA. Note that these correction coefficients are applied in both the AS training phase and the data transmission phase, and is equivalent to replacing the ones in F_Aor T_A(m) by the corresponding correction coefficients.

[0068] The above calculation can be repeated N_B_—_SStimes, with n_B=1 . . . N_B_—SS. The resultant N_B_—SSsets of correction coefficients can then be averaged over a time period to reduce the impact from channel estimation errors.

[0069] Calibration for Receive Antenna Selection

[0070] FIG. 4 shows the calibration procedure when the receiver (STA B) 120 conducts antenna selection. All the possible receiver RF responses corresponding to the connections of receive RF chain and receive antennas are:

[0071]

\begin{matrix} \begin{matrix} β_{11}^{(Rx)} & β_{22}^{(Rx)} & \dots & β_{N_{B_SS} N_{B_SS}}^{(Rx)} \\ β_{12}^{(Rx)} & β_{23}^{(Rx)} & \dots & β_{N_{B_SS} (N_{B_SS} + 1)}^{(Rx)} \\ ⋮ & ⋮ & ⋱ & ⋮ \\ β_{1 (N_{B} - N_{B_SS} + 1)}^{(Rx)} & β_{2 (N_{B} - N_{B_SS} + 2)}^{(Rx)} & \dots & β_{N_{B_SS} N_{B}}^{(Rx)} . \end{matrix} & (2) \end{matrix}

[0072] STA B 120 sends a calibration training request (CTRQ) 401 to STA A 110, which contains the number of training frames 410 (N_B−N_B_—_SS+1) the receiving station requires.

[0073] STA A responds by transmitting the (N_B−N_B_—_SS+1) training frames 410 from a predetermined set of transmit antennas. STA B receives each training frame by using the receive RF chain and receive antenna connections defined in the corresponding row of expression (2), and estimates 420 the corresponding subchannel matrix.

[0074] After receiving all the (N_B−N_B_—_SS+1) training frames 410, STA B 120 determines 430 the RF imbalance receive (Rx) correction coefficients based on all the estimated subchannel matrices. This determination is similar as for the transmit antenna selection case: for receive antenna j and for lεL_j,

[0075]

λ_{lj} = \frac{{\tilde{h}}_{A -> B, {jn}_{A}}^{' (\min {L_{j}} j)}}{{\tilde{h}}_{A -> B, {jn}_{A}}^{' (lj)}} = \frac{β_{\min {L_{j}} j}^{(Rx)} h_{A -> B, {jn}_{A}} α_{l_{A} n_{A}}^{(Tx)}}{β_{lj}^{(Rx)} h_{A -> B, {jn}_{A}} α_{l_{A} n_{A}}^{(Tx)}} = \frac{β_{\min {L_{j}} j}^{(Rx)}}{β_{lj}^{(Rx)}},

where {tilde over (h)}′_A→B,jn_A^(lj)is the estimated channel coefficient from any transmit antenna n_Aconnected to transmit RF chain l_Ato receive antenna j corresponding to the connection of receive RF chain l to receive antenna j at STA B, and L_jis the set of RF chain indices that are to be connected to antenna j as defined in expression (2). The correcting coefficient λ_ijis multiplied on the baseband stream received by the l^threceiving RF chain, whenever it is connected to antenna j at STA B, and consequently any reception by antenna j, in the AS training phase or in the data transmission phase, leads to a corresponding receive RF response of β_min{L_j_}j^(Rx). The equivalent complete channel matrix can be expressed as:

[0076]

{\tilde{H}}_{A \to B} = \begin{matrix} diag {β_{11}^{(Rx)}, β_{12}^{(Rx)}, \dots, β_{\min {L_{j}} j}^{(Rx)}, \dots, β_{N_{B_SS} N_{B}}^{(Rx)}} \cdot \\ H_{A \to B} C_{A, Tx} . \end{matrix}

[0077] Hence, there is no distortion between the AS training phase and the data transmission phase at STA B. These correction coefficients are applied in both the AS training phase and the data transmission phase, and is equivalent to replacing the ones in F_Bor T_B(m) by the corresponding correction coefficients.

[0078] The above calculations can be repeated N_A_—_SStimes, with n_A=1 . . . N_A_—_SS. The resultant N_A_—_SSsets of correction coefficients can be averaged over a time period to reduce the impact from channel estimation errors.

[0079] When both stations perform antenna selections, their calibrations can be conducted one after the other. After the first station, either a transmit or receive station, completes calibration, the first station uses a predetermined subset of antennas each connecting to a predetermined RF chain to assist the calibration of the second station. As a result, after the defined calibration procedure(s), the equivalent complete channel matrix is:

[0080]

{\tilde{H}}_{A \to B} = \begin{matrix} diag {β_{11}^{(Rx)}, \dots, β_{\min {L_{j}} j}^{(Rx)}, \dots, β_{N_{B_SS} N_{B}}^{(Rx)}} \cdot \\ H_{A \to B} \cdot diag {α_{11}^{(Tx)}, \dots, α_{\min {L_{i}} i}^{(Tx)}, α_{N_{A_SS} N_{A}}^{(Tx)}} \end{matrix}

which always contains fixed transmit and receive RF responses. Then, antenna selection training can be done in the conventional way without distortions. Only M training frames are required in each AS training phase, in which the training signals are transmitted from and/or received by antenna subsets exclusive with each other.

[0081] The above transmit or receive AS calibration procedure can also be conducted in a normal transmit/receive AS training phase, in which multiple consecutive AS training frames are transmitted from, or received by disjointed antenna subsets without considering RF imbalance. One example is the AS training scheme described in PCT Patent Application number PCT/US2005/042358, “Method for Selecting Antennas and Beams in MIMO Wireless LANs,” filed in the U.S. Receiving Office by Gu et al. on Nov. 21, 2005 and incorporated herein by reference. To do so, the number of the consecutive training frames is modified to be equal to that required for calibration calculation. For example, if N_B=N_B_—_SS, with the RF chain/antenna connection constraints of Equation (1) for transmit AS calibration, this number is N_A−N_A_—SS, instead of

[0082]

⌈ \frac{N_{A}}{N_{A_SS}} ⌉ .

[0083] In the case of transmit AS calibration, STA A in FIG. 3 sends N_A−N_A_—_SS+1 calibration training frames from the antenna subsets with the RF chain connections according to expression (1). STA B treats these frames as normal AS training frames, i.e., STA B estimates subchannel matrices from them and feeds the complete channel matrices back. Finally STA A calculates the transmit correction coefficients.

[0084] In the case of receive AS calibration, STA B in FIG. 4 informs STA A of the required number of calibration training frames by a normal AS training request. STA A then responds by sending the corresponding number of AS training frames with no knowledge of whether these training frames are used for normal receive AS training or for receive AS calibration. STA B receives these frames by the antenna subsets with the RF chain connections according to expression (2), and estimates the subchannel matrices. Finally, STA B determines the receive correction coefficients.

[0085] In such a way, the calibration procedure does not introduce a new signaling protocol.

[0086] Variations

[0087] The above processes can be applied to the cases where the system is frequency-selective, such as orthogonal frequency division multiplexing (OFDM) networks. In OFDM networks, the above training or calibration procedures are conducted for each subcarrier, or in each group of subcarriers.

[0088] In the case that each selected antenna subset is connected to the RF chains with predetermined ordering, the AS training process has more training frames in each AS training phase to estimate the subchannel matrix corresponding to each possible RF chain and antenna connection. The calibration procedure also sends sufficient training frames to compensate for each possible connection, e.g., if any RF chain can connect to any antenna, in the calibration of the transmitter at STA A, L_i={1, 2, . . . , N_A_—_SS}, ∀i, and therefore

[0089]

α_{2 i}^{(Tx)} \dots α_{N_{A_SS} i}^{(Tx)}

all need to be corrected so that the connection of any RF chain with antenna i always leads to a transmit RF response α_li^(Tx).

[0090] In the calibration procedures, the compensation for the RF chains connected to a particular antenna can also be applied with respect to a coefficient other than α_min{L_i_}^(Rx)or β_min{L_j_}j^(Rx). Any calibration procedure by which any transmit or receiver antenna brings a fixed RF response to the equivalent channel can be used.

[0091] The calibration procedure is based on the fact that different RF chains and antenna connections, i.e., switch positions, lead to different amplitude and phase responses. On the other hand, if the switch does not introduce any distortions depending on its positions, i.e., for any particular RF chain its connections to different antennas result in substantially identical RF responses, i.e., all antennas have similar matching properties with the RF chain, then the calibration procedure can be simplified.

[0092] Specifically, this is so because the RF imbalance is caused only by different responses of the RF chains. If the connection ordering constraint described above is applied or not, then it is sufficient to calibrate every RF chain only once: in the first training frame connect the first RF chain to antenna i, then connect the second RF chain to antenna i for transmitting or receiving the second training frame, . . . , connect the last RF chain to the same antenna in the last training frame.

[0093] Finally,

[0094]

α_{1 i}^{(Tx)} \dots α_{N_{A_SS} i}^{(Tx)}

are compensated with respect to one fixed value at the transmitter after feeding back all the measured subchannel matrices, each with a single transmit antenna, or,

[0095]

β_{1 i}^{(Rx)} \dots β_{N_{{B_SS}^{i}}}^{(Rx)}

are directly compensated with respect to one fixed value at the receiver, based on all the measured subchannel matrices, each with a single receive antenna. In this scenario, the RF imbalance problem is caused only by the difference among RF chains, the compensated RF chains, with antenna i, also lead to no distortion when connected to all the other antennas.

[0096] Although the invention has been described by way of examples of preferred embodiments, it is to be understood that various other adaptations and modifications can be made within the spirit and scope of the invention. Therefore, it is the object of the appended claims to cover all such variations and modifications as come within the true spirit and scope of the invention.

Claims

We claim:

1. A method for selecting antennas in a multiple-input, multiple-output wireless communications network including a plurality of stations, each station including a set of transmit RF chains and a set of receive RF chains, wherein each RF chain is connectable to a set of antennas, comprising the steps of:

transmitting, from a first station to a second station during training, a training frame for each possible subset of the set of antennas, and wherein the set of transmit RF chains are connected to the subset of antennas according to a connection mapping rule while transmitting each training frame;

estimating, in the second station, a subchannel matrix for each frame;

feeding back, from the second station to the first station, the subchannel matrix for each frame;

determining, in the first station, radio frequency (RF) imbalance transmit correction coefficients based on the subchannel matrices fed back by the second station, wherein the first station A has the set of N_Aantennas with N_A_—_SStransmit RF chains, and the second station B has the set of N_Bantennas with N_B_—_SSreceive RF chains, and wherein the RF imbalance receive correction coefficients are determined as

{\tilde{H}}_{A \to B}^{'} (1) = [\begin{matrix} {\tilde{h}}_{A \to B, 11}^{(11)} & {\tilde{h}}_{A \to B, 12}^{(22)} & \dots & {\tilde{h}}_{A \to B, 1 N_{A_SS}}^{(N_{A_SS} N_{A_SS})} \\ {\tilde{h}}_{A \to B, 21}^{(11)} & {\tilde{h}}_{A \to B, 22}^{(22)} & \dots & {\tilde{h}}_{A \to B, 2 N_{A_SS}}^{(N_{A_SS} N_{A_SS})} \\ ⋮ & ⋮ & ⋱ & ⋮ \\ {\tilde{h}}_{A \to B, N_{B_SS} 1}^{(11)} & {\tilde{h}}_{A \to B, N_{B_SS} 2}^{(22)} & \dots & {\tilde{h}}_{A \to B, N_{B_SS} N_{A_SS}}^{(N_{A_SS} N_{A_SS})} \end{matrix}],

where

{\tilde{h}}_{A \to B, n_{B} i}^{(li)} = β_{l_{B} n_{B}}^{(Rx)} h_{A \to B, n_{B} i} α_{li}^{(Tx)}

represents an equivalent channel involving all RF responses, and h_A→B,n_B_iis an actual physical channel coefficient from transmit antenna i to receive antenna n_Bconnected to receive RF chain l_B, and β_l_B_n_B^(Rx)is a corresponding receive RF response at the second station B, and α_li^(Tx)is a transmit RF response at the first station A with a connection of transmit antenna i and transmit RF chain l; and

multiplying a baseband stream transmitted by selected transmit antennas at the first station by the RF imbalance transmit correction coefficients.

2. The method of claim 1, further comprising:

determining, in the second station, RF imbalance receive correction coefficients based on the subchannel matrices estimated by the second station; and

multiplying the baseband stream received by selected receive antennas at the second station by the RF imbalance receive correction coefficients.

3. The method of claim 1, further comprising:

transmitting, from the second station to the first station, a calibration training request.

4. The method of claim 1, further comprising:

averaging the RF imbalance transmit correction coefficients over a time period.

5. The method of claim 2, further comprising:

averaging the RF imbalance receive correction coefficients over a time period.

6. The method of claim 1, further comprising:

performing a calibration procedure to reduce an RF imbalance at the first station.

7. The method of claim 1, in which the RF chains have RF chain indices and the antennas have antenna indices, and a RF chain with smaller RF chain index always connects to an antenna with smaller antenna index.

8. The method of claim 1, further comprising:

correcting an RF imbalance using the RF imbalance transmit correction coefficients.

9. The method of claim 8, wherein the selected antennas are connected to different RF chains during data transmission and the training to cause the RF imbalance.

Patent Citations (4)

Patent	Date	Inventor	Cited By
US6804312(B1)	2004-10-01	Win et al.	Examiner
WO3/034614	2003-04-01		Applicant
WO2005/034387	2005-04-01		Applicant
WO2005/034387(A2)	2005-04-01		Examiner

Non-Patent Literature (1)

Molisch et al. “MIMO Systems with Antenna Selection,” Mar. 2004.Applicant