In closed-loop multiple input multiple output (MIMO) beamforming, a subscriber station (SS) quantizes the ideal beamforming matrix and sends the quantization index back to a base station (BS). The BS reconstructs the beamforming matrix according to the feed back index and conducts the beamforming. It is well known that the beamforming increases the link performance and system throughput.
Although the present invention is not limited in this respect, in 802.16e (WiMAX), the ideal beamforming matrix is quantized by a constant quantization codebook. The codebook is optimized for a single channel scenario, where the transmit antenna correlation at the BS is zero. However, the transmit antenna correlation is not constantly zero in reality and varies with several factors such as the antenna spacing at the BS, the BS antenna height, LOS/NLOS condition, BS and SS separation. Furthermore, the optimal quantization codebook varies with the antenna correlation, and thus it is desirable to adapt the codebook to the correlation. For example, the discrete Fourier transform (DFT) codebook and the 802.16e codebook are optimized for either the high or the low antenna correlations but not both. Fortunately, the antenna correlation varies very slowly as compared to the short-term channel fading, and there is a feedback mechanism for long term information in 802.16e.
Thus, a strong need exists for techniques utilizing adaptive codebooks for beamforming in wireless networks.
The subject matter regarded as the invention is particularly pointed out and distinctly claimed in the concluding portion of the specification. The invention, however, both as to organization and method of operation, together with objects, features, and advantages thereof, may best be understood by reference to the following detailed description when read with the accompanying drawings in which:
a provides a quantization constellation for the off-diagonal entry of Rt, which has a point at (0,0) for uncorrelated channels of embodiments of the present invention;
a provides a quantization constellation for the off-diagonal entry of Rt, which is dedicated for correlated channels of embodiments of the present invention;
It will be appreciated that for simplicity and clarity of illustration, elements illustrated in the figures have not necessarily been drawn to scale. For example, the dimensions of some of the elements are exaggerated relative to other elements for clarity. Further, where considered appropriate, reference numerals have been repeated among the figures to indicate corresponding or analogous elements.
In the following detailed description, numerous specific details are set forth in order to provide a thorough understanding of the invention. However, it will be understood by those skilled in the art that the preset invention may be practiced without these specific details. In other instances, well-known methods, procedures, components and circuits have not been described in detail so as not to obscure the present invention.
Although embodiments of the invention are not limited in this regard, discussions utilizing terms such as, for example, “processing,” “computing,” “calculating,” “determining,” “establishing”, “analyzing”, “checking”, or the like, may refer to operation(s) and/or process(es) of a computer, a computing platform, a computing system, or other electronic computing device, that manipulate and/or transform data represented as physical (e.g., electronic) quantities within the computer's registers and/or memories into other data similarly represented as physical quantities within the computer's registers and/or memories or other information storage medium that may store instructions to perform operations and/or processes.
Although embodiments of the invention are not limited in this regard, the terms “plurality” and “a plurality” as used herein may include, for example, “multiple” or “two or more”. The terms “plurality” or “a plurality” may be used throughout the specification to describe two or more components, devices, elements, units, parameters, or the like. For example, “a plurality of stations” may include two or more stations.
Embodiments of the present invention provide an adaptive codebook varying with the antenna correlation. Both a BS and a SS transform the 802.16e codebooks synchronously to generate a new quantization codebook optimized for the BS antenna correlation. Simulation results demonstrate that the codebook of embodiments of the present invention uniformly outperforms the competitor codebooks for all antenna correlation. Furthermore, the codebook of the present invention has almost the same performance of the optimal codebooks that are optimized for each given antenna correlation. Finally, the transformation of embodiments of the present invention provides a backward compatible solution to 802.16e and further the 802.16e codebook.
In closed-loop MIMO beamforming, a subscriber station (SS) quantizes the ideal beamforming matrix and sends the quantization index back to a base station (BS). The BS reconstructs the beamforming matrix according to the feed back index and conducts the beamforming. As mentioned in above, it is well known that the beamforming increases the link performance and system throughput.
In IEEE 802.16e (WiMAX), the ideal beamforming matrix is quantized by a constant quantization codebook. The codebook is optimized for a single channel scenario, where the transmit antenna correlation at the BS is zero. However, the transmit antenna correlation is not constantly zero in reality and varies with several factors such as the antenna spacing at the BS, the BS antenna height, LOS/NLOS condition, BS and SS separation. Furthermore, the optimal quantization codebook varies with the antenna correlation, and thus it is desirable to adapt the codebook to the correlation. For example, present DFT codebooks and 802.16e codebooks are optimized for either one of the high and low antenna correlations but not both. Fortunately, the antenna correlation varies very slowly as compared to the short-term channel fading. The feedback of the correlation costs little system overhead because (unlike beamforming matrix feedback varying for each subband) it remains the same for the whole frequency band and is fed back infrequently say in every 100 ms.
Embodiments of the present invention provide an adaptive codebook varying with the antenna correlation. Both BS and SS transform the 802.16e codebooks synchronously to generate a new quantization codebook optimized for the BS antenna correlation.
The 802.16e codebook is designed for zero antenna correlation, where the ideal beamforming matrix is uniformly distributed. Consequently, the quantization codeword matrix (or vector) is uniformly distributed over quantization spaces. Since the beamforming in an 802.16e system is only for a single user MIMO in long distance (and NLOS) case, the antenna correlation is typically close to zero. However, for WiMAX 2, more transmit antennas (up to 8) and closer antenna mounting (half wavelength) are considered, although the present invention is not limited in this respect. Therefore, the antenna correlation is not zero for some cases while it is close to zero for the others. For example, multi-user MIMO (or downlink SDMA) works well in small antenna spacing while single user MIMO prefers large antenna spacing.
It is undesirable to define multiple codebooks for the different antenna correlations. Embodiments of the present invention provide a universal solution and can transform codebooks adaptively for all antenna correlations. The transformation is a simple function of the antenna correlation.
The present inventions adaptive codebook can be viewed from a differential perspective. The antenna correlation matrix varies very slowly and it carries the average beamforming directions of the strongest received signals. Furthermore, the distribution of the instantaneous beamforming directions concentrates at the average beamforming directions as the correlation increases. Instead of uniformly spreading the quantization codewords, embodiments of the present invention transform the uniform codebook so that the codewords concentrate at the average directions for reducing quantization error.
The idea of the adaptive codebook is illustrated generally as 100 of
due to the constant modulus constraint and leaves big holes in the quantization space. For an uncorrelated channel, the input to the quantizer, i.e. the ideal beamforming matrix, uniformly distributes over the semicircle and thus matches the codeword distribution of the 802.16e codebook. For a highly correlated channel, the channel response magnitude of each transmit antenna has almost the same value and therefore the entry magnitude of the ideal beamforming matrix also has almost the same value. This matches the codeword distribution of the DFT codebook. This explains how two codebooks 110 and 120 work for the two extreme scenarios but not both. Besides the magnitude information used in the DFT codebook, the adaptive codebook 130 further exploits the phase information obtained from the antenna correlation. The adaptive codebook 130 transforms the quantization codewords to the preferable direction where the input beamforming matrix will mostly likely be present. The relocated codewords match the input distribution of the beamforming matrix and thus reduce the quantization error.
Signal Model
The baseband signal model is given by
y=HVs+n, (1)
where n is the complex AWGN with variance N0; s is the Ns by 1 transmitted vector with unit power; Ns is the number of spatial streams; y is the received vector; H is the channel matrix of size Nr by Nt; V is the beamforming matrix (or vector) of size Nt by Ns. In Rayleigh fading channel, the correlated channel matrix H is generated from the channel matrix Hw with independent, identically distributed (i.i.d.) entries as
H=Rr1/2HwRt1/2, (2)
where Rr is the Nr by Nr receive covariance matrix and Rt is the Nt by Nt transmit covariance matrix. The transmit covariance matrix may be defined as
in theory or be simply computed as the empirical mean of HHH over channel matrix samples Hs in practice. Since the SS antenna height is low with respect to the surrounding objects, Rr can be approximated by the identity matrix. Therefore, (2) can be simplified as
H=HwRt1/2. (3)
Let the singular value decomposition of Rt be
Rt=QΣ2QH, (4)
where Q is an Nt by Nt unitary matrix i.e. QHQ=I; Σ is the diagonal matrix with the square roots of the singular values σi for i=1, . . . , Nt in decreasing order σi≧σi+1. Note that the power of Rt in (3) can be written as Rt1/2=QΣlQH. Equation (2) is a simplified channel model of correlated channels, where the correlations at the transmitter and the receiver are decoupled and are modeled by two separate matrixes Rr and Rt. For the case where the decoupling is invalid, the correlation matrix of channel matrix entries is defined as RH=E(vec(H)vec(H)H), where vec(X) stacks the columns of matrix X to make a long column vector.
In Ricean fading channel, a line-of-sight (LOS) component
H=
The transmit covariance matrix is defined as
Codebook Transformation
Denote the quantization codebook with uniformly distributed codewords as Cunif and its codewords as Vi for i=1, . . . , Nc. We would like to transform the uniform codebook to generate a new codebook for the correlated channels.
The present invention derived the distribution of the ideal beamforming matrix for correlated channels with known Rt. Using the derived input distribution, for very high resolution codebooks with many codewords, we can derive the asymptotically optimal transformation that minimizes the channel capacity loss. However, since the transformation requires high complexity functions such as hypergeometric function, it is not practical for the implementation on the mobile device. We simplify the transformation and maintain a performance very close to the optimal. The transformation takes the form of
{tilde over (V)}i=orth(FVi), (7)
where {tilde over (V)}i is the i-th codeword of the new codebook; orth(X) converts the input matrix (or vector) X to an orthogonal matrix with orthonormal column(s) that span the same subspace as X's columns; F is a Nt by Nt transformation matrix. orth(X) is essentially the orthogonalization of X and can be simply implemented by various methods such as Grant-Schmidt and QR decomposition. The transformation matrix F may be a function of RH, Rt, Rr,
Embodiments of the present invention provide three representations of transformation matrix F.
For practicality, F takes the polynomial form
where
F=Rt (9)
may be used for all cases.
The l-th power of Rt has a general form
Rtl=QΣQa, (9.1)
where Qa can be any Nt by Nt unitary matrix. Because of this, (8) and (9) have various forms for different choices of Qas. When Qa=QH, the decomposition of Rt in (4) gives another way to write (8) as
where Q contains the global and local maximums of the ideal beamforming directions for Rayleigh fading channels. If the original codebook for the transformation is uniformly distributed, then the rotation of QH in (10) and (11) can be dropped i.e. Qa=I and the simplified transformation F is
It should be noted that Rt1/2 has a general form Rt1/2=QΣQa, where Qa can be any Nt by Nt unitary matrix. The Cholesky decomposition form of Rt1/2, which is a triangular matrix, is a special case of the general form.
In sum, the simplified codebook transformation can be written as
{tilde over (V)}i=orth(RtlVi) (14)
or equivalently
{tilde over (V)}i=orth(QΣ2lVi), (15)
where Q and Σ are computed from the long term CSI; {tilde over (V)}i is the transformed codeword; and Vi is codeword of the original codebook. Q consists of the favorable, long term beamforming directions and Σ specifies the concentration to those directions. For implementation simplicity, since l=1 delivers good performance for all cases and it doesn't require computing the rational power of Rt, the simplest form of transformation is
{tilde over (V)}i=orth(RtVi). (16)
For feedback reduction, the receiver may not feed back the correlation matrix Rt. Instead, it may feed back only part of Q and Σ. For example, it may feed back the first Ns columns of Q and the first Na eigenvalues of Σ. The transmitter can add complementary and orthogonal columns to the fed back columns to approximate Q and use small values e.g. the minimum of the fed back eigen values to replace the not fed back eigenvalues of Σ. Furthermore, differential technique can be applied to the feedback of Rt or its equivalents.
In the practical system the suitable representation could be chosen based on the balance between the performance, feedback overhead and complexity.
The adaptive codebook can be applied to both one-shot and differential feedbacks. The description above is for one-shot feedback. The differential feedback has the form [3][4]
V(t+1)=Q(V(t))Di, (17)
where V(t) and V(t+1) are the beamforming matrices at time t and t+1; Q(V(t)) is a square rotation matrix that rotates V(t) to V(t+1) using Di. Di is selected from a so called “differential codebook” and the index of Di is fed back from the receiver to the transmitter. The codewords of the differential codebook usually do not uniformly distribute. Instead, they center on somewhere e.g. the identity matrix. The concentration of the codewords increases as the correlation. Therefore, the differential codebook can adapt to the correlation as the one-shot codebook did using (7). A simple example of the transformation is
{tilde over (D)}i=orth(ΣlDi), (18)
where Σ is computed from the long term CSI; l is some number depending on Nt, Ns, Nr, γ, and mobile speed; {tilde over (D)}i is the transformed codeword; and Di is the codeword of the original codebook.
Variants with Low Complexities
Since the orth( ) operation increases complexities at the subscriber station, we devise some variants of the original scheme above. The first variant is the simplest. The SS simply removes the orth( ) operation and directly uses the unorthogonalized matrix FVi to select codeword and feeds back codeword index.
The second variant is as follows. The SS doesn't transform the codebook. Instead, it transforms the channel matrix H and uses the uniform codebook (e.g. 16e codebook) to quantize the ideal beamforming matrix of the transformed channel matrix {tilde over (H)} as
{tilde over (H)}=HTH. (19)
When TH=Rt−1/2, the correlation in H expressed in (3) is removed and {tilde over (H)} becomes uncorrelated, whose distribution matches the uniform codebook. In general, we may let TH=QΣ−l
{tilde over (V)}i=orth(TH−HVi). (15)
Now, most of the computation burden is shifted to the BS.
Quantization of Transformation Matrix
The receiver may feed back the transformation matrix in various forms. For example, it may send back the upper triangle of Rtl
and the maximum number on the diagonal is normalized to unity. Since the diagonal of Rt is real and positive, 1 quantization bit may be sufficient and the diagonal entry may be quantized to either 1 or 0.8. The off-diagonal entry of Rt is usually complex and requires 4 quantization bits per entry. The 16 points of the quantization constellation may be allocated within the unit circle. One example, although not limited to this example, is shown in
is used and
is computed by Cholesky decomposition, similar scalar quantization scheme can be applied to the upper triangle of
that is an upper triangular matrix.
Simulation Results
Some of the link level results are shown generally as 200 of
Some of the system level results are shown as 400 of
While certain features of the invention have been illustrated and described herein, many modifications, substitutions, changes, and equivalents may occur to those skilled in the art. It is, therefore, to be understood that the appended claims are intended to cover all such modifications and changes as fall within the true spirit of the invention.
This application claims the benefit under 35 U.S.C. §119(e) to U.S. Provisional Patent Application No. 61/093,699, filed Sep. 2, 2008 and is hereby incorporated by reference in its entirety.
Number | Name | Date | Kind |
---|---|---|---|
6650881 | Dogan | Nov 2003 | B1 |
7069050 | Yoshida | Jun 2006 | B2 |
7181246 | Lo | Feb 2007 | B2 |
7430168 | Yamaura et al. | Sep 2008 | B2 |
7778826 | Lin et al. | Aug 2010 | B2 |
20030185309 | Pautler et al. | Oct 2003 | A1 |
20070174038 | Wang et al. | Jul 2007 | A1 |
20070206626 | Lee et al. | Sep 2007 | A1 |
20080080449 | Huang et al. | Apr 2008 | A1 |
20080094281 | Teng et al. | Apr 2008 | A1 |
20080240274 | Han et al. | Oct 2008 | A1 |
20090274230 | Heath et al. | Nov 2009 | A1 |
20090298424 | Liu et al. | Dec 2009 | A1 |
20090323849 | Bala et al. | Dec 2009 | A1 |
20110064156 | Kim et al. | Mar 2011 | A1 |
Number | Date | Country |
---|---|---|
101127582 | Jul 2007 | CN |
101677251 | Mar 2010 | CN |
2321911 | Apr 2011 | EP |
2011516126 | May 2011 | JP |
2011525080 | Sep 2011 | JP |
10-2001-0036597 | May 2001 | KR |
10-2002-0089094 | Nov 2002 | KR |
201021455 | Jun 2010 | TW |
WO 2007022330 | Feb 2007 | WO |
WO 2007024959 | Mar 2007 | WO |
2007050861 | May 2007 | WO |
2009157715 | Dec 2009 | WO |
2010027647 | Mar 2010 | WO |
2010027647 | Mar 2010 | WO |
2010027647 | Oct 2010 | WO |
Entry |
---|
International Preliminary Report on Patentability received for PCT Patent Application No. PCT/US2009/054136, mailed on Mar. 17, 2011, 6 pages. |
International Search Report and Written Opinion received for PCT Patent Application No. PCT/US2009/054136, mailed on Dec. 30, 2009, 11 pages. |
Kim, et al. “Optimizing Codebook for Quantized Precoder Using Steepest Descent Method Considering Codeword Quantization”, Korea Information & Communications Society, vol. 33 , Jul. 2006, pp. 564-567. |
Adaptive codebook for beamforming in limited feedback MIMO systems, Information Sciences and Systems, CISS 2008. 42nd Annual Conference, 994-998, Kiarash, A, et al., Mar. 21, 2008. |
Office Action received for Korean patent application No. 10-2011-7005021 mailed on Jun. 27, 2012, 8 pages of Office Action including 5 pages of English translation. |
Office Action received for Japanese patent application No. 2011-525080 mailed on Sep. 18, 2012, 9 pages of Office Action including 5 pages of English translation. |
Office Action received for European patent application No. 09811943.1 mailed on Apr. 11, 2011, 2 pages. |
Office Action received for Chinese patent application No. 200910173609.8 mailed on Feb. 29, 2012, 8 pages of Office Action including 3 pages of English translation. |
Office Action Received for Application Serial No. 200910173609.8 on Sep. 2, 2009, 13 pages (8 pages of Translation), 13 pgs. |
International Preliminary Report on Patentability received for PCT/US2009/054136 on Mar. 17, 2011, 8 pages. |
International Search Report and Written Opinion received Dec. 30, 2009, for PCT Application Serial No. PCT/US2009/054136, 11 pages. |
Office Action received for Korean Patent Application 10-2011-7005021, mailed on Dec. 21, 2012, 9 pages of Office Action including 4 pages of English translation. |
Office Action received for Taiwan Patent Application 98128665, mailed on Jan. 10, 2013, 6 pages of Office Action including 4 pages of English translation. |
Office Action received for Japan Patent Application No. 2011-525080, mailed on Mar. 12, 2013, 4 pages of Office Action, including 2 pages of English translation. |
Office Action received for Korean Patent Application No. 10-2011-7005021, mailed on May 20, 2013, 5 pages of Office Action including 2 pages of English translation. |
Office Action Received for Taiwan Patent Application No. 98128665, mailed on Dec. 19, 2013, 6 pages of Office Action including 1 page of Search Report. |
De Francisco, et al.; Beaforming for Correlated Broadcast Channels with Quantized Channel State Information; Signal Processing Advances in Wireless Communications; Jul. 2008; pp. 161-165; SPAWC 2008 IEEE 9TH Workshop on Signal Processing Advances in Wireless Communications 2008; IEEE. |
Mondal et al.; Channel Adaptive Quantization for Limited Feedback MIMO Beaforming Systems; IEEE Transactions on Signal Processing; Dec. 2006; pp. 4717-4729; vol. 54, No. 17; IEEE. |
Samanta et al.; Codebook Adaptation for Quantized MIMO Beaforming Systems; Sigmnals, Systems and Computers; Oct. 2005; pp. 376-380; Conference Record of the Thirty-Ninth Asilomar Conference on Signals, Systems and Computers 2005; IEEE. |
EP application 09811943.1; filed May 12, 2011; Intel Corporation; European Search Report datd Oct. 28, 2014. |
Number | Date | Country | |
---|---|---|---|
20100056216 A1 | Mar 2010 | US |
Number | Date | Country | |
---|---|---|---|
61093699 | Sep 2008 | US |