Beamforming mimo pdf

On-demand beamforming is typically performed between two devices e. The beacon portion includes a quasi-omni section and a directional section.

The quasi-omni section may comprise a plurality of identical quasi-omni Q-omni sub-beacons S-beacons , also referred to as transmission segments, covering different and possibly overlapping geographical areas around the piconet controller.

Each Q-omni S-beacon is transmitted using a different Q-omni beamforming pattern selected from a Q-omni codebook. One Q-omni beamforming vector is used per Q-omni sub-beacon transmission. The listening period also comprises a plurality of receiving segments.

During the l th sub-CAP, the piconet controller is in receiving mode, and it employs the same Q-omni beamformer vector used for transmission during the l th Q-Omni beacon. The quasi-omni transmissions convey information about the structure of the directional training sections, and the directional training sections enable channel state information CSI acquisition and tracking.

The directional section comprises a plurality of repetitions of a training sequence which may also be referred to as transmission segments , where each repetition is transmitted by the piconet controller with a different orthogonal or quasi-orthogonal beamforming vector from an orthogonal or quasi-orthogonal codebook. In another embodiment of the invention, a subscriber device in a piconet is configured for selecting beamforming and combining weights.

The subscriber device and the piconet controller both comprise an antenna array. The subscriber device receives a signal comprising a plurality of transmission segments transmitted by the piconet controller. Each of the plurality of transmission segments is transmitted with a different beam pattern from a predetermined beamforming codebook. The subscriber device receives at least a subset of the plurality of transmission segments and estimates a preferred beamforming vector therefrom.

The subscriber device also estimates a preferred combining vector for processing it receives. At least the preferred beamforming vector is sent back to the piconet controller during a listening period. The applicant recognizes that the frame formats and methods described with respect to the subscriber device communicating with the piconet controller may also be employed by the piconet controller communicating with one or more subscriber devices.

In a further embodiment of the invention, a quasi-omni acquisition signaling protocol comprises a first transceiver transmitting a number L of quasi-omni packets followed by L listening periods ACKs until it receives an ACK in one of the L listening periods e.

The first transceiver selects the l th Q-omni direction for transmission from the Q-omni codebook. The second transceiver records its best Q-omni receiving direction and uses it for any future Q-omni reception.

Embodiments of the invention may also provide for a frame format for directional training employing periodic transmissions from the first transceiver to the second transceiver. For example, one cycle of directional training sequences transmitted by the first transceiver may correspond to all J orthogonal quasi-orthogonal beamforming vectors from a subset of the selected codebook.

Each cycle is followed by a listening period ACK to listen to any feedback from the second transceiver. The second transceiver estimates w 1 and c 2 and couples at least the w 1 estimate to the first transceiver during the listening ACK period. These estimates may be updated during a subsequent tracking step. Furthermore, this procedure may be performed for uplink e. Although particular embodiments are described herein, many variations and permutations of these embodiments fall within the scope and spirit of the invention.

Although some benefits and advantages of the preferred embodiments are mentioned, the scope of the invention is not intended to be limited to particular benefits, uses, or objectives. Rather, embodiments of the invention are intended to be broadly applicable to different wireless technologies, system configurations, networks, and transmission protocols, some of which are illustrated by way of example in the figures and in the following description of the preferred embodiments.

The detailed description and drawings are merely illustrative of the invention rather than limiting, the scope of the invention being defined by the appended claims and equivalents thereof. Embodiments according to the present invention are understood with reference to the following figures.

Accordingly, while the embodiments of the present disclosure are susceptible to various modifications and alternative forms, specific exemplary embodiments thereof are shown by way of example in the drawings and will herein be described in detail. It should be understood, however, that there is no intent to limit the invention to the particular forms disclosed, but on the contrary, the invention is to cover all modifications, equivalents, and alternatives falling within the spirit and scope of the invention.

Like numbers may refer to like elements throughout the description of the figures. For example, two blocks shown in succession may in fact be executed substantially concurrently or the blocks may sometimes be executed in the reverse order, depending upon the functionality and procedures involved. A transceiver that employs the same antenna s for both transmission and reception is referred to as a Symmetric Antenna System SAS.

A transceiver that employs one set of antennas for transmission and another set of antennas for reception such as shown in FIG. A first transceiver employs M T transmit antennas and M R receive antennas.

A second transceiver employs N T transmit antennas and N R receive antennas. The channel models may be used to express any of the possible antenna configurations that may be employed in the related art. Furthermore, the channel models may be used to express different transmission protocols. In such cases, it is typical to assume that the cyclic prefix is longer than any multipath delay spread between any transmit-receive pair of antenna elements.

The discrete channel model between the first transceiver's transmitter and the second transceiver's receiver is expressed by a Single Input Single Output SISO channel,. Embodiments of the invention may provide for one or more beamforming algorithms configured to select the beamforming vectors w 1 and w 2 and the combining vectors c 1 and c 2 that maximize at least one signal-quality parameter, such as the ESNR. In the general AAS case, the first transceiver may transmit known information to the second transceiver , which then derives matrices characterizing the channel state information.

This enables estimates of w 1 and c 2 to be calculated. The second transceiver may transmit known information to the first transceiver to provide channel state information that allows for estimates of w 2 and c 1 to be calculated. Some embodiments of the invention may employ known data symbols, pilot signals, or other training information to be transmitted for acquiring channel state information.

Alternative embodiments may employ blind adaptive processing or other techniques utilizing unknown transmitted data to derive channel state information. In AAS, both directions of the link are used to estimate the vectors w 1 , w 2 , c 2 , and c 1. For SAS, the beamforming vectors w 1 and w 2 and the combining vectors c 2 and c 1 in a particular direction should be equal.

For example, one transceiver may be a piconet controller and the other transceiver may be a piconet subscriber device. A channel state information CSI acquisition step enables the second transceiver to acquire a first CSI matrix, which is used to estimate the first transceiver's optimal or preferred beamforming vector w 1 and the second transceiver's optimal or preferred combining vector c 2.

The CSI acquisition step may comprise configuring the first transceiver to transmit a subset of a beamforming codebook Furthermore, the second transceiver may be configured to employ a subset of a combining codebook to acquire the first CSI matrix.

An estimation step comprises producing the optimal beamforming vector w 1 and the optimal combining vector c 2. Other constraints may apply. A feedback step provides for sending the optimal beamforming vector w 1 and, optionally, the optimal combining vector c 2 to the first transceiver Thus, an optimal beamforming vector w 2 and an optimal combining vector c 1 are estimated.

A tracking step provides for tracking the beamforming and combining vectors. The tracking step is similar to the acquisition step, except that the first transceiver transmits a subset of the beamforming codebook at a rate that is lower than the rate employed during acquisition Similarly, lower-rate updates are made to the optimal beamforming vector w 1 and the optimal combining vector c 2 , and the values w 1 and c 2 are fed back to the first transceiver Thus, the estimates for the optimal beamforming vector w 2 and the optimal combining vector c 1 are updated.

For a uniformly spaced linear antenna array with N elements, the array factor is defined by. The antenna array directivity is given by. The array factor of a two-dimensional array is given by.

The array factor of a two-dimensional array that is separable into one-dimensional x-axis and y-axis array components is expressed as. In FIG. These beam patterns are characterized by a codebook of vectors given by columns of the following weight matrix.

Orthogonal codebook vectors are given by columns of the following weight matrix. In an alternative embodiment, the following weight matrix may be employed. The beam patterns correspond to codebook vectors given by columns of the weight matrix. The beam patterns correspond to codebook vectors given by columns of the following weight matrix.

The codebook for this case comprises the set of vectors given by the columns of the following weight matrix W. The codebook comprises the set of vectors given by the columns of the following weight matrix.

The codebook comprises the set of four orthogonal vectors given by the columns of the following weight matrix. Beam patterns corresponding to an extended codebook for a four-element array are shown in FIG. The codebook comprises the set of eight vectors given by the columns of the following weight matrix.

The codebook comprises the set of six vectors given by the columns of the following weight matrix. The codebook comprises the set of eight orthogonal vectors given by the columns of the following weight matrix. Each codebook may contain multiple options for quasi-omni patterns. Depending on the polar and azimuthal antenna gain patterns, one or more quasi-omni patterns may be used.

Depending on the polar and azimuthal antenna gain pattern, one or both complementary patterns can be used. A quasi-omni codebook for a two-element array comprises two orthogonal beamforming or combining vectors given by the columns of the weight matrix.

The codebook vectors are the columns of the following weighting matrix. The codebook comprises three quasi-orthogonal beam-forming vectors, which are columns of the following weighting matrix. The first pattern has a HPBW of The second pattern has a HPBW of The third pattern has a HPBW of A quasi-omni codebook for an eight-element array comprises two orthogonal beamforming or combining vectors given by the columns of the weight matrix. A quasi-omni codebook for a two-element array employing quadrature weights is identical to the case in which binary weights are employed.

Thus, the codebook comprises two orthogonal beamforming or combining vectors given by the columns of the weight matrix. The codebook comprises the pair of quadrature vectors derived from the pair of columns in the following weight matrix. One pattern has a HPBW of Both patterns have a maximum directivity of 2. The resulting quasi-omni codebook for this case is given by vectors represented by the columns of the following weight matrix.

The quasi-omni beam patterns shown in FIGS. The vectors are components of a quasi-omni codebook and include the columns of the following weight matrix. The maximum directivity of the first two patterns is 1. An eight-element array can employ a codebook comprising a pair of quadrature weight vectors expressed by columns of the following weight matrix to produce two complementary beam patterns shown in FIG. One embodiment of the invention provides for a unified messaging protocol that is independent of antenna configuration and estimation algorithms employed for beamforming i.

The messaging protocol may be configured to support a variety of antenna configurations used for transmitting and receiving. Such configurations may include beamforming antenna arrays, such as phased arrays.

Antenna configurations may include sectored and switched antenna arrays. Antenna configuration may be defined by a variety of beam patterns, including omni, quasi-omni, or directional single antenna. The messaging protocol also supports SAS and AAS configurations, and it may be configured to support proactive and on-demand beamforming.

In one embodiment, the messaging protocol is further configured to support a variety of link models, including but not limited to per-packet beamforming between a piconet controller and multiple subscriber devices, a link between a piconet controller and a single subscriber device, and peer-to-peer links between subscriber devices. Embodiments of the invention may provide for a plurality of beamforming protocols to be implemented, including a proactive beamforming protocol and an on-demand beamforming protocol.

Proactive beamforming is performed using the beacon portion of a superframe such as shown in FIG. For example, the piconet controller may be a Kiosk, STB, or Laptop computer, and it is configured to send each packet in at least one of a plurality of different directions to the destination device. On-demand beamforming may be employed for transmissions between two subscriber devices or between a piconet controller and one subscriber device. In both beamforming protocols, quasi-omni transmissions convey information about the structure of the directional training sections, and the directional training sections enable CSI acquisition and tracking.

The piconet controller beacon comprises at least one quasi-omni Q-omni section and at least one directional section. In one embodiment, the Q-omni section has L identical Q-omni sub-beacons S-beacons covering different and possibly overlapping geographical areas around the piconet controller. The aggregated coverage of the Q-omni S-beacons covers the target space around the piconet controller.

In the embodiment shown in FIG. In this case, the directional training segments are sent back-to-back, except for a small guard interval. However, alternative embodiments of the invention may provide for interleaving or otherwise positioning Q-omni training segments within the directional section.

For example, a Q-omni training segment may have the same format as a directional training segment, but it is sent omni-directionally. In one embodiment, a Q-omni training segment follows each directional training segment.

A subscriber device uses the Q-omni training segments to help compensate for timing and frequency drift, as such compensation is necessary for generating accurate estimates of the CSI. During the l th S-CAP, the piconet controller is in a receiving mode using the same Q-omni beamformer vector it used for transmission during the l th Q-Omni beacon. The number L of Q-omni S-beacons may be reduced in order to reduce overhead.

For SAA, L is the number of sector or switched antennas. In beamforming or phased-array configurations, L equals 1 or 2, but it may be more. During transmission, the piconet controller employs L Quasi-omni beamforming vectors from a corresponding quasi-omni codebook, and one Q-omni beamforming vector is used per Q-omni sub-beacon transmission.

The directional section of the frame shown in FIG. The sequences denoted by vectors u , v , and s are described in Provisional Application Ser. The training sequence includes N s repetitions where N s may be zero of a sync sequence s followed by N c e.

Each training sequence transmitted by a piconet controller employs a different orthogonal or quasi-orthogonal beam pattern selected from an orthogonal or quasi-orthogonal codebook. In the case where OFDM and single-carrier employ short training sequences, the two sync sequence may be used for automatic gain control AGC.

A matched-filter receiver may align and add the outputs of a Golay matched filter configured to correlate the received signal with the modified Golay sequences u and v to produce a perfect channel estimate. Also, the receiver may produce a difference signal, which provides a perfect noise estimate. For example, the first training sequence is sent in the direction corresponding to beamforming matrix W xy,11 , the second training sequence is sent in the direction corresponding to beamforming matrix W xy,12 , etc.

In some embodiments, the piconet controller may be configured to employ only a subset of the available beamforming matrices. For example, the piconet controller may transmit over a restricted angular range e. The piconet controller employs a directional codebook, which is the subset of possible beamforming matrices that the piconet controller could use to train the subscriber devices.

If the directional codebook is of size J, a transmission of J training sequences in the corresponding J directions is referred to as a cycle. In some embodiments, the L identical Q-omni S-beacons may include indices of the codebook vectors selected by the piconet controller. A subscriber device is configured to listen to the Q-omni transmissions from a piconet controller. Upon detection, the subscriber device decodes the content of the Q-omni S-beacon to obtain the structure of the directional section.

The subscriber device selects a first codebook vector to steer its antenna to a first direction. The subscriber device selects a second codebook vector to steer its antenna to a second direction, and it may repeat this procedure for each codebook vector. Alternatively, the subscriber device may select codebook vectors from a subset of the codebook. The subscriber device calculates the CSI matrix, H, from which it estimates the optimal beamforming weights for the piconet controller and the optimal combining vector for itself.

In the SAS case, the subscriber device may listen to the Q-omni transmissions until it determines a combination of weights that it deems to provide adequate link quality. The resulting beamforming weights are transmitted back to the piconet controller.

The antenna type may include information about the piconet controller antenna e. The fields of the beamforming information element may be adapted for different antenna configurations.

For example, the training sequence information and the antenna array information may be omitted for a single antenna configuration or a switched antenna array. This information may include the number N x of antennas along the x-axis, ID of the codebook used along the x-axis, the number N y of antennas along the y-axis, and ID of the codebook used along the y-axis in the case of a two-dimensional array.

Some embodiments may include the size J x and ID of the subset of the beamforming vectors to be used along the x-axis, and the size J y and ID of the subset of the beamforming vectors to be used along the y-axis. A piconet controller transmits a number L of Q-omni S-beacons and a number N of directional training sequences per superframe. A subscriber device listens to and decodes at least one of the Q-omni S-beacons, from which it acquires information related to the directional section. In one embodiment of the invention, the subscriber device may listen to the entire set of Q-omni S-beacons.

The subscriber device selects an appropriate subset of an orthogonal or quasi-orthogonal codebook and begins a scanning procedure using the selected combining vectors. In one embodiment, when the subscriber device steers to a particular direction using a vector from the codebook and listens to a transmitted cycle, it may store a Link quality Factor LQF.

This process is repeated until the subscriber device finds an l th LQF above a predetermined threshold, or until it has finished listening to all the codebook vectors and acquires the CSI matrix.

The subscriber device estimates its optimal combining vector c 2 and an optimal beamforming vector w 1 for the piconet controller. The estimated optimal beamforming vector w 1 and optionally, the optimal combiner vector c 2 are fed back to the piconet controller during the l th S-CAP. The subscriber device may track the beamforming and combining vectors by periodically scanning the beacon. The subscriber device may periodically feed back any updates to w and c.

In the AAS case, steps - are bypassed. Instead, on-demand beamforming may be used during the CTAP allocated to the communication link in order to complete bi-directional beamforming. As shown in FIG. The update rate tracking rate depends on the maximum Doppler that the system can tolerate. On-demand beamforming which is performed between two subscriber devices or between a piconet controller and one subscriber device employs the Channel Time Allocation CTA part of the superframe shown in FIG.

Antenna information is exchanged during association such that each transceiver knows the antenna array processing capabilities e. Quasi-omni acquisition is performed as a first transceiver transmits a Q-omni S-beacon to a second transceiver. Directional acquisition is performed as the first transceiver transmits a directional training sequence to the second transceiver.

Once acquisition is complete, the system may perform tracking as the first transceiver transmits data to the second transceiver. Antenna information is exchanged during association. Quasi-omni acquisition is performed as the second transceiver transmits a Q-omni S-beacon to the first transceiver.

Similarly, directional acquisition is performed as the second transceiver transmits a directional training sequence to the first transceiver. The first transceiver continues repeating this structure until it receives an ACK in one of the L listening periods e. From this point on, the first transceiver selects the l th Q-omni direction for transmission which is also the reception direction in the SAS case from the Q-omni codebook.

The second transceiver records its best Q-omni receiving direction which is also the best transmitting direction in SAS case and uses it for any future Q-omni reception. In the AAS case, the process is repeated as the second transceiver transmits to the first transceiver.

A directional acquisition period may comprise an optional Q-omni packet followed by a listening period. One cycle of directional training sequences transmitted by the first transceiver corresponds to all J orthogonal quasi-orthogonal beamforming vectors from a subset of the selected codebook.

The second transceiver estimates preferred w 1 and c 2 and couples at least the w 1 estimate to the first transceiver during the listening ACK period. The preferred beamforming vector w 1 may be one of the transceiver's codebook vectors, or it may be a linear combination of the codebook vectors, or it may be unrelated to the transceiver's codebook vectors. In one embodiment employing a four-element array with a codebook represented by the following weight matrix.

If quadrature weights are employed i. Another embodiment may employ weights comprising a range of amplitudes and phases. The transceivers will employ the w 1 and c 2 estimates for beamforming and combining during data communications. Either or both transceivers may update the estimates w 1 and c 2 when the first transceiver transmits a directional acquisition period.

The first transceiver estimates w 2 and c 1 and couples at least the w 2 estimate to the second transceiver. The estimates w 1 and c 2 are employed as beamforming and combining weights during data transmission, and these estimates w 1 and c 2 may be updated periodically when the second transceiver transmits a directional acquisition period.

In [7] we introduced a heuristic to C. The combination of the channel h In this section we drop the user index k for simplicity. The non-separable prior covariance should A. However, to simplify beamformer design, we shall force carrier of a given OFDM symbol can be written as [8], [9] a separable prior model by considering all of B to be unknown Np with elements that are i. In the case of distributed antenna systems, the array responses become a function of all position parameters In this paper we shall consider mostly a per stream approach of the path scatterers.

In an IBC formulation, one stream per user can be nominal path represents in fact a superposition of paths with expected to be the usual scenario.

In the development below, in similar parameters correspond to the fast fading. All the other the case of more than one stream per user, treat each stream as parameters vary on a slower time scale and correspond to slow an individual user. So, consider again an IBC with C cells with fading.

We shall consider a system-wide numbering of the users. User k is served by BS bk. Averaging of the path k. More specifically, consider the dependence of 1 WSR on Qk alone. Since a linear function is simultaneously convex and Rk , Rk are the total and interference plus noise Rx covariance concave, consider the first order Taylor series expansion in Qk 0 0 0 0 matrices resp.

Now, dropping constant A. If we multipliers could be zero. Note that given the f and g or an arbitrary number of streams per user by determining mul- in other words the euk , the WSUMSE decomposes into the w1 tiple dominant generalized eigenvectors, and then let the WF and w2 dependencies, which can be optimized in parallel.

The operation decide how many streams can actually be sustained. Now, as long V. Tx signal xk , we require VI. Now we consider various deterministic approxima- u u tions for the EWSR. Hence consider optimizing the expected weighted sum the term for user k.

Alternating plus noise. Negro, S. Prasad Shenoy, I. Ghauri, and D. In [12] S. Kim and G. Theory, May Negro, I. Negro, D. Slock, and I. Lejosne, D. Slock, and Y. Wireless Days, Valencia, Spain, Nov.