Patent Application
20170222702
The invention relates to MIMO (multiple input, multiple output) systems and methods.
In MIMO (multiple input multiple output) OFDM (orthogonal frequency division multiplexing) systems, there are multiple transmit antennas and multiple receive antennas and a plurality of sub-carriers that are available for transmission between the transmit antennas and the receive antennas for either one or multiple users. New advances in MIMO OFDM systems are taught in Applicant's co-pending application <attorney docket 71493-1320> entitled “Pilot Design For OFDM Systems With Four Transmit Antennas” filed Mar. 15, 2005, and in Applicant's co-pending application <attorney docket 71493-1330) entitled “Wireless Communication Methods, Systems, And signal Structures” filed Apr. 4, 2005, both hereby incorporated by reference in their entirety. With open loop implementations, the transmitter transmits on the multiple transmitter antennas and sub-carriers without the benefit of channel information fed back from the receivers.
Efforts have been made to facilitate wireless closed-loop MIMO communications including broadband closed-loop MIMO, which might for example be based on OFDM modulation schemes, and narrowband closed-loop MIMO. Broadband closed-loop MIMO includes many sub-bands. Each of these sub-bands requires MIMO channel feedback for a closed-loop implementation. As a result the feedback resources required for broadband closed-loop MIMO can become quite large. Narrowband closed-loop MIMO, by comparison, includes one or a few sub-bands and requires a relatively smaller amount of feedback resources. Broadband and narrowband MIMO, therefore, have different applications.
According to one broad aspect, the invention provides a MIMO system comprising: a transmitter having multiple transmit antennas; at least one receiver, each receiver having at least one receive antenna; each receiver being adapted to transmit at least one type of feedback information selected from a group consisting of: information for use in performing beam-forming; antenna selection/grouping information.
In some embodiments, a transmission format to each receiver is selected from a group of transmission formats consisting of: spatial multiplexing; vector weighted spatial multiplexing; matrix weighted spatial multiplexing; K-stream spatial multiplexing employing more than K transmit antennas; single stream STTD; single stream STTD with proportional weighting and antenna selections multi-stream STTD; multi-stream STTD with layer weighting; multi-stream STTD with a combination of layer weighting and proportional weighting; and hybrid beam-forming and spatial multiplexing.
In some embodiments, a defined sub-set of available formats is made available for a given receiver, and wherein the given receiver feeds back a selection of one of the defined sub-set of available formats.
In some embodiments, each receiver performs respective channel measurements and feeds back information for use in performing beam-forming based on the respective channel measurements.
In some embodiments, the information for use in performing beam-forming is selected from a group consisting of: a) elements of a measured channel matrix; b) elements of a V matrix of a SVD decomposed channel matrix; p) parameters of a Givens decomposition of a V matrix of a SVD decomposed channel matrix; d) parameters of a truncated Givens decomposition of a V matrix of a SVD decomposed channel matrix, where one or more eigen-vectors are discarded; e) differentially encoded elements of a measured channel matrix; f) differentially encoded elements of a V matrix of a SVD decomposed channel matrix; g) differentially encoded parameters of a Givens decomposition or truncated Givens decomposition of a V matrix of a SVD decomposed channel matrix, h) Householder decomposition; i) full soalar quantization of any of the information types of a) through h); j) partial scalar quantization of any of the information types a) through g); k) scalar quantization of any one of the information types a) through h) where varying resolution is used to quantize parameters; l) vector quantization of any of the information types of a) through h); m) a combination of scalar quantization and differential quantization for any of the information types a) through h); n) using a Delta Sigma quantizer for any of the information types a) through h); o) binary beam-forming weights; p) a differential index into a set of vector quantizations; and q) pre-defined codebook.
In some embodiments, beam-forming feedback is performed by each receiver as a function of receiver specific criteria.
In some embodiments, the receiver specific criteria is selected from a group consisting of: Max SNR; b) Max Shannon capacity; and c) True receiver operational process.
In some embodiments, antenna selection/grouping information is at lease one information type selected from a group consisting of: a) selection between SM (spatial multiplexing) and STTD (space time transmit diversity) transmission format; b) selection of particular antennas for SM transmission; c) selection and grouping of particular antennas for STTD transmission; and d) eigen-mode selection information.
In some embodiments, the system further comprises the receiver determining the antenna selection/grouping information by performing a step selected from a group of steps consisting of: performing SVD decomposition and discarding weak eigen-modes; selecting antennas using determinants of sub-MIMO channel matrices.
In some embodiments, feed back and beam-forming and/or antenna selection/grouping is performed for sub-carriers of a multi-carrier system to a resolution selected from a group consisting of: a) for every sub-carrier individually; b) for groups of consecutive sub-carriers; c) for an entire set of sub-carriers; d) for sets of groups of sub-carriers.
In some embodiments, transmission matrices and feedback are in accordance with one of
In some embodiments, the transmitter transmits pilots on each transmit antenna for use in performing channel estimation.
In some embodiments, at least some of the pilots are punctured pilots.
In some embodiments, at least some of the pilots comprise un-coded pilots for use by multiple receivers.
In some embodiments, the pilots comprise user specific pre-coded pilots for use by particular receivers receivers.
In some embodiments, the pilots comprise user specific pre-coded pilots for use by particular receivers receivers and un-coded pilots for use by multiple receivers.
In some embodiments, the pilot patterns are as shown in any one of
In some embodiments, the pilot patterns are as shown in one of
In some embodiments, feedback information is transmitted using a feedback channel having the structure of one of
In some embodiments, at least one receiver has a plurality of receive antennas.
In some embodiments, the at least one receiver comprises a plurality of receivers.
In some embodiments, sub-channels are defined using at least one of: AMC sub-channels, where respective adaptive modulation and coding is defined for each AMC sub-channel; PUSC sub-channels.
In another embodiment, a receiver is provided that is adapted to implement receiver functionality as summarized above.
In another embodiment, a transmitter is provided that is adapted to implement transmitter functionality as summarized above.
Other aspects and features of the present invention will become apparent to those ordinarily skilles in the art upon review of the following description of specific embodiments of the invention in conjunction with the accompanying figures.
Preferred embodiments of the invention will now be described with reference to the attached drawings in which:
The following provides a glossary of some of the terms used in this application:
For purposes of providing context for the embodiments described below, an example OFDM system will now be described with reference to
An example of a high level overview of the mobile terminals 16 and base stations 14 that may be used with embodiments of the present invention is provided prior to delving into the structural and functional details of the preferred embodiments. With reference to
The baseband processor 22 processes the digitized received signal to extract the information or data bits conveyed in the received signal. This processing typically comprises demodulation, decoding, and error correction operations. As such, the baseband processor 22 is generally implemented in one or more digital signal processors (DSPs) or application-specific integrated circuits (ASICs). The received information is then sent across a wireless network via the network interface 30 or transmitted to another mobile terminal 16 serviced by the base station 14.
On the transmit side, the baseband processor 22 receives digitized data, which may represent voice, data, or control information, from the network interface 30 under the control of control system 20, and encodes the data for transmission. The encoded data is output to the transmit circuitry 24, where it is modulated by a carrier signal having a desired transmit frequency or frequencies. A power amplifier (not shown) will amplify the modulated carrier signal to a level appropriate for transmission, and deliver the modulated carrier signal to the antennas 28 through a matching network (not shown). Modulation and processing details are described in greater detail below.
With reference to
The baseband processor 34 processes the digitized received signal to extract the information or data bits conveyed in the received signal. This processing typically comprises demodulation, decoding, and error correction operations, as will be discussed on greater detail below. The baseband processor 34 is generally implemented in one or more digital signal processors (DSPs) and application specific integrated circuits (ASICs).
For transmission, the baseband processor 34 receives digitized data, which may represent voice, data, or control information, from the control system 32, which it encodes for transmission. The encoded data is output to the transmit circuitry 36, where it is used by a modulator to modulate a carrier signal that is at a desired transmit frequency or frequencies. A power amplifier (not shown) will amplify the modulated carrier signal to a level appropriate for transmission, and deliver the modulated carrier signal to the antennas 40 through a matching network (not shown). Various modulation and processing techniques available to those skilled in the art are applicable to the present invention.
In OFDM modulation, the transmission band is divided into multiple, orthogonal carrier waves. Each carrier wave is modulated according to the digital data to be transmitted. Because OFDM divides the transmission band into multiple carriers, the bandwidth per carrier decreases and the modulation time per carrier increases. Since the multiple carriers are transmitted in parallel, the transmission rate for the digital data, or symbols, on any given carrier is lower than when a single carrier is used.
OFDM modulation typically employs the performance of an Inverse Fast Fourier Transform (IFFT) on the information to be transmitted. For demodulation, a Fast Fourier Transform (FFIT) is typically performed on the received signal to recover the transmitted information. In practice, the IFFT and FFT are provided by digital signal processing carrying out an Inverse Discrete Fourier Transform (IDFT) and Discrete Fourier Transform (DFT), respectively. Accordingly, the characterizing feature of OFDM modulation is that orthogonal carrier wave are generated for multiple bands within a transmission channel. The modulated signals are digital signals having a relatively low transmission rate and capable of staying within their respective bands. The individual carrier waves are not modulated directly by the digital signals. Instead, all carrier waves are modulated at once by IFFT processing.
In the preferred embodiment, OFDM is used for at least the downlink transmission from the base stations 14 to the mobile terminals 16. Each base station 14 is equipped with N transmit antennas 28, and each mobile terminal 16 is equipped with M receive antennas 40. Notably, the respective antennas can be used for reception and transmission using appropriate duplexers or switches.
With reference to
The scheduled data 44, which is a stream of bits, is scrambled in a manner reducing the peak-to-average power ratio associated with the data using data scrambling logic 46. A cyclic redundancy check (CRC) for the scrambled data is determined and appended to the scrambled data using CRC adding logic 48. Next, channel coding is performed using channel encoder logic 50 to effectively add redundancy to the data to facilitate recovery and error correction at the mobile terminal 16. The channel coding for a particular mobile terminal 16 may be based on the CQI. The channel encoder logic 50 uses known Turbo encoding techniques in one embodiment. The encoded data is then processed by rate matching logic 52 to compensate for the data expansion associated with encoding.
Bit interleaver logic 54 systematically reorders the bits in the encoded data to minimize the loss of consecutive data bits. The resultant data bits are systematically mapped into corresponding symbols depending on the chosen baseband modulation by mapping logic 56. Preferably, Quadrature Amplitude Modulation (QAM) or Quadrature Phase Shift Key (QPSK) modulation is used. The degree of modulation is preferably chosen based on the CQI for the particular mobile terminal. The symbols may be systematically reordered to further bolster the immunity of the transmitted signal to periodic data loss caused by frequency selective fading using symbol interleaver logic 58.
At this point, groups of bits have been mapped into symbols representing locations in an amplitude and phase constellation. When spatial diversity is desired, blocks of symbols are then processed by space-time block code (STC) encoder logic 60, which modifies the symbols in a fashion making the transmitted signals more resistant to interference and more readily decoded at a mobile terminal 16. The STC encoder logic 60 will process the incoming symbols and provide N outputs corresponding to the number of transmit antennae 28 for the base station 14. The control system 20 and/or baseband processor 22 will provide a mapping control signal to control STC encoding. At this point, assume the symbols for the N outputs are representative of the data to be transmitted and capable of being recovered by the mobile terminal 16. See A. F. Naguib, N. Seshadri, and A. R. Calderbank, “Applications of space-time codes and interference suppression for high capacity and high data rate wireless systems,” Thirty-Second Asilomar Conference on Signals, Systems & Computers, Volume 2, pp. 1803-1810, 1998, which is incorporated herein by reference in its entirety.
For the present example, assume the base station 14 has two antennas 28 (N=2) and the STC encoder logic 60 provides two output streams of symbols. Accordingly, each of the symbol streams output by the STC encoder logic 60 is sent to a corresponding IFFT processor 62, illustrated separately for ease of understanding. Those skilled in the art will recognize that one or more processors may be used to provide such digital signal processing, alone or in combination with other processing described herein. The IFFT processors 62 will preferably operate on the respective symbols to provide an-inverse Fourier Transform. The output of the IFFT processors 62 provides symbols in the time domain. The time domain symbols are grouped into frames, which are associated with a prefix by like insertion logic 64. Each of the resultant signals is up-converted in the digital domain to an intermediate frequency and converted to an analog signal via the corresponding digital up-conversion (DUC) and digital-to-analog (D/A) conversion circuitry 66. The resultant (analog) signals are then simultaneously modulated at the desired RF frequency, amplified, and transmitted via the RF circuitry 68 and antennas 28. Notably, pilot signals known by the intended mobile terminal 16 are scattered among the sub-carriers. The mobile terminal 16, which is discussed in detail below, will use the pilot signals for channel estimation.
Reference is now made to
Initially, the digitized signal is provided to synchronization logic 76, which includes coarse synchronization logic 78, which buffers several OFDM symbols and calculates an auto-correlation between the two successive OFDM symbols. A resultant time index corresponding to the maximum of the correlation result determines a fine synchronization search window, which is used by fine synchronization logic 80 to determine a precise framing starting position based on the headers. The output of the fine synchronization logic 80 facilitates frame acquisition by frame alignment logic 84. Proper framing alignment is important so that subsequent FFT processing provides an accurate conversion from the time to the frequency domain. The fine synchronization algorithm is based on the correlation between the received pilot signals carried by the headers and a local copy of the known pilot data. Once frame alignment acquisition occurs, the prefix of the OFDM symbol is removed with prefix removal logic 86 and resultant samples are sent to frequency offset correction logic 88, which compensates for the system frequency offset caused by the unmatched local oscillators in the transmitter and the receiver. Preferably, the synchronization logic 76 includes frequency offset and clock estimation logic 82, which is based on the headers to help estimate such effects on the transmitted signal and provide those estimations to the correction logic 88 to properly process OFDM symbols.
At this point, the OFDM symbols in the time domain are ready for conversion to the frequency domain using FFT processing logic 90. The results are frequency domain symbols, which are sent to processing logic 92. The processing logic 92 extracts the scattered pilot signal using scattered pilot extraction logic 94, determines a channel estimate based on the extracted pilot signal using channel estimation logic 96, and provides channel responses for all sub-carriers using channel reconstruction logic 98. In order to determine a channel response for each of the sub-carriers, the pilot signal is multiple pilot symbols that are scattered among the data symbols throughout the OFDM sub-carriers in a known pattern in both time and frequency. Continuing with
The frequency domain symbols and channel reconstruction information, which are derived from the channel responses for each receive path are provided to an STC decoder 100, which provides STC decoding on both received paths to recover the transmitted symbols. The channel reconstruction information provides equalization information to the STC decoder 100 sufficient to remove the effects of the transmission channel when processing the respective frequency domain symbols
The recovered symbols are placed back in order using symbol de-interleaver logic 102, which corresponds to the symbol interleaver logic 58 of the transmitter. The de-interleaved symbols are then demodulated or de-mapped to a corresponding bitstream using de-mapping logic 104. The bits are then de-interleaved using bit de-interleaver logic 106, which corresponds to the bit interleaver logic 54 of the transmitter architecture. The de-interleaved bits are then processed by rate de-matching logic 108 and presented to channel decoder logic 110 to recover the initially scrambled data and the CRC checksum. Accordingly, CRC logic 112 removes the CRC checksum, checks the scrambled data in traditional fashion, and provides it to the de-scrambling logic 114 for de-scrambling using the known base station de-scrambling code to recover the originally transmitted data 116.
In parallel to recovering the data 116, a CQI, or at least information sufficient to create a CQI at the base station 14, is determined and transmitted to the base station 14. As noted above, the CQI in a preferred embodiment is a function of the carrier-to-interference ratio (CR), as well as the degree to which the channel response varies across the various sub-carriers in the OFDM frequency band. For this embodiment, the channel gain for each sub-carrier in the OFDM frequency band being used to transmit information are compared relative to one another to determine the degree to which the channel gain varies across the OFDM frequency band. Although numerous techniques are available to measure the degree of variation, one technique is to calculate the standard deviation of the channel gain for each sub-carrier throughout the OFDM frequency band being used to transmit data.
It is to be understood that the
The embodiments set forth below represent the necessary information to enable those skilled in the art to practice the invention and illustrate the best mode of practicing the invention. Upon reading the following description in light of the accompanying drawing figures, those skilled in the art will understand the concepts of the invention and will recognize applications of these concepts not particularly addressed herein. It should be understood that these concepts and applications fall within the scope of the disclosure and the accompanying claims. While the embodiments can be applied in a system such as exemplified in
Various Transmission formats are provided for closed loop MIMO air-interface designs. MIMO transmission formats and signalling apparatus are generalized to allow a variety MIMO schemes to operate using the same air-interface design. According to one aspect of the invention, basic transmission formats include: (1) spatial multiplexing (SM) and (2) space-time transmit diversity (STTD), with vector or matrix weighted full MIMO or sub-MIMO transmission based on various transmit antennas configurations. These formats can be used in single or multi-carrier configurations. The schemes can also be generalized to cover multiple base station transmission.
Some embodiments feature a receiver selecting between various available transmit formats. The particular formats to include in a given system are implementation specific.
BLAST is also called spatial multiplexing (SM). This transmission format allows maximum parallel transmission of data streams and can achieve a theoretical capacity limit.
For an open loop M×N MIMO SM transmission (M transmit antennas and N receive antennas), basic transmission matrices are presented below in Table 1 for up to four transmit antennas. It is readily apparent how similar transmission matrices can be constructed for numbers of transmit antennas larger than four.
For M transmit antenna SM, there can be N receiver antennas with N≧M. Where N<M for a given receiver, antenna selection can be performed to identify a subset of the M transmit antennas for use in BLAST transmission to the receiver. Mechanisms for performing antenna selection are described below.
For M×N MIMO vector weighted SM transmission according to an embodiment of the invention, the transmission matrices are presented below in Table 2 for up to four transmit antennas. It is readily apparent how vector weighted SM matrices can be extended to handle more than four transmit antennas. With vector weighted closed loop transmission, each symbol is transmitted on a single transmit antenna and a single weight is applied to each such symbol. Other users can then be supported using the remaining antenna(s).
Vector weighted SM allows, for example, per antenna power allocation and balancing, SINR optimization from the receiver, and joint transmit-receive closed loop optimization as a Weiner solution. For M transmit antenna vector weighted SM, there can be N receiver antennas with N≧M. Where N<M for a given receiver, antenna selection can be performed to identify a subset of the M transmit antennas for use in BLAST transmission to the receiver. Mechanisms for performing antenna selection are described below. Other users can then be supported using the remaining antenna(s)
M×N MIMO matrix weighted SM transmission according to an embodiment of the invention has associated transmission matrices as shown in Table 3, where 2, 3 and 4 transmit antenna matrices are shown. As in the previous examples, it is readily apparent how matrix weighted SM transmission matrices can be extended to handle more than four transmit antennas. With matrix weighted spatial multiplexing, each symbol is represented on multiple transmit antennas with a respective weight. However, depending upon the weight selected, a given symbol may not necessarily be represented on all transmit antennas. Different examples of this are presented below.
The matrix weighted M×N SM transmission can be applied to a single user reception application where N≧M or to the multi-user concurrent transmission cases, such as 2×2×1, 4×4×1, 4×2×2, for example. In the above examples, 2×2×1 for example, means two transmit antennas, two users, and each user having one receive antenna. Where N<M for a given receiver, antenna selection can be performed to identify a subset of the M transmit antennas for use in BLAST transmission to the receiver. Mechanisms for performing antenna selection are described below. Other users can then be supported using the remaining antenna(s).
Tables 4 to 6 and
The first configuration 100 is for a single user. The user input is indicated at 102, this consisting of two symbols x1, x2 at a given instant. These two inputs are then multiplied beam-forming weights in a beam-former 104 that implements beam-forming matrix G1 104. The output of the beam-former 104 is then fed to a pair of transmit antennas 106. In the event of a matrix weighting, the beam-forming matrix corresponds to that of matrix-A for Table 3 above. A similar structure exists in the event of vector weighting (not shown), in which case the beam-forming matrix would correspond to matrix-A of Table 2 described above.
For the second configuration, generally indicated at 110, there are two separate user inputs indicated at 112, 114 respectively. Each of the two user inputs are multiplied by a respective vector with the first input 112 being multiplied by V1 113, and the second input being multiplied by V2 115. The outputs of these vector multiplications are summed on a per element basis in adders 116,118. The output of the first adder 116 is fed to the first transmit antenna 120 and the output of the second adder 118 is fed to the second transmit antenna 122. The two vector multiplications V1 113, V2 115 collectively implement the beam-forming operation of matrix-A of Table 3. Given that it is a multi-user format, it also corresponds to the transmission format-2 of Table 4 above.
Table 5 below provides similar information to that described in Table 4 above, but for a three transmit antenna case. In other words, this is elaborating the details of transmit formats available for the matrix weighted SM matrix-B described above in Table 3. The options available with a three transmit case are to use all three streams for a single user as indicated at format-1 in the first row, to use one of the three symbol streams for a first user and the remaining two symbol streams for a second user as shown in format-2 in the second row, and to assign a different user to each of the three streams as shown in format-3 in the third row. In all cases, the transmit beam-forming matrix has the same structure. The feedback options are again listed in the right hand three columns of Table 5.
Similarly, the configurations available for a four transmit antenna configuration with matrix weighting are shown in Table 6 below. A single user format is shown in the first row; two user formats are shown in the second and fourth rows, a three user format is shown in the third row, and a four user format is shown in the fifth row. It is to be understood that not all of the potential formats have been exhaustively shown in Table 6. In all cases, the transmit beam-forming matrix has the same structure, and the feedback options for the different formats are shown in the right hand three columns of Table 6.
The second configuration shown in
The above described BLAST vector and weighted transmission configurations are generally applicable for closed loop MIMO systems with 2, 3 and 4 antennas. It should be readily apparent how these can be extended to handle a larger number of antennas.
In a multi-carrier environment, these transmission configurations can be applied either on a per-carrier basis, for blocks of contiguous carriers, sets of non- or partially-contiguous carriers, or for an entire transmit bandwidth. Thus, for the OFDM particular implementations, in some implementations, a different beam-forming matrix/user assignment can be applied on a per-carrier basis; in other implementations, the available OFDM bandwidth is divided into sub-bands, and the same beam-forming matrix/user assignment is used on each sub-band; and yet a further available option, the entire OFDM bandwidth is used to transmit with the same beam-forming configuration. In yet another option, the entire available OFDM bandwidth divided into small sub-channels, with each user being assigned a respective set of such sub-channels.
Spatial Multiplexing when there are More Transmit Antennas than Receive Antennas
When there are extra transmit antennas, antenna selection and signal pre-processing can be used to significantly enhance link level performances. The SNR relation is: γpp>γas>>γfa, where γpp, γas, and γfa are the SNRs of pre-processing, antenna selection, and fixed antenna configuration (using M fixed transmit antennas from a pool of N). The gain of signal pre-processing over antenna selection is mainly from the use of beam-forming. Details of how feedback can be performed for beam-forming are given below. The gain of antenna selection over fixed antenna configuration is from two areas: diversity gain and average SNR gain (over STC). Preferably, the water-filling principle (the principle of using the most energy where conditions are the best) is applied to both antenna selection and beam-forming. Antenna selection is the most simple and effective way to achieve diversity and SNR gain.
With antenna selection, the receiver analyzes conditions on the channel and selects antennas that are best for that receiver. This antenna selection information is then fed back to the transmitter for use in performing the actual transmissions.
A single stream selective transmit diversity technique is represented in Table 7 below. This example assumes that there are four transmit antennas available to be selected for a single receive antenna user. The four different transmit options are shown in the table to include: the user's symbol s1 being transmitted either using the first transmit antenna Tx1, or the second, third or fourth transmit antenna similarly labelled. In order for the transmitter to know which antenna to use, preferably a feedback signal from the receiver is employed. In the particular example shown, a two bit feedback signal is shown in the first column that can be used to identify one of four different possible transmit antennas. Other formats for the feedback information can be employed. When a single antenna is used, there is no spatial multiplexing.
Once again for a multi-carrier implementation, antenna selection can take place to any appropriate sub-carrier resolution. The particular feedback signalling is one particular example of how to signal back an identity of one of four antennas. An appropriate signalling mechanism can be employed. Furthermore, the antenna selection mechanism can be extended to differing numbers of transmit antennas.
An embodiment of the invention proposes closed loop sub-MIMO selection for SM. The term “sub-MIMO” is used to describe the case where there are multiple transmit antennas and a fewer number of receive antennas, and antenna selection is used to select a subset of the transmit antennas for use by the given receiver with the remaining antennas being available for other receivers. This will be described in the context of a 4×2 sub-MIMO system, in which the channel matrix to a given receiver has the following format:
Selecting two of the four antennas results in a smaller 2×2 channel matrix for the given receiver. There are six different possible channel matrices depending upon which of the two of the four transmit antennas are selected. The six possible channel matrices are as follows:
These six potential channel matrices can then be analyzed to determine which sub-MIMO channel is best, i.e., which of the two four transmit antennas are best for the particular user. Any appropriate way of selecting the two antennas can be implemented. In the particular example that follows, antenna selection is performed on the basis of determinants calculated for the different channel matrices. The particular pair of antennas that is best for a given receiver is determined by selecting the sub-system Hij that satisfies
|det(Hij)|=max{|det(H12)|,|det(H13)|,|det(H23)|,|det(H14)|,|det(H24)|,|det(H34)|}.
This is depicted in
More generally, in a K receive antenna system, K columns from channel matrix H can be selected, where K is the size of the subset antenna group. There are a total of CKM choices. For each possible selection, calculate det(H′H). Finally, select the antenna group with max(det(H′H)).
As in previous discussions, for multi-carrier applications, antenna selection can be used at whatever frequency resolution may be determined for a given application. The best antenna group may differ for different sub-carriers. If no other users are simultaneously being transmitted to, null sub-carriers are preferably fed into non-selected antennas. This antenna selection process is typically performed at the receiver. Once the best two transmit antennas have been selected at the receiver, the selection is fed back to the transmitter. An example of how this feedback can be performed is shown in Table 8 below. The transmit contents that are possible are shown for each of the four transmit antennas. It can be seen that there are six different permutations depending upon which two transmit antennas are selected. The first column is used to show how three bits can be used to transmit a selection of two of the six antennas. Other feedback formats are possible. In some implementations feed back is for every sub-carrier. If three bits are fed back on a per sub-carrier basis, then for L sub-carriers, L×3 bits would be required. In some implementations, feed back is for a sub-channel of contiguous sub-carriers. If three bits are fed back for a contiguous block of sub-carriers, under the assumption that channel conditions are substantially constant across the block, obviously fewer feedback bits are required.
In this case, the sub-MIMO channel selection is based on the sub-MIMO channel condition and the antenna power allocation for the SM data can be adjusted by the feedback information pertaining to the weights.
Antenna group index (illustratively 3 bits), and proportional weighting vector (illustratively 2 real elements (or alternatively complex elements), each element being of 4-bits, 6-bits, or 8-bits) are preferably provided as feedback signals. It should be appreciated, however, that weights may be derived from feedback information and need not necessarily be directly provided in feedback information.
Closed Loop Sub-MIMO Selection with Beam-Forming
For a DL 4×2 MIMO system, consider six sub-MIMO systems H1,2, H1,3, H1,4, N2,3, H2,4, and H3,4. Assuming that Hij, Hlk and Hil are the sub-MIMO systems that satisfy:
|det(Hij)|+|det(Hlk)|+|det(Hil)|=max{|det(Hij)|+|det(Hlk)|+|det(Hll)|,|det(Hij)|+|det(Hjk)|+|det(Hjl)}
then by beam-forming with the jth and kth columns of H, and setting the weights to
respectively, where the particular selection of the three determinants maximizes the denominator in the above-weight calculation we have
Table 9 is an illustrative example of this. A single antenna is used to transmit a stream without any weighting; another antenna is used to transmit the same stream with weighting; and the remaining two antennas are used to transmit the other second stream, both with weighting. The particular antenna used to transmit either the un-weighted symbol or the weighted symbols is information that is preferably fed back to the transmitter. The weighting vector or information from which the weights can be derived is also preferably fed back from the receiver to the transmitter. Thus, the illustrated example, this would require the equivalent of three bits to identify one of the six permutations, and additional feedback for the three weights w1, w2 and w3. This format is a hybrid of beam-forming and spatial multiplexing. In this example, there are two data streams mapped onto four antennas. Each symbol is transmitted on two different antennas with the weights as shown in the table below. Antenna selection is performed to select which antennas go together to transmit the same symbol, and beam forming is used to apply weights to three of the four antennas. Four weights could be applied alternatively. It should be readily apparent how this concept can be extended to handle larger numbers of antennas overall and/or larger numbers of antennas per stream and/or larger numbers of streams. This approach will be referred to as K-layer SM employing more than K transmit antennas. In the example, we have 2-layer SM with four transmit antennas.
STTD Transmission formats
The transmission formats described thus far have all been based on the BLAST approach in which there is a single transmitted symbol for each information symbol, or on the BLAST variant described by way of example with reference to Table 9 above in which each K-stream blast is employed on more than K antennas, so at least some of the symbols are transmitted twice. Additional transmission formats are based on STTD (space time transmit diversity) in which there are at least two transmitted symbols based on each information symbol, and typically sets of two information symbols are transmitted using 2×2 Alamouti block. However, other STTD formats are also contemplated.
Single Stream Weighted STTD (SSTD) for a Single User with Antenna Selection and Proportional Weighting
In another transmission format, for a single data steam, transmit diversity is “proportionally weighted” STTD. This is a middle approach between conventional STTD (no weighting) and antenna switching. One advantage of this scheme is to safeguard against possible feedback error, feedback delay, and possible channel change between feedback updates for the antenna selection. The channel matrix condition (CMC) does not affect STTD when orthogonal codes are used. For quasi-orthogonal codes, CMC does affect signal quality.
In the case of single stream transmit diversity, we also have the two options of beam-forming and antenna selection that can be performed similarly to that described above for the BLAST transmission formats. Beam-forming involves applying weights to the elements of the STTD groups, and antenna selection involves selecting a sub-set of a set of available transmit antennas for transmitting the stream.
For an SSTD system with four transmit antennas, a hybrid proportionally weighted STTD approach can be used. In one embodiment, the two antennas with the strongest combined SNR are selected, and proportional weighting on these two selected antennas is then applied. The advantages of this scheme include: diversity and SNR gain due to antenna selection, increased robustness provided by proportional weighting, and the employment of efficient Alamouti codes. An example of SSTD with antenna selection and proportional weighting is given in Table 10, where two antennas out of four are selected and weighted.
In this case, feedback signals preferably include antenna group index (illustratively 3 bits), proportional weighting vector or other information allowing the determination of the weighting vector (illustratively 2 real elements, each element might for example be of 4-bits, 6-bits, or 8-bits). It should be readily apparent how the single stream case can be extended to larger numbers of transmit antennas.
With double stream transmit diversity, two streams are transmitted to a single user with each stream being sent using a respective weighted Alamouti code structure (or some other STTD structure). More generally, with an M transmit antenna system, M/2 STTD sub-groups can be formed and transmitted to a different receiver, although multiple or all of the sub-groups might be sent to one receiver. This is a hybrid case of STTD and space-multiplexing (SM). Because of inter-code interference, antenna grouping becomes more important. Preferably, layer based water-filling is employed to provide additional gain to the system. With layer-based water-filling, more power is transmitted to the layer having the better channel. Thus, with a double STTD system, a respective pair of antennas is assigned to each stream, and the respective “layer” is transmitted on that pair of antennas. Preferably, more power is given to the layer having the better channel conditions.
An example of a set of possible layer weighting permutations is shown in Table 11 for the four transmit antenna case. In the first row, transmit antennas 1 and 2 are grouped together, and antenna 3 and 4 are grouped together, to form two constituent STTD codes that can be assigned to respective receivers or to a single receiver. In the example of Table 11 below, symbols s1 and s2 are of one layer, while symbols s3 and s4 are of a second layer.
The feedback signals in this case are used to select one of the three different permutations, and this can take the form of an antenna group index (2 bits in the example shown), and weights or information relating to weights. In this implementation, there are two STTD water-filling weights and each can be fed back to an implementation specific resolution, for example, 4-bits, 6-bits, or 8-bits. The water-filling weights determine the relative amount of power used on the two STTD groups.
In summary, the options for STTD transmissions for a single user include a pair of antennas to transmit a single layer, or multiple pairs of transmit antennas each used to transmit a respective layer, in each of these oases each layer being transmitted using STTD. Preferably, for multi-carrier applications, antenna selection/grouping is performed on a per sub-carrier basis. Alternatively, it can be performed for some other sub-carrier breakdown. In the event an entire OFDM band, this would be referred to as “fixed” STTD, while in the event of antenna grouping on a per sub-carrier basis, this is referred to as antenna grouping D-STTD.
In addition to antenna grouping, antenna selection can also be used to eliminate some of the antennas and only use a single antenna per STTD block. This will make a 4×2 double STTD system degenerate to a 2×2 BLAST system. In fact, this can be viewed as applying antenna selection to each STTD sub-code-after the rule of antenna grouping has been applied.
For example, if antenna selection is also used with the W-STTD format in the second row of Table 11 to eliminate the first and fourth transmit antennas, the result is something that is effectively equivalent to a 2×2 BLAST system as shown in Table 12 below.
In another embodiment, proportional weighting is applied to each constituent STTD code after antenna grouping, rather than or in addition to for the STTD codes as a whole as was the case in Table 11 above. An example of weighted double STTD with antenna grouping and proportional weighting is given in Table 13 below. There are two levels of weighting in this design: water-filling is applied across layers (i.e. across constituent STTD codes), and proportional weighting is applied within each constituent STTD code. In other words, weighting is sued to get the best allocation of the Alamouti pairs among the antennas, and weighting is used to weight the antennas of a given pair differently. In the table below, both of these weights are included within the single weight wi applied to each symbol. In the example below, antennas 1 and 3 have been grouped to form a first STTD group, and antennas 2 and 4 have been grouped to form a second STTD group.
Referring now to
Only a subset of antennas/sub-channels are selected for transmitting signals to a given receiver. Receivers only need to feedback antennas/sub-channels group indexes to the transmitter. In the event weighting is also employed, then this case, weighting information might be fed back for each sub-carrier, or for groups of six sub-carriers.
In multi-user applications, antenna/sub-channel selection is based on multi-user diversity; different users may select different antenna groups. Quasi-water-filling over is preferably employed in selecting antennas/sub-channels: transmitting power on good channels and antennas, while avoiding poor channels. Power balancing can be achieved through multi-user diversity. Since different receivers experience different channel environments, their antenna group selection will not necessarily the same. When different receivers select different antenna groups, the average power transmitted by each transmit antenna can be somewhat balanced.
The above STTD and SM options have all featured weighting, either in the form of water-filling across different STTD groups/SM antennas and/or proportional weighting within STTD groups. Antenna selection/antenna grouping can be combined with these weighting options. In another embodiment, antenna selection/antenna grouping alone is employed. Several examples of pure antenna selection will now be described.
Example STTD formats for three transmit antenna applications will now be described. In the example of
Another antenna grouping example is shown in
Also shown is a two stream example where the receiver must select two of the three possible antennas, and feedback to the transmitter which two of the three antennas it would like to receive its data on. In this case, a power boosting factor of 1/√{square root over (2)} such that the total amount of power is divided across the two antennas for the given user.
With the two stream example, the matrices B1, B2, B3, B4, B5, B6 show how two streams consisting of s1, s2, s3, s4 and s5, s6, s7, s8 can be spread across four transmit antennas in six different ways. Space time coding is again employed between sets of two symbols within each stream.
For two streams, two of the four antennas are selected, and there are six different options as set out in the table. A power boosting factor of 1/√{square root over (2)} is applied for this case. For a three stream example, there are four different ways to select three of the four antennas as set out in the table. A power boosting factor of 1/√{square root over (3)} is applied. Also shown in the table is an example set of bits that can used to feedback on the channel quality indicator channel selection of an appropriate one of the potential transmission matrices depending upon whether there are 1, 2 or 3 streams.
It is noted that while the above examples show antenna selection feedback, binary beam-forming can be used to similar effect. With binary beam-forming, the beam-forming weights are binary, meaning that a given antenna is either on or off. This is analogous to selection. Two specific examples of using beam-forming to perform antenna selection are provided in
Referring now to
Channel estimates can be interpolated for the non-pilot sub-carrier locations.
Referring first to
Referring now to
While a particular staggering arrangement has been shown in
Referring now to
Referring now to
Referring now to
Referring now to
Referring now to
As indicated above, for the example pilot patterns of
Pilot Transmission Schemes with Pilot Pre-Coding
Systems and methods employing pre-coding of MIMO pilots in accordance with an embodiment of the invention will now be described. In the above described embodiments, it has been assumed that pilot channel information is transmitted without any weights being applied. In this manner, any receiver can look at the transmitted pilot information and recover a respective channel for the particular receiver. In some embodiments, a pre-coding of the pilot is performed.
For a pre-coded pilot, the equivalent channel measured at the receiver becomes G-HW, where H is the actual channel, and W are the weights applied to the pilots. The feedback beam-former W, or information from which it can be derived, may be computed from the channel H for a particular user. Many examples of how this feedback can be performed are detailed below.
In the event pre-coding of the pilot is performed, the channel can be recovered according to:
H=GW′(WW′)−1
However, the pilots need to be user-specific for such a case because beam-forming weights are receiver-specific. This is in contrast to the case where the pilots are not pre-coded and the channel can be recovered by each user.
1. current H analysis and future feedback; and
2. multiple users (possibly including non-MIMO) to assist in channel interpolation.
The non-pre-coded channel H can be re-encoded to coherently demodulate current data for each user.
In contrast, with the example of
Advantageously, by using pre-coded pilots, a smaller number of pilots can be used. The number of pilots can be reduced in either the time or frequency dimension. Preferably, for cases in which the receiver is mobile, the density in time is kept the same and the density is reduced in the frequency domain. Furthermore, for band adaptive modulation and coding, it is preferable to select the most flat part of the band or use larger FFT size, and/or to use an interlaced antenna pilot mapping.
For the case where a receiver is nomadic, i.e., it is not highly mobile, but does move periodically, it is advantageous to keep the frequency density the same, and to reduce the density in the time domain. In this case, preferably a block antenna pilot mapping is employed.
The location of the receiver-specific pilots is an implementation specific decision. For the purpose of completeness, several examples are presented below.
The example of
Referring now to
The example of
In accordance with another embodiment of the invention a pre-coded pilot design for the PUSC (partial use sub-carrier applications) is provided. For PUSC the symbol is first divided into basic clusters of consecutive sub-carriers. Pilots and data carriers are allocated within each cluster of sub-carriers. According to IEEE 802.16-2004, a permutation is used when allocating the data carriers to a sub-channel, a permutation consisting of a remapping of the a cluster sub-carriers into other allocations (that can be almost random). The data carriers in each cluster may be assigned to different MSSs, and therefore, pilots may be used by multiple MSSs. To support the dedicated pilots to closed-loop MSSs operating in PUSC mode, the permutation procedure to partition the sub-carriers into sub-channels is disabled or is allowed to apply to the PUSC sub-channel within a single user. Each sub-channel includes 48 data carriers from two clusters.
Referring now to
The transmission formats discussed above rely on antenna selection/grouping feed back and/o beam-forming weights.
The first option indicated at 600 is to perform MIMO/STC set up based on closed loop feedback. This may involve selection between SM and STTD, and/or antenna selection and/or antenna grouping, as defined for a given application. Given a set of options that are to be made available to a given receiver, the feedback information allows the receiver to select between the options and inform the transmitter of the selection.
The next option indicated at 602 is to perform beam-forming based on feedback information. The feedback information may come in a variety of forms described in detail below. The information that is fed back allows a transmitter to generate weights for the particular receiver. The feedback information can be the weights per se, or other information.
The next option indicated at 604 involves performing MIMO/STC setup based on feedback and feeding the output of this into a beam-former that might be static in nature.
In another embodiment, indicated at 606, the output of a MIMO/STC matrix is input into a beam-former that undergoes beam-forming as a function of feedback.
Finally, the next option indicated at 608 involves performing MIMO/STC set up based on feedback, and feeding the output of this into a beam-former that also performing beam-forming a function of feedback.
In summary, the options are to apply feedback for MIMO/STC set up and/or applying feedback to the beam-former. The beam-former feedback information may be for a unitary beam-former structure, or some other beam-former. Detailed examples have been given previously as to how MIMO/STC set up information can be fed back. Many options for beam-former feedback are given in the details below.
In accordance with various embodiments of the invention there is provided methods of facilitating closed loop MIMO pre-coding and feedback in a communications network that might, for example be operating in accordance with the IEEE 802.16(e) and IEEE 802.11(n) standards. As will be apparent to one of skill in the art the various embodiments can be implemented in software, firmware or an ASIC (application specific integrated circuit) or any suitable combination of these or other components.
SVD (singular value decomposition) is theoretically optimal in terms of the best channel matching transmission and achieving the link level Shannon capacity. However, SVD typically requires a large amount of computing and a large amount of CSI (channel state information) feedback in the TDD (frequency division duplexing) case. With SVD, the channel matrix H is decomposed according to
H=UΛV
where U and V are unitary, and Λ is a diagonal matrix containing the eigenvalues. Typically, information defining V is fed back (although H could be fed back and the decomposition re-executed at the transmitter). Then the transmitter sends VTS, where S is the information vector. The receiver receives this, as modified by the channel, i.e.,
{hacek over (S)}=HV
T
S=UΛVV
T
S(+noise).
The receiver multiplies this by U1 (conjugate transpose of U, known at the receiver) to yield
U
−1
ΛVV
S
T
=ΛS
due to the properties of unitary matrices. The diagonal matrix Λ is also known at the receiver and S can then be recovered. Various methods of efficiently feeding back V or information allowing the transmitter to reconstruct V are given below.
H can be further expressed as follows:
H=λ
1
u
1
v
1+λ2u2v2+λ3u3v3+λ4u4v4
the U matrix consists of columns equaling the vectors u1,u2,u3,u4 and the V matrix consists of rows equaling the vectors v1,v2,v3,v4 all the vectors being unitary.
Thus, the matrix VT used in the transmitter consists of a set of unitary vectors as columns, each associated with a respective eigenvalue. For example, a 4×4 matrix is [v1,v2,v3,v4] where v1,v2,v3,v4 are unitary vectors in columns of the matrix. Eigen-mode selection involves setting one or more of the vectors in V to zero, typically the vectors associated with the weak eigenvalue(s). In some embodiments, eigen-mode selection information is fed back to the receiver.
Antenna selection/grouping is based on the selection of sub-set of antennas from a larger set of available antennas by the terminal based on a simple criterion. The terminal generally needs to feedback very little information to the base station. Mechanisms for transmitting this feedback information from the receiver to the transmitter have been presented above.
Antenna grouping criterion can be based on determinant as described previously. Antenna grouping can also be based on an eigenvalue approach. This approach may avoid selection being dominated by weak layers. Typically degradation due to antenna selection is smaller when the group size is larger.
For a MIMO system, the eigenvalues of the channel matrix are often unevenly distributed. Removing the layers associated with small eigenvalues will have little effects on channel capacity. With a reduced number of layers, more sophisticated MIMO decoders can be used to improve link performance. The power saved from layers that are not used for this a particular receiver can be used by other users on other sub-bands.
This expression is for the channel capacity. It can be seen that larger lambdas make a larger contribution to the channel capacity. While a default would be to assign equal power to all eigenvalues, in another embodiment, water filling according to eigenvalue is performed, and/or some eigenvalues are not used at all.
The V matrix of an SVD decomposed channel matrix consists of a set of horizontal vectors each of which are multiplied by a respective eigenvalue in the product UDV. In the transmitter, through the use of a beamforming matrix VT a respective stream can be transmitted in association with each eigenvalue/column of VT=row of V. Thus, if VT=[v1,v2,v3,v4], then what is transmitted can be expressed as v1×s1+v2×s2+v3×s3×v4×s4 for a 4×4 system. Transmission using such a vector will be referred to as an “eigen-mode”. Different AMC can be applied in the transmitter for each such eigen-mode. For example, in a 4×4 system, there will be 4 eigenvalues, and four beamforming vectors each for use with a respective stream. In the event water filling is employed, the power allocated to the streams is not equal, and in some cases one or more of the layers is simply not used at all, and the power that would have been used for that stream is available for the other layers. Feedback in such systems can consist of an index of the eigen-modes to employ, with the remaining eigen-modes shut down, and/or with channel quality readings for some or all of the selected eigen-modes. In some embodiments, the channel quality defines an associated AMC for use on the eigen-mode at the transmitter.
These two approaches are summarized in
SVD generally provides optimal performance when adaptive modulation and coding (AMC) can be performed on each individual layer. When AMC is not layer based, the performance difference is generally smaller due to imbalanced layer SNR distribution. Preferably; therefore, AMC is implemented on per layer basis.
In some embodiments, when SVD is used the 4-th layer can generally be eliminated, meaning only 3 columns of the V-matrix generally need to be fed back to the transmitter. In systems with larger numbers of antennas, it may be possible to eliminate more layers.
Various antenna grouping algorithms will now be described. Any of these can be employed on a per sub-carrier, per block of sub-carriers, per group of blocks of sub-carriers or for an entire OFDM band. They can be applied for a single or multiple user system.
An antenna grouping algorithm in accordance with an embodiment of the invention will now be described. With channel estimates from the pilots: the receiver computes criterion for use in selecting between STTD and SM transmission modes as follows, where if the STTD criteria is greater than the SM criteria, then STTD transmission is employed and alternatively, SM is employed. Other criteria can be used.
and g1 is the row of matrix G, where G≈(H′H)−1 H′, and H is the channel matrix. For a MISO arrangement the weights may be provided as follows:
According to another embodiment one may exploit the benefits of both. For example, when a channel changes slowly, beam-forming provides a SNR gain, while STTD provides diversity gain (on the symbol level) which improves FEC performance (especially for higher code rates). Then, when a channel changes quickly one can degenerate to an STTD system which will preserve protection against fading.
An antenna grouping algorithm in accordance with an embodiment of the invention will now be described for an STTD mode. The antenna group selection can be based on criterion including the following:
For the SM mode, the antenna group selection can be based on criterion including the following:
An antenna grouping algorithm in accordance with an embodiment of the invention will now be described. For the STTD mode, in accordance with an embodiment of the invention, define
and select a set of antennas according to:
And wherein transmission on an antenna is stopped when:
L
i
<αL
x
In accordance with embodiment of the invention, for the SM mode, if
then a new subset of antennas then can be selected according to criterion including maximum determinant criterion:
S
α=arg max(det(Hα′Hα))
where Hα is the channel matrix of the new sub-MIMO size, and α and β are predefined thresholds, and where λmax and λmin are the largest and smallest eigenvalues respectively.
The following is a particular example of an antenna grouping procedure:
Step 1: Decide antenna group size for STTD and SM mode;
Step 2: Select transmit antenna sub-sets based on selected size;
Step 3: Decide transmit mode: STTD vs. SM;
Step 4: If SM is selected and the antenna group size is larger than SM layers, then use Matrix B, otherwise, use Matrix C.
A multi-user pre-coding algorithm in accordance with an embodiment of the invention will now be described. Consider a MIMO broadcast channel transmit antennas to transmit to K users each with N receive antennas. The base band model of this system is defined by:
y
k
=H
k
s+w
k 1≦k≦K
where HkεCN×M denotes the channel matrix from the base station to the kth user, sεCM×1 represents the transmitted vector, and ykεCN×1 signifies received vector by the kth user. The vector wkεCN×1 is white Gaussian noise with zero-mean and unit-variance.
In the proposed method, the transmitted vector a carries information for M users, defined as follows.
where π(J), π(j), j=1 . . . M are the indexes of a subset of users so-called active user vπ(j)εCM×1, j=1 . . . . M are a set of orthogonal vectors, and dπ(j) includes information for the user π(j). In addition, dirty-paper pre-coding is used on top of the system such that if i>j, the interference of the user π(i) over user π(j) is zero.
For demodulation, the user π(j) multiplies the received vector to a normal vector u′π(j) where (•)′ denotes transpose conjugate operation. In the next section, we specify a method to select the set of active users, modulation vectors vπ(j)′ and demodulation vectors uπ(j) for j=1 . . . M.
In this part, it is assumed that the channel state information is available at the transmitter. Later, this algorithm is modified such that only partial channel state information is required.
Each stage of the algorithm includes two optimizing operations. First, for each user, finding a direction in which that user has maximum gain. Second, selecting the best user in terms of the larger gain. This optimization is performed in the null space of the former selected coordinates. In the following, the proposed algorithm is presented.
1) Set j=1 and the condition matrix Geff=OM×M.
2) Find σ2π(j) where
σ2π(j)=maxrmaxxx′HrH′rx
x′x=1
G
eq
x=0
set π(j) and vπ(j) to be equal to the optimizing parameter r and x, respectively.
4) Set gj=vπ(j), where gj is the jth column of the matrix Geq.
5) J←j+1. If j≦M go to step two, otherwise stop.
As it has been mentioned before, in this algorithm, a major part of the processing can be accomplished at the receivers. Therefore the perfect channel state information is not required at the transmitter which results a significant decreasing in the rate of feedback.
1) Set j=1 and Geq=QM×M.
2) Each user calculates σ2r(j), defined as follows:
σ2r(j)=maxrx′HrH′rx
x′x=1
G
eq
x=0
vr(j) represents the optimizing parameter x.
3) Each user calculates
4) Each user sends σrr(j) and Vr(j) to the base station, if σr(j)2≧th(j). th(j) is a threshold which is predetermined by the base station.
5) Base station selects the user with the largest σ2r(j). Let π(j) be the index of the selected user. The corresponding gain and coordinate of that user. The corresponding gain and coordinate of that user are σ2π(j) and vπ(j), respectively.
6) The π(j)th user sends Uπ(j)Hπ(j)wπ(j) i=1 . . . i−1 to the base station.
7) Base station send vπ(j) to all users. All include vπ(j) in Geq as the jth column.
8) J←j+1. If j≦M go to step two, otherwise stop.
Feedback MIMO channel and CQI Separately
Referring now to
At the transmitter, there is a transmit beam-former 202 that uses matrix VT this being the transpose of the matrix V in the singular value decomposition of the channel H. The output of the beam-former is transmitted on M transmit antennas 204. This is received at the receiver by N receive antennas 206, and the MIMO channel H is measured as indicated at 208. There are many existing methods of measuring the channel. This may for example be done using pilot symbols transmitted at known locations in time and/or frequency, possibly using one of the formats introduced previously and performing interpolation if necessary. This may include receiver specific pilots and or generic pilots. Next, block 212 performs a singular value decomposition of the channel matrix H into the matrix product UDV where U and V are unitary matrices and D is a diagonal matrix containing the eigenvalues.
At 220, AMC/eigen assignment according to according to eigenvalue is performed for any appropriate frequency resolution, for example for a given sub-channel. A sub-channel may for example be a contiguous block of sub-carriers that have substantially the same eigenvectors/eigenvalues. As discussed previously, weak eigen-modes can be discarded. In some embodiments, all that needs to be fed back is an index of the eigen-modes that are to be used. Then, the receiver, after it reconstructs the H or V matrix from the fed back channel information, can re-construct the eigen-modes, and use each eigen-mode for a respective stream. This information is then sent back to the transmitter on the CQI feedback channel 222 or using some other mechanism.
For the channel information, preferably the difference between the new channel matrix H and the previous channel matrix H is measured and a differential is determined between the two most recent channel matrices. This amount is encoded at 210 and transmitted back to the receiver on the MIMO feedback channel 214. Advantageously, the differential of two consecutive channel matrices will have a significantly smaller dynamic range than would the actual channel matrix H or V matrix. This enables the MIMO feedback channel 214 to use fewer bits to transmit back effectively the same amount of channel information.
At the transmitter, the channel matrix H is reconstructed at 216 using the fed back channel differential and the previous channel matrix H. Singular value decomposition is performed at 218 to recover the V matrix component of the channel matrix H, and the transpose of this matrix is then used in the beam-former 202. Furthermore, the CQI feedback channel 222 feeds into the per stream AMC function 200 that performs adaptive modulation of coding on each stream using the channel quality information as sent back from the receiver. In the above-described embodiment, a differential of the channel matrix H is fed back. Equivalently, the unitary matrix V or a differential of the unitary matrix V could be fed back. Various detailed methods of feeding back the channel matrix or the unitary matrix V are provided below.
The embodiment of
Referring now to
The two above-described embodiments employ the feedback of differentials in either the channel matrix H or the unitary matrix V.
While the illustrated example uses +1, −1 quantization, this requiring one bit per element to be transmitted on the feedback channel 264, higher resolution quantization could alternatively be employed. Furthermore, some embodiments may not necessarily feature the scalar quantization step. Note that for a 4×4 channel matrix, each channel element is complex and therefore has an I and Q component. Therefore, to feedback quantized information for a 2×2 channel matrix requires eight bits, two bits per channel element, one bit per I and Q components.
In accordance with another referred embodiment of the invention a ΔΣ modulator 1-bit quantizer is employed to determine the information to be fed back. An example of a 1st-order ΔΣ modulation encoder is shown in
The differential encoder output may be grouped together to map to the existing CQI feedback channel or MIMO feedback channel.
In another embodiment of the invention, the V matrix determined through singular value decomposition is fed back from the receiver to the transmitter using Givens feedback. Unitary matrices have the property that they can be decomposed into a product of Givens matrices. In particular, an n×n V matrix can be decomposed into Givens matrices containing n2−n complex parameters. The parameters can be fed back directly, or differentially and/or quantized to potentially reduce the amount of feedback required. Then, at the transmitter, the Givens matrices can be reconstructed, and then the V matrix reconstructed as the product of the reconstructed Givens matrices.
An example of a system employing Givens feedback is shown
Then, the quantized parameters θ and C are fed back over the MIMO feedback channel 336, this requiring two parameters to be sent back for each Givens matrix. At the transmitter, the fed back parameters θ and C are reconstructed at 338, and then the beam-forming matrix V is reconstructed by determining the product of the Givens matrices generated using the reconstructed parameters, at 340. Finally, the transpose of the matrix V is used by the beam-former 320. The parameters may be directly fed back or differentially fed back. 1
Preferably, at the same time AMC/eigen assignment is performed on the basis of eigenvalues of the channel matrix of the SVD composition performed at 328 as described previously. This step is indicated at 332. The AMC/eigen assignment information is fed back over the CQI channel at 342. This information is fed back to the transmitter where per stream AMC is performed is indicated at 344.
Advantageously, for V matrix feedback, the base station is able to verify the integrity of the received matrix V by exploiting the orthogonality of the matrix.
By decomposing the SVD-based unitary V matrix into Givens matrices, the V matrix can be represented by n2−n independent complex parameters.
In accordance with an embodiment of the invention, 2(n2−n) bits (one bit for I and Q of each parameter) are used for the differential quantization of the unitary V matrix, although other methods of feeding back the parameters can be used.
In accordance with this embodiment an n stream CQI channel may also be used, preferably 4 bits each. Combining the Given's feedback and the CQI yields the total amount for the unitary matrix feedback of 2n2+2n bits. However, for the direct differential encoding approach the total feedback used is 2n2 including the CQI information, since it is embedded in H.
The following is an SVD based Givens transform algorithm in accordance with an embodiment of the invention suitable for 2-Transmit Antennas. With 2-antenna Givens transformation, 22−2=2 complex parameters are required to feedback the V matrix. There are multiple known formats for Givens transformations any of which can be used. The Givens rotation for 2 transmit antennas (Format-A) includes:
The parameter space includes θ=(−π+π) and c={0,1} s=(1−c2)1/2.
An alternative format is the following Givens rotation for 2 transmit antenna (Format-B) that takes the form:
The parameter space includes θ={−π+π} and η={−π+π}
An SVD based Givens transform algorithm in accordance with an embodiment of the invention will now be described for 3-Transmit Antennas. With 3-antenna Givens transforms, n2−n=6 complex parameters can be used to completely represent the V matrix. These can be represented in the following Givens rotation for 3 transmit antenna (Format-A) includes G=G1G2G3 represented as follows where each of the three matrices has a respective parameter c (c1,c2,c3) and s (θ1,θ2,θ3) which differs between the three matrices although this is not shown.
The parameter space includes
θ1→2 bit and c1→1 bit
θ2→2 bit and c2→1 bit
θ3→2 bit and c3→1 bit
It can be seen in the above example, the quantization performed for the parameter θ is more precise than that performed for the parameter C. It is generally the case that different quantization accuracies may be performed for the different parameters being fed back using given speed back. Through experimentation, it has been found that acceptable performance can be achieved with a less accurate C parameter. In some embodiments described below, certain C parameters can be ignored.
An SVD based Givens transform SQ algorithm in accordance with an embodiment of the invention will now be described for 4-Transmit Antennas. 4-antenna Givens transformation requires 42−4=12 complex parameters. These can be represented in the Givens rotation for 4 transmit antenna (Format-A) as G=G1G2G3G4G5G6 represented as follows where each of the six matrices has a respective parameter c (c1,c2,c3,c4,c5,c6) and s (θ1, θ2, θ3,θ4, θ5,θ6):
The parameter space includes
An SVD based Givens transform algorithm in accordance with an embodiment of the invention will now be described for n-Transmit Antennas. The Givens rotation for n transmit antennas (Format-A) includes:
V=Π
k=1
n=1Πf=1n-kG(k,i)
Preferably, scalar quantization is performed for each of the parameters as discussed previously.
An SVD based Givens transform algorithm in accordance with an embodiment of the invention will now be described that employs truncation of the Givens expansion. Preferably, weak eigen-modes may be discarded with application of water filling in the eigen domain. Each Given matrix is associated with one or more eigen-modes and respective eigenvalue. As such, if a given weak eigenvalue is to be discarded, then the associated Givens matrices can be discarded. In the receiver, the V matrix is re-generated, and then only the selected eigen-modes used to transmit streams as in the above example for eigen-modes selection.
For example, where the full Givens expansion for the 3×3 case is as follows:
this can be truncated by dropping the “weakest” eigen-mode. For example if the third matrix corresponds to the weakest eigen-mode, the truncated expansion is:
The full expansion above corresponds to all three eigenvectors of the channel. In contrast, the truncated expansion corresponds to only a single eigenvector. The correspondence of eigenvector to Givens matrices is a known math property of Givens expansions.
The following is another example of truncation of Givens expansion for a 4-antenna case. The Givens expansion for eigenvector #1 of a 4×4 channel matrix is as follows:
The Givens expansion for eigenvectors 1 and 2 for a 4×4 system is as follows:
The Givens expansion for all four eigenvectors of the 4×4 system is as follows:
It can be seen that discarding two eigen-modes reduces the Givens expansion by one of the six matrices while discarding three of the four eigen-modes will reduce the Givens expansion by three of the six matrices. Given that each matrix requires two parameters to be fed back, a corresponding reduction in the quantity of feedback information that needs to be transmitted is realized. Generally, it can be seen that in an arbitrary n×n case, a Givens expansion can be employed, and the amount of feedback can be optionally reduced by discarding one or more eigen-modes.
An SVD based Givens transform algorithm in accordance with an embodiment of the invention will now be described in which full scalar quantization is performed. As discussed above, it can be advantageous to use different quantizations for the various parameters being fed back using Givens feedback. Based on the pre-coding QoS requirement, the bit allocation may provide a tradeoff between performance and feedback penalty.
For the Givens parameter in Format-A, θ is uniformly distributed resulting in more bits being needed for accurate feedback, whereas C is non-uniformly distributed resulting in less bits being needed. In some implementations, some C have only minimum impact on performance and can be treated as constant without a substantial performance penalty.
Using a Lloyd-MAX quantizer the following bit allocations can be employed to realize a full soalar quantizer.
An SVD based Givens transform algorithm in accordance with an embodiment of the invention will now be described in which partial scalar quantization is performed. With partial scalar quantization, a differential is not fed back for each and every parameter. In an example solution, a 1 bit or 0 bit is used for each parameter c and a differential 1-bit is used for each parameter theta. Another approach includes using less bits or zero bits for the non-significant Givens parameters.
The following is an example of bit allocation with partial scalar quantization for 2, 3 and 4 antennas. It can be seen that in a particular example, for the 3-antenna case for example, the o parameter is only fed back for the first of three Givens matrices, and a three bit value is sent back for theta for the first of the three matrices, and a one bit differential is used for theta for the other two matrices. This has the potential of reducing the number of feedback bits required from nine bits from the full quantization example given above to five bits for the example in the table.
The following table shows an example set of feedback allocations for different MIMO configurations using Givens feedback with full scalar quantization, versus partial scalar quantization with a 1-bit differential for theta and 1-bit for c. It can be seen that the savings in the number of bits required for feedback increase as the complexity of the MIMO configuration increases. Also shown in the table are the number of bits required for a Hausholder based feedback mechanism.
The table below shows examples of the feedback allocations Zor various MIMO configurations comparing full scalar feedback to partial feedback in which a 1-bit differential for theta is employed, and 1-bit or zero bits are used for each c parameter. Again, the requirements for Hausholder feedback are also shown for the sake of comparison.
With scalar quantization, be it full scalar or partial scalar, each feedback element pertains to one parameter of interest. This parameter can be a Givens parameter, or a channel matrix parameter from H or V for example, or some other value. In another embodiment, vector quantization is performed, in which each element fed back represents a vector, either in absolute terms or differential terms, and in respect of a set of parameters of interest. Each potential state for the output of the vector quantization operation is represented by a code word identifier/index which is fed back, and used by the transmitter to look up the vector. The vectors thus looked up are then used to generate the beam former in the transmitter. The vector might be a vector of a channel matrix, a vector of a V matrix (this would be a unitary vector), or a set of Givens parameters for example.
In one embodiment, an SVD based Givens transform vector quantization algorithm is provided that employs Grassmann Subspaoe Packing. However, other forms of vector quantization may alternatively be employed.
Each individual Givens matrix is a unitary matrix. The product of Givens matrices is a unitary matrix. The partial product of two or more Givens matrices is a unitary matrix. After determining a truncated or full Givens expansion, the resulting of unitary matrix may be quantized by using Grassmann sub-space packing (see for example D. J. Love, R. W. Heath Jr., and T. Strohmer, “Grassmannian Beamforming for Multiple-Input Multiple-Output Wireless Systems,” IEEE Transactions on Information Theory, vol. 49, pp. 2735-2747, October 2003). In this case, the code book for quantization of a unitary matrix may be considered as sub-space packing in a Grassmann manifold.
Two codebooks that might be used include the uniform distributed Grassmann space, and Block circulant DFT, but other codebooks may alternatively be employed.
An example of encoding using shape and gain quantizer according to an embodiment of the invention is shown in
An SVD based Givens transform vector quantization algorithm in accordance with an embodiment of the invention will now be described that employs a Spherical Code Based Quantizer. This method begins with putting the Givens parameters into a vector and then encoding the vector using a spherical code based quantizer. The following steps are performed:
1. Given k element vector form the vector X→Rk
2. Compute g=∥X∥ and S=X/g
3. Use the gain codebook to quantize g as g
4. Find i such that αi=<sin−1xk<αi+1 and compute
h
l(S)=X/∥X∥*(∥XLj∥−∥Xl−X∥)
5. Find the nearest neighborhood hi(S) to hi(S)
6. Compute hi1(hi(S)) to identify the quantized shape S
7. Compute the index gS and transmit
In accordance with an embodiment, Leech Lattice is used as a codebook. Alternatively, a trellis-coded quantizer can also be employed.
A receiver based Givens transform in accordance with an embodiment of the invention will now be described in which search criteria are established in the receiver.
In accordance with this embodiment of the invention channel filling is based on criteria including the receiver criteria. Based on the QoS requirements, the receiver determines the minimum feedback needed adaptively. The process may include the following:
determining the Givens truncation level;
based on the Scalar quantizer structure assigning parameter values; and/or
performing combinatorial searches of combinations.
One example search criterion may include receiver MSE. Others may include Max SNR, Max Shannon capacity, or True receiver operational process defining as follows:
Based on the receiver criterion, exhaustive search the code book and maximize the given received based criterion to determine the best pre-code matrix. These equations represent exhaustively computing beam-forming weights to see which one is the best for a given criteria.
Referring now to
In any of the examples above, the Givens decomposition can be computed using any appropriate method. Iterative approaches may for example be employed such as Cholsky factorization and reverse order multiplication. These methods are known in the art and will not be described further here. It is noted that Givens decomposition is about 10% of the complexity of SVD computing complexity. This can be seen from the following table comparing the complexity of SVD versus Givens as a function of the number of transmit antennas, and the complexity is measured in terms of multiply and add operations.
The following table shows a comparison of the complexity of Givens based feedback as opposed to Hausholder feedback as a function of the number of the transmit antennas. It can be see that the Hausholder approach is significantly more complex.
An example of Vector Pre-coding (code-book construction) for purpose of context and comparison will now be described. The cross-correlation of the codeword in this examples has a block circulant structure. The diagonal rotation matrix Q is defined as:
Matrix P., is selected from sub-matrixes of DFT matrix as
The codebook can be constructed as:
P
1
=Q
l
P
1
l=2,3 . . . L
The codebook can be optimized by choosing rotation matrix Q indexes
c=[c
1
. . . c
M
]
The code book can be optimized by choosing rotation matrix Q indexes
u=[u
1
. . . C
N
]
By exhaustive search of the codebook we have
An overview of Matrix Pre-coding (column by column vector quantize channel-1) for purposes of context and comparison will now be described. Preferably, the V matrix is quantized column by column and recursively.
STEP-0: Denote the beam-forming matrix as
STEP-1: Quantize the first column of V denoted as V1 as follows: {circumflex over (V)}1=arg maxUE
Where C1 is a codebook containing unit 4-vectors for quantization. {circumflex over (V)}1 has the maximum inner product among all unit vectors in the codebook.
STEP-2: Compute Hausholder reflection matrix as follows
Where φ1 is the phase of v11
where two properties are employed to get the result, i.e. {circumflex over (v)}11 is real and the unitary property of V Since both F1 and V are unitary, V2 is unitary. From STEP-2, we see that size of V2 is 3 by 2 and it is reduced from that of V1 by one on both the row and column dimensions.
STEP-3: Quantize the first column of V2 denoted as v2, using another codebook or unit 3-vectors, whose first element of each codeword is real.
STEP-4: Construct a Hausholder reflection matrix F2
STEP-5: Multiply F2 with V2 as follows.
The reconstruction of the beam-forming matrix V is as follows:
STEP-0: Two vectors, v3 and v2, are reconstructed using the feedback quantization indexes and the corresponding 2-vector and 3-vector codebooks.
STEP-1: Compute a Hausholder matrix using the reconstructed V2 as
where w={circumflex over (v)}2−e1 and {circumflex over (v)}2 is the reconstructed 3-vector; F2 can be stored beforehand to reduce computation.
STEP-2: V2 can be reconstructed as
STEP-3: we reconstruct the first column of V using the quantization index and compute a Hausholder matrix as
where w={circumflex over (v)}1−e1 and {circumflex over (v)}1 is the reconstructed first column of V.
STEP-4 the beam-forming matrix V is given by
The codebook is constructed such that the codeword vectors distribute on the n-dimension complex unit sphere uniformly. Additionally, the first element of each codeword is set to be real for the next step.
The Hausholder matrix can be computed and stored beforehand for small codebooks. Even in the case that there is no quantization error, the reconstructed matrix could be different from the original V by a global phase on each column and this is fine with closed loop MIMO.
Pre-design the rotation matrix for 2 transmit antennas, and then the rotation matrix is parameterized. A set of parameterized rotation matrixes serves as codebook.
According to an embodiment of the invention a differential index feedback is provided as illustrated pictorially in
A MIMO feedback channel ageing algorithm in accordance with an embodiment of the invention will now be described, in which receiver ageing beam-former correction is utilized.
For a mobile MIMO channel the channel matrix may be time-varying:
Time-0→H0=U0D0V0′
Time 1→H1=U1D1V1′
The beam forming matrix V0 sent to the transmitter may already be old by the time it is utilized for transmission. However, the receiver may still be able to compute the latest receiver beam forming matrix U0 as follows:
U
1
′y=U
1
′U
1
D
1
V
1
′V
0
′s+U
1
′n=D
1
V
1
′V
0
s+n
This may prevent the ageing impact at receiver side, the ageing impact potentially causing the inter-antenna interference. In the above example, at the receiver the received sequence y has been multiplied by the latest beam-forming matrix U1, rather than use U0 that would have been used with the channel information at time 0. In the result, the product V1′V0 or likely be close to the identity matrix and can be ignored; alternatively given that both V1′ and V0 are known at the receiver, the effect of this can also be divided out at the receiver. This would increase the complexity however.
A CQICH Support of Differential Encoding Algorithm in accordance with an embodiment of the invention will now be described. For OFDM systems, preferably differential encoding is used to cross multiple sub-carriers to feedback the vector index or other feedback information. The following is an example of a mini-tile modulation scheme for sending back a vector index of zero or one. Preferably this would be transmitted on two sub-carriers. In this case, two different phases are transmitted for vector index zero, and a different arrangement of the same phases is sent for vector index one.
Another example is shown in
Another example is shown in
In yet another example, shown in
Each embodiment is generalizable to an arbitrary number of sub-carriers and/or an arbitrary number of transmit antennas/receive antennas as will be apparent to one skilled in the art. Embodiments provide transmitters adapted to generate signals containing the disclosed transmit code-sets/sub-carrier allocations, methods of transmitting such signals, receivers adapted to receive such transmissions, and methods of receiving and decoding such signals.
Numerous modifications and variations of the present invention are possible in light of the above teachings. It is therefore to be understood that within the scope of the appended claims, the invention may be practised otherwise than as specifically described herein.
This application claims the benefit of U.S. Provisional Patent Application No. 60/581,356 filed on Jun. 22, 2004, U.S. Provisional Patent Application No. 60/582,298 filed on Jun. 24, 2004, U.S. Provisional Patent Application No. 60/601,178 filed on Aug. 13, 2004, Provisional Patent Application No. 60/614,621 filed on Sep. 30, 2004, Provisional Patent Application No. 60/619,461 filed on Oct. 15, 2004 and Provisional Patent Application No. 60/642,697 filed on Jan. 10, 2005, all of which are hereby incorporated by reference in their entirety.
| Number | Date | Country | |
|---|---|---|---|
| 60642697 | Jan 2005 | US | |
| 60619461 | Oct 2004 | US | |
| 60614621 | Sep 2004 | US | |
| 60601178 | Aug 2004 | US | |
| 60582298 | Jun 2004 | US | |
| 60581356 | Jun 2004 | US |
| Number | Date | Country | |
|---|---|---|---|
| Parent | 15000542 | Jan 2016 | US |
| Child | 15486060 | US | |
| Parent | 11630391 | Dec 2006 | US |
| Child | 15000542 | US |
