Patent Grant
8831127
Multiple Input Multiple Output (MIMO) communication is becoming an integral part of current and future wireless communication standards. Using multiple transmit and receive antennas, MIMO communications enable multiple data streams to be conveyed simultaneously and independently between the transmitter and the receiver without consuming additional bandwidth or other radio resources. To that end, the transmitter and/or receiver comprise an antenna array having multiple antennas, each associated with a variable antenna weight, where the antenna weights at the transmitter are generally referred to as pre-coders. Through the use of the weighted transmitter and/or receiver antennas, different patterns may be formed for different data streams. If the wireless radio channel exhibits rich scattering, e.g., low correlation or a small singular value spread, then multiple possible propagation paths exist between the transmitter and receiver, allowing different data streams to be transmitted by orthogonal mapping over the different paths.
The receiver must process the received composite signal to separate and decode each of the transmitted data streams. To that end, conventional systems use linear receivers, e.g., minimum mean square error (MMSE) receivers, or non-linear receivers, e.g., maximum likelihood (ML) receivers. The ability of either type of receiver to separate the transmitted data streams present in a received composite signal depends on the orthogonality between the channels of the individual data streams. In general, the separation will not be perfect, leading to inter-stream interference, which limits the achievable signal-to-noise ratio (SNR) or signal-to-interference plus noise ratio (SINR) for each signal stream. The more the data stream channels resemble each other, the more difficult it will be for the receiver to separate the data streams. Channel similarity may be expressed based on the cross-correlation of the channels, through an alternative measure known as the singular value spread (which is derived based on the channel). A large singular value spread indicates highly similar channels, and thus, a difficult receiver problem. Therefore, the best conditions for MIMO communications occur when the SNR or SINR is high and the wireless channel experiences rich scattering, as indicated by low correlation or a small singular value spread.
Unfortunately, to some extent the beneficial channel conditions for MIMO are mutually exclusive, meaning the highest SNR or SINR conditions often occur at the same time as the lowest experienced channel richness, and vice versa. This problem may be exacerbated when one or more dominant data streams overpower weaker multi-path data streams. As used herein, a dominant data stream or a dominant signal path is defined as the data stream or path associated with a dominant mode, a dominant eigenmode, and/or a line-of-sight (LOS) path. For example, a large singular value spread or a large amplitude difference between the data streams in the received composite signal (e.g., due to a dominant LOS data stream) may cause some of the weaker data streams to end up with low SNRs. In response, the receiver may try to optimize the throughput by requesting a lower rank transmission (i.e., to reduce the number of data streams) to avoid wasting power on data streams where little to no throughput is expected, and by requesting a power increase for the data streams where the SNR gain will translate into improved throughput.
Requesting a power increase, however, can exacerbate noise conditions proportional to or dependent on the signal strength, i.e., multiplicative noise, particularly when such noise conditions limit the throughput conditions. Further, the use of fewer data streams leads to lower peak data rates over the wireless connection, which is expected to become even more problematic as standards and technology trend towards transmitters and receivers capable of handling larger numbers of signal streams. For example, both LTE release 10 and IEEE 802.11 ac have recently standardized up to 8×8 MIMO transmissions. Thus, there remains a need for improving MIMO throughput conditions limited by multiplicative noise.
The MIMO method and apparatus disclosed herein improve throughput conditions limited by multiplicative noise by reducing the gain of the data streams associated with one or more dominant signal paths between MIMO communication nodes. As used herein, the term multiplicative noise refers to any noise dependent on or proportional to a signal strength at a transmitting node and/or a receiving node of a wireless communication network. Also as used herein, a dominant signal path comprises any path associated with a data stream that overpowers other data streams, and includes but is not limited to a path (or data stream) associated with a dominant mode, a dominant eigenmode, and/or a LOS signal path.
An exemplary method comprises determining that multiplicative noise limits the throughput conditions, and reconfiguring one or more antennas of a MIMO communication node to change the gain of the reconfigured antennas for the dominant signal path(s). In particular, the antennas are reconfigured to redirect a radiation pattern of each of the reconfigured antennas to reduce a gain of the reconfigured antennas for a dominant signal path between the MIMO communication node and a remote communication node.
An exemplary MIMO communication node comprises an antenna array having a plurality of antennas, a noise processor, and an antenna configuration unit. The noise processor is configured to determine that multiplicative noise limits the throughput conditions for the MIMO communication node. The antenna configuration unit is configured to reconfigure one or more of the antennas to redirect a radiation pattern of each of the reconfigured antennas to reduce a gain of the reconfigured antennas for a dominant signal path between the MIMO communication node and a remote communication node.
Another exemplary embodiment comprises a method of characterizing MIMO throughput conditions relative to multiplicative noise in a MIMO wireless communication system. The method comprises evaluating throughput conditions between a transmitting MIMO node and a receiving MIMO node based on at least one of a signal quality parameter and a MIMO utilization parameter. The method further comprises generating a multiplicative noise evaluation characterizing the MIMO throughput conditions relative to multiplicative noise based on the evaluated throughput conditions.
An exemplary noise processor is configured to characterize MIMO throughput conditions relative to multiplicative noise in a MIMO wireless communication system. The noise processor comprises a parameter unit and a characterization unit. The parameter unit is configured to evaluate throughput conditions between a transmitting MIMO node and a receiving MIMO node based on at least one of a signal quality parameter and a MIMO utilization parameter. The characterization unit is configured to generate a multiplicative noise evaluation characterizing the MIMO throughput conditions relative to multiplicative noise based on the evaluated throughput conditions.
The methods and apparatuses disclosed herein improve the achievable MIMO throughput in conditions where the MIMO throughput is limited by multiplicative noise, e.g., phase noise or quantization errors in the equipment, rather than by additive noise, e.g., thermal noise or interference. To that end, one or more antennas are reconfigured when multiplicative noise limits the throughput conditions to reduce a gain of data signals associated with one or more dominant signal paths. The following describes the invention in terms of a line of sight (LOS) signal path. It will be appreciated, however, that the present invention is applicable for MIMO conditions having any type of dominant signal paths, including but not limited to a path or data stream associated with a dominant mode, a dominant eigenmode, and/or a LOS signal path. Before discussing this further, the following first provides details regarding MIMO communication systems and the associated problems.
More particularly, multipliers 113 apply a first column vector of weights w1=[w11; w12; w13; . . . ; w1N], which may be frequency dependent, to a first data stream x1 for transmission from antenna array 110. Different combinations of weights correspond to different radiation patterns through the antenna array response a(θ,φ)=[a1; a2; a3; . . . ; aN], where:
an(θ,φ)=gn(θ,φ)eik·r
In Equation (1), k represents the wave vector 2π/λn where n represents a unit vector in the direction (θ,φ), rn represents the antenna displacement vector, and gn (θ,φ) represents the per-antenna gain vector. While gn (θ,φ) is shown here as a scalar, it will be appreciated that gn (θ,φ) may be generalized to polarized antennas and channels. The shape G1(θ,φ) of the radiation pattern used to transmit data stream x1 becomes:
G1(θ,φ)=w1Ta(θ,φ). (2)
A second data stream x2 may be transmitted using the same antenna array 110 using the weights w2=[w21, W22, W23, . . . , w2N]. In general, up to N data streams can be transmitted simultaneously. The total transmitted signal may be represented by the superposition:
WTXx=w1x1+w2x2. (3)
The application of weights WTX=(w1, w2) at the transmitter node 100 is generally referred to as pre-coding.
The receiver node 200 also comprises an antenna array 210 of M antennas 212, where the radio channel between the nth transmitting antenna 112 and the mth receiving antenna 212 may be expressed as hnm, where hnm may be time-dependent and/or frequency-dependent. While not shown, the receiver node 200 may also apply different weight vectors to the antenna array 210 for different received signals. In matrix notation, the received signal y=[y1, y2, y3, . . . , yM] may be expressed as:
y=(I+Σrx)H(I+Σtx)WTXx+Σ+Q, (4)
where the diagonal matrixes Σrx and Σtx represent multiplicative noise, including but not limited to phase noise and quantization errors, H represents a matrix of the channel estimates, Σ represents additive noise, including but not limited to thermal noise, and Q represents the interference.
The receiver node 200 is thus tasked with the problem of separating the received composite signal y into the individual data streams x1, x2, . . . , xN in order to successfully decode each data stream. This problem is generally complex, but it has been shown that in the absence of interference, phase noise, and dynamic range limitations, the optimal pre-coders and receive weight vectors are given by the singular value decomposition (SVD) of the wireless channel. In particular, it has been shown that the receive weight vectors under these conditions may be given by H=U·S·V, where U represents the unitary pre-coding matrix, V represents the unitary receiver weight matrix, and S represents a diagonal matrix where each diagonal element represents the signal amplitude that a particular mode of communication will experience. Thus, for an M×N channel H, the diagonal matrix S will be of size M×N. In other words, the number of independent channels that can be transmitted over this channel is bound by min (M, N). If the channel H is rank-deficient, then some of the elements of S will be zero, further limiting the available number of data streams. In a noise-limited scenario, the total capacity R of the channel is known to be the sum of the Shannon capacity for each sub-stream, assuming equal and independent noise level σ2 over the different receiver radio chains, as given by:
where Pn represents the power allocated to the nth data stream and fBW represents the bandwidth.
In general, finding and utilizing the SVD requires full channel knowledge at the transmitter node 100, something which is costly to obtain in practical circumstances. In particular, full feedback of all channel coefficients hnm may require a prohibitive amount of reverse link capacity, especially when hnm is time- or frequency variant and the numbers N and/or M are large. Therefore, different lower-complexity schemes have been devised. One such scheme relies on quantized feedback of preferred pre-coders based on a previously agreed codebook, known as closed-loop pre-coding. Closed-loop pre-coding is a part of standards including but not limited to 3GPP UMTS and 3GPP LTE. The closed-loop codebook consists of a finite number of available pre-coding matrixes WTX for each rank (e.g., for each number of data streams to be transmitted). The receiver node 210 estimates the channel H, typically using reference symbols or pilots transmitted from each of the transmit antennas 112, as well as the noise and interference level, using any known technique. Next, the receiver node 200 evaluates all possible pre-coding matrixes and searches for the one that will result in the best utilization of the available radio resources, which is typically the rank and pre-coder that results in the highest data rate throughput given the estimated SNR or SINR. Once this is found, the receiver node 200 signals the desired rank and pre-coder on the reverse link to the transmitter node 100 to indicate the receiver node's preference, typically using a Rank Indicator (RI) and a pre-coding matrix indicator (PMI). In addition, the receiver node 200 may signal back the perceived channel quality using a Channel Quality Indicator (CQI) that it expects to experience with the selected pre-coder, to allow the transmitter node 100 to adjust the coding and modulation scheme optimally, a process known as link adaptation. The transmitter node 100 may in some cases be required to use the preferred pre-coding indicated by the receiver node 200. In other cases, the transmitter node 100 may override the recommendation at its discretion, e.g., due to circumstances unknown to the receiver node 200, e.g., the existence of additional receivers, scheduling priorities, non-full data buffers, etc. While possible, such an override may obsolete the CQI feedback because was generated based on the receiver's preferred pre-coder, which typically renders link adaptation more challenging. Another MIMO scheme included in the same wireless communication standards is the so-called open-loop pre-coding. In this scheme, no channel or preferred pre-coder information is fed back to the transmitter node 100, although CQI and rank feedback are still used. Instead the transmitter node 100 will typically send information with a fixed pre-coder or a pre-coder that cyclically varies in time and/or frequency. Additionally, when multiple data streams are transmitted, the pre-coders are usually mixed in order to provide the same channel quality for all data streams. Open-loop pre-coding is suboptimal compared to closed-loop pre-coding. At high signal to noise ratios, however, the difference between the two diminishes. In particular, the closed-loop scheme will essentially find good pre-coders that result in good data stream separation and high per-stream SNRs at the receiver node 200, while the open-loop scheme instead relies on a high likelihood of using sufficiently good pre-coders and the ability of the receiver node 200 to separate the streams. The closed-loop scheme is sensitive to noisy channel estimates, which might result in non-optimal pre-coders. Also the reported pre-coder might be outdated by the time the transmitter node 100 uses it for the actual transmission. The open-loop scheme, on the other hand, relies on diversity and uses a wide range of different pre-coders. Thus, the open-loop scheme will not be optimal, but it is less sensitive to noise and timing issues.
Regardless of whether the system uses a closed-loop or open-loop pre-coding scheme, the receiver node 200 processes the received composite signal to decode each of the data streams. Because the channel will typically mix the data streams, as is evident from Equation (4), signal processing is required to separate the data streams. Many different MIMO receiver structures are known in the art. Broadly, these fall in to two categories: linear receivers and non-linear receivers. The operation of a linear receiver may be expressed by a linear operator operating on the received signal vector y according to {circumflex over (x)}=WRXy, where WRX represents the receiver weight matrix. An example of a linear receiver is the minimum mean square error (MMSE) receiver, which selects weights WRX such that the mean square error between the transmitted and the estimated symbols is minimized in the presence of additive noise and interference. The MMSE receiver is equivalent to a zero-forcing receiver in the case where no noise or interference exists. The operation of a non-linear receiver relies on the use of more complex operations, e.g., multi-stage operations. An example of a non-linear receiver is the Maximum Likelihood (ML) receiver or the Successive Interference Cancellation (SIC) receiver.
The ability of receiver node 200 to separate the data streams depends on the orthogonality between the data stream channels. In general the separation will be non-perfect, leading to inter-stream interference which limits the achievable SINR for each stream. The more the data stream channels resemble each other, the more difficult the separation will be, leading to lower effective SINRs on each sub-stream. Channel similarity is often expressed by the cross correlation, though an alternative measure is the singular value spread. The cross correlation coefficient of two channel coefficients h1 and h2, for example, may be defined by:
and is typically estimated by the sample cross correlation:
where {tilde over (h)}1(k) and {tilde over (h)}2(k) represent sequences of (typically noisy) channel estimates of the channels h1 and h2. The singular value spread is derived from the singular value matrix S (derived from the channel H). A simple measure of this spread is the ratio of the largest and the smallest singular value, e.g.,
Large correlations between the elements of the channel matrix H implies a large singular value spread and hence a difficult receiver problem.
In the case that multiple streams are transmitted, the power per data stream will be lower than if fewer or a single data stream is transmitted. Successful link adaptation thus requires finding the optimal number of data streams to transmit, and also the power to use for each data stream. This optimum will be SNR-dependent. At low SNRs it is typically better to allocate all power to one data stream, while at higher SNRs the available transmit power may be equally shared across data streams while still maintaining a sufficiently high per-stream SNR to allow a high order of modulation and coding. It has been shown that an optimal per-stream power allocation, in the absence of interference, phase noise, and dynamic range limitations, is given by a “water filling” solution in which power is allocated proportional to the per-stream SNR, but only to those streams that have an SNR exceeding a certain threshold. However existing cellular standards tend to share transmit power equally across the data streams. As seen in Equation (5), data streams with poor conditions (weak sn) will not contribute as much to the total throughput as data streams with good conditions (strong sn) due to the log2 expression. The best conditions for MIMO communications thus occur when the SNR or SINR is high and the wireless channel experiences rich scattering, e.g., low correlation or a small singular value spread.
In cellular communication systems, where multiple transmitters and receivers in different cells (or even the same cell) re-use the same radio resources, e.g., time slots and/or frequency bands, there will be interference between the transmissions. In addition, the further the receiver node 200 is from the transmitter node 100, and the more obstacles in between that block the direct radio path, the weaker the received signal will be. Thus, the signal levels tend to be highest when there is line of sight (LOS) path between the desired transmitter-receiver pair, and no LOS path for the interfering transmitters.
However, the channel gain of the data stream utilizing the LOS path and the others differ substantially as the scattered paths are much weaker, as shown in
In addition, various transmitter and receiver impairments will further exacerbate the problems of channel richness and of inter-stream interference. For example, non-linearities in the transmitter node 100 may cause a power-dependent error floor of the transmitted symbols. This error is commonly characterized by the Error Vector Magnitude (EVM), which is defined as the error of a complex modulation symbol divided by the amplitude of that symbol. Typical sources of EVM include thermal and phase noise, the dynamic range of the Digital-to-Analog (D/A) converter, quantization errors in the digital representation of the transmitted signals, and saturation or clipping in the power amplifiers. Similarly, the receiver node 200 may also suffer from various impairments that can be characterized by a receiver EVM. The EVM in the transmitter node 100 and/or receiver node 200 may be reduced by using more expensive, high quality components and complex circuitry. However, the cost-benefit trade-off in commercial and mass-market communication equipment usually leads to EVM values of at best around 3% or −30 dB. As a result, the SNR that the receiver experiences on its channel estimates will be upper limited by ˜30 dB. A large singular value spread or data stream amplitude spread will cause some of the weaker of the potential MIMO data streams to end up with low or negative (in dB) estimated SNRs. As the receiver node 200 tries to optimize the throughput it will most likely request a lower rank transmission to avoid wasting power on streams where little or no throughput is expected, and instead increase the power of the stronger data streams where the SNR gain will translate into a larger throughput gain. The fact that the channel estimates for the weaker data streams will be noisier also has consequences for the capabilities of the MIMO receiver node 200 to suppress the inter-stream interference, which further discourages the use of many streams.
The use of fewer MIMO data streams leads to lower peak data rates over the wireless connection, because the data rate per data stream is typically limited by the highest modulation and coding scheme for which equipment and standard is prepared. This effect has been observed in deployed systems, typically in situations where there LOS conditions exist between the transmitter node 100 and receiver node 200. The degradation can be quite large; the throughput can drop by a factor of 2 or even 3 compared to non-LOS (NLOS) conditions. Examples include situations where the terminal passes into a LOS path at a street corner; thus, the effect can be very rapid. The problem increases the higher the number of transmit and receive antennas there are in the MIMO link. As both LTE release 10 and IEEE 802.11ac have recently standardized up to 8×8 MIMO configurations, the problem is expected to become even more evident as 8-antenna products become available.
Antennas 112, 212 may be reconfigured using any known means, including through the use of phase shifters 116, motors 118, and/or switches 119. For example, in one exemplary embodiment the antenna array 110, 210 comprises a phase shifter 116 operatively coupled to one antenna element of each antenna 112, 212 (see
In one exemplary embodiment, antenna configuration unit 130 reconfigures the antenna(s) 112, 212 such that a first antenna 112, 212 receives the signal in a LOS direction, while steering the remaining antennas 112, 212 away from the LOS direction such that the remaining antennas 112, 212 receive signals from the NLOS directions while suppressing the signal from the LOS direction. It will be appreciated that when antenna array 110, 210 comprises dual polarization reconfigurable antennas 112, 212, each having dual polarized antenna elements, where one reconfigurable antenna for each polarization may be directed towards the LOS direction. In any event, antenna configuration unit 130 configures a “LOS” antenna 112, 212 to increase a gain of the LOS antenna 112, 212 for LOS direction while reconfiguring the remaining antennas 112, 212 to reduce the gain for the LOS direction. Such a configuration brings out the data streams associated with the previously hidden NLOS paths. Further, by keeping at least one antenna 112, 212 configured for the LOS direction, the LOS signal path may still be used for data transmission.
In another exemplary embodiment, the antenna configuration unit 130 reconfigures the antenna(s) 112, 212 to steer the antenna gain of all of the antennas 112, 212 away from the LOS direction, as shown in
The antenna configuration unit 130 may determine how to reconfigure the antennas 112, 212 based on any number of techniques. For example, antenna configuration unit 130 may evaluate different antenna array configurations based on a measured MIMO operating parameter, e.g., by iteratively changing the antenna array configuration according to a predetermined pattern and evaluating the MIMO operating parameter for each iterative change. Subsequently, the antenna configuration unit 130 selects the antenna array configuration providing the best MIMO operating parameter. In this example, the antenna configuration unit 130 reconfigures the antenna(s) according to the selected antenna array configuration. In another example, antenna configuration unit 130 reconfigures the antenna(s) 112, 212 based on an antenna configuration report received from a remote reporting node, e.g., the receiver node 200 or another remote network node.
Referring again to
According to another exemplary embodiment, path unit 150 includes an optional power unit 156 and the processor 152, where the power unit 156 is configured to determine a power associated with each of the antennas 112, 212. The processor 152 is configured to identify the LOS path based on a relative comparison between the determined powers. For example, for wireless links having throughput conditions limited by multiplicative noise, it is of interest to know which of the B data streams or pre-coders gives the highest received power, because it is the power and multiplicative noise associated with this data stream that causes the degradation for the weaker data streams. In some situations, such as for an LOS link with co-polarized antennas, there might be a single dominating path corresponding to one dominant pre-coder and data stream. In other situations, there might be two or more dominant paths/data streams of similar power, e.g., in a LOS link with dual-polarized antennas.
The power unit 156 may determine the power per data stream according to:
Pj=|H·wj|2, (8)
where wj represents the jth pre-coder vector of weights. This power may be determined for all possible pre-coder vectors, or for only those pre-coder vectors that correspond to the preferred or recommended pre-coder vectors (the number depending on the preferred or recommended rank). A MIMO node 100, 200 with limited channel state information, e.g., the transmitter node 100 in an FDD system, may instead utilize PMI and CQI feedback to determine dominating pre-coders. In some cases, CQI values are coupled directly to corresponding pre-coders, while in other cases one CQI value is coupled to a codeword that is mapped to multiple data streams. The CQI value, which is indicative of the estimated SNR (or power), may therefore uniquely identify one pre-coder, or it may identify the sum of two or more pre-coders. Either case provides beneficial information. Finally, a MIMO node 100, 200 with limited channel state information, e.g., a transmitter node 100 in an FDD system, may also utilize second order channel statistics measured on the reverse link, such as antenna correlations or direction estimation via various methods as known in the art, in order to determine which direction or pre-coder will give the highest received power in the receiver node 200.
As discussed herein, one or more of the antennas 112, 212 are reconfigured when multiplicative noise limits the throughput conditions, i.e., when the noise proportional to or dependent on the received (or transmitted) signal power experienced by the receiver node 200 on, e.g., its channel estimates, dominates. Thus, under such conditions, reducing the received/transmitted signal power corresponding to the strongest signal paths (e.g., the strongest channel eigenvalues) while maintaining the power corresponding to the weaker signal paths (e.g., the weaker channel eigenvalues) effectively results in significantly reduced multiplicative noise in the weaker signal paths. This is because the cross-talk from the strongest signal paths has been reduced, which in turn improves the per-data-stream SNR/SINR and subsequently the throughput over the wireless MIMO channel. If, on the other hand, the thermal noise and interference is stronger than the multiplicative noise then no such improvements are possible because any reduction in received (or transmitted) signal power will only degrade the SNR or SINR levels for all data streams.
Because the antenna reconfiguration disclosed herein depends on the knowledge of whether multiplicative noise limits throughput conditions, it is also beneficial to provide methods and apparatuses to detect when multiplicative noise limits the throughput. Multiplicative noise Nπ limits the throughput conditions when:
where Psig represents the signal power, σ2 represents the thermal noise power, and I represents the interference power. The multiplicative noise Nπ may be expressed as a combination of the multiplicative noise at the transmitter node 100Σtx2
and the multiplicative noise at the receiver node 200
Σrx2
according to:
Nπ=Σtx2
+
Σrx2
. (10)
The determination of whether multiplicative noise limits the throughput conditions may be made in the MIMO node 100, 200 reconfiguring the antennas 112, 212, or in a network node remote from the reconfiguring MIMO node 100, 200 that subsequently sends a quantitative or qualitative multiplicative noise evaluation to the reconfiguring MIMO node 100, 200. Further, multiplicative noise determination may be made along with the dominant path determination and separate from the antenna reconfiguration, e.g., in a separate node from the antenna reconfiguration. In this scenario, the multiplicative noise determination and/or the dominant path determination are reported to the reconfiguring node. For example, the noise processor 120 may report the multiplicative noise determination and the path unit 150 may report the dominant path determination.
Parameter unit 122 evaluates the throughput conditions based on a signal quality parameter and/or a MIMO utilization parameter (block 410). Characterization unit 128 generates a multiplicative noise evaluation RN
For example, in one embodiment, the parameter unit 122 may evaluate the throughput conditions by evaluating past observations, e.g., regarding whether multiplicative noise limited conditions previously existed for a particular cell, mobile device, location, etc. Such past observations may, for example, be stored in memory 129. In another embodiment, the parameter unit 122 comprises a signal quality unit 124 and a MIMO utilization unit 126. The signal quality unit 124 is configured to evaluate the throughput conditions by evaluating a signal quality parameter associated with the signals being communicated between two network nodes, while the MIMO utilization unit 126 is configured to evaluate the throughput conditions by evaluating a MIMO operating parameter associated with the communicated signals. Exemplary signal quality parameters include, but are not limited to, signal power, SNR, SINR, CQI report, etc. The signal quality unit 124 may estimate the SNR and/or SINR values during operation of the receiver node 200 using reference symbols, training sequences, or pilot signals, and/or a priori knowledge of the thermal noise level. The signal power and interference level can be measured directly using known techniques, or they could be inferred from predictions or measurements on the reverse link. Typically, the SNR or (SINR) estimates γ will contain contributions from all noise (and interference) sources, including additive noise σ2 and multiplicative noise Nπ, e.g.,:
Exemplary MIMO operating parameters include, but are not limited to, the throughput, rank indicator (RI), channel quality indicator (CQI), etc. The following provides multiple examples of different ways to determine whether multiplicative noise limits the throughput conditions. It will be appreciated that the claimed invention is not limited to these specific examples.
In a first exemplary embodiment, the characterization unit 128 compares the multiplicative noise Nπ with an SNR or SINR value γ estimated by the signal quality unit 124. The multiplicative noise Nπ might be known beforehand, e.g. from the design or specifications of the radio equipment, or it may have been characterized during manufacturing and programmed into the corresponding network device. Alternatively, Nπ may be measured directly in conditions with high received signal power. For example, the highest estimated SNR value over a range of channel and interference conditions typically represents a good estimate of the multiplicative noise. In any event, the comparison may be performed in the noise processor 120 of the receiver node 200, in the noise processor 120 of the transmitter node 100 (using feedback information provided by the receiver node 200, reverse-link measurements, or utilizing channel reciprocity), or in a third network node. If needed, existing or new feedback or signaling mechanisms might be required to ensure that the Nπ and SNR/SINR estimates are available at the node performing the comparison.
In any event, when a multiplicative noise parameter (1/Nπ) is less than or equal to γ, the throughput conditions are limited by multiplicative noise, as shown by Equation (9). Alternatively, a lower threshold may be used, e.g., if γ is within some predetermined range, e.g., 3 dB, of 1/Nπ, then there is a high likelihood that the MIMO throughput will be limited by multiplicative noise. In yet another alternative, a γ estimate that includes the effect of the multiplicative noise according to Equation (11) may be corrected according to:
where {circumflex over (γ)} represents the corrected estimate corresponding to the additive receiver noise only. In this case, it should be noted that:
It is obvious that this expression is positive if:
In other words, the expression of Equation (13) is positive if |γ| is within 3 dB of 1/Nπ. Multiplicative noise therefore limits the throughput conditions when Equation (14) holds true.
As noted above, the multiplicative noise evaluation for this exemplary embodiment may simply comprise a qualitative yes/no indication. Alternatively, multiplicative noise evaluation may be quantitative by indicating how much γ exceeds 1/Nπ, e.g., by representing the results of Equation (14) with a certain numerical precision. Such quantitative information provides an understanding of how much the total received (transmitted) power can be changed without changing the Nπ-dominance conditions.
In some cases, channel and Nπ estimates may be unreliable or even unavailable. In these cases, noise processor 120 in general, and MIMO utilization unit 126 in particular, uses secondary information to indirectly deduce whether multiplicative noise limits the MIMO throughput conditions. Because the primary difference between an Nπ-limited MIMO link and a non-Nπ-limited MIMO link is that only the latter will benefit from an increased received (transmitted) power level, this fact can be utilized to determine if the link is Nπ-limited or not. Thus, the MIMO utilization unit 126 of one exemplary embodiment measures at least one MIMO operating parameter (throughput, RI, CQI, etc.) for at least two different power settings. Characterization unit 128 determines whether multiplicative noise limits the throughput conditions by comparing the measured MIMO operating parameters for the different power settings. For example, assume MIMO utilization unit 126 measures two operating parameters M1 and M2 at two different power settings P1 and P2, respectively, where P1>P2. If M1 and M2 are substantially equal, or if M1≦M2, then characterization unit 128 indirectly determines that multiplicative noise limits the throughput conditions. To optimize the accuracy of this technique, the characterization unit must compare M1 and M2 measurements made under as similar channel and interference conditions as possible to avoid erroneous conclusions due to varying channel richness or interference. One simple way to achieve this is to first estimate M1 for a particular time slot where the received (or transmitted) power is P1 and subsequently estimate M2 for another time slot where the received (or transmitted) power equals P2. Alternatively, different frequency resources may be utilized, or even a combination of time slots and frequency slots. More generally, the channel and interference variations may be equalized between the two measurements by utilizing multiple time and frequency slots, or by utilizing diversity or coding techniques as known in the state of the art.
The powers may be varied either directly or indirectly. For example, the power may be directly varied at the transmitter using analog or digital methods, where the former includes attenuators or power amplifier settings while the latter may include changing the digital representation of the signals to be transmitter. At the receiver, the receiver preferably directly varies the power using analog methods because digital methods may reduce the additive noise by the same amount as the multiplicative noise. Alternatively, the naturally occurring power variation due to multipath fading may be utilized. Multipath fading causes the channel coefficients hnm to vary in time and/or frequency due to the constructive and destructive superposition of multiple radio waves. Therefore, parameter unit 122 may identify time instants or frequency bins where the power is closer to P1 or P2, where characterization unit 128 compares the measures M1 and M2 for these time instants or frequency bins. Because the channel richness also strongly depends on the multipath fading, however, such a method will be less precise. To compensate for this imprecision, some form of averaging over multiple observations may be used to improve the accuracy.
In another exemplary embodiment, the noise processor 120 may use a combination of signal quality parameters and MIMO utilization parameters. For example, low MIMO utilization may be caused by poor channel richness and/or Nπ-limitations. However, the higher the signal quality or power, the more unlikely it is that MIMO utilization is limited by additive noise or interference. Therefore, a combination of high signal power and low MIMO utilization provides a strong indication of multiplicative noise limited conditions. Thus, for this third exemplary embodiment, MIMO utilization unit 126 provides a MIMO utilization measure MMIMO to the characterization unit 128, and signal quality unit 124 provides a signal quality measure MQ to the characterization unit 128. The characterization unit 128 combines MMIMO with MQ to determine whether multiplicative noise limits the throughput conditions. For example, the characterization unit 128 may determine that multiplicative noise limits the throughput conditions when MMIMO<MMIMO,ref and MQ>MQ,ref, where MMIMO,ref and MQ,ref represent reference values. In some embodiments, MQ,ref depends on a MIMO utilization parameter. The reference values may be determined from simulations or from measurements under controlled conditions. For example, when MMIMO is the preferred rank RI while MQ is the CQI, a very high MQ indicates a low interference and noise condition, while a low MMIMO indicates a low MIMO utilization. The reference values could in this case be the CQI and RI values that, e.g., in simulations without Nπ, are appropriate in channel conditions with typical channel richness. It will be appreciated that a network node 100 may calibrate the reference values MQ,ref and MMIMO,ref using both direct observations and indirect observations of Nπ-limited conditions exist in parallel. These tuned reference values can then be used for the benefit of MIMO communications where only indirect observations are available.
In another exemplary embodiment, noise processor 120 may use past observations to predict whether a current transmitter-receiver link is likely to be subject to throughput conditions limited by multiplicative noise. The past observations may be from measurements with the same and/or different transmitter or receiver nodes, e.g. different terminals within the same cell area, and the observations may have been obtained with any of the methods described herein. For example, if multiple wireless terminals consistently report throughput conditions in a certain cell are limited by multiplicative noise, then noise processor 120 may use these observations to conclude that wireless terminals within the same cell are likely to experience throughput conditions limited by multiplicative noise. Similarly, if wireless terminals of a certain brand and/or type have reported multiplicative noise limited conditions at a particular received signal strength level, noise processor 120 may use this information to predict when other wireless terminals will experience the multiplicative noise limited throughput conditions. Further, positioning information may be used to further improve the detection of multiplicative noise limited conditions. For example, observations of multiplicative noise limited conditions may be tied to particular geographical areas with the help of positioning information available via numerous methods known in the art, e.g., GPS.
It will be appreciated that any one or more network nodes may implement the various operations disclosed here. For example, a first network node may determine whether multiplicative noise limits the throughput conditions, the same or a second network node may determine which pre-coder results in the high received power, and the same or a third network node may reconfigure the antennas. Whenever two or more nodes are involved, the information regarding the multiplicative noise limited conditions and/or the relevant pre-coders is communicated to the reconfiguring node, either directly or via one or more additional network nodes. Thus, different aspects of the embodiments disclosed herein may be implemented in different network nodes. Each such network node should therefore also be able to communicate with the other network nodes as needed.
The present invention may, of course, be carried out in other ways than those specifically set forth herein without departing from essential characteristics of the invention. The present embodiments are to be considered in all respects as illustrative and not restrictive, and all changes coming within the meaning and equivalency range of the appended claims are intended to be embraced therein.
| Number | Name | Date | Kind |
|---|---|---|---|
| 7505529 | Kwak et al. | Mar 2009 | B2 |
| 7729232 | Wang et al. | Jun 2010 | B2 |
| 8094744 | Song et al. | Jan 2012 | B1 |
| 8446318 | Ali et al. | May 2013 | B2 |
| 8588324 | Mondal et al. | Nov 2013 | B2 |
| 20050088358 | Larry et al. | Apr 2005 | A1 |
| 20070115180 | Kish et al. | May 2007 | A1 |
| 20090086648 | Xu et al. | Apr 2009 | A1 |
| 20090186658 | Jiang et al. | Jul 2009 | A1 |
| 20100002801 | Jia et al. | Jan 2010 | A1 |
| 20100291865 | Gore et al. | Nov 2010 | A1 |
| 20110051834 | Lee et al. | Mar 2011 | A1 |
| 20120120997 | Park et al. | May 2012 | A1 |
| 20120157001 | Chiu et al. | Jun 2012 | A1 |
| 20140029509 | Murakami et al. | Jan 2014 | A1 |
| Entry |
|---|
| Baum, D. et al., “Impact of Phase Noise on MIMO Channel Measurement Accuracy”, IEEE 60th Vehicular Technology Conference, Sep. 26, 2004, pp. 1614-1618, vol. 3, IEEE. |
| Lozano, A. et al., “Mercury/Waterfilling: Optimum Power Allocation with Arbitrary Input Constellations”, International Symposium on Information Theory, Proceedings, Sep. 4, 2005, pp. 1773-1777, IEEE. |
| Yoo, T. et al. “Multi-Antenna Downlink Channels with Limited Feedback and User Selection.” IEEE Journal on Selected Areas in Communications, vol. 25, No. 7, Sep. 2007, pp. 1478-1491. |
| Grant, S. et al. “Performance Enhancement Through Joint Detection of Cochannel Signals Using Diversity Arrays.” IEEE Transactions on Communications, vol. 46, No. 8, Aug. 1998, pp. 1038-1049. |
| Number | Date | Country | |
|---|---|---|---|
| 20140169430 A1 | Jun 2014 | US |
