MiMo
Italian record label
≈ 13 min read
MiMo. encyclopegia.org, snapshot mimo-444017-741f628c3ddf6f6d. /a/mimo
Names in other languages
- EnglishMiMo
- EspañolMiMo
- ItalianoMiMo
Article bodies are English and are not machine-translated. A name in another language is not a translation of the article.
Mimoň ( ) is a town in Česká Lípa District in the Liberec Region of the Czech Republic. It has about 6,300 inhabitants. The town is located on the Ploučnice River in the Ralsko Uplands.
Mimoň became a town at the beginning of the 16th century. The main landmark of the town is the Church of Saints Peter and Paul.
MIMO ( ), or multiple-input multiple-output, is a wireless technology that multiplies the capacity of a radio link using multiple transmit and receive antennas. MIMO has become a core technology for broadband wireless communications, including mobile standards—4G WiMAX (802.16 e, m), and 3GPP 4G LTE and 5G NR, as well as Wi-Fi standards, IEEE 802.11n, ac, and ax.
MIMO uses the spatial diversity to increase link capacity. The technology requires multiple antennas at both the transmitter and receiver, along with associated signal processing, to deliver data rate speedups roughly proportional to the number of antennas at each end. MIMO starts with a high-rate data stream, which is demultiplexed into multiple, lower-rate streams.
Each of these streams is then modulated and transmitted in parallel with different coding from the transmit antennas, with all streams in the same frequency channel. These co-channel, mutually interfering streams arrive at the receiver's antenna array, each having a different spatial signature—gain phase pattern at the receiver’s antennas.
These distinct array signatures allow the receiver to separate these co-channel streams, demodulate them, and re-multiplex them to reconstruct the original high-rate data stream. This process is sometimes referred to as spatial multiplexing. The key to MIMO is the sufficient differences in the spatial signatures of the different streams to enable their separation.
This is achieved through a combination of angle spread of the multipaths and sufficient spacing between antenna elements. In environments with a rich multipath and high angle spread, common in cellular and Wi-Fi deployments, an antenna element spacing at each end of just a few wavelengths can suffice.
However, in the absence of significant multipath spread, larger element spacing (wider angle separation) is required at either the transmit array, the receive array, or at both. \mathbf{y} = \mathbf{H}\mathbf{x} + \mathbf{n} where \mathbf{y} and \mathbf{x} are the receive and transmit vectors, respectively, and \mathbf{H} and \mathbf{n} are the channel matrix and the noise vector, respectively. ) and ergodic open-loop (channel is unknown, no CSI) capacities. Number of transmit and receive antennas is 4 ( N_r=N_t=4 ).
Referring to information theory, the ergodic channel capacity of MIMO systems where both the transmitter and the receiver have perfect instantaneous channel state information is C_\mathrm{perfect-CSI} = E\left[\max_{\mathbf{Q}; \, \mbox{tr}(\mathbf{Q}) \leq 1} \log_2 \det\left(\mathbf{I} + \rho \mathbf{H}\mathbf{Q}\mathbf{H}^{H}\right)\right] = E\left[\log_2 \det\left(\mathbf{I} + \rho \mathbf{D}\mathbf{S} \mathbf{D} \right)\right] where ()^H denotes Hermitian transpose and \rho is the ratio between transmit power and noise power (i.e., transmit SNR). The optimal signal covariance \mathbf{Q}=\mathbf{VSV}^H is achieved through singular value decomposition of the channel matrix \mathbf{UDV}^H \,=\, \mathbf{H} and an optimal diagonal power allocation matrix \mathbf{S}=\textrm{diag}(s_1,\ldots,s_{\min(N_t, N_r)},0,\ldots,0) .
The optimal power allocation is achieved through waterfilling, that is s_i = \left(\mu - \frac{1}{\rho d_i^2} \right)^+, \quad \textrm{for} \,\, i=1,\ldots,\min(N_t, N_r), where d_1,\ldots,d_{\min(N_t, N_r)} are the diagonal elements of \mathbf{D} , (\cdot)^+ is zero if its argument is negative, and \mu is selected such that s_1+\ldots+s_{\min(N_t, N_r)}=N_t . If the transmitter has only statistical channel state information, then the ergodic channel capacity will decrease as the signal covariance \mathbf{Q} can only be optimized in terms of the average mutual information as C_\mathrm{statistical-CSI} = \max_{\mathbf{Q}} E\left[\log_2 \det\left(\mathbf{I} + \rho \mathbf{H}\mathbf{Q}\mathbf{H}^{H}\right)\right].
The spatial correlation of the channel has a strong impact on the ergodic channel capacity with statistical information. If the transmitter has no channel state information it can select the signal covariance \mathbf{Q} to maximize channel capacity under worst-case statistics, which means \mathbf{Q}=1/N_t \mathbf{I} and accordingly C_\mathrm{no-CSI} = E\left[\log_2 \det\left(\mathbf{I} + \frac{\rho}{N_t}\mathbf{H}\mathbf{H}^{H}\right)\right].
Depending on the statistical properties of the channel, the ergodic capacity is no greater than \min(N_t, N_r) times larger than that of a SISO system.
MIMO detection
The MIMO system can be described by: \mathbf{y} = \mathbf{H} \mathbf{x} + \mathbf{n} , where \mathbf{y} is the received vector, \mathbf{H} is the channel matrix, \mathbf{x} is the transmitted vector, and \mathbf{n} is the noise vector. The goal of MIMO detection is to estimate \mathbf{x} from \mathbf{y} given knowledge of \mathbf{H} .This can be posed as a statistical detection problem, and addressed using a variety of techniques including zero-forcing, successive interference cancellation a.k.a. V-blast, maximum likelihood estimation and recently, neural network MIMO detection.
Such techniques commonly assume that the channel matrix \mathbf{H} is known at the receiver. In practice, in communication systems, the transmitter sends a pilot signal and the receiver learns the state of the channel (i.e., \mathbf{H} ) from the received signal Y and the pilot signal X . Recently, there are works on MIMO detection using deep learning tools which have shown to work better than other methods such as zero-forcing. \hat{x}_{ZF} = G_{ZF} z, where G_{ZF} is the pseudo-inverse of matrix H and is given by: G_{ZF} = H^{\dagger} = (H^H H)^{-1} H^H. Despite its simplicity, this approach suffers from noise enhancement.
After decoupling by Equation, the ZF solution \hat{x}_{ZF} is either quantized and demapped to binary bits or used to compute the LLR. Note that such an approximation introduces negligible error rate degradation and significantly reduces the computation needed.
As the ZF detection decouples the multiple correlated streams into independent streams, the extrinsic LLR of the j th bit of the current symbol in the p th stream resembles the soft-output equalization, and is given by: L_{j}^{(p),E} = \frac{1}{\|g_p\|^2} \left( \max_{X^{(p)} \in \chi_{1,j}} \left[ -|\hat{X}^{(p)}_{ZF} - X^{(p)}|^2 \right] - \max_{X^{(p)} \in \chi_{-1,j}} \left[ -|\hat{X}^{(p)}_{ZF} - X^{(p)}|^2 \right] \right), where g_p denotes the p th column vector of matrix G_{ZF}^T , \hat{X}^{(p)}_{ZF} is the p th element of the symbol vector \hat{x}_{ZF} , and \chi_{b,j} indicates the subset of constellation points whose j th bit has value b . \tilde{\mathbf{x}} = \mathbf{G}_{\mathrm{MMSE}} \mathbf{z} \quad \text{(1.1)} where \mathbf{G}_{\mathrm{MMSE}} = \mathbf{R}_{\mathbf{xz}} \mathbf{R}_{\mathbf{zz}}^{-1} \quad \text{(1.2)} Note that the cross-correlation matrix \mathbf{R}_{\mathbf{xz}} is computed as: \mathbf{R}_{\mathbf{xz}} = \mathbb{E}\{\mathbf{x} \mathbf{z}^H\} = \mathbf{R}_{\mathbf{xx}} \mathbf{H}^H = \sigma_x^2 \mathbf{H}^H \quad \text{(1.3)} whereas the auto-correlation matrix is given by: \mathbf{R}_{\mathbf{zz}} = \mathbb{E}\{\mathbf{z} \mathbf{z}^H\} = \sigma_x^2 (\mathbf{H} \mathbf{H}^H) + \sigma_v^2 \mathbf{I}_Q \quad \text{(1.4)} where \sigma_x^2 and \sigma_v^2 are the signal energy and the noise variance, respectively. Combining the above three equations, one obtains: \mathbf{G}_\mathrm{MMSE} = \mathbf{H}^H \left( \mathbf{H} \mathbf{H}^H + \frac{\mathbf{I}_Q}{\rho} \right)^{-1} = \left( \mathbf{H}^H \mathbf{H} + \frac{\mathbf{I}_P}{\rho} \right)^{-1} \mathbf{H}^H with the SNR \rho = \sigma_x^2 / \sigma_v^2 .
The effective SINR of the signal in the p th stream of the MMSE detection output can be formulated as: \alpha^{(p)} = \mathbf{h}_p^H \left( \mathbf{H}_{(-p)} \mathbf{H}_{(-p)}^H + \frac{\mathbf{I}_Q}{\rho} \right)^{-1} \mathbf{h}_p \quad \text{(1.5)} where \mathbf{H}_{(-p)} represents the matrix \mathbf{H} with the p th column removed, and \mathbf{h}_p is the p th column vector of \mathbf{H} . Equation (1.1) is referred to as the biased MMSE detector because the detected signal power is smaller than the transmitted signal power by a factor of \alpha^{(p)} / (\alpha^{(p)} + 1) .
To avoid this degradation, an unbiased MMSE detector has been proposed: \hat{\mathbf{x}}_{\mathrm{MMSE}} = \mathbf{B} \mathbf{G}_{\mathrm{MMSE}} \mathbf{z} \quad \text{(1.6)} where \mathbf{B} is a diagonal matrix with the p th diagonal element equal to (\alpha^{(p)} + 1)/\alpha^{(p)} . The unbiased MMSE detection solution has better BER performance than the biased MMSE detection solution.
Interestingly, this phenomenon implies that minimizing the MSE does not necessarily minimize the BER. The soft-output unbiased MMSE detection is similar to the soft-output ZF detection, thus: L_j^{(p),E} = \alpha^{(p)} \left( \max_{X^{(p)} \in \chi_{1,j}} \left[ -\left| \hat{X}^{(p)}_{\mathrm{MMSE}} - X^{(p)} \right|^2 \right] - \max_{X^{(p)} \in \chi_{-1,j}} \left[ -\left| \hat{X}^{(p)}_{\mathrm{MMSE}} - X^{(p)} \right|^2 \right] \right) \hat{X}^{(p)} = Q\left( \mathbf{g}_{p(i)}^T \mathbf{z}^{(i)} \right) \quad where Q(\cdot) denotes the quantizer and \mathbf{g}_{p(i)} is the column vector of \mathbf{G}^{(i)T} .
Then, the interference from \hat{X}^{(p)} is removed: \mathbf{z}^{(i+1)} = \mathbf{z}^{(i)} - \mathbf{h}_p \hat{X}^{(p)} Note that OSIC is known to perform better than linear detectors at high SNR, but it is worse at low SNR. Therefore, adequately switching between linear detection and OSIC can further improve the error rate performance. \mathbf{H} = \begin{bmatrix} H(0,0) & H(0,1) \\ H(1,0) & H(1,1) \end{bmatrix} The OSIC scheme checks rows of matrix \mathbf{G} , and if: |H(0,0)|^2 + |H(1,0)|^2 then the detected signal can be computed by: \tilde{X}^{(1)} = \frac{1}{\det(\mathbf{H})} \begin{bmatrix} - H(1,0) & H(0,0) \end{bmatrix} \begin{bmatrix} Z(0) \\ Z(1) \end{bmatrix}, \quad \hat{X}^{(1)} = Q(\tilde{X}^{(1)}) and \tilde{X}^{(0)} = \frac{1}{|H(0,0)|^2 + |H(1,0)|^2} \begin{bmatrix} H(0,0)^* & H(1,0)^* \end{bmatrix} \left( \begin{bmatrix} Z(0) \\ Z(1) \end{bmatrix} - \hat{X}^{(1)} \begin{bmatrix} H(0,1) \\ H(1,1) \end{bmatrix} \right), \quad \hat{X}^{(0)} = Q(\tilde{X}^{(0)}) Otherwise, if: |H(0,0)|^2 + |H(1,0)|^2 \geq |H(0,1)|^2 + |H(1,1)|^2, then: \tilde{X}^{(0)} = \frac{1}{\det(\mathbf{H})} \begin{bmatrix} - H(0,1) & H(1,1) \end{bmatrix} \begin{bmatrix} Z(1) \\ Z(0) \end{bmatrix}, \quad \hat{X}^{(0)} = Q(\tilde{X}^{(0)}) and \tilde{X}^{(1)} = \frac{1}{|H(1,1)|^2 + |H(0,1)|^2} \begin{bmatrix} H(1,1)^* & H(0,1)^* \end{bmatrix} \left( \begin{bmatrix} Z(1) \\ Z(0) \end{bmatrix} - \hat{X}^{(0)} \begin{bmatrix} H(1,0) \\ H(0,0) \end{bmatrix} \right), \quad \hat{X}^{(1)} = Q(\tilde{X}^{(1)}) \hat{\mathbf{x}}_{ML} = \arg\min_{\mathbf{x} \in \chi^{N_t}} \|\mathbf{y} - \mathbf{H} \mathbf{x}\|^2 Although ML provides optimal performance, its complexity grows exponentially with the number of transmit antennas and modulation order, making it impractical for large MIMO systems. \mathbf{x}_{\mathrm{ML}} = \arg\min_{\mathbf{x} \in \chi^P} \left\| \mathbf{z} - \mathbf{H} \mathbf{x} \right\|^2 = \arg\min_{\mathbf{x} \in \chi^P} M(\mathbf{x}) \quad where \chi^P is the P -fold Cartesian product over constellation set \chi , and M(\mathbf{x}) is the metric value of a symbol vector.
The ML detector must search all possible combinations of P symbols, thus the complexity grows exponentially with P . In light of this huge complexity, the sphere decoder (SD) was proposed to reduce the search space in an ML MIMO detector. The SD only searches those constellation points lying within a P -dimensional hypersphere.
This is effective only when the radius d is large enough to include the ML solution: M(\mathbf{x}) The QR decomposition (QRD) is typically applied to convert the exhaustive search into a constrained tree search: \mathbf{x}_{\mathrm{ML}} = \arg\min_{\mathbf{x} \in \chi^P} \left\| \tilde{\mathbf{z}} - \mathbf{R} \mathbf{x} \right\|^2, \quad \text{where} \quad \tilde{\mathbf{z}} = \mathbf{Q}^H \mathbf{z} \quad Let the (i, j) th element in \mathbf{R} be R_{i,j} and the p th element in \tilde{\mathbf{z}} be \tilde{Z}(p) . Then the metric M(\mathbf{x}) can be expressed as: M(\mathbf{x}) = \sum_{p=0}^{P-1} T(p) \quad where the partial distance (PD) is defined as: T(p) = \left| \tilde{Z}(p) - \sum_{j=p}^{P-1} R_{p,j} X(j) \right|^2 . The resulting sphere decoding process becomes a P -level tree search.
In level L , only child nodes from parent nodes satisfying: \sum_{p=L+1}^{P-1} T(p) are considered. Once the accumulated partial distance: T(L) + \sum_{p=L+1}^{P-1} T(p) exceeds d^2 , all nodes in the subtree rooted at that child node are removed from the search space.
When nodes at the bottommost layer are visited, the ML solution: \mathbf{x}_{\mathrm{ML}} = \left[ \hat{X}(0), \hat{X}(1), \dots, \hat{X}(P-1) \right] is the path with the minimum metric value. For example, one starts from variable X(P-1) and discards all the nodes where: T(P-1) > d^2 .
Then, for surviving X(P-1) nodes, the SD procedure proceeds to examine all the underlying X(P-2) and again discards those partial vectors [X(P-1), X(P-2)] for which: T(P-1) + T(P-2) > d^2 . Since T(p) > 0 , the accumulated PDs increase monotonically, and more nodes are pruned at lower layers.
With careful design of the radius and search strategy, sphere decoding can approach ML performance with significantly lower average complexity. Different tree search algorithms significantly affect the sphere decoder's efficiency. In algorithm design, tree search strategies are commonly categorized into three major types: depth-first search, breadth-first search, and best-first search.
Text derived from https://en.wikipedia.org/w/index.php?oldid=1363900413 under CC BY-SA 4.0; modifications have been made.