Ferkans — Interactive Telecom Tutor

Trading Diversity for Multiplexing

Space-time codes (Section 16.1) focus on diversity gain. An alternative strategy is spatial multiplexing: sending independent data streams from each antenna to maximise throughput. The Bell Labs Layered Space-Time (BLAST) architectures pioneered this approach, demonstrating that MIMO channels can support multiple parallel data streams without additional bandwidth or power.

Definition:
Spatial Multiplexing

Spatial multiplexing transmits $N_s \leq \min(N_t, N_r)$ independent data streams (layers) simultaneously from $N_t$ antennas. The transmitted vector is:

$\mathbf{x} = [x_1, x_2, \ldots, x_{N_t}]^T$

where each $x_i$ is an independently encoded symbol. The received signal is:

$\mathbf{y} = \mathbf{H}\mathbf{x} + \mathbf{n}$

The achievable rate scales linearly with $\min(N_t, N_r)$ at high SNR, compared to the logarithmic scaling of single-antenna systems.

Spatial multiplexing requires $N_r \geq N_t$ for the system to be (over-)determined. When $N_r < N_t$ , the system is underdetermined and some form of precoding or code design is needed to separate the layers.

,

Definition:
V-BLAST (Vertical BLAST)

V-BLAST (Vertical Bell Labs Layered Space-Time) is a spatial multiplexing architecture where:

The input bit stream is demultiplexed into $N_t$ sub-streams.
Each sub-stream is independently encoded and modulated.
The $i$ -th sub-stream is transmitted from the $i$ -th antenna.

At the receiver, the layers are detected using successive interference cancellation (SIC): the strongest layer is detected first, its contribution is subtracted from the received signal, and the process repeats for the remaining layers.

The key innovation is the detection ordering: layers are detected in order of decreasing post-detection SNR, not in the natural antenna order.

V-BLAST does not use temporal coding across antennas — each antenna simply transmits its own independent data. The "vertical" refers to the fact that each layer corresponds to a single antenna (a column of $\mathbf{H}$ ).

,

Definition:
D-BLAST (Diagonal BLAST)

D-BLAST (Diagonal Bell Labs Layered Space-Time) distributes each data stream diagonally across antennas and time, so that each stream experiences all $N_t$ transmit-antenna channels:

$\text{Stream } k \text{ uses antenna } ((k + t - 1) \bmod N_t) + 1 \text{ at time } t$

This diagonal structure ensures that every stream sees all spatial sub-channels, providing better average performance per stream. However, D-BLAST requires a more complex receiver and incurs an initial space-time wastage of $(N_t - 1)$ time slots for the diagonal ramp-up.

D-BLAST was Foschini's original 1996 proposal. V-BLAST was developed later as a simpler implementation that sacrificed some theoretical elegance for practical feasibility.

V-BLAST Detection Performance

Compare the BER of V-BLAST with different detection orderings and receiver types. Observe how optimal ordering dramatically improves performance over natural (fixed) ordering.

Parameters

N_t

(transmit antennas)4

N_r

(receive antennas)4

Modulation

Detector type

Example: V-BLAST Detection Ordering for a 2x2 System

Consider a $2 \times 2$ MIMO system with channel matrix:

$\mathbf{H} = \begin{bmatrix} 1 & 0.5 \\ 0.3 & 1.2 \end{bmatrix}$

and QPSK symbols $\mathbf{x} = [x_1, x_2]^T$ . Using ZF-based V-BLAST, determine the optimal detection order.

Solution

Compute ZF filter rows

The ZF filter is $\mathbf{G} = (\mathbf{H}^{H}\mathbf{H})^{-1}\mathbf{H}^{H}$ .

$\mathbf{H}^{H}\mathbf{H} = \begin{bmatrix} 1.09 & 0.86 \\ 0.86 & 1.69 \end{bmatrix}$

$(\mathbf{H}^{H}\mathbf{H})^{-1} = \frac{1}{1.09 \times 1.69 - 0.86^2} \begin{bmatrix} 1.69 & -0.86 \\ -0.86 & 1.09 \end{bmatrix} = \begin{bmatrix} 1.971 & -1.003 \\ -1.003 & 1.271 \end{bmatrix}$

Determine detection order

The post-detection noise enhancement for layer $k$ is $\|[\mathbf{G}]_k\|^2 = [(\mathbf{H}^{H}\mathbf{H})^{-1}]_{kk}$ .

Layer 1: noise enhancement $= 1.971$
Layer 2: noise enhancement $= 1.271$

Since layer 2 has lower noise enhancement (higher post-detection SNR), we detect layer 2 first.

Summary

The optimal V-BLAST ordering is: detect $x_2$ first, subtract its contribution $\hat{x}_2 \mathbf{h}_2$ from $\mathbf{y}$ , then detect $x_1$ from the interference-cancelled signal. This ordering minimises the probability that error propagation corrupts subsequent layers. $\blacksquare$

Quick Check

In a $4 \times 4$ V-BLAST system with MMSE-SIC detection, which layer is detected first?

The layer transmitted from antenna 1 (natural order)

The layer with the highest post-detection SINR

The layer with the lowest post-detection SINR

All layers are detected simultaneously

Correction:

The layer with the highest post-detection SINR

The optimal ordering detects the layer with the highest post-detection SINR first, minimising error propagation to subsequent layers.

Historical Note: Bell Labs and the BLAST Revolution

1996-1998

Gerard J. Foschini at Bell Labs published the D-BLAST concept in 1996, showing that MIMO spectral efficiency grows linearly with the minimum number of antennas — a revolutionary insight at the time. His colleague Emre Telatar independently derived the MIMO capacity formula in 1999. The V-BLAST variant was demonstrated experimentally at Bell Labs in 1998 by Wolniansky, Foschini, Golden, and Valenzuela, achieving 40 bits/s/Hz spectral efficiency in a laboratory $8 \times 12$ system — a world record at the time. This experimental demonstration transformed MIMO from a theoretical curiosity into a practical technology pursued by the entire wireless industry.

V-BLAST

Vertical Bell Labs Layered Space-Time: a spatial multiplexing architecture where each antenna transmits an independent data stream, detected at the receiver using ordered successive interference cancellation.

Related: D-BLAST, Spatial Multiplexing

D-BLAST

Diagonal BLAST: a layered space-time architecture where each data stream is diagonally distributed across antennas and time slots to average the spatial sub-channel qualities.

Related: V-BLAST, Spatial Multiplexing

Spatial Multiplexing

The technique of transmitting multiple independent data streams over the parallel spatial sub-channels of a MIMO system, with throughput scaling linearly with $\min(N_t, N_r)$ at high SNR.

Related: V-BLAST, D-BLAST

V-BLAST and D-BLAST