Multi-Satellite Beamspace Transmission

• Multi-Satellite Multi-Stream beamspace transmission is a distributed MIMO framework that leverages multiple satellites to deliver independent data streams using beamspace (DFT) precoding.
• It employs beamspace channel representations and low-complexity, geometry-based precoding to maximize spectral efficiency under limited CSI and synchronization constraints.
• Practical implementations and field trials show near-centralized performance with effective clustering, beam selection, and signal covariance optimization for enhanced multiplexing gains.

Multi-Satellite Multi-Stream (MSMS) beamspace transmission refers to the cooperative use of multiple satellites—typically in low Earth orbit (LEO) or geostationary orbit (GEO)—to form a distributed multiple-input multiple-output (MIMO) system capable of delivering multiple independent data streams to single or multiple user terminals (UTs). MSMS leverages beamspace (DFT-based or codebook-defined) channel representations and beam-domain precoding methods, enabling high spectral efficiency in the presence of strict synchronization, limited channel state information (CSI), and power constraints. The MSMS paradigm generalizes classical terrestrial massive MIMO, exploiting the unique geometry and deterministic propagation of space-based distributed antenna arrays to achieve multiplexing gains unattainable by single-satellite or traditional multibeam satellite systems (Röper et al., 2021, Wang et al., 26 Dec 2025).

1. System Models and Channel Representation

In MSMS transmission, $N$ distributed satellites, each with $N_{t,\ell}$ transmit antennas (or single feeds in GEO cluster systems), form the transmit side. The satellites orbit Earth in a defined formation, often one-dimensional "trails" at orbital radius $r_0 = r_E + d_0$ (Earth radius $r_E \approx 6371$ km, orbital altitude $d_0$, e.g., 600 km in LEO). Adjacent satellites are separated by $D_S$, yielding angular separation $$\Delta\vartheta = \arccos(1 - \frac{D_S^2}{2 r_0^2})$$. Each satellite $\ell$ occupies polar coordinates $$(r_0, \vartheta_\ell) = (r_0, \vartheta_1 + (\ell-1) \Delta\vartheta)$$ (Röper et al., 2021).

Ground stations (GS) or user terminals are equipped with a ULA of $N_r$ antennas, each with half-wavelength spacing. The predominance of line-of-sight (LoS) propagation in space channels simplifies the electromagnetic channel modeling: each satellite–GS/UT link can be approximated by a single dominant path. The channel matrix from satellite $\ell$ is

$$H_\ell \approx \alpha_\ell\, a_r(\theta_\ell)\, a_t(\phi_\ell)^H,$$

where $\alpha_\ell$ is the large-scale path gain and phase, $a_r(\theta_\ell)$ and $a_t(\phi_\ell)$ are GS and satellite ULA steering vectors for angles of arrival (AoA) and departure (AoD), respectively.

The full aggregate channel from all cooperating satellites is

$$H = [ H_1, H_2, \ldots, H_N ] \in \mathbb{C}^{N_r \times N_t}=A_r\,\textrm{Diag}(\alpha_1, \ldots, \alpha_N)\,A_t^H,$$

with $A_r$ comprising receive array responses and $A_t$ the block-diagonal stacking of the transmit steering vectors.

In beamspace representation, this channel aligns with a singular value decomposition

$H = U_r\,\Sigma\,U_t^H,$

where $U_r$ and $U_t$ approximate DFT matrices ("DFT beams"), and $\Sigma$ contains $N$ significant singular values. For GEO field trials with two satellites and two single-antenna UTs ($N=K=2$), a $2 \times 2$ complex channel matrix $H$ models slow phase drifts and LoS gains (Storek et al., 2020).

2. Distributed and Beamspace Precoding Schemes

MSMS transmission employs distributed linear precoding, where each satellite transmits its allocated data stream using only local position and GS/UT direction knowledge. Geometry-based transmit beams are formed as

$$w_\ell = \sqrt{ \rho_\ell / N_{t,\ell} }\, a_t(\phi_\ell), \quad W = \textrm{diag}(w_1, \ldots, w_N),$$

with $\rho_\ell$ as the per-satellite power budget. Crucially, no instantaneous CSI exchange or inter-satellite coordination is required; deterministic geometric parameters suffice (Röper et al., 2021).

At the GS/UT, detection employs beamspace linear equalization (zero forcing, MMSE, or matched filter), exploiting the angular orthogonality of incoming beams: $$y = HWs + n, \quad G_{\text{ZF}} = (H_{\text{geom}} W)^{+}.$$ Beam selection (truncated DFT/beamspace projection) is applied when $N_r \gg N$ to suppress sidelobe interference, reducing effective processing complexity.

In scenarios without instantaneous CSI, statistical CSI (sCSI) and codebook-based beams are used. The MSMS beamspace model defines the multi-satellite, multi-user channel as

$$\bar{\mathbf H}_k^{\textrm{MS}} = [\,\bar{\mathbf H}_{s,k}\,]_{s\in\mathcal S_k} \in \mathbb{C}^{N_R \times \sum_s B_s},$$

where each satellite $s$ serves UT $k$ with $B_s$ selected DFT beams (Wang et al., 26 Dec 2025).

3. Optimization: Clustering, Beam Selection, and Precoder Design

Designing optimal MSMS systems entails three main coupled tasks:

• Satellite clustering: Determining which satellites serve each UT, based on large-scale gains $\gamma_{s,k}$, Rician $\kappa_{s,k}$, and per-satellite user capacity constraints.
• Beam selection: Choosing the subset of DFT beams/steering directions per satellite to maximize effective channel power, using low-complexity two-stage heuristics.
• Precoding: Computing the transmit precoder (per-satellite, per-user) under power constraints, often via convex or non-convex optimization.

With only sCSI, the sum-rate upper-bound is approximated as

$$\bar R_k = \log_2\det\left( I + \bar{\mathbf R}_{\rm other,k}^{-1} \bar{\mathbf R}_{\rm sig,k} \right),$$

where $\bar{\mathbf R}_{\rm sig,k}$ and $\bar{\mathbf R}_{\rm other,k}$ are desired signal and interference-plus-noise covariances, computable from location, fading, and phase-error statistics. Joint optimization is typically posed as a covariance decomposition weighted MMSE (CDWMMSE) problem.

Closed-form decomposition of signal covariances and iterative algorithms, as detailed in (Wang et al., 26 Dec 2025), enable practical precoder computation at typical system scales. Heuristic closed-form precoders—such as covariance-decomposition MMSE (MS$^2$CDM) and location-information-based (LIB) schemes—provide nearly optimal performance at significantly reduced complexity.

The table below summarizes key algorithmic approaches in MSMS optimization:

Aspect	Method/Class (per (Wang et al., 26 Dec 2025))	Complexity
Precoding	Iterative CDWMMSE (Alg. 1)	High
Precoding	MS$^2$CDM (covariance MMSE, closed-form)	Medium
Precoding	Location-info-based (LIB)	Low
Clustering	Enhanced competition-based (per-user satellite quota)	Low
Beam Selection	Two-stage LCMS (effective power-based)	Low

4. Performance Analysis and Optimal Array Geometry

MSMS architectures can approach or match the spectral efficiency of ideal, full-CSI, centralized MIMO systems, even while operating with only local ephemeris- or sCSI-based control and no inter-satellite exchange (Röper et al., 2021, Wang et al., 26 Dec 2025).

Spectral efficiency: The sum-rate depends on satellite spacing $D_S$, receive array size $N_r$, the number of cooperating satellites $N$ or $S$, transmit power, and the degree of orthogonality in AoA at the GS/UTs. The achievable post-equalizer SINR per stream $\ell$ is: $$\text{SINR}_\ell = \frac{ |g_\ell^H H_{\text{geom},\ell} w_\ell|^2 }{ \sum_{i \neq \ell} |g_\ell^H H_{\text{geom},i} w_i|^2 + \sigma_n^2 \|g_\ell\|^2 }.$$ The sum-rate is maximized when uplink beams are nearly orthogonal at the receiver, requiring proper design of $D_S$. Analytical expressions yield optimal inter-satellite spacing

$$D_{S,\text{opt}} \approx r_0 \cdot \arccos\left( 1 - \frac{1}{2}(2k/N_r)^2 \right)$$

to ensure AoA orthogonality (Röper et al., 2021).

Simulation and field trial results:

• LEO simulations with $N = 2-6$, $N_r$ fixed at 100, total $N_t = 60$ distribute one stream per satellite, showing that linear MSMS (no instantaneous CSI) attains 99.8% of the centralized SVD performance across full SNR regimes. Throughput peaks at $D_S \approx 60-70$ km for typical orbit geometry, consistent with theory.
• GEO field trials over two co-located satellites demonstrated nearly $2 \times$ sum-rate gain in real video streaming with only minor technical overhead for synchronization and CSI (Storek et al., 2020).

5. Channel State Information and Synchronization

MSMS operation is highly sensitive to time-varying phases, oscillator offsets, and propagation delays across the distributed elements. Multiple strategies are adopted:

• LEO deterministic geometry eliminates small-scale fading, enabling all precoding to rely on satellite ephemeris and GS/UT location.
• GEO MSMS requires accurate real-time estimation of complex channel coefficients via orthogonal Zadoff-Chu pilots and pilot-aided least-squares estimation. Estimates are aggregated (e.g., five consecutive averages at 0.2 Hz), then returned to the gateway for precoder computation (Storek et al., 2020).
• Carrier frequency and phase drift (Doppler and oscillator effects) are compensated in GEO via narrow-band reference tone transmission and all-digital PLL tracking, ensuring residual inter-satellite phase errors remain within ZF tolerance thresholds (e.g., $\sigma_{\Delta\phi}=5^\circ$).

Residual errors, reference signal and pilot overhead, and synchronization bandwidths constitute key practical considerations in field deployments.

6. Practical Realizations and Impact

MSMS beamspace transmission has been demonstrated both in over-the-air field trials and in detailed simulations:

• In the first field trial over two GEO satellites, a sum-rate improvement from $5.5$ to $10.3$ bit/s/Hz was observed for two single-antenna users using ZF beamspace precoding, with real-time, independent MPEG video streams decoded on COTS receivers. Overhead levels and residual phase errors were found to be compatible with robust MSMS operation under QPSK 5/6 modulation (Storek et al., 2020).
• Simulations for massive MIMO LEO clusters confirm that beamspace MSMS with a limited number of DFT/codebook beams (e.g., 48 out of 256 per satellite) delivers more than 90% of the rate of full-dimension MIMO with less than 1% of the computational complexity, provided the number of data streams does not exceed $\min\{ S_k, N_R\ }$ (Wang et al., 26 Dec 2025).

MSMS frameworks thus efficiently combine distributed MIMO multiplexing, beamspace channel modeling, and sCSI-driven joint optimization, thereby offering scalable, low-overhead solutions for next-generation satellite downlinks in dense UT environments.

7. Research Directions and Limitations

Ongoing work extends MSMS theory and design to:

• Enhanced synchronization and compensation for higher-order modulations and lower SNR regimes.
• Advanced user scheduling, adaptive clustering, and beam selection under mobility or traffic variability.
• Integration with terrestrial networks and joint management of link heterogeneity.

A plausible implication is that the complexity–performance tradeoff inherent to beamspace MSMS can be further optimized by leveraging emerging hardware and AI-accelerated sCSI estimation, especially for large $K$, $S$ systems and sub-THz carrier frequencies.

Limitations include increased sensitivity to synchronization errors for high-order constellations, potentially stricter pilot/reference bandwidth requirements, and demand for accurate ephemeris/localization data. Fielded MSMS systems remain constrained by downlink user hardware and backend network interfaces, though successful demonstrations with standard infrastructure indicate high practicality.

---

References:

(Röper et al., 2021): "Beamspace MIMO for Satellite Swarms" (Storek et al., 2020): "Multi-Satellite Multi-User MIMO Precoding: Testbed and Field Trial" (Wang et al., 26 Dec 2025): "Multi-Satellite Multi-Stream Beamspace Massive MIMO Transmission"