MSSA Regenerative NTN Payload Architecture 2.0 Explained
The Mobile Satellite Services Association (MSSA) released MSS Reference Architecture Version 2.0 in September 2026, marking a significant architectural evolution for non-terrestrial networks (NTN).
While Version 1.0 focused primarily on transparent, or bent-pipe, payloads, Version 2.0 introduces a comprehensive regenerative payload architecture. The new architecture moves selected or complete RAN processing functions onboard the satellite and provides a framework for simultaneously supporting 5G-NR broadband and NB-IoT narrowband services across a shared satellite constellation.
The reference architecture is intended to give satellite operators, semiconductor vendors, RAN equipment providers, and mobile network operators a common framework for evaluating payload architectures and accelerating commercial Direct-to-Cell (D2C) and integrated space-ground networks.
🚀 From Bent-Pipe to Regenerative Payloads #
MSSA Reference Architecture 2.0 defines four primary payload models across a spectrum of onboard processing, feeder-link requirements, power consumption, and implementation complexity.
| Payload Model | Onboard Components | Ground Components | Key Advantages | Major Trade-offs / Limitations |
|---|---|---|---|---|
| Transparent (Bent-Pipe) | Analog RF amplification and frequency conversion | Complete gNB (RU/DU/CU) | Lowest payload power consumption and complexity | Very high feeder-link bandwidth requirements; higher control-plane latency; limited onboard storage-and-forward capabilities |
| Regenerative Option 1 | Radio Unit (RU) | Distributed Unit (DU) + Centralized Unit (CU) | Relatively low onboard processing complexity | Feeder links must transport large volumes of raw IQ samples, creating extreme bandwidth requirements |
| Regenerative Option 2a | RU + Distributed Unit (DU) | Centralized Unit (CU) | Substantially reduces feeder-link bandwidth | High LEO orbital velocity can result in frequent DU-CU mobility and handoff procedures |
| Regenerative Option 2b | Complete gNB (RU + DU + CU) | 5G Core Network (5GC) | Lowest latency and strong support for inter-satellite Xn handovers | Highest payload complexity, power consumption, thermal requirements, and R&D cost |
Architectural Trade-Offs #
The fundamental architectural decision is where the RAN processing boundary should reside.
A transparent payload keeps virtually all intelligence on the ground. The satellite performs RF forwarding, while the feeder link carries the traffic between the satellite and a terrestrial gNB. This minimizes satellite complexity but places significant pressure on feeder-link capacity and terrestrial coordination.
Regenerative Option 1 moves the RU onboard. Although this provides a more integrated RF architecture, transporting raw or minimally processed IQ data over the feeder link can become prohibitively bandwidth-intensive.
Regenerative Option 2a moves both the RU and DU onboard. This substantially reduces feeder traffic because more baseband processing occurs in orbit. However, the separation between the onboard DU and ground-based CU introduces additional mobility-management challenges for fast-moving LEO satellites.
Regenerative Option 2b places the complete gNB onboard, leaving the 5GC as the primary ground-side network function. This maximizes autonomy and minimizes latency while creating the most demanding requirements for satellite compute, power, thermal management, software lifecycle management, and hardware qualification.
🛰️ LEO Resource Management and Beam Hopping #
Resource management becomes a fundamental design constraint in LEO constellations because satellite power, antenna resources, and digital beamforming capacity cannot necessarily support all available spot beams simultaneously.
For a representative 600 km LEO satellite, the available power and beamforming budget may permit only a fraction of the total configured spot beams to operate concurrently. Beam hopping therefore becomes an essential mechanism for dynamically allocating RF and baseband resources.
Hierarchical Resource Allocation #
The MSSA architecture separates satellite mobility awareness from conventional terrestrial RAN scheduling through a hierarchical resource-management model.
Satellite Tracker #
The Satellite Tracker decouples satellite orbital dynamics from the terrestrial RAN and provides the relevant satellite-state information to the 5G system.
One important interface is the generation of SIB19 system information, which allows UEs and the terrestrial RAN to account for NTN-specific satellite and orbital information.
Beam Arbitrator #
The Beam Arbitrator functions as a higher-level multi-cell resource scheduler.
Instead of replacing the mature scheduling mechanisms already implemented in terrestrial gNB software, it generates beam-hopping time-slot patterns that can be consumed by existing Layer 2 scheduling functions. This approach allows NTN implementations to maximize reuse of established 5G baseband software while introducing satellite-specific resource orchestration above the conventional scheduler.
SSB and Control-Plane Efficiency #
The architecture can also take advantage of 3GPP Release 19 extended SSB broadcasting periods to reduce the amount of resources consumed by inactive or low-utilization cells.
This becomes increasingly important when a satellite simultaneously manages a large number of logical cells but has insufficient instantaneous resources to activate all of them.
📡 Co-Deploying 5G-NR Broadband and NB-IoT #
Supporting 5G-NR broadband and NB-IoT on the same satellite payload introduces a fundamentally different resource-allocation problem.
5G-NR typically requires significantly wider bandwidth and high-throughput processing, while NB-IoT is optimized for narrowband operation, low device complexity, extended coverage, and potentially large numbers of connected devices.
A shared digital beamforming architecture therefore needs to balance fixed hardware overhead against the marginal computational cost of additional NR and NB-IoT streams.
Digital Beamforming Architectures #
Three implementation approaches illustrate the primary trade-offs.
Architecture 1: Frequency-Domain Beamforming #
Frequency-domain beamforming uses a shared IFFT structure on the antenna-port side.
The architecture introduces relatively high fixed hardware overhead, but the marginal cost of adding additional NB-IoT streams can be very low. This makes it particularly attractive for deployments expecting large-scale IoT traffic.
The key trade-off is that the architecture must carry the fixed cost of the shared IFFT and associated processing even when the instantaneous workload is relatively small.
Architecture 2: Time-Domain Beamforming #
Time-domain beamforming maps streams directly onto the subcarrier grid without imposing the same fixed antenna-port overhead.
This reduces fixed costs but causes NB-IoT traffic to traverse much of the same broadband processing pipeline as 5G-NR. Consequently, the incremental cost of an NB-IoT stream can approach that of a 5G-NR stream.
This architecture is attractive when hardware flexibility and a common processing pipeline are more important than maximizing NB-IoT density.
Architecture 2-a: Native Narrowband Processing #
The native narrowband architecture maintains a dedicated narrowband processing pipeline for NB-IoT.
This avoids forcing narrowband traffic through the complete 5G-NR processing chain. However, the resulting signals require upsampling for distribution across antenna ports, introducing additional computational overhead as the number of antenna ports increases.
The architecture therefore trades processing specialization for a scaling penalty at high antenna-port counts.
64-Antenna-Port Case Study #
For a 64-antenna-port implementation, the three architectures illustrate different cost curves:
| Architecture | Fixed Cost | Marginal NB-IoT Cost | Primary Strength | Primary Limitation |
|---|---|---|---|---|
| Frequency-Domain | High | Very low | Efficient scaling for large IoT workloads | Significant fixed processing overhead |
| Time-Domain | Low | High | Flexible shared pipeline | NB-IoT streams consume broadband-class processing resources |
| Native Narrowband | Moderate | Moderate | Dedicated NB-IoT processing | Upsampling cost increases with antenna-port count |
The optimal architecture therefore depends heavily on the expected NR-to-IoT traffic ratio, antenna configuration, onboard compute budget, and target constellation economics.
🌐 Multi-Operator Deployment and Global Roaming #
Commercial Direct-to-Cell networks are expected to operate across multiple terrestrial mobile network operators (MNOs), countries, and regulatory domains. MSSA Reference Architecture 2.0 therefore addresses network sharing, subscriber routing, and multi-PLMN operation.
Gateway Multiplexing #
Gateway multiplexing is primarily applicable to transparent payload architectures.
A Beam Arbitrator can interface with multiple independent terrestrial base stations and coordinate their access to satellite resources. The approach provides operator separation but increases orchestration and integration complexity.
MOCN and MORAN #
MOCN (Multi-Operator Core Network) and MORAN (Multi-Operator Radio Access Network) provide a more integrated RAN-sharing model.
A shared onboard gNB can serve multiple terrestrial MNOs while their respective core networks remain logically separated. Subscriber routing can use mechanisms such as IMSI ranges to determine the appropriate operator domain.
For regenerative payloads, this model can significantly improve utilization of onboard RAN resources while avoiding unnecessary duplication of satellite-side infrastructure.
Network Slicing and RAN Sharing #
Network slicing extends sharing beyond basic RAN resource pooling by providing logical end-to-end isolation across the RAN and 5G Core (5GC).
A slice-aware architecture allows different operators or services to receive differentiated resource policies, quality-of-service treatment, and operational isolation.
The approach requires corresponding support from UEs, RAN functions, and core networks, making interoperability a key implementation consideration.
Multi-PLMN and Roaming #
Satellites serving international footprints must support multi-PLMN broadcasting, potentially advertising both the Satellite Network Operator (SNO) identity and participating terrestrial MNO identities.
Integration with IPX (IP Exchange) networks is also relevant for aligning signaling, charging, and roaming processes with established GSMA inter-operator frameworks.
The result is an architecture in which satellite coverage can become an extension of existing mobile operator footprints rather than an isolated satellite-only network.
🧩 3GPP and O-RAN Alignment #
MSSA Reference Architecture 2.0 is designed around the broader evolution of 3GPP NTN standards, with particular relevance to Release 17 through Release 19 capabilities for FR1 NTN deployments in bands such as L- and S-band.
The architecture also adapts O-RAN principles to satellite-specific resource and mobility requirements.
RIC4NTN Functional Model #
A satellite-oriented RAN Intelligent Controller model can divide optimization responsibilities according to their timescales.
Non-Real-Time RIC #
The Non-Real-Time RIC can handle longer-timescale optimization tasks such as inter-satellite beam planning, cell configuration, policy optimization, and constellation-level resource planning.
Near-Real-Time RIC #
The Near-Real-Time RIC operates on shorter timescales and can coordinate dynamic load balancing, mobility management, and inter-satellite resource decisions within a constellation.
Local gNB Layer 2 Scheduler #
The onboard or local gNB Layer 2 scheduler remains responsible for fine-grained intra-cell resource allocation, including scheduling individual UEs and managing instantaneous radio resources.
This separation allows satellite-specific intelligence to be introduced without discarding the established RAN scheduling model.
🔭 Architectural Implications #
The central shift in MSSA Reference Architecture 2.0 is not simply moving more processing onto the satellite. It is the emergence of a software-defined, regenerative NTN architecture in which satellite compute, beam scheduling, RAN functions, mobility management, and terrestrial network integration become tightly coupled.
The four payload models provide different points on the architecture spectrum:
- Transparent payloads minimize onboard complexity but maximize feeder-link dependence.
- RU-regenerative architectures reduce RF-related ground dependencies but can remain feeder-bandwidth intensive.
- RU+DU architectures substantially reduce feeder requirements while introducing more complex mobility management.
- Full onboard gNB architectures maximize autonomy, latency performance, and ISL-based mobility capabilities at the cost of significantly greater onboard compute, power, thermal, and lifecycle requirements.
At the same time, supporting 5G-NR and NB-IoT together means that digital beamforming and baseband architecture can become just as important as the satellite RF design itself.
For commercial D2C deployments, the resulting architecture is likely to be determined by a combination of payload power budget, feeder-link capacity, onboard compute, antenna-port count, spectrum availability, UE density, operator-sharing requirements, and constellation mobility characteristics.
MSSA Version 2.0 therefore provides a useful architectural framework for evaluating these trade-offs before committing to a particular regenerative NTN implementation.