Edge-Side Residual Timing and Frequency Control for Software-Defined Ground Stations in 5G NTN Uplinks

Abstract: This paper studies a ground-segment implementation problem in 5G non-terrestrial networks (NTN): once UE-side geometric pre-compensation has produced a coarse timing/frequency prior, can an edge-side residual loop keep the uplink inside an NR-feasible operating region under rapid LEO dynamics? We examine this question with a software-defined ground station (SDGS) design that keeps the coarse prior at the UE and closes the residual timing-advance (TA) / carrier-frequency-offset (CFO) loop at the ground-station edge. This paper takes a systems-and-control view rather than proposing a full-stack intelligent architecture. Its evidence base consists of a March 2026 hardware-in-the-loop (HIL) campaign and a companion uncertainty analysis. The HIL campaign includes same-window reference runs collected on the same platform with edge residual control disabled, but it does not include a cloud-loop benchmark. The placement claim is therefore architectural and control-oriented rather than a head-to-head cloud-versus-edge proof. In the Shenzhen steady-state tracking interval, the edge-controlled mode lowers mean RTT from 70.51 +/- 2.34 ms to 32.84 +/- 2.56 ms and, within the retained Layer-3 transport mapping, improves artifact-level goodput from 80.14 +/- 0.14 Mbps to 196.04 +/- 1.87 Mbps relative to that reference configuration. Across four ground-station locations, the closed-loop controller keeps residual TA P95 at 0.49 us and residual CFO P95 within 76-77 Hz. Together with the uncertainty analysis, these observations support a bounded claim: an edge-side residual timing/frequency loop can keep the SDGS uplink in a more stable NR-feasible operating regime under the assumptions retained in the current HIL artifact.

• The paper introduces a control architecture that decouples UE-side geometric prediction from edge-tier residual refinement to stabilize 5G NTN uplink transmissions.
• It employs a fast PID-based correction loop that cuts mean RTT by over 50% and nearly doubles the artifact-level goodput as validated by HIL experiments.
• The HIL campaign demonstrates microsecond-level timing accuracy and sub-hundred-Hz CFO control across multiple geographic locations under dynamic LEO conditions.

Edge-Side Residual Timing and Frequency Control in SDGS Uplinks for 5G NTN

Context and Motivation

The integration of non-terrestrial networks (NTN) with 5G, especially using LEO constellations, imposes stringent demands on timing and frequency synchronization for uplink transmissions. The rapid geometry variation associated with LEO satellites creates residual timing advance (TA) and carrier frequency offset (CFO) errors that are not fully compensated by UE-side pre-compensation. Standard NR-NTN procedures motivate a division between user equipment (UE) prediction and ground-segment refinement, but the operational realization, particularly in software-defined ground stations (SDGS), is under-specified.

This manuscript directly addresses the systems-level gap: after coarse geometric prediction at the UE, can an edge-side residual timing and frequency control loop maintain the uplink within NR-feasible bounds under rapid LEO dynamics? By focusing on SDGS architectures, the authors emphasize the value of NFV/SDN virtualization and control-placement flexibility. Rather than a universal architectural claim, the manuscript establishes a bounded control-placement argument using a hardware-in-the-loop (HIL) campaign and uncertainty analysis.

System Model and Control Architecture

The proposed SDGS solution partitions the control path into explicit UE-side geometric derivation and edge-tier residual refinement. The UE computes an open-loop TA/CFO prior using satellite TLE/SGP4 and GNSS, while the SDGS edge absorbs a fast PID-based residual correction loop. This control path is formalized as a prediction–refinement pipeline: geometric prior narrows the residual range, edge observability closes the loop, and local adaptation logic maintains session stability. The SDGS implementation emphasizes latency-sensitive placement to minimize control-delay ($d_\mathrm{fb}$), critical for high Doppler slope and frequent LEO handovers.

Channel modeling adopts a simplified free-space paradigm, focusing on Doppler variation governed by relative satellite–UE velocity. Residual TA is budgeted against NR cyclic prefix (CP), and residual CFO against subcarrier spacing (SCS). The output is not a perfect open-loop prediction claim, but an operational assertion that the combined UE-side prior and edge residual loop achieves microsecond-level timing accuracy and sub-hundred-Hz frequency residuals within CP/SCS feasibility.

Hardware-in-the-Loop Experimental Validation and Uncertainty Analysis

A March 2026 HIL campaign underpins all key system claims, using geographically distributed ground stations (Shenzhen, Beijing, Tokyo, Los Angeles). Each station underwent controlled ON/OFF runs with the edge residual loop enabled and disabled, forming a direct artifact-level comparison.

The results are robust:

• Edge-controlled mode reduces mean RTT from $70.51 \pm 2.34$ ms to $32.84 \pm 2.56$ ms, a 53.4% decrease.
• Artifact-level goodput improves from $80.14 \pm 0.14$ Mbps to $196.04 \pm 1.87$ Mbps, a 144.6% increase.
• Closed-loop residual TA P95 stabilizes at $0.49\,\mu$s, and CFO P95 within $76$–$77$ Hz across all stations.

These metrics are consistent across the four stations with no retuning, indicating geographic robustness. Model-based uncertainty analysis further corroborates practical resilience, showing residual TA P95 of $0.45\,\mu$s and CFO P95 of $90$ Hz under realistic ephemeris, GNSS, and clock noise. Delay-sensitivity sweeps confirm stable operation under moderate feedback delays and quantization loss; performance degrades gracefully rather than catastrophically as $70.51 \pm 2.34$0 or $70.51 \pm 2.34$1 increase.

Transient handover window rows are archived for transparency but excluded from headline comparison; the current artifact does not elevate these to a full transient-control claim.

Practical and Theoretical Implications

The findings directly inform SDGS network operators and system architects. The demonstrated control path allows for incremental adoption: UE-side geometric prediction remains, while fast edge refinement is implemented locally at ground stations, decoupling stabilization from remote/cloud dependency. This placement is compelling when bottlenecks are dominated by residual impairment and tracking instability, rather than spectrum scarcity or downlink-only performance.

The paper situates its claims within the broader NR-NTN and O-RAN literature: prior works focus either on UE-side prediction and receiver synchronization, or on intelligent orchestration. The current contribution is narrower but more precise: evidence supports the operational value of localized, latency-sensitive residual control in SDGS uplinks, using transparent measurement and explicit error budgeting.

Theoretical implications extend to future SDGS architectural explorations—control delay, closed-loop observability, and threshold-triggered adaptation must be optimized jointly. The explicit separation of geometric prior and feedback refinement enables modular ablation studies and adaptation to non-ideal channel conditions. Placement arguments are motivated by control-theoretic delay margins rather than absolute throughput benchmarking.

Limitations and Opportunities for Future Work

Several explicit limitations are acknowledged:

• Evaluation is restricted to Layer-3 artifact-level mapping; waveform-level transport proofs and RF-front-end validations are deferred.
• Channel modeling omits ionospheric/tropospheric effects, phase noise, urban blockage, and multipath. Future artifacts will incorporate TR 38.811/38.821-aligned channel components.
• No direct cloud-loop or core-network delayed-feedback baseline is included; placement claims arise from control-delay sensitivity, not direct benchmarking.
• Only data-transmission is covered; voice/video service models, economic assessment, and transient handover analysis are omitted.

Opportunities for future work include expanding campaign scope with live satellite links, incorporating full RF instrumentation, widening disturbance models (scintillation, oscillator instability), and performing rigorous ablations between local adaptation and residual control. A stronger control paper would include detailed time-series transient response and validate transport–PHY mapping with explicit SINR/BLER relationships.

Conclusion

This manuscript advances the understanding of SDGS uplink stabilization in 5G NTN by providing bounded, explicit evidence for edge-side residual timing and frequency control. The separation between UE-side geometric prediction and edge-tier closed-loop refinement, validated by artifact-level HIL results, demonstrates a practical path towards robust uplink performance under challenging LEO dynamics. Future research will expand the measurement scope to waveform-level validation and address unmodeled channel and service-layer uncertainties. The artifact repository linked in the manuscript supports reproducibility and transparency while not overstating current empirical claims.