Timing-over-Air Protocols (TAP)

• Timing-over-Air Protocols (TAP) are precision network timing methods that distribute high-granularity time references over the radio interface using advanced protocol extensions and hardware enhancements.
• They employ star or hierarchical synchronization models with techniques like timing advance estimation, PHY timestamping, and boundary-clock procedures for deterministic operation in mission-critical systems.
• TAP integrates comprehensive error budgeting, statistical filtering, and specialized enhancements for 5G, TSN, and LEO satellite networks to achieve ultra-reliable low-latency performance.

Timing-over-Air Protocols (TAP) refer to a class of precision network timing methods that distribute high-granularity time references from a master source—typically a 5G base station or core grandmaster—over the radio interface to slave devices such as sensors, actuators, gateways, and TSN end points. TAP is designed to provide sub-microsecond, and even nanosecond, synchronization accuracy in ultra-reliable low-latency communication (URLLC) scenarios, industrial automation, smart grid protection, and emerging satellite-based networks. By leveraging protocol extensions and hardware enhancements within the 3GPP framework, TAP enables wireless domains to achieve deterministic and isochronous operation on par with, or exceeding, wired time-distribution standards.

1. Synchronization Requirements and Use Cases

Mission-critical applications in factory automation, power systems, and converged 5G/TSN environments impose strict device-level time-synchronization constraints. Hard and isochronous real-time classes in industrial automation require cycle times ≤1 ms, end-to-end jitter ≤1 μs, and clock alignment at the ±0.5 μs level, with reliability approaching 99.999%. Fault protection in smart grids (e.g., line differential relays) targets synchronization accuracy <20 μs, while phasor measurement units (PMUs) and time-sensitive measurement demand sync errors below ±1 μs. Control and monitoring can tolerate looser synchronization (±10–100 μs) (Mahmood et al., 2018). In TSN-5G integration, the requirement for absolute end-to-end error is generally ≤1 μs at the 99.9th percentile (Shi et al., 2021).

2. System Architectures and Reference Models

TAP implementations typically employ a star or hierarchical synchronization model. A global time baseline (e.g., UTC via GPS or grandmaster clocks) is maintained in the core network. 5G base stations (BS or gNB) synchronize to this baseline and act as local masters for attached user equipment (UEs) or gateways. UEs/TSN endpoints, as slaves, align their clocks to the BS reference, potentially propagating time further through gateways for legacy or hybrid wired/wireless networks (Mahmood et al., 2018). The 3GPP Release 16 boundary-clock model provides a “bridge” where the gNB acts as a PTP boundary clock, translating time from the TSN grandmaster to a 5G frame reference, then disseminating it over the air (e.g., enhanced SIB16) to UEs (Shi et al., 2021). For satellite networks (notably LEO), the system must address large Doppler shifts and cell radii, extending beyond standard terrestrial assumptions (Zhu et al., 2024).

3. TAP Synchronization Procedures and Mathematical Framework

TAP protocols rely on a combination of one-way and two-way timestamped exchanges. The basic two-way handshake resembles IEEE 1588: the master transmits a SYNC frame at $t_1$, received by the slave at $t_2$; the slave replies with a DELAY_REQ at $t_3$, received by the master at $t_4$. The clock offset and round-trip delay estimates are: $$\hat{\theta} = \frac{1}{2}[(t_2 - t_1) - (t_4 - t_3)],\quad \hat{\delta} = \frac{1}{2}[(t_2 - t_1) + (t_4 - t_3)]$$ The slave updates its local time as $C_S'(t) = C_S(t) + \hat{\theta}$ (Mahmood et al., 2018). Total synchronization error is budgeted as: $$\epsilon_{total} \leq \epsilon_{clk\_drift} + \epsilon_{tstamp} + \epsilon_{link\_asym} + \epsilon_{queue} + \epsilon_{TA\_quant} + \epsilon_{sched}$$ Deterministic and stochastic sources (oscillator drift, timestamping, propagation asymmetry, scheduling jitter, quantization) contribute, with specific mechanisms to minimize each (Zhang et al., 2024).

Allan variance modeling provides a statistical framework for analyzing phase deviations and optimal averaging intervals to minimize timing error: $$\sigma^2_\theta(T) = \frac{1}{2T} E\left[\left( \int_0^T \frac{d\theta}{dt}(t)dt - \int_0^T \frac{d\theta}{dt}(t+T)dt \right)^2\right]$$ This supports advanced statistical filtering (e.g., Kalman filters) to further suppress phase noise (Zhang et al., 2024).

4. Protocol Mechanisms and Enhancements

TAP implementations draw upon several mechanisms:

• Timing Advance (TA) + SIB16 Broadcast: TA estimates and corrects one-way propagation delay (quantized in steps, e.g., $16T_s \approx 520$ ns). SIB16 provides the UTC reference but with typical millisecond granularity in legacy R15; enhanced versions support symbol-level granularity (down to 10 ns). Error arises predominantly from TA quantization (±260 ns) and SIB broadcast periodicity (Mahmood et al., 2018, Shi et al., 2021).
• Physical Layer Time-Stamping: Timestamping at PHY egress/ingress boundaries reduces MAC scheduling jitter, essential in high-load scenarios (Mahmood et al., 2018).
• RIBS and Dedicated RRC/PTP-Style Handshakes: Using dedicated radio interface signals or dual-timestamp RRC messages allows direct offset and delay measurement with minimized MAC/PHY uncertainty (Mahmood et al., 2018).
• Boundary-Clock Procedures in TSN Integration: The gNB translates and broadcasts precise TSN timestamps to UEs. UEs compensate for propagation, time alignment, reference time granularity, and time-of-arrival errors according to explicit mathematical models, maintaining consistency with gPTP and IEEE 802.1AS (Shi et al., 2021).
• Enhanced TA in LEO Satellite Environments: UE-centric time-frequency pre-compensation leverages multiple satellite synchronization signal blocks to jointly estimate one-way delay and frequency offsets, reducing residual TA and CFO to within standard PRACH command ranges. New cyclic-prefix-free preamble formats with flexible order and differential power allocation further counteract multi-UE partial-period interference (Zhu et al., 2024).

Suggested enhancements include finer TA quantization, increased SIB / reference broadcast periodicity, PHY-level SIB time-stamping, message integrity for anti-spoofing, and oscillator upgrades to bound drift. These enable sub-microsecond, and in optimized configurations, sub-200 ns synchronization (Mahmood et al., 2018, Zhang et al., 2024).

5. Error Sources and Performance Evaluation

A comprehensive error model for TAP encompasses:

• Reference Clock Instability ($e_{src}$): Sourced from GNSS or atomic-tier local oscillators.
• Granularity Error ($e_{Gr}$): Determined by time-stamp quantization, e.g., 10 ms in legacy; 10 ns in R16+.
• Air-Channel Delay Estimation Error ($e_{air}$): Due to propagation variability, multipath, noise; under SRS/PRS methods, errors are as low as ±4 ns.
• Scheduling and Processing Delays ($\epsilon_{sched}$, $t_0$): MAC/PHY timing, internal UE processing (Zhang et al., 2024).

In field and simulation studies:

• Coarse SIB16 + TA yields end-to-end UE-UE skews of ~1.2 μs (1σ), with worst-case jitter up to 3 μs.
• Enhanced SIB16 and finer TA (~4·$T_s$ resolution) deliver 1σ skew ~300 ns, max error <1 μs for UEs within 100 m of the BS (Mahmood et al., 2018).
• Dedicated RRC/PTP handshakes with PHY-timestamping and semi-static grants achieve 1σ skew <100 ns, worst-case jitter <250 ns even with 90% subframe load (Mahmood et al., 2018).
• RIBS/UE-based asymmetry correction drives UE-BS sync error below 150 ns.
• In TSN-5G integration scenarios, overall absolute mean path/ToA errors decrease with larger subcarrier spacings (SCS), ranging from 192 ns (15 kHz SCS) to 24 ns (120 kHz SCS). At SCS ≥30 kHz and reference granularity ≤20 ns, sub-μs error at the 99.999th percentile is achieved, given UE drift ≤10 ppm and TAP broadcast intervals ≤60 ms (Shi et al., 2021).
• 5G TAP prototypes employing Allan-variance optimized statistical filtering and hardware compensation exhibit empirical error below 200 ns (99.9th percentile), and long-term mean errors converging to 1–2 ns (Zhang et al., 2024).

Performance of the enhanced LEO satellite TAP solution: under –6 dB SNR, missed-detection rates are <1% even with 64 simultaneous UEs; TA estimation errors for all active UEs remain below 25 sampling intervals (≈0.8 ms for SCS = 30 kHz) (Zhu et al., 2024).

6. Protocol Realization, Interoperability, and Deployment

TAP is realized with minimal protocol or hardware changes:

• Protocol extensions utilize nonCriticalExtension fields in SIB9/SIB16, or small control fields in RRC/PHY messages (Zhang et al., 2024). No new RLC/MAC protocols required.
• FPGA-based clock triggering and PTP→PPS synchronization at the BS eliminate within-period jitter. Factory-calibrated hardware minimizes UE-specific processing delays.
• Use of low-cost, mass-market oscillators is compatible if statistical compensation (Kalman filtering, Allan variance-based averaging) is deployed.
• TAP is compatible with 3GPP Release 16+ and can serve as a boundary clock for IEEE 802.1AS TSN systems without changes on the TSN transport side, hiding air-interface complexity from edge devices (Shi et al., 2021, Zhang et al., 2024).
• In the LEO context, TAP accommodates highly dynamic propagation, large cell radii, and high user densities by integrating frequency pre-compensation and robust preamble formats (Zhu et al., 2024).

7. Security, Resilience, and Limitations

Replay and half-CP forwarding attacks present concrete threats for TAP, especially at the time-broadcast and preamble levels. Enhancements such as CRC protection, dual-threshold processing (guarding against ARQ/HARQ messaging replay), and randomization in preamble order increase attack complexity but do not eliminate all vectors (Zhang et al., 2024). For sustained sub-μs accuracy, environmental conditions (e.g., multipath in factories, LoS conditions for industrial/motion control) and hardware provisioning (oscillator class, timestamping resolution) remain critical.

TAP achieves two to three orders of magnitude greater accuracy than conventional NTP/PTP-over-radio (~200 ns vs. ~10 μs), with long-term stability readily reaching the single-digit nanosecond range using advanced filtering and hardware compensation. Nevertheless, convergence time, protocol security, sensitivity to hardware nonidealities, and scalability in high-load/multi-domain environments remain active topics for ongoing optimization (Mahmood et al., 2018, Shi et al., 2021, Zhang et al., 2024, Zhu et al., 2024).