Abstract: GPS denied timing is no longer an edge case for tactical SDR networks; it is the design point. This post explains how timing loss breaks TDMA/OFDM waveforms, how to set a practical holdover budget, and how to combine oscillators, disciplining sources, and network distribution to keep sync in contested RF. We also cover current industry moves—from next-gen chip-scale atomic clocks to altGNSS—and where Novocomms Space & Defence fits in integration and ruggedisation.
In a lab, GPS is a convenience. In theatre, it is a dependency your adversary will try to turn into a weakness. GPS denied timing is now a routine operating condition for tactical radios—whether the cause is deliberate jamming, spoofing, or simply operating under canopy, in urban corridors, or inside vehicles where sky view is poor. For waveform and timing engineers, the question is not “can we reacquire GNSS quickly?” but “how long can we keep the network coherent when we cannot trust GNSS at all?”
Holdover is where tactical reality bites. It is the quiet interval between “GNSS went bad” and “the mission is over”, and it is defined by oscillator physics, thermal management, network architecture, and your waveform’s tolerance to time and frequency error. Get it right and your MANET stays usable. Get it wrong and you will watch perfectly good RF front ends fail because time fell apart.
Why GPS denied timing breaks tactical waveforms (and not always immediately)
Most tactical SDR waveforms don’t just use time for logging and message timestamps; they use it to stay on the air without colliding. TDMA slot boundaries, guard times, hopping schedules, TDD switching, and coherent demodulation all assume that nodes share a common notion of time and frequency. Lose that reference and three failure modes tend to appear, often in this order:
1) Slot creep: your transmit window slides relative to everyone else until you hit guard time limits and begin to interfere. Even a small fractional frequency error accumulates into large time error over minutes and hours.
2) Acquisition pain: frequency and timing uncertainty expand the search space for synchronisers, lengthening reacquisition and reducing link availability exactly when the channel is hostile.
3) Network fragmentation: nodes drift into incompatible timing islands. Routing may still “work”, but throughput collapses as contention and retransmissions spike.
The uncomfortable detail: you can be “only a few microseconds off” and still be operationally broken, depending on slot size, guard time, and PHY design. That is why holdover must be treated as a first-class waveform requirement, not a footnote in the GNSS receiver datasheet.
Setting a holdover budget: translate ppm into mission minutes
Holdover starts with a budget that is stated in time error versus outage duration. The simplest back-of-the-envelope relationship is:
Time error ≈ fractional frequency error × elapsed time
As a sanity check: a 1 ppb frequency error (1×10-9) corresponds to about 1 ns per second drift, or roughly 3.6 μs per hour. A 0.1 ppm error (1×10-7) is ~0.1 μs per second, i.e. 360 μs per hour. Those numbers quickly collide with TDMA guard times and coherent processing assumptions.
When you set the budget, include the real-world contributors:
• Temperature sensitivity (tempco): vehicle cabins, mastheads, and soldier-worn radios see rapid thermal swings. The oscillator is often the thermometer you forgot to spec.
• Ageing: frequency drift over weeks/months matters for “cold start” performance and long deployments.
• Retrace: the frequency offset after power cycling is not the same as before. Tactical radios power-cycle in ways lab benches do not.
• Servo history: a well-trained disciplining loop (when GNSS was good) can materially improve holdover; an untrained loop can make it worse.
GPS denied timing holdover: oscillator choices from TCXO to CSAC
There is no universal “best clock”; there is only a best trade for SWaP-C, environment, and holdover duration. Practically, you will see four tiers:
TCXO: attractive for SWaP and cost; often adequate for short outages if your waveform has generous guard times. The problem is not the nominal ppm—it is the thermal gradient and how quickly the platform moves through it.
OCXO: a step change in stability, with a penalty in warm-up time and power. If your radio spends long periods powered, OCXO can be a very sensible “engineering” choice.
Rubidium (Rb): strong holdover over hours to days in exchange for size, cost, and power. Useful at vehicle/command-post class nodes acting as timing anchors.
Chip-Scale Atomic Clock (CSAC): when you need atomic-grade stability at portable SWaP. Industry is pushing hard here. DARPA’s ACES programme (now complete) explicitly targeted next-generation, battery-powered CSACs with 1000× improvement in key stability parameters (notably temperature sensitivity, ageing, and retrace) to reduce reliance on GNSS timing. The significance for tactical radios is straightforward: better CSACs turn “minutes of tolerance” into “hours of coherence” for nodes that cannot carry a full-size atomic reference.
Strategic opinion: for most tactical SDR networks, the winning architecture is heterogeneous. Put the best clocks at nodes that can afford them (vehicle, relay, gateway), and use disciplined mid-tier oscillators elsewhere—then distribute timing cleverly rather than trying to make every handheld an atomic timekeeper.
Disciplining without GNSS: multi-source timing is becoming the baseline
When GNSS is denied or untrusted, you need alternative references to discipline frequency/time and to detect when you are being lied to. Two industry shifts are worth baking into new designs:
1) Resilient PNT is moving from single-source to multi-source. Across the PNT community, the direction of travel is clear: combine constellations, signals, and independent checks so you can keep operating and also know when you should stop trusting an input.
2) altGNSS via LEO is being productised. Low Earth Orbit timing services (for example, Iridium-based STL approaches) are being promoted as higher power, harder-to-jam complements to GNSS for denied and degraded environments. For tactical systems, treat these as additional references—not magic bullets—then use them to reduce drift and improve confidence scoring when GNSS integrity is suspect.
Other non-GNSS references can include network-derived timing (from a better-clocked node), inertial-aided frequency steering, or opportunistic signals of opportunity. The engineering discipline is to assign each reference a trust model and a measurement model, then fuse them rather than “switching” blindly.
PTP-style thinking for tactical SDRs: don’t ignore the servo
Even if you are not running IEEE 1588 end-to-end, the same control concepts apply: a servo loop estimates offset and drift, then steers a local oscillator. Telecom timing work has shown that holdover performance is not only oscillator-limited; it can be greatly improved when the system has strong frequency assist and a well-tuned servo history. The takeaway for waveform teams is that you should design holdover as a control problem, not just a parts selection problem.
Implementation details that decide whether holdover works in the field
Holdover failures are usually death by a thousand cuts. A few pragmatic design points that repeatedly matter on tactical platforms:
Thermal engineering is timing engineering. Mounting an OCXO next to a hot PA, or routing airflow past it inconsistently, will erase the spec advantage. Model gradients, not just ambient temperature.
Phase noise and spurs still matter. Improving long-term drift while degrading close-in phase noise can harm demodulation and EVM. Evaluate the whole chain, including PLL choices and reference distribution.
Power cycling is a scenario, not a corner case. If your CONOPS includes aggressive duty cycling, you must characterise retrace and warm-up behaviour, and decide what “time-valid” means during warm-up.
Network roles should be timing-aware. Elect cluster heads/relays with better clocks when you can, and treat timing quality as a routing or scheduling metric. If your network can self-heal, let it heal around timing as well as RF.
Plan for spoofing, not just jamming. A jammed receiver is obvious; a spoofed receiver can be confidently wrong. Holdover logic should include sanity checks, hysteresis, and cross-validation between sources.
Where Novocomms Space & Defence fits: from RF front ends to resilient timing architectures
At Novocomms Space & Defence we spend our time in the uncomfortable overlap between RF, antennas, and the system constraints that make timing fail in real deployments: vibration, temperature cycling, EMI, and contested spectrum. For GPS-denied tactical radios, our work typically supports programmes in three practical ways:
• Rugged RF and antenna integration: designing and integrating antenna systems and RF front ends that maintain performance under MIL-STD environmental stress, helping your timing and comms subsystems avoid self-inflicted interference and sensitivity loss.
• System engineering for contested environments: assisting teams with architecture trade-offs across oscillators, filtering, distribution, and enclosure constraints—because holdover is rarely solvable by a single component change.
• Test-informed design: supporting verification approaches that reflect operational reality (thermal ramps, vibration, power cycling, and interference exposure), so “holdover” is measured as an end-to-end network outcome rather than a bench-top oscillator plot.
Conclusion: design for GPS denied timing as a default mode
If your tactical SDR network only meets its timing spec when GNSS is clean, it does not meet its timing spec. The modern baseline is to assume outages, assume deception, and engineer holdover as a system capability: a budget tied to waveform tolerance, an oscillator strategy matched to platform SWaP, disciplined by multiple references, and validated under field-like stress.
If you are developing or upgrading a tactical radio, gateway, relay, or timing distribution approach for contested operations, we can help you turn holdover from a risk into a design feature. Contact Novocomms Space & Defence at https://novocomms.space/contact-us/.