Communication system, device and method for system timing synchronization

Abstract

A method at a communication device in a communication system comprising a plurality of such devices, and a central processing hub. The communication device measures timing from a received Global Navig

Description

Technical Field

[0001] The technical field relates generally to a system and method for system timing synchronization. In particular, in some examples, the technical field relates to a system and method to provide position, navigation and timing of associated communication devices to improve resilience to jamming and (hostile) radio frequency (RF) interference. Background

[0002] In recent years, third generation (3G) wireless communications have evolved to the longterm evolution (LTE™) cellular communication standard, sometimes referred to as 4th generation (4G) wireless communications. Both 3G and 4G technologies are compliant with third generation partnership project (3GPP™) standards. 4G networks and phones were designed to support mobile internet and higher speeds for activities, such as video streaming and gaming. The 3 GPP™ standards are now developing a fifth generation (5G) of mobile wireless communications, which is set to initiate a step change in the delivery of better and faster data rate communications, for example powering businesses, improving communications within homes and spearheading advances such as driverless cars.

[0003] One of the potential technologies targeted to enable future cellular network deployment scenarios and applications is the support for wireless backhaul and relay links that enable flexible and very dense deployment of 5G-new radio (NR) cells without a need for densifying the transport network proportionately.

[0004] Due to the expected larger bandwidth available for NR compared to long term evolved (LTE™) (e.g., mmWave spectrum) along with the native deployment of massive multiple-in / multiple-out (MIMO) or multi-beam systems in NR creates an opportunity to develop and deploy integrated access and backhaul (IAB) links. It is envisaged that this may allow easier deployment of a dense network of self-backhauled NR cells in a more integrated manner, by building upon many of the control and data channels / procedures defined for providing access to UEs. It is known that cellular mesh-networks exist that can extend the range of 5G and implement IAB parts of the 5G specification by leveraging multiple nodes to distribute the signal, making it ideal for rural and hard-to-reach areas. This allows for multiple individual nodes to be connected to each other simultaneously and being dynamically allocated to one or more other nodes at any time in order to grow or contract the network as ‘business need’ dictates.

[0005] Referring now to FIG. 1, a simplified block diagram of a known 5G base station 100 is illustrated. The 5G base station 100 (often referred to as a gNodeB (gNB)), includes hardware 150 functionality and software 105 functionality. The 5Gbase station 100 is configured to communicate with another communication network and other 5G base stations and wireless communication devices, such as user equipment (UEs). In this illustration, the software 105 functionality includes four functions, referred to as: Centralized Unit (CU) 120 and IAB Distributed Unit (DU) 115 and a Radio Unit (RU) 110. In some examples, the CU 120 provides support for the higher layers of the protocol stack such as Service Data Adaption Protocol (SDAP), Packet Data Convergence Protocol (PDCP) and Radio Resource Control (RRC) whilst, the DU 115 provides support for the lower layers of the protocol stack such as Radio Link Control (RLC) and Medium Access Control (MAC) layers, whereas the RU 110, sometimes referred to as radio head, is configured to handle the Physical layer of communications in the protocol stack. The 5G base station 100 also includes a 5G core 125, in essence network gateway software that fundamentally controls the network's operation. The 5G core 125 is configured to provide signalling control (such as user authentication and data session creation and tear-down and typically known as the ‘Control Plane’), plus transport of data across those sessions (typically known as the ‘User Plane’). Time synchronization data distribution 117 is passed between the DU 115 and the CU 120. Network-wide time synchronization data 122 is passed between the CU 120 and the 5G core 125.

[0006] The hardware 150 functionality includes GNSS timing 155 that provides timing data / signals 157, typically at Ipps, to a field programmable gate array (FPGA) device with a software defined radio (SDR) card 160, which performs signal processing and timing capture. A single board computer 170 (i.e., a primary computing platform) is connected 162 to the FPGA with SDR card 160. The single board computer 170 is connected to the 5G core 125 via a hardware-software interface 172. The FPGA with SDR card 160 provides synchronization data 164 to the 5G core 125, timing adjustments data 166 to the DU 115 and time-stamped RF data 168 to the RU 110.

[0007] In modern communication systems and devices, an accurate clock (or clocks) is essential for both exact radio frequencies and for precise timing of signals. When available, Global Navigation Satellite System (GNSS) satellites offer a world-wide ‘shared clock’ (that is - the same space-borne clock that is visible to many users on Earth), similar to the Global Positioning System (GPS). A shared clock allows users to establish a common view of time. In fact, GNSS flies atomic clocks, which are themselves regulated to the Coordinated Universal Time (UTC) standard, thus propagating the terrestrial standard from reference laboratories to every GNSS receiver.

[0008] GPS / GNSS offer reliability, frequency, bandwidth, but are known to cause significant issues with malicious interference, such as denial jamming, which is a daily occurrence in most conflict zones. Hence, particularly in conflict zones, continuous GNSS-based synchronization cannot be relied upon. At this point the drift rate of each clock defines its ‘hold-up time’, i.e., the period after which its drift from true starts to compromise system performance. For 5G systems, time drift is the critical error, since time drift is cumulative (whereas frequency drift is not).

[0009] Typically, an accurate clock is needed as it is important to determine position, navigation and timing (PNT) of communication devices. Positioning is the ability to accurately and precisely determine a device's location and orientation two-dimensionally (or three-dimensionally when required) referenced to a standard system. Navigation is the ability to determine a current and a desired position (relative or absolute) and apply corrections to a course, orientation, and speed to attain a desired position anywhere around the world, from subsurface to surface and from surface to space. Timing is the ability to acquire and maintain accurate and precise time from a standard (say, UTC), anywhere in the world and within user-defined timeliness parameters. Timing typically also includes time transfer.

[0010] Alternatives to PNT are known and these tend to vary between use cases. Each of these alternatives to PNT systems offers strengths but also weaknesses, and the technology depends on factors such as the specific application, environmental conditions, accuracy requirements, and cost considerations. Such alternative technologies include: Laser Range Finding, which uses laser beams to measure the distance to a target. This technology is commonly used in military applications for target acquisition, artillery guidance, and reconnaissance. Whilst it provides accurate distance measurements, it's limited by line-of-sight constraints and atmospheric conditions. Inertial Navigation Systems (INS) is a further alternative technology, but has issues with reliability, and rotational and cartesian drift. Further it relies on sensors such as accelerometers and gyroscopes to continuously calculate an object's position, velocity, and orientation. Furthermore, INS suffers from drift over time and requires periodic recalibration. Celestial Navigation (which is still used today in marine and aviation navigation) is a yet further alternative technology but involves using celestial bodies such as stars, planets, and the sun to determine a device's position on the Earth. Celestial navigation requires accurate timekeeping and knowledge of celestial bodies' positions. Use of Radio Frequency (RF) Beacons is a still yet further alternative technology, where RF beacons emit signals that can be used for navigation. These signals can be triangulated to determine the receiver's position. Whilst not as accurate as GPS, RF beacons are useful in indoor environments, underground tunnels, and urban canyons where GPS signals may be weak or unavailable.

[0011] Vision-Based Navigation is a still yet further alternative technology and relies on cameras and computer vision algorithms to analyse visual data and determine an object's position and orientation. This technology is commonly used in robotics, autonomous vehicles, and augmented reality applications. Vision-based navigation can complement other PNT systems or serve as a standalone solution in certain environments. Magnetic Navigation is a still yet further alternative technology and uses Earth’s magnetic field to determine orientation. Magnetic compasses have been used for centuries for navigation at sea and on land. However, magnetic navigation is sensitive to magnetic anomalies and requires calibration for accurate results. It is known that hybrid Systems may encompass a combination of many modern PNT solutions for enhanced accuracy, reliability, and resilience. For example, integrating GPS with inertial sensors (GPS-aided INS) or combining GPS with RF beacons can provide more robust navigation solutions that are less susceptible to jamming or signal loss.

[0012] In conflict zones, where it is known that various interference mechanisms exist, maintaining reliable and secure communication is critical for military and defence operations. Therefore, unwanted interference, which includes GNSS jamming and adversarial RF disruptions, can significantly degrade communication networks, such as 5G and 6G systems

[0013] The inventors have recognised and appreciated that various threats, ranging from intentional to environmental, can directly impact global positioning system (GPS™) signal reception. Examples of potential threats include: intentional interference that targets the unavailability of the system (typically referred to as ‘jamming’); sophisticated attacks that provide inaccurate information about both location and time by generating fake GPS signals (typically referred to as ‘deception jamming / spoofing’); (end-point) attacks launched from one or more computers against another computer, multiple computers or networks (typically referred to as ‘cyber attacks’); loss of signal due to obstruction in the line of sight (LOS) to a satellite; unintentional signal interference due to environmental factors or other RF signal interference; and / or (cyber) attacks from hostile states that attack (target) the supply chain, i.e., availability of chips and capabilities.

[0014] In addition, the inventors have recognised and appreciated that reference clocks have several types of drift. In this context, it is assumed that only changes in frequency with time, not absolute accuracy, are important and it is assumed that the clock(s) has fully warmed up (noting that for lower-grade evened quartz oscillators this period can be several days).

[0015] It is known that an accurate reference time is essential for the correct operation of a 5G network. Typically, an accurate reference time is sourced from a fixed fibre source. However, if this is not available, as often happens with an IAB network, devices (such as base stations) have to use a satellite link in order to obtain reference time. A satellite-sourced reference time cannot be guaranteed. If absolute timing, e.g., through GPS, GNSS, etc. is denied or jammed there is an issue that there is no source of reference time for communication devices. Furthermore, in current (and planned future) communication systems, the inventors have recognised and appreciated that there is a resilience and reliability issue in the dependability of accurately determining a position of a communication device due to such jamming.

[0016] There is therefore a need to support defence initiatives and security personnel in determining accurate positional information, alongside the provision of high bandwidth communication. Thus, a system, device and method are needed for improving the resilience of wireless communication networks, and particularly increasing resilience to jamming and (hostile) RF interference in GPS and satellite communications, whilst reducing PNT errors and potential positioning spoofing risks, removing single point failure and reliance within current GPS and satellite technologies Summary In a first aspect, a system is described that comprises: a central processing hub; a plurality of communication devices connected to the central processing hub. Each of the plurality of communication devices comprise: a receiver arranged to receive a Global Navigation Satellite System, GNSS, shared clock; a plurality of clocks; a transmitter; and a processor, operably coupled to the receiver, transmitter and the plurality of clocks, wherein the processor is arranged to generate derived clock information from: measured timing of clock signals obtained from the plurality of clocks, measured timing of the GNSS shared clock, and a comparison of the timing of clock signals obtained from the plurality of clocks and the GNSS shared clock. The transmitter is arranged to transmit the derived clock information to at least one of: the central processing hub, others of the plurality of communication devices.

[0017] In a second aspect, a method for timing synchronisation in a communication system that has a plurality of communication devices is described. The method comprises, at a communication device: receiving a Global Navigation Satellite System, GNSS, shared clock; receiving timing information from a clock of the communication device; and deriving clock information from: measured timing of clock signals obtained from the clock; measured timing of the GNSS shared clock; and a comparison of the measured timing of clock signals obtained from the clock and the GNSS shared clock. The method further comprises transmitting the derived clock information to at least one of: a central processing hub, others of the plurality of communication devices.

[0018] In a third aspect, a communication device is described according to the first aspect.

[0019] In a fourth aspect, central processing hub is described. The central processing hub comprises: a receiver configured to receive derived clock information from a plurality of communication devices, wherein the derived clock information comprises a comparison of measured timing of clock signals from multiple clocks in the communication devices and measured timing of a Global Navigation Satellite System, GNSS, shared clock: a learning processor operably coupled to the receiver and having at least one input and at least one output, the at least one input arranged to receive the received derived clock information and at least one of: temperature sensitivity or environmental impacts, wherein the learning processor is configured to predict and output a timing adjustment signal comprising at least one of: clock trends, timing anomalies, drift, jitter, a timing adjustment of a communication device clock; and a transmitter, operably coupled to the learning processor and arranged to transmit the timing adjustment signal to at least one communication device of the plurality of communication devices. Brief Description of the Drawings

[0020] Further details, aspects and embodiments will be described, by way of example only, with reference to the drawings. In the drawings, similar reference numbers are used to identify like, or functionally similar, elements. Elements in the figures are illustrated for simplicity and clarity and have not necessarily been drawn to scale.

[0021] FIG. 1 illustrates a simplified block diagram of a known 5G base station.

[0022] FIG. 2 illustrates a simplified block diagram of a self-learning clock synchronization system 200, adapted to leverage 5G base stations, a master base station, and AI-ML processing for drift detection and synchronization, and adapted to provide a synchronized distributed network of local time sources, according to some example embodiments.

[0023] FIG. 3 illustrates a simplified block diagram of one example of a 5G base station, according to some example embodiments.

[0024] FIG. 4 illustrates a block diagram of an IAB base station (or node), adapted in accordance with some example embodiments.

[0025] FIG. 5 illustrates a simplified overview of the data involved in timing operations and calculations of a central processing hub / master base station in a communication system, in accordance with some example embodiments.

[0026] FIG. 6 illustrates a simplified flowchart of a PNT circuit timing system flow (such as PNT circuit 340 in FIG. 3), in accordance with some example embodiments.

[0027] FIG. 7 illustrates a graph of example results of a preliminary simulation according to some example embodiments.

[0028] FIG. 8 illustrates an example of a neural network that may be employed as an artificial intelligence-based learning processor architecture to analyse an identification of radio frequency (RF) space availability / suitability, say within a contested RF environment, according to some examples of the present invention.

[0029] FIG. 9 illustrates a simplified flowchart of timing synchronisation in a communication system, in accordance with some example embodiments.

[0030] FIG. 10 illustrates a simplified flowchart of some operations AI / ML operations of a central processing hub / master base station in a communication system, in accordance with some example embodiments.

[0031] Skilled artisans will appreciate that elements in the figures are illustrated for simplicity and clarity and have not necessarily been drawn to scale. For example, the dimensions and / or relative positioning of some of the elements in the figures may be exaggerated relative to other elements to help to improve understanding of various embodiments herein-described. Also, common but well-understood elements that are useful or necessary in a commercially feasible embodiment are often not depicted in order to facilitate a less obstructed view of these various example embodiments. It will be further appreciated that certain actions and / or steps may be described or depicted in a particular order of occurrence while those skilled in the art will understand that such specificity with respect to sequence is not actually required. It will also be understood that the terms and expressions used herein have their ordinary technical meaning as is accorded to such terms and expressions by persons skilled in the technical field as set forth above except where different specific meanings have otherwise been set forth herein. Detailed Description

[0032] In order to improve the resilience of wireless communication networks to jamming and (hostile) radio frequency (RF) interference, particularly in conflict situations, examples herein-described propose a solution that integrates a number of (typically three) clocks, for example Chip Scale Atomic Clocks (CSAC) on a 5G base station (e.g., 5G gNodeB), bonded to a distributed 5G network as a bonded atomic PNT solution, that does not need fibre optic links. In some examples, accurate timing (‘T’) may be achieved using only one base station, whereas to provide positioning (‘P’) or navigation (‘N’) multiple (i.e., 2 or more) base stations (nodes) are needed. In particular, a clock synchronization process is designed to provide robust, accurate, and resilient timekeeping across a distributed network of ‘local reference’ (LR) time sources. Here, for example, defence and security personnel who want to be able to accurately know where they are located, whilst concurrently providing high bandwidth communication is provided. Additionally, an increased resilience to GPS and satellite communication drifts, jamming and errors is provided, whilst reducing PNT errors and potential positioning spoofing risks. The accurate source of time also provides the ability to generate accurate frequency signals. A plurality of connected 5G base stations (e.g., gNodeBs) are configured to detect and reduce local reference (LR) drift at a lower ‘cost’, where ‘cost’ includes one or more of: space, weight, electrical power and monetary cost.

[0033] 5G networks require very accurate timing to function properly, as part of this where GPS is denied, a single onboard CSAC can provide up to 2 weeks resiliency to GPS, which is the known limit of accommodating atomic clock drift with a CSAC. By bonding three CSAC’s together using a 5G base station, each 5G base station is able to both broadcast Position, or Time to User Equipment (UE’s such as phones and other devices), which enables Navigation, and is able to determine any future base station geographical location by relay of time, whilst also concurrently supporting network communications. In this manner, it is possible to deliver low-power, safe, secure, high performance, robust and reliable operational timing.

[0034] Although examples are described herein with respect to a 5G implementation, it is envisaged that the concepts are equally applicable to 6G systems and many other communications systems that use a satellite-sourced reference time. Thus, herein, the term ‘local reference point’ may be used interchangeably for a specific example of a 5G base station that includes multiple clocks, to encompass say 6G and other envisaged communications systems that may benefit from an alternative (albeit temporary, in some instances) solution to a satellite-sourced reference time.

[0035] In some examples, a significant benefit in a resilience to GPS interference may be achieved, which is advantageous in certain regions where GPS may currently be denied, such as Baltic countries. In some examples, the 5G base stations are configured to also receive all four timing constellations, i.e., GPS (US), BEIDOU (China), GALILEO (EU), and GLONASS (Russia), in order to generate an understanding of whether a single CSAC is starting to slip (e.g., drift mitigation), or if a timing constellation is exhibiting errors (e.g., due to spoofing, timing signal checking). By bonding these elements together, it is possible to generate machine learning (ML) calibrations on each 5G base station. Thus, in this manner, a network of CSAC atomic clocks located onboard 5G base stations and synchronised through a 5G radio system may be used to detect and mitigate jamming and deception jamming (i.e., inaccuracy of atomic clock time in terms of drift and / or jitter).

[0036] Some examples herein-described provide high capability PNT, with GPS denial, deception resilience and mitigation. In this manner, the 5G base stations and associated 5G network offer an accessible, low power, minimal maintenance, affordable solution, thereby allowing for a reliable alternative to GPS with high precision (e.g., as a calibrated drift time (‘t’)) and with limited drift risks between 5G base stations over a large area of coverage.

[0037] This presents an opportunity to set up an effective network architecture to facilitate a single information environment, referred to herein as ‘assured data’, within a forward tactical ad hoc network to allow for the dispersed and disaggregated C2 of future operations, alongside ensuring secure C4ISTAR (Command, Control, Communications, Computers (C4) Intelligence, Surveillance and Reconnaissance (ISR)). For example, this is the ability to use the connectivity in a defence environment, whereby there's a need to link the full scope of the environment, where the person with the sensor, through to the data, through to the decision maker, is often dispersed.

[0038] By use of bonding of multiple data sources via data fusion to the onboard PNT systems enables a determination as to whether jamming exists, or the node is an active target. In this context, data fusion encompasses taking all time inputs, and forming an accurate time. Therefore, constellation time inputs are received, plus the CSAC times with drift curves, and the next GnodeB times.

[0039] In some examples, this may provide an integrated PNT system, where there is data source redundancy (e.g., a loss in one of the connected 5G nodes), even during extended periods where the system may or may not be in use or contributing. By using the simple formula of Speed = Distance / Time and with a known radio frequency propagation speed coupled with having accurate reliable calibrated time both in a 5G gNodeB, it is possible, by means of message time delay, to calculate a range and then triangulate messages between two different 5G gNodeB’s.

[0040] In some examples, at a system level, it is envisaged that it is possible that a real-time kinematic (RTK) layer could be overlaid on top of the receive time; this is where the gNodeB’s are configured to look at the exact movement the carrier wave signal starts to rise, removing the time uncertainty from the data packet, rather than just the data packet arrival time or the message time. In this manner, this may provide an increase in accuracy far above GPS depending on the carrier wave frequency. For example, this could be at the military spectrum band of 24.35Ghz within the 5G mm wave spectrum, where it is possible to have a positioning accuracy of + / 12.3mm from each gNodeB. Thus, in this manner, the bonded atomic clocks provide the time references at each node, which may be broadcast via 5G or in this example by adding a RTK layer where it is possible to obtain a far more accurate distance measurement for ‘P’ and / or ‘N’.

[0041] In some examples, the level of accuracy may be changed rapidly, depending on the capability requirement, such as if it was required for Autonomous Underwater Vehicles (AUV), Anti-Submarine Warfare (ASW), Mine Countermeasure Vessel (MCMV), and Littoral Strike operations etc. In some examples, the level of accuracy may be changed rapidly, with changing of operational frequency bands, (active methods possible) to avoid detection, as each 5G base station (e.g., 5G gNodeB) has a multi-band capability. For example, by changing the 5G band from, say, a low FR1 band such as 700MHz to an FR2 5G frequency say 52.6GHz, it is possible to change the accuracy of the RTK element from +1- 42cm to +1- 5.7mm, albeit that this example change occurs with a reduction in range. In some examples, time signal message encryption may be provided and, as such, would reduce risks of a false messaging perspective.

[0042] In some examples, it is envisaged that the frequency of operation of the 5G base stations may be arranged to rapidly change, unlike GPS-reliant existing 5G base stations. As such, examples described herein may be less likely to be contested, as the timing control is located much closer to the user, as compared to GPS satellite-control that uses higher transmit power. Thus, as GPS is a fixed frequency with a wide-area coverage, examples herein-described are configured to employ beamforming (i.e., generate a narrow beam) for the distributed PNT signal, thereby making the PNT source very hard to detect with also a potentially very high-power beam that is hard to jam. Alongside this, it is envisaged that examples described herein may be configured as point-to-point and have a very low EW footprint with beamforming, unlike GPS.

[0043] In some examples, it is envisaged that the system with three 5G base stations with bonded atomic clocks, using mm wave frequencies, may be capable of mm level precision at multiple km, thereby enabling accuracy that is far higher than GPS and notably without a need for a significant number of base stations / nodes. Advantageously, it is envisaged that a system with four to five 5G base stations with bonded atomic clocks, using mm wave frequencies, may be capable of an order of magnitude higher in accuracy than many alternatives currently available.

[0044] Referring now to FIG. 2, a simplified block diagram of a self-learning clock synchronization system 200, adapted to leverage for example 5G base stations, a central processing device or master base station, and AI-ML processing for drift and / or jitter detection and synchronization, and adapted to provide a synchronized distributed network of local time sources, is illustrated, according to some example embodiments. In some examples, the selflearning clock synchronization system 200 may ensure continuous operation, even in GNSS failure scenarios. The synchronization process provided by the 5G system 200 provides robust, accurate, and resilient timekeeping. As will be appreciated by a skilled artisan, an accurate source of time implies the ability to generate accurate frequency signals, if needed. Although described with reference to a 5G system 200, it is envisaged that other systems that use an accurate timing reference may also benefit from the concepts described herein.

[0045] In this example 5G system 200, it is envisaged that the system may leverage 5G Integrated Access Backhaul (IAB) communication links to enhance time transfer capabilities between 5 G base stations acting as local reference points, thereby providing high-speed, low-latency connections that enable precise synchronization. Furthermore, 5GIAB communication links provide symmetry and stability of the communication link, thereby ensuring compatibility with the stringent requirements of time transfer protocols like PTP. A 5G infrastructure allows the 5G system 200 to scale easily, supporting additional reference points or devices without degrading synchronization accuracy.

[0046] In examples herein-described, the distributed network of local time sources in the 5G system 200 is represented as a network of 5G base stations each of which preferably contains multiple clocks of the same type. A processor in each 5G base station, e.g., a FPGA with SDR that has a bidirectional link with each clock, is configured to measure and compare their own timekeeping accuracy of their own 5G base station clocks against a monitored shared clock from GNSS 210.

[0047] Global Navigation Satellite Systems (GNSS), such as GPS, Galileo, GLONASS, and BeiDou, provide highly accurate time references, but their signals can be affected by ionospheric distortions. The ionosphere is a region of the Earth's upper atmosphere that can bend and delay GNSS signals as they travel through it. This distortion varies depending on satellite elevation (the angle between the satellite and the observer on Earth) and can introduce timing errors. By implementing the concepts described herein the system can identify errors within a single GNSS constellation, whether due to natural disturbances, accidental misconfigurations, or intentional interference (e.g., spoofing or jamming attacks). Since multiple GNSS constellations are available, comparing signals across two or more constellations allows inconsistencies to be detected and mitigated. Additionally, a dual-frequency GNSS receiver significantly corrects for ionospheric effects by analyzing signal delays at two different frequencies, improving overall accuracy.

[0048] In examples herein-described, the time derived from the monitored shared clock from GNSS 210 serves as the ultimate long-term stable reference, linking all local clocks back to a Coordinated Universal Time (UTC). This ensures that the system remains synchronized with an internationally recognized standard, providing a stable, accurate, and globally accepted time reference for all 5G base stations.

[0049] The 5G system 200 is designed to handle outages in GNSS 210, not continuous denial of GNSS service. In examples herein-described, the local clocks in the local time sources, e.g., the distributed network of 5G base stations, are monitored against GNSS 210 signals as well as each other to detect and correct deviations. This allows the 5G system 200 to identify anomalies such as drift and / or jitter, step changes, or environmental impacts. In the event of GNSS 210 signal loss, it is envisaged that the local clocks maintain timekeeping using the most recent corrections and guidance. In some examples, it is envisaged that the 5G system 200 dynamically calculates and updates error bounds for local clocks based on one or more of: (i) Time elapsed since the last GNSS 210 update; (ii) Known drift rates and / or jitter and environmental factors; (iii) Statistical estimates of possible step changes (if these cannot be detected against other local clock). In this manner, the error bounds, when re-synchronizing on-board clocks, indicate a growing uncertainty over time. In some examples, it is envisaged that local references / 5G base stations use these error bounds to determine when a clock’s reliability is insufficient, prompting it to be excluded from the system. In some examples, it is envisaged that local references may be configured to synchronize with nearby nodes using the described time transfer methods, thereby reducing reliance on a single GNSS 210 connection, and thereby allowing quicker detection and rejection of a compromised constellation.

[0050] Thus, once the multiple 5G base stations have measured and compared their clock against the monitored shared clock from GNSS 210, the multiple 5G base stations are configured to share their respective measurements across the multiple 5G base stations. With the knowledge of the various measurements of the monitored shared clock from GNSS 210, one or more of observations of aging, GNSS perceived errors and environmental drift can usefully be shared by and between the 5G base stations.

[0051] In one example, the 5G base stations may be configured to aggregate, and store in a repository, clock data from multiple sources (GNSS and local timing circuits) Therefore, in one example, the above scenario provides a real-time GNSS deviation detection system that identifies (and in some instances flags) irregularities, e.g., GNSS system malfunctions, for example due to enemy jamming / spoofing activities. In some examples, it is envisaged that the base station(s), e.g., 5G base stations have AI-ML processing capabilities (as described in FIG. 4 and FIG. 8) that are arranged to use real-time RF input data to determine and analyse an RF in order to detect inconsistencies in GNSS signals. Thereafter, the base station(s), e.g., 5Gbase stations, share (and in some instances securely share) the GNSS deviation irregularities (and / or local time irregularities) with trusted entities for situational awareness. In this manner, an AIdriven GNSS integrity monitoring system is able to learn from global timing anomalies and provides a secure, decentralized approach to sharing GNSS integrity warnings, for example based on collective data.

[0052] Where available, in some examples, the 5G system 200 uses precision communication protocols such as Precision Time Protocol (PTP) or specialized RF links to enable accurate time transfers between local reference points / 5G bases stations 220, 230, 240. As illustrated, local references / base stations 220, 240 receive an accurate timing reference from the GNSS, whereas local reference / base station 230 receives a broadcast PNT signal via 5G, and not via the GNSS. In some examples, the communication link between local reference points / 5G bases stations is arranged to ensure symmetric latency in both directions, an important factor for accurate time transfer. In some examples, the communication link between local reference points / 5G bases stations is arranged to have low jitter in order to minimize unreported variability in latency and thereby ensure consistent and reliable synchronization. In some examples, the communication link between local reference points / 5G bases stations is arranged to increase the available clock pool without requiring additional physical devices. In one such example, it is envisaged that two distant reference points equipped with rubidium clocks may be arranged to exchange timing data, providing mutual validation and improving overall reliability. In another such example, it is envisaged that local reference points / 5G bases stations with limited electrical power may be employed to avoid needing to host the more power-hungry types of clock (e.g. rubidium atomic).

[0053] In some examples, at least two reliable clocks are required for effective operation in cases where the GNSS is operational. It is envisaged that a third clock provides additional redundancy, thereby enabling majority voting to isolate and identify faulty clocks. Using each local clock in the local time sources, e.g., the distributed network of 5G base stations, the 5G system 200 continuously compares the outputs of the local clocks (e.g., quartz, rubidium, or atomic clocks). Notably, each base station performs measurements and identifies any deviation beyond a defined threshold triggers a cross-check against GNSS (or other reference clocks) to confirm the anomaly's source. In some examples, it is envisaged that local time sources, e.g., the distributed network of 5Gbase stations, may combine precision clocks, including quartz, rubidium, and chip-scale atomic devices to provide resilience against individual clock failures or anomalies.

[0054] In examples herein-described, the 5G system 200 also includes a central processing hub 260, for example a base station hosting system, that is configured to run core processes and take measurements of parameters. Each 5Gbase station records its clock’s performance, including parameters such as drift and / or jitter, environmental changes, and step variations and transmits this data to the central processing hub 260, typically periodically and via a secure network connection. In some examples, the central processing hub 260 includes AI-ML processing circuit 262 / capabilities (as described in FIG. 4 and FIG. 8) and referred to as a learning processor. The central processing hub 260 is arranged to aggregate data from multiple reference points across the network and, say, using artificial intelligence and statistical methods (as illustrated in FIG. 8), analyses the aggregated data to perform one or more of: (i) Identify patterns in clock behaviour (including GNSS constellations); (ii) Correlate anomalies with environmental factors (e.g., temperature, magnetic fields, clock orientation against gravity, barometric pressure); (iii) Predict aging trends and potential step changes; and / or (iv) Generate corrective models and adjustments. In some examples, the central analysis of such data allows emerging patterns to be determined by the central processing hub 260, even if newer more accurate clock types are developed.

[0055] Based on the analysis, the central processing hub 260 is configured to issue updates to local reference points, such as the 5G base stations. In some examples, it is envisaged that these updates may include one or more of: (i) Corrections for known drift and / or jitter and aging effects; (ii) Predictions for likely step changes under varying conditions; (iii) Guidance on which clocks (in which local reference points / 5G bases stations) to prioritize or de-prioritize based on reliability. Thereafter, it is envisaged that the local reference points / 5G bases stations apply the most recent corrections that they have received or possess, ensuring robust alignment with the most accurate and current timing data.

[0056] In some examples, it is envisaged that the central processing hub 260 in the 5G system 200 may continuously refine its performance through iterative learning using AI-ML processing circuit 262. In this manner, in some examples, it is envisaged that, by aggregating data from multiple locations, the central processing hub 260 in the 5G system 200 is able to provide a comprehensive view of clock behaviour under varying conditions. With Al-driven analysis, the central processing hub 260 in the 5G system 200 detects trends and anomalies in AI-ML processing circuit 262, and instigates procedures (and instructions) to mitigate them, thereby improving the system’s understanding of factors such as temperature sensitivity or environmental impacts. In some examples, the model will be able to predict drift, jitter and time with accurate measurement of CSAC time and environmental factors that are used in the model these include the CSAC temperature, voltage, gravity, acceleration seen. Furthermore, by employing periodic updates, central processing hub 260 in the 5G system 200 ensures that the 5G system 200 evolves to address emerging challenges and / or new clock technologies. In some examples, it is envisaged that the central processing hub 260 in the 5G system 200 may also predict when a clock is to be replaced (e.g., nearing the end of its useful life), for example based on observed drift rates and other indicators, such as observed drift, jitter, or susceptibility to environmental factors monitored and assessed by the AI-ML processing circuit 262.

[0057] In some examples, it is envisaged that the central processing hub 260 in the 5G system 200 may apply statistical techniques to evaluate clock performance. For example: (i) If multiple clocks show a sudden frequency step change simultaneously, it may indicate an external disturbance (e.g., environmental or power-related); (ii) The central processing hub 260 in the 5G system 200 may then identify the most reliable clocks for timekeeping, ensuring accuracy despite localized disruptions.

[0058] In some examples, it is envisaged that the central processing hub 260 in the 5G system 200 may apply statistical techniques to evaluate clock performance ( based on the atomic clock information obtained, e.g., CSAC) received using time division duplex (TDD) and IAB to determine a range of base stations . In some examples, it is envisaged that the central processing hub 260 in the 5G system 200 may apply statistical techniques to identify GNSS spoofing. In some examples, it is envisaged that the central processing hub 260 in the 5G system 200 may apply statistical techniques to maintain PNT operations despite GNSS spoofing / denial. In some examples, it is envisaged that the central processing hub 260 in the 5G system 200 may apply statistical techniques using RTK to improve a determined accuracy of a location of base stations. RTK processing refines GNSS positioning by correcting for atmospheric distortions and leveraging high-precision carrier-phase measurement.

[0059] In scenarios where two 5G base stations have a direct connection that has very low jitter on latency, it is envisaged that it may be possible to transfer the measured time information from a (local) clock / timer in one 5G base station to the other 5G base station - thus in effect increasing the population of local clocks. As the two 5G base stations are able to communicate timing to each other it is possible to effectively reset the clock drift if a counter can be ‘set to zero’ if the two 5G base stations are able to re-synchronize.

[0060] Thus, in examples described herein a novel, self-learning clock synchronization system uses multiple clocks (quartz, rubidium, chip-scale atomic) contained within a 5G base station at various locations. Each 5G base station sends data to the central processing hub 260 for processing by AI-ML processing circuit 262. This central processing hub 260 may be configured to issue corrections based on detected drift and / or jitter and environmental factors. Advantageously, the system does not require real-time, low-latency connections but benefits greatly from them.

[0061] Although the central processing hub 260 is illustrated as being a separate entity to the base stations, and arranged to communicate with each of a number of base stations, it is envisaged in some examples that a central processing hub may be configured as part of a master base station 245 with a learning processor 247 (ALML processor), say, as illustrated in FIG. 2.

[0062] Thus, examples herein-described provide an adaptive AI and the use of a range of timing solutions bonded to accurate clock types that can be monitored for environmental changes in order to detect and compensate for anomalous changes.

[0063] In some examples, it is envisaged that a combination of the aforementioned technologies, enables the deployed 5G network of base stations to be able to be continuously updated as GNSS or RF spectrum interference changes. In some examples, it is envisaged that updates may dynamically redeploy (either automatically or via manned intervention) the geographical configuration of the network and RF signature that it uses.

[0064] In the context of the examples herein-described, the inventors have recognised and appreciated the following aspects related to the accuracy of clocks. Short-term variation due to clock noise (‘phase noise’) is important for RF purposes, but by definition the long-term average of clock noise is ‘zero’. Aging of clocks may cause permanent changes in frequency due to slow physical changes in the clock. Environmental effects, such as temperature, supply voltage, mechanical accelerations, may also affect the accuracy and reliability of the clock. Furthermore, at high-precision levels, smaller environmental effects may occur, such as the connected load impedance, orientation in Earth’s gravitational field, orientation and strength of Earth’s magnetic field, barometric pressure, perhaps even humidity. In general a majority of these effects may be reversible using the concepts described herein. Other effects that may be reversible using the concepts described herein include random ‘fast’ changes in a clock’s that cause a permanent step change in frequency and ‘retrace’ errors, i.e., when a clock is powered off it cools to an ambient temperature and is then powered on, where it may show a small change in frequency (lasting at least many hours).

[0065] FIG. 3 illustrates a simplified drawing of a 5G base station 300 according to examples herein-described. The 5Gbase station 300 (often referred to as a gNodeB (gNB)), includes hardware 350 functionality and software 305 functionality. The 5G base station 300 is configured to communicate with another communication network and other 5G base stations and wireless communication devices, such as user equipment (UEs). In this illustration, the software 305 functionality includes four functions, referred to as: Centralized Unit (CU) 320 and IAB Distributed Unit (DU) 315 and a Radio Unit (RU) 310. In some examples, the CU 320 provides support for the higher layers of the protocol stack such as Service Data Adaption Protocol (SDAP), Packet Data Convergence Protocol (PDCP) and Radio Resource Control (RRC) whilst, the DU 315 provides support for the lower layers of the protocol stack such as Radio Link Control (RLC) and Medium Access Control (MAC) layers, whereas the RU 310, sometimes referred to as radio head, is configured to handle the Physical layer of communications in the protocol stack.

[0066] The 5G base station 300 also includes a 5G core 325, in essence network gateway software that fundamentally controls the network's operation. The 5G core 325 is configured to provide signalling control (such as user authentication and data session creation and tear-down and typically known as the ‘Control Plane’), plus transport of data across those sessions (typically known as the ‘User Plane’). Time synchronization data distribution 317 is passed between the DU 315 and the CU 320. Network-wide time synchronization data 322 is passed between the CU 320 and the 5G core 325.

[0067] In accordance with some examples, the hardware 350 functionality comprises a number of interconnected hardware functions, as shown. Such hardware 350 functionality includes a field programmable gate array (FPGA) device with a software defined radio (SDR) card 360, which performs signal processing and timing capture. The hardware 350 functionality also includes a plurality of sensors 380, which provide measurements on various parameters, for example including (but not limited to): temperature, pressure, location, supply voltage, orientation to gravity, magnetic fields, etc. In some examples, it is envisaged that a use of modern cheap sensors may be exploited. In some examples, such sensor data is used as input sources of data to an AI-ML model in order to understand the impact of the sensed data and corresponding sensed circuits / devices on the CSAC performance. In this manner, the AI-ML model is able to predict future changes over time, for example if the base station is moved or something occurs where these sensed devices are affected.

[0068] In accordance with some examples, the hardware 350 functionality comprises a low-cost sensor array 380 for clock stability monitoring of the base station. The sensor array 380 may be configured to measure one or more (a plurality or all) of: pressure, temperature (for drift and / or jitter correction); orientation (via accelerometer / gyroscope, hall effect, etc.); magnetic field alignment (compass for position confirmation); supply voltage 450 (to detect potential power anomalies affecting the clock stability); and location data (GPS or other positioning inputs). In some examples, the sensor array may be embedded in local clock hardware as a self-monitoring sub-system. In some examples, suitable, known data may be input into AI-ML model (or processor) in order to detect pattern anomalies that may affect the base station’s clock drift and jitter. In this manner, the sensor output from the AI-ML model may be used for self-calibration routines to predict faults and thereafter correct timing inconsistencies.

[0069] A single board computer 370 (i.e., a primary computing platform) is connected 362 to the FPGA with SDR card 360. The FPGA with SDR card 360 provides time stamped RF data 368 to the RU 310. The single board computer 370 is connected to the 5G core 325 via a hardwaresoftware interface 372.

[0070] In accordance with examples herein described, such hardware 350 functionality also includes a graphic processing unit (GPU) 390, which contains an AI-ML based processing engine (sometimes referred to as a ‘learning processor’), namely a processor of time data and associated data that impacts the CSAC performance. In accordance with some examples, the AI-ML based processing engine GPU 390 is configured to receive real-time sensed data from sensors 380 for immediate signal processing. In this manner, the use of such input data sources together with the CSAC data helps the AI-ML based processing to understand drift and / or jitter and apply corrections. For example, taking one or more of: temperature, voltage, magnetic field, pressure, etc. and inputting this sensed information with the PPS timing signal into the GPU 390. Once the sensor data is matched to the observed drift and / or jitter, it is possible to obtain an accurate time, even when the (satellite) reference time is removed.

[0071] The AI-ML based processing engine GPU 390 is configured to provide re-configuration signals / outputs to SBC 370, which in turn recommends environmental-based adjustments for the sensors 380, in response to AI-ML based timing predictions (sensors 380 to GPU 390).

[0072] The AI-ML based processing engine GPU 390 is configured to ingest raw CSAC data as well as the environmental parameters and run the neural network. The Al-based engine GPU 390 is also connected to a PNT circuit 340, arranged to process time data and associated data that impacts a CSAC performance. The AI-ML based processing engine GPU 390 provides drift and / or jitter detection information and synchronization data 392 to the 5G core 325, AI-ML-based timing predictions 394 to the CU 320. The AI-ML based processing engine GPU 390 provides timing adjustments data 396 to the DU.

[0073] FIG. 4 illustrates a block diagram of a 5G base station 400, which in some examples may support IAB, adapted in accordance with some example embodiments. In some examples, it is envisaged that the central processing hub 260 may be of a similar architecture to the 5G base station 400 of FIG. 4, as would be understood by a skilled artisan. In some examples, the 5G base station 400 may be simplified to function as a repeater, without the suite of control functions that are contained in a fully-functional base station. The 5G base station 400 contains an antenna array 402, for receiving transmissions, coupled to an antenna switch 404 that provides isolation between receive and transmit chains within the 5G base station 400. One or more receiver chains, as known in the art, include receiver RF front-end (RFFE) circuitry 406 (effectively providing reception, filtering and intermediate or base-band frequency conversion). The receiver front-end circuitry 406 is coupled to a signal processing module 408 (generally realized by a digital signal processor (DSP)). A skilled artisan will appreciate that the level of integration of receiver circuits or components may be, in some instances, implementationdependent.

[0074] The controller 414 maintains overall operational control of the 5G base station 400. The controller 414 is also coupled to the receiver RFFE circuitry 406 and the signal processing module 408. In some examples, the controller 414 is also coupled to a frequency generation circuit 417 and a memory device 416 that selectively stores operating regimes, such as decoding / encoding functions, synchronization patterns, code sequences, and the like. A PNT circuit 418 is operably coupled to the controller 414 to control the timing of operations (e.g., transmission or reception of time-dependent signals) within the 5G base station 400. In accordance with examples described herein, the PNT circuit 418 is responsive to control signals issued by the controller 414 to adjust the timing of operations in the 5G base station 400 according to the concepts described herein. In this example, PNT circuit 418 is operably coupled to a plurality of clocks 430. The plurality of clocks, which in some examples includes three clocks, may be quartz, rubidium, and / or chip-scale atomic clocks (CSACs). Ideally, the plurality of clocks are the same type of clock.

[0075] As regards the transmit chain, this essentially includes an input circuit 420 configured to receive information from one or more sensors connected to signal processor 408. In some examples, the one or more sensors may include at least a sensed temperature to detect external influences on a local clock (i.e., in a timer in PNT circuit 418). The signal processor 408 is coupled in series through transmitter / modulation circuitry 422 and a power amplifier 424 to the antenna array 402, which in some examples comprises a plurality of antenna arrays that are configured as a steerable beam-forming antenna. The transmitter / modulation circuitry 422 and the power amplifier 424 are operationally responsive to the controller 414. In accordance with some example embodiments, the signal processor 408 and transceiver (e.g., transmitter / modulation circuitry 422) of the 5G base station 400 may be configured to communicate with another 5G base station.

[0076] The signal processor 408 in the transmit chain may be implemented as distinct from the signal processor in the receive chain. Alternatively, a single processor may be used to implement a processing of both transmit and receive signals, as shown in FIG. 4. Clearly, the various components within the 5G base station 400 can be realized in discrete or integrated component form, with an ultimate structure therefore being an application-specific or design selection.

[0077] In some examples, the signal processor 408 includes an artificial intelligence circuit 432 (sometimes referred to as a learning processor and with some aspects described with reference to FIG. 8), which may be coupled to a predictive model circuit 434, which is configured to consider the environmental state of the antenna structure, to anticipate changes in the absolute timing of signals within the 5G base station 400 and between the 5G base station 400 and other base stations, such as base stations 220, 230, 240 in FIG. 2. The predictive capability provided by the predictive model circuit 434, coupled to antenna beam-form controller 435 ensures that adjustments are made in a timely manner, thereby aligning the antenna and timing circuits for optimal communication.

[0078] In some examples, for example when the base station is a master base station (or in some examples the central processing hub 260 of FIG. 2) the signal processor 408 may include an artificial intelligence (AI) circuit 432 (such as learning processor 247 or 262 in FIG. 2), which works alongside a predictive model circuit 434. The predictive model circuit 434 is designed to analyze the environmental conditions of the antenna structure and anticipate potential changes in the absolute timing of signals within the 5G base station 400 and between 5G base station 400 and other base stations (e.g., base stations 220, 230, 240 in FIG. 2).

[0079] This predictive capability allows the system to proactively adjust for timing variations, which can be caused by environmental factors such as temperature changes, signal propagation delays, or network congestion. The predictive model circuit 434 is also linked to an antenna beam-form controller 435, enabling real-time adjustments to antenna alignment and timing circuits. These adjustments optimize signal transmission and reception, ensuring synchronization and efficient communication between base stations.

[0080] By integrating Al-driven predictive modelling with beam-forming control, this system enhances network stability, timing precision, and overall communication efficiency in a dynamic 5G environment.

[0081] In accordance with examples described herein, the PNT circuit 418 comprises three or more types of clock (so that random changes can be detected and aging will not be correlated). Here, multiple instances of a single type of clock, distributed across multiple 5G base stations, are useful to study that clock type. In some examples, it is envisaged that, ideally, the respective clock type distributed across multiple 5G base stations, may be from different manufacturing batches in order to decorrelate aging effects. Here, it is envisaged that the central processing hub may track batch numbers for clocks so that it is possible to identify behaviour common to all batch members of a particular clock type.

[0082] In accordance with examples described herein, the PNT circuit 418 may be connected to one or more timing-grade GNSS circuit 436 in order to provide a long-term stable one part per second (1PPS) source. Here, multi-constellation, dual frequency for a most robust performance; surveyed-in to a fixed location to reduce the effect of switching satellites; second receiver to assess inter-constellation correlation (and hence spot potential interference); monitor internet sources for solar activity - especially CME arrivals (to assess when step-changes in GNSS time are likely); disciplining a PRS-10 rubidium standard to provide the local ‘master’ clock to compare against; if needed, disciplining a well-aged OXCO to distinguish random changes in the PRS-10 from random changes in GNSS time.

[0083] Referring now to FIG. 5, a simplified overview 500 of the data involved in timing operations and calculations of a central processing hub / master base station in a communication system is illustrated, in accordance with some example embodiments. In this example overview 500, timing information from an accurate standard 510, such as a rubidium standard (or GNSS, GPS or GLONAS) is illustrated, showing a time ‘A’ 512. In some examples, each base station includes three CSAC clocks 520, 524, 528, respectively providing three times: Time ‘B’ 522, Time ‘C’ 526, Time ‘D’ 530. At 540, all the respective times are compared. At 545, drift curves of the CSAC clocks 520, 524, 528 are created and measured. At 550, the AI / ML model in the central processing hub / master base station measures various parameters over a period of time. At 555, if, say, a GPS time is deleted, a determination of where the respective base station’s timing is made by assessing the drift curve, to determine whether the GPS clock is drifting or whether the CSACs are drifting. In one example, a determination may be made as to whether all satellite constellations are correct. At 560, predictions / estimates of time are made by comparing the timing information of time ‘A’ 512 from the accurate standard 510, with each of the CSAC clocks 520, 524, 528. In one example, a gNB 565 receives improved predictions estimates of time 560 and gNB 565 uses these as it's source for time keeping.

[0084] A graphical representation of the above is shown in the graph 570 of ‘real time’ 572 versus recorded time 574. In this example, the GNSS stops working (or being received) between Tx 576 and Ty 578. However, in accordance with the examples described herein, the 5G base station (or a local reference point) was able to monitor drift vs GNSS. It is anticipated that they will drift at the same rate; as such it is possible to estimate real time based on differentials between drifting clocks.

[0085] Referring now to FIG. 6, a simplified flowchart 600 illustrates the timing system flow of a PNT circuit 610 (such as PNT circuit 340 in FIG. 3 and PNT circuit 418 in FIG. 4), in accordance with some example embodiments. In this system, the PNT circuit 610 serves as the central processing unit for maintaining highly accurate positioning, navigation, and timing (PNT) data. To achieve this in this example, the PNT circuit 610 receives timing data from two primary sources: (i) A Timing-Grade GNSS Circuit 620 (such as timing-grade GNSS circuit 436 in FIG. 4); and (ii) A PRS-10 Rubidium Standard 680. The Timing-Grade GNSS Circuit 620 plays an important role in synchronizing time and position information, as it processes satellite signals and enables the following: • Multi-Constellation GNSS Signals 630 - Supports multiple satellite navigation systems (e.g., GPS, GLONASS, Galileo, BeiDou) to improve accuracy and redundancy; • Dual Frequency Signals 640 - Uses dual-frequency GNSS signals to correct ionospheric errors and enhance precision; • Fixed Location Calibration 650 - Ensures accurate time synchronization by referencing a precisely known location, minimizing timing errors; • Second GNSS Receiver 660 - Provides redundancy and cross-verification to detect errors, spoofing, or anomalies in primary GNSS signals; and • Solar Activity Monitoring 670 - Monitors solar activity, such as geomagnetic storms, that can interfere with GNSS signals, allowing for adaptive timing corrections.

[0086] Together, these components allow the PNT circuit 610 to generate a robust and resilient timing solution, critical for applications that require ultra-precise synchronization, such as communications networks, financial systems, and scientific research.

[0087] FIG. 7 illustrates a graph 700 illustrating some example results of a preliminary simulation according to some example embodiments. Graph 700 illustrates an intersect of the potential locations from triangulation of three message times, with error bands. The hued dots 730 in FIG. 7 represent the centre points of three equally spaced PNT base stations. A first set of lines 740 in the graph 700 illustrates the position of the wave adding the error. A second set of lines 750 shows the position of that same wave subtracting the error. The axis 710, 720 illustrates a relative distance between the PNT base stations. As illustrated with the lines 740, 750 in the graph 700, a greater area is provided than if the point of interest is in the middle of the three equally spaced PNT base stations. Given the actual distances involved (e.g., cm compared with km) the area was approximated to a square. These values therefore have approximations that are highly accurate, due to the very small area (see Table 1 below). This is based on a three-node array of interconnected 5G NodeB's. Frequency Point to Point Precision (cm) Triangulated positional area Triangulated positional approximation / 11^ (cm 2) approximation (m 2) 350 MHz 85.5 7310 0.731 900 MHz 33.3 1110 0.111 2.4 GHz 12.5 156 0.0156 3.6 GHz 8.3 68.9 0.0069 5.0 GHz 6 36 0.0036 24 GHz 1.2 1.4 0.00014 Table 1: Accuracy results from the simulations run. The above results illustrate the accuracy achieved relative to the 5G frequency band used for ‘P’ and ‘N’

[0088] FIG. 8 illustrates an example of a neural network 800 that may be employed as an artificial intelligence / machine learning (AI-ML)-based learning processor architecture to analyse jitter, drift, temperature, gravity and voltage changes at the CSAC to predict accurate time.

[0089] In some examples described herein any reference to Al encompasses a mathematical methodology to extract information from signals and / or images that may be similar to that used in cognitive functions in humans associated with learning or pattern recognition, in a supervised or unsupervised deep learning setting. Here, flexible adaption is able to extract information previously not necessarily recognized or hypothesized or based on manually derived algorithms specific to the measurement of defined parameters. In the former case, Al in the setting of unsupervised learning is able to exploit serendipity in much the same way as scientific discoveries are made by chance.

[0090] In some examples described herein, any reference to ML, as distinct from Al, may be used as a mathematical methodology that refers to mathematical and statistical methods, which can without supervision, analyse and identify a specific outcome related to, in the case of examples described herein, a performance of analysing an identification of radio frequency (RF) space availability / suitability, say within a contested RF environment. In some examples, it may apply to algorithms and iterative methodologies in order to arrive at an optimal identification of a set task. In some examples, it may be used for data mining. For simplicity of terminology, however, any reference to AI examples of the invention described herein encompass supervised and unsupervised learning, as well as Machine Language.

[0091] In some examples, an artificial neural network is a network of artificial neurons and nodes meant for solving learning processor-type problems, such as (AI) problems. The connections of the neuron are weighted to model excitatory or inhibitory connections and summed. An activation function defines the amplitude of the output. It is envisaged in some applications of the invention that such artificial neural networks may be used for predictive modelling and adaptive control. In some examples, these neural networks may also be trained via a dataset, such as the examples described herein. Self-learning or unsupervised learning, resulting from analysis of datasets, can occur within the network, which allows the network to derive conclusions from a seemingly unrelated information set.

[0092] In some examples, the dataset that the Al algorithms will use as a learning platform will include, but is not limited to, data obtained from analysing a timing of RF signals, for example within a contested RF environment and a non-contested environment. The analysed timing of RF signal variables derived from the techniques discussed can be derived using conventional analysis or derived using Al methodologies. The timing parameters of RF signal system may be used as input variables and these variables / parameters may then be used as a learning set for Al to evaluate the various example embodiments described herein.

[0093] Referring now to FIG. 8, an example of a neural network 800 that may be employed as a learning processor, such as an artificial intelligence (Al)-based architecture to analyse the timing of RF signals is illustrated according to some examples of the present invention. In some examples, the example neural network 800 may comprise a convolutional neural network 800, which applies a series of node mappings 880 to an input 810, which ultimately resolves into an output 830 consisting of one or more values, from which at least one of the values is used by the neural network 800, for example an Al-based architecture. The example convolutional neural network 800 comprises a consecutive sequence of network layers (e.g. layers 840), each of which consists of a series of channels 850. The channels are further divided into input elements 860. In this example, each input element 860 may store a single value. Some (or all) input elements 860 in an earlier layer are connected to the elements in a later layer by node mappings 880, each with an associated weight. The collection of weights in the node mappings 880, together, form the neural network model parameters 847. For each node mapping 880, the elements in the earlier layer are referred to as input elements 860 and the elements in the output layer are referred to as the output elements 870. An element may be an input element to more than one node mapping, but an element is only ever the output of one node mapping function 820.

[0094] In order to calculate the output 830 of the convolutional neural network 800 the system first considers the input layer as the earlier layer. The layer(s) to which the earlier layer is connected by a node mapping function 820 are considered in turn as the later layer. The value for each element in later layers is calculated using the node mapping function 820 in equation [1], where the values in the input elements 860 are multiplied by their associated weight in the node mapping function 820 and summed together. Node mapping function 820: d = A(wad x a + wbd xb + wcd x c) [1]

[0095] The result of the summing operation is transformed by an activation function, ‘A’ and stored in the output element 870. The convolutional neural network 800 now treats the previously considered later layer(s) as the earlier layer, and the layers to which they are connected as the later layers. In this manner the convolutional neural network 800 proceeds from the input layer 840 until the value(s) in the output 830 have been computed.

[0096] In examples of the invention, the convolutional neural network 800 may be trained. In some examples of the invention, the training of the convolutional neural network 800 may entail repeatedly presenting timing of RF signal data as the input 810 of the convolutional neural network 800, in order to analyse the performance of the 5G system. In some examples of the invention, an optimisation algorithm may be used to reduce a loss function, for example by measuring how much each node mapping 880 weight contributed to the loss, and using this to modify the node mapping functions 820 in such a way as to reduce the loss. Each such modification is referred to as an iteration. After a sufficient number of iterations, the convolutional neural network 800 can be used to analyse the function from an input of timing of RF signals data.

[0097] In some examples of the invention, the large number of model parameters 847 used in the convolutional neural network may require the device to include a memory 890. The memory 890 may be used to store the training data 815, the model parameters 847, and any intermediate results 893 of the node mappings.

[0098] Thus, in the proposed 5G system, input data (a training dataset, clinical dataset, model parameters or intermediate results) is fed to the learning processor neural network in a format that fits the input matrix. In order to obtain accurate AI models a reference time source of any suitable clock may be used. However, the more accurate clock the better, and ideally three clocks are used so that timing variations can be monitored, determined and activated upon. Nodes are mapped in a specific way that is adapted to the purpose of the system (forming e.g. a convolutional neural network). The information is gradually reduced through a series of interconnected input / output elements to generate the final output.

[0099] This synchronization process creates a robust, adaptive, and scalable timekeeping system capable of maintaining high accuracy across a distributed network, even under challenging conditions or in the absence of GNSS. It combines advanced data analysis, diverse clock technologies, and innovative communication methods to ensure precision and resilience. With this technology there is a clear technically practical way to provide robust PNT off grid, when GPS has been denied.

[00100] Referring now to FIG. 9, a simplified flowchart 900 of timing synchronisation in a communication system that has a plurality of communication devices is illustrated, in accordance with some example embodiments. The simplified flowchart 900 starts at 910 with a communication device (e.g., a 5Gbase station) receiving a Global Navigation Satellite System, GNSS, shared clock. At 920, the communication device receives or monitors timing information from the communication device’s clock. At 930, the communication device generates clock information from at least one of: measured timing of clock signals obtained from the clock, measured timing of the GNSS shared clock, a comparison of the timing of clock signals obtained from the clock and the GNSS shared clock. At 940, the communication device transmits the derived clock information to at least one of: a central processing hub (such as central processing hub 260 of FIG. 2), others of the plurality of communication devices.

[00101] In some examples, at 950 and at the central processing hub, the flowchart includes aggregating, processing and analysing received clock information from multiple devices of the plurality of communication devices and determine therefrom: at least one emerging pattern related to at least one of: (i) the GNSS shared clock, a clock employed by a transmitting communication device; (ii) correlate anomalies with environmental factors that comprise at least one of: temperature, magnetic field, clock orientation against gravity, barometric pressure; (iii) predict a clock aging trend and potential step change; (iv) a corrective model and transmit a timing adjustment to at least one communication device based thereon.

[00102] Referring now to FIG. 10, a simplified flowchart 1000 of a self-learning clock synchronization, leveraging 5G base stations, a master base station, and AI-ML processing for drift and / or jitter detection and synchronization is illustrated, in accordance with some example embodiments. Advantageously, the self-learning clock synchronization ensures continuous operation, even in GNSS failure scenarios. At 1010, 1020, each 5Gbase station performs data collection clock from a plurality of its clock sources, for example including (but not limited to) one or more of: Quartz oscillators, Chip-scale atomic clocks (CSACs) and other precision timekeeping devices. In some examples, each clock transmits measurement data, including batch numbers, operational metrics, and environmental conditions. At 1030, the clock data from one or multiple 5G base stations is sent over a data transmission network to a central processing hub / master base station. In some examples, for example in addition to the clock data, the 5G base stations may transmit clock frequency drift and / or jitter measurements, environmental conditions, current synchronization status

[00103] At 1040, the central processing hub, which may be located at a base station or a network operations centre, receives and processes the aggregated data. This central processing hub performs initial analysis and anomaly detection, for example identifying trends in clock behaviour across different environments. In some examples, the latest AI-ML models are transmitted to 5G base stations equipped with GPUs capable of executing advanced predictive algorithms. These models enhance time synchronization accuracy and serve as an independent timekeeping source in cases where GNSS signals are unavailable or compromised. At 1050 AI-ML model(s) process clock data and, for example, detected drift and / or jitter patterns, optimize real-time corrections, etc., notably operating independently of real-time GNSS connectivity. Thus, the AI-ML models generate insights that are transmitted back to the centralized processing hub for further refinement. A feedback loop from 1050 via feedback loop 1080 to 1030 enables continuous learning and optimization of the timekeeping models, thereby improving accuracy and resilience over time.

[00104] At 1060, the central processing hub may generate and issue / distribute corrective updates to relevant stakeholders, for example each of the base stations, following data analysis, including via feedback loop 1080 to 1030. The updates may be periodic updates and may include calibration instructions, performance insights, and refined AI-ML models for enhanced timekeeping accuracy. At 1080, the AI-ML model continues to process feedback information in a feedback loop from 1050 to 1030 or from 1060 to 1030, in order to detect drift and / or jitter patterns, optimize real-time corrections, notably operating independently of real-time GNSS connectivity. At 1070, in some examples, in the event of GNSS failure, the AI-ML models stored on the base station GPU provides an alternative means of maintaining accurate time synchronization, ensuring continued network stability.

[00105] In alternative examples of concepts described herein, a system, device and method comprise the central processing hub being arranged to: receive from base stations (or other appropriate clock-based devices) batch numbers of deployed clocks alongside their associated drift and / or jitter patterns, recorded over a period of time, as determined by a learning processor (AI / ML algorithm) in the base stations (or other appropriate clock-based devices); process (correlate) the batch numbers of deployed clocks and associated drift and / or jitter patterns; and identify commonalities that are indicative of manufacturing defects or environmental influences. In some optional examples, the system, device and method may apply the same methodology for other timing standards, such as with an integration with GNSS-based timing circuits (e.g., PNT Circuit (418)) to log batch-specific performance. In some optional examples, the system, device and method may include storing such data in a networked clock tracking system. Thus, in some optional examples, the system, device and method may include a machine-learning-based diagnostic tool located in the central processing hub that predicts failure trends based on batch data. In some optional examples, the system, device and method may include a determination to perform or request an automated recalibration or early replacement notifications of one or more clocks based on observed batch behaviours.

[00106] In alternative examples of concepts described herein, a system, base station (e.g., 5G base station) and method include the base station comprising: a receiver arranged to receive GNSS signals; an AI-ML processor capability (a ‘learning processor’, as described in FIG. 4 and FIG. 8) that is arranged to use real-time RF data that is input to an AI-ML, and is arranged to determine and analyse the RF data in order to detect a deviation irregularity / inconsistency in received GNSS signals (and / or internal clock / timing signals); wherein a transmitter is arranged to share (and optionally securely share) the GNSS deviation irregularity / inconsistency (and / or local clock / time irregularities) with trusted entities for situational awareness. In some optional examples, 5G base stations may be configured to aggregate, and store in a repository, clock data from multiple sources (GNSS and local timing circuits). In this manner, examples provide a realtime GNSS deviation detection system that identifies and flags GNSS irregularities. In this manner, an Al-driven GNSS integrity monitoring system is able to identify and learn from identified global timing anomalies and provide a secure, decentralized approach to sharing GNSS integrity warnings, for example where the GNSS deviation irregularity / inconsistency may be due to enemy jamming / spoofing activities.

[00107] In alternative examples of concepts described herein, a system, base station (e.g., 5G base station) and method include the base station comprising a plurality of sensors 380 that provide measurements on various parameters, for example temperature, pressure, location, supply voltage 450, orientation to gravity, magnetic field alignment, etc. that are used as input sources of data to an AI-ML model that is also arranged to receive clock data, wherein the AI-ML model is arranged to determine a relationship between the sensed data and corresponding sensed circuits / devices and the clock data (e.g., that identifies a performance of a plurality of CSACs). In this manner, the AI-ML model is able to predict future changes over time of one or more of the plurality of CSACs based on the sensed data, for example if the base station is moved or something occurs where these sensed devices are affected. In an optional example, the plurality of sensors may be a sensor array configured to measure one or more (a plurality or all) of: temperature (for drift and / or jitter correction); magnetic field alignment (compass for position confirmation); supply voltage 450 (to detect potential power anomalies affecting the clock stability; and location data (GPS or other positioning inputs). In some examples, suitable, known data may be input into AI-ML model (or processor) in order to detect pattern anomalies that may affect the base station’s clock drift and jitter. In this manner, the sensor output from the AI-ML model may be used for self-calibration routines to predict faults and thereafter correct timing inconsistencies.

[00108] It will be appreciated that, for clarity purposes, the above description has described example embodiments with reference to different functional units and processors. However, it will be apparent that any suitable distribution of functionality between different functional units or processors, for example with respect to the signal processor may be used without detracting from the example embodiments described. For example, functionality illustrated to be performed by separate processors or controllers may be performed by the same processor or controller. Hence, references to specific functional units are only to be seen as references to suitable means for providing the described functionality, rather than indicative of a strict logical or physical structure or organization.

[00109] Aspects may be implemented in any suitable form including hardware, software, firmware or any combination of these. The example embodiments may optionally be implemented, at least partly, as computer software running on one or more data processors and / or digital signal processors or configurable module components such as field programmable gate array (FPGA) devices. Thus, the elements and components of an embodiment may be physically, functionally and logically implemented in any suitable way. Indeed, the functionality may be implemented in a single unit, in a plurality of units or as part of other functional units.

[00110] Although examples have been described in connection with some embodiments, it is not intended to be limited to the specific form set forth herein. Rather, the scope of the present invention is limited only by the accompanying claims. Additionally, although a feature may appear to be described in connection with particular embodiments, one skilled in the art would recognize that various features of the described embodiments may be combined. In the claims, the term ‘comprising’ does not exclude the presence of other elements or steps.

[00111] Furthermore, although individually listed, a plurality of means, elements or method steps may be implemented by, for example, a single unit or processor. Additionally, although individual features may be included in different claims, these may possibly be advantageously combined, and the inclusion in different claims does not imply that a combination of features is not feasible and / or advantageous. Also, the inclusion of a feature in one category of claims does not imply a limitation to this category, but rather indicates that the feature is equally applicable to other claim categories, as appropriate.

Claims

1. A communication system comprising:a central processing hub;a plurality of communication devices connected to the central processing hub, wherein each of the plurality of communication devices comprise:a receiver arranged to receive a Global Navigation Satellite System, GNSS, shared clock;a plurality of clocks;a transmitter; anda processor, operably coupled to the receiver, transmitter and the plurality of clocks, wherein the processor is arranged to generate derived clock information from: measured timing of clock signals obtained from the plurality of clocks, measured timing of the GNSS shared clock, a comparison of the timing of clock signals obtained from the clock and the GNSS shared clock;wherein the transmitter is arranged to transmit the derived clock information to at least one of: the central processing hub, others of the plurality of communication devices.

2. The communication system of Claim 1 wherein the receiver of at least one of the plurality of communication devices (220, 240) is arranged to additionally receive a timing adjustment signal from the central processing hub, in response to an analysis of the transmitted derived clock information.

3. The communication system of Claim 2 wherein the received timing adjustment signal comprises positioning, ‘P’ information, and navigation, ‘N’ information, and timing, ‘T’ information for the at least one communication device.

4. The communication system of Claim 3 wherein the central processing hub (260) applies statistical techniques using a real-time kinematic, RTK, process to derive at least one of:positioning, ‘P’ information, navigation, ‘N’ information, for a communication device of the plurality of communication devices.

5. The communication system of any of preceding Claims 2 to 4, wherein the received timing adjustment signal comprises an observation of a communication device clock aging.

6. The communication system of Claim 5, wherein the observation of the communication device clock aging comprises a prediction of when the communication device clock is to be replaced.

7. The communication system of Claim 6, wherein the prediction of when thecommunication device clock is to be replaced is based on a determination from the transmitted derived clock information of at least one of: observed drift rate, jitter, susceptibility to environmental factors.

8. The communication system of any preceding Claim, wherein the plurality of communication devices are base stations and the plurality of clocks comprise Chip Scale Atomic Clocks (CSAC) configured to form a bonded atomic clock arrangement.

9. The communication system of Claim 8 wherein the derived clock information is derived from the bonded atomic clock arrangement and the derived clock information provides time references at a received communication device of the plurality of communication devices.

10. The communication system of any preceding Claim, wherein the transmitter comprises a beamformer antenna array and the transmitter is arranged to employ beamforming to transmit the derived clock information.

11. The communication system of Claim 1 or Claim 2 wherein the derived clock information that is transmit to the central processing hub (260) comprises at least one of: clock timing, atleast one perceived error of the GNSS shared clock, an environmental drift measurement, a time elapsed since a last GNSS shared clock update.

12. The communication system of any preceding Claim wherein the central processing hub (260) is arranged to aggregate, process and analyse received clock information from multiple devices of the plurality of communication devices and determine therefrom at least one timing pattern related to at least one of: the GNSS shared clock, a clock employed by a transmitting communication device.

13. The communication system of any preceding Claim wherein the central processing hub (260) is arranged to aggregate, process and analyse received clock information from multiple devices of the plurality of communication devices and correlate anomalies with environmental factors that comprise at least one of: temperature, magnetic field, clock orientation against gravity, barometric pressure.

14. The communication system of any preceding Claim wherein the central processing hub (260) is arranged to aggregate, process and analyse received clock information from multiple devices of the plurality of communication devices and predict a clock aging trend and potential step change.

15. The communication system of any preceding Claim wherein the central processing hub (260) is arranged to aggregate, process and analyse received clock information from multiple devices of the plurality of communication devices and generate a corrective model and transmit a timing adjustment to at least one communication device based thereon.

16. The communication system of any preceding Claim wherein the transmitted clock information is transmitted periodically via a secure network connection.

17. The communication system of any preceding Claim wherein the processor is arranged, in response to a determination of a GNSS signal loss, to apply a recent correction to the clock to maintain timekeeping.

18. The communication system of any preceding Claim wherein the plurality of clocks comprise three clocks of a same clock type.

19. The communication system of any preceding Claim wherein the communication system is one of: a fifth generation, 5G, communication system and the plurality of communication devices are 5G base stations, a sixth generation, 6G, communication system and the plurality of communication devices are 6G base stations.

20. The communication system of any preceding Claim wherein the central processing hub (260) is one of: the plurality of communications devices, a base station.

21. A method (900) for timing synchronisation in a communication system that has a plurality of communication devices, the method comprising, at a communication device: receiving (910) a Global Navigation Satellite System, GNSS, shared clock, receiving (920) timing information from a clock of the communication device; deriving clock information (930) from:measured timing of clock signals obtained from the clock;measured timing of the GNSS shared clock; anda comparison of the measured timing of clock signals obtained from the clock and the GNSS shared clock; andtransmitting (940) the derived clock information to at least one of: a central processing hub (260), others of the plurality of communication devices.

22. The method of Claim 21, further comprising at the central processing hub, aggregating, processing and analysing (950) received clock information from multiple devices of the plurality of communication devices and determine therefrom:(i) at least one emerging pattern related to at least one of: the GNSS shared clock, a clock employed by a transmitting communication device;(ii) correlate anomalies with environmental factors that comprise at least one of: temperature, magnetic field, clock orientation against gravity, barometric pressure;(iii) predict a clock aging trend and potential step change;(iv) a corrective model and transmit a timing adjustment to at least one communication device based thereon.

23. A communication device comprises:a central processing hub;a plurality of communication devices connected to the central processing hub, wherein each of the plurality of communication devices comprise:a receiver arranged to receive a Global Navigation Satellite System, GNSS, shared clock;a plurality of clocks;a transmitter; anda processor, operably coupled to the receiver, transmitter and the plurality of clocks, wherein the processor is arranged to generate derived clock information from: measured timing of clock signals obtained from the plurality of clocks, measured timing of the GNSS shared clock, a comparison of the timing of clock signals obtained from the clock and the GNSS shared clock;wherein the transmitter is arranged to transmit the derived clock information to at least one of: a central processing hub, at least one other communication device.

24. A central processing hub (260) comprising;a receiver configured to receive derived clock information from a plurality of communication devices, wherein the derived clock information comprises a comparison of measured timing of clock signals from multiple clocks in the communication devices and measured timing of a Global Navigation Satellite System, GNSS, shared clock:a learning processor operably coupled to the receiver and having at least one input and at least one output, the at least one input arranged to receive the received derived clock information and at least one of: temperature sensitivity or environmental impacts, wherein the learning processor is configured to predict and output a timing adjustment signal comprising at least one of: clock trends, timing anomalies, drift, jitter, a timing adjustment of a communication device clock; anda transmitter, operably coupled to the learning processor and arranged to transmit the timing adjustment signal to at least one communication device of the plurality of communication devices.

Images