Explicit Ensemble Mean Clock Synchronization for Optimal Atomic Time Scale Generation
Takayuki Ishizaki, Takahiro Kawaguchi, Yuichiro Yano, Yuko Hanado · 2025
Read Explicit Ensemble Mean Clock Synchronization for Optimal Atomic Time Scale Generation by Takayuki Ishizaki, Takahiro Kawaguchi, Yuichiro Yano, Yuko Hanado (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2504-15540.txt.
Introduction
Takayuki Ishizaki and co-authors set out a fresh theory for how atomic clocks agree on one shared time. Their 2025 paper joins three tasks in one frame: making a time scale, syncing clocks, and tuning each clock's rate. The math splits a clock group's state into a part that can be seen and a part that cannot.
Within that split, the standard filter used today does more than track the seen part, the authors prove. It also estimates the hidden part, and it counts as a special case of their frame tuned for long-term steadiness. Adding feedback control on both parts then gives the best joint result. The frame aims to make timing systems easier to explain and trust.
Catalog record
| CREATOR | Takayuki Ishizaki, Takahiro Kawaguchi, Yuichiro Yano, Yuko Hanado |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 86 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2504-15540
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Explicit Ensemble Mean Clock Synchronization for Optimal Atomic Time Scale Generation Takayuki Ishizaki, Takahiro Kawaguchi, Yuichiro Yano, Yuko Hanado Abstract This paper presents a novel theoretical framework, called explicit ensemble mean (EEM) synchronization. This framework unifies time scale generation, clock synchronization, and oscillator frequency regulation within the systems and control theory paradigm. By exploiting the observable canonical decomposition of a standard atomic ensemble clock model, the system is decomposed into two complementary components: the observable part, which represents the synchronization error, and the unobservable part, which captures the synchronization destination. Within this structure, we mathematically prove that standard Kalman filtering, which is widely used in current time scale generation, not only performs observable state estimation, but also significant unobservable state estimation, and it can be interpreted as a special case of the proposed framework that optimizes long-term frequency stability in terms of the Allan variance. Furthermore, applying state feedback control based on Kalman filtering to each component achieves optimal time scale generation, clock synchronization, and oscillator frequency regulation in a unified manner. The proposed framework provides a foundation for developing explainable timing systems. I Introduction I-A Background Time scale generation from an ensemble of atomic clocks has long been a cornerstone technique in time and frequency metrology, particularly in the construction of national and international time references such as Coordinated Universal Time (UTC) [ 1 , 2 , 3 ] . In addition, this approach is critical to realizing resilient timing infrastructures that can operate robustly even without reference time derived from Global Navigation Satellite System (GNSS) signals. Examples include composite clocks and the continuously operating primary reference time clock, the Coherent Network Primary Reference Time Clock (cn-PRTC) [ 4 , 5 ] . In practice, a time scale, or standard time reference, called the atomic time (TA), is generated and maintained by physically regulating the frequency of local oscillators and digitally adjusting clock readings based on the ensemble time generated from multiple pairwise clock phase difference measurements [ 6 ] . While a number of studies have examined methods for computing TA as a reference, few have addressed integrated system design, including the regulation and steering of physical oscillators. In particular, a unified theoretical framework based on systems and control theory that systematically addresses time scale generation, clock synchronization, and oscillator frequency regulation yet to be established. Furthermore, little attention has been given to mathematically proving the working principles and convergence properties of such an integrated system. These issues are critical to the stability of reliable timing infrastructures in the short and long term. Kalman filtering is widely used in this context, such as in the GPS composite clock [ 7 ] , due to its ability to provide optimal state estimation under Gaussian noise assumptions [ 8 ] . The Kalman filtering has been used to generate a time scale by implicitly computing the ensemble mean weight of the clock states, known as the “implicit” ensemble mean. However, the direct application of Kalman filtering has certain limitations. In particular, the error covariance matrix diverges over time, even though the Kalman gain converges to a constant value. This divergence implies the presence of indeterminate operations in the algorithm, which can lead to numerical instability. Several error covariance reduction schemes have been proposed to address this issue [ 7 , 9 ] , but it has been observed that optimal Kalman filtering often results in suboptimal time scales in terms of short-term frequency stability. An ad hoc method involving the modification of the Kalman gain has been introduced to mitigate this problem [ 10 ] , called the Kalman Plus Weights (KPW) algorithm, though the underlying mechanism has not been theoretically elucidated. In addition to approaches that implicitly determine the ensemble mean weight of atomic clocks through Kalman filtering, there are also approaches that use Kalman filtering solely to estimate the clock state or noise. These approaches assign the ensemble mean weight according to a different policy. One example is the KAS-2 algorithm [ 11 , 12 ] . This method avoids error covariance divergence by applying the Kalman filtering only to estimate relative or differential clock states. The ensemble mean weight of the clocks is indirectly determined by the technical assumption that the weighted average of the stochastic clock noise is always zero. Another example is the AT1 algorithm and its extended version, the AT2 algorithm [ 13 , 14 ] . These methods update the ensemble mean weights based on the magnitude of the state estimation error during operation. The weights are known to be updated in a manner that improves short-term stability. Kalman filtering is used for clock frequency estimation in the AT2 algorithm. Note that these time scale generation algorithms use Kalman filtering to digitally adjust clock readings rather than physically regulating the frequency of local oscillators. In other words, they perform so-called paper clock calculations. I-B Contributions In this paper, we present a unified and rigorous framework called “explicit” ensemble mean (EEM) synchronization. This framework integrates time scale generation, clock synchronization, and oscillator frequency regulation within a coherent systems and control theory paradigm. This framework provides an explicit understanding of how the Kalman filtering for the GPS composite clock implicitly computes the ensemble mean weight of clocks. In particular, we prove that the Kalman filtering is significant not for estimating the observable relative clock state, but rather for estimating the “unobservable” weighted average state. From a systems and control theory perspective, it is counterintuitive that the Kalman filtering is significant for estimating unobservable states. The literature lacks such a clear explanation of the working principles of implicit ensemble mean for time scale generation. The main contributions are summarized as follows. • We derive an equivalent formulation of the Kalman filtering algorithm that completely eliminates indeterminate operations leading to the error covariance divergence by applying a specially tailored observable canonical decomposition to a standard atomic ensemble clock model. • We construct a physical oscillator frequency regulation algorithm that synchronizes all clocks in the ensemble to a free-running ensemble mean dynamics, the weighting of which can be explicitly specified by the user. • We mathematically prove that the standard Kalman filtering is a special case of the proposed EEM synchronization, which achieves the least Allan variance only in the long term. • We design a novel synchronization algorithm that explicitly balances both short-term and long-term frequency stability of the generated time scale by incorporating intermittent feedback control of the synchronization destination. These contributions provide a rigorous foundation for resilient timing infrastructure and open new avenues for atomic time scale generation, clock synchronization, and oscillator frequency regulation via advanced systems and control theory. I-C Mathematical Preliminaries I-C 1 Symbols Throughout the paper, we use the following symbols. \begin{array}[]{ll}\mathbb{R}&\mbox{the set of real numbers}\\ \mathbb{Z}&\mbox{the set of integers }\\ \mathds{1}_{N}&\mbox{the $N$-dimensional all-ones vector}\\ I_{N}&\mbox{the $N$-dimensional identity matrix}\\ A^{\sf T}&\mbox{the transpose of a matrix $A$}\\ \end{array} \begin{array}[]{ll}\sfim A&\mbox{the image of a matrix $A$}\\ \sfker A&\mbox{the kernel of a matrix $A$}\\ \boldsymbol{\Lambda}(A)&\mbox{the set of the eigenvalues of a square matrix $A$}\\ \mathbb{E}[X]&\mbox{the mean of a random variable $X$}\end{array} Other symbols will be explained when used. We use the equality “ := ” in the sense that a new symbol on the left