⚡ electrical engineering

Explicit Ensemble Mean Synchronization for Time Scale Generation with Mixed Atomic Clock Ensembles

This paper proposes an explicit ensemble mean synchronization algorithm based on a divergence-free Kalman filter derived from observable canonical decomposition to generate an optimized time scale from mixed cesium and hydrogen maser atomic clock ensembles, allowing users to tune frequency stability over different intervals by regulating unobservable state dynamics.

Original authors: Priyanka Dey, Takahiro Kawaguchi, Yuichiro Yano, Yuko Hanado, Takayuki Ishizaki

Published 2026-05-26

📖 4 min read☕ Coffee break read

CC BY 4.0

Original authors: Priyanka Dey, Takahiro Kawaguchi, Yuichiro Yano, Yuko Hanado, Takayuki Ishizaki

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine you are trying to keep a perfect time for a massive orchestra. You have two types of musicians:

The "Steady Eddies" (Cesium Clocks): These are like old-school metronomes. They might wobble a little bit every second, but over the course of a day or a week, they are incredibly reliable and don't drift off course.
The "Flashy Freds" (Hydrogen Maser Clocks): These are like sprinters. They are incredibly precise for the first few seconds—they barely wobble at all—but if you let them run for a long time, they start to drift and lose their rhythm.

In the real world, scientists use groups (ensembles) of these clocks to create a single, "perfect" time signal for things like GPS and financial networks. The goal is to mix the "Steady Eddies" and "Flashy Freds" so that the final time signal is stable in the short term and the long term.

The Problem: The "Ghost" in the Machine

The paper explains that when you try to mathematically combine these different clocks using standard methods (called Kalman filtering), the math starts to break down.

Think of it like trying to balance a scale where you can only see the difference between the weights, not the weights themselves. Because you can't see the absolute weight of every clock, the computer's "error calculation" (a safety check) starts to grow infinitely large, like a balloon that keeps inflating until it pops. This makes the computer unstable and the time signal inaccurate.

The Solution: A New Way to Look at the Data

The authors, Priyanka Dey and her team, came up with a clever trick to fix this. Instead of trying to track every single clock individually, they reorganized the math to split the problem into two parts:

The Visible Part: This tracks the differences between the clocks (the part we can actually measure). This part is stable and doesn't explode.
The Invisible Part: This tracks the "average" behavior of the whole group (the part we can't measure directly).

By separating these two, they created a new algorithm that never suffers from that "exploding balloon" problem. It's like realizing you don't need to know the exact weight of every apple in a basket to know if the basket is balanced; you just need to know how the apples relate to each other.

The "Explicit Ensemble Mean" Strategy

Once the math is stable, the team proposes a new way to generate the final time signal, which they call the Explicit Ensemble Mean Synchronization.

Imagine the "Flashy Freds" and "Steady Eddies" are all trying to march in step. The algorithm acts like a conductor who tells every musician: "Don't just march to your own beat; march to the beat of the group average."

But here's the magic: The conductor can decide how much weight to give to each type of musician.

For Short-Term Stability: If you need perfect timing for the next few seconds (like for a high-speed trade), the algorithm tells the "Flashy Freds" to lead the way because they are so precise in the short term.
For Long-Term Stability: If you need to keep time for the next few days, the algorithm tells the "Steady Eddies" to take the lead because they don't drift over time.

The user (the conductor) can choose a "weighting" to decide which clock type dominates the final signal.

The Results: A Better Timekeeper

The team tested this with a simulated group of 10 clocks (7 "Steady Eddies" and 3 "Flashy Freds").

Short-Term Test: When they tuned the system to prioritize short-term precision, the resulting time signal was much smoother and more accurate than any single clock could be on its own.
Long-Term Test: When they tuned it for long-term stability, the signal stayed on track for days without drifting, again beating any single clock.

The Catch

The paper notes one limitation: You can't have your cake and eat it too. You can't optimize for both the short-term and long-term perfectly at the exact same moment with this specific method. You have to choose which one is more important for your current needs.

In summary: The paper provides a new mathematical recipe to mix different types of atomic clocks without the computer crashing. It allows scientists to create a "super-clock" that can be tuned to be either a sprinter (for short-term precision) or a marathon runner (for long-term stability), depending on what the situation requires.

Technical Summary: Explicit Ensemble Mean Synchronization for Time Scale Generation with Mixed Atomic Clock Ensembles

Problem Statement
This paper addresses the challenge of generating a reliable time scale from a mixed ensemble of atomic clocks, specifically combining Cesium-type (Cs-type) and Hydrogen maser-type (Hm-type) clocks. While conventional Kalman filtering (CKF) is widely used for estimating clock deviations, it faces a critical limitation in this context: the mixed ensemble model is undetectable because only clock reading differences are measurable. This undetectability causes the error covariance matrix in the CKF to diverge, leading to numerical instability and inaccurate estimations. Furthermore, existing methods for time scale generation often rely on implicit ensemble means or are restricted to homogeneous ensembles (either only Cs-type or only Hm-type), failing to fully exploit the complementary stability characteristics of mixed ensembles. The authors identify two primary problems: (1) developing a Kalman filtering algorithm for mixed ensembles that is free from diverging terms, and (2) generating a time scale that synchronizes clock deviations to a user-defined weighted average while optimizing frequency stability (measured by Hadamard variance) for either short-term or long-term intervals.

Methodology
The proposed solution involves a two-stage approach combining state-space transformation and explicit control design:

Observable Canonical Decomposition and Transformed Kalman Filter (TKF):
The authors first apply an observable canonical decomposition to the mixed ensemble model. By transforming the state space into observable and unobservable components, they derive a new Kalman filtering algorithm. In this transformed domain, the error covariance matrices associated with the unobservable states (which cause divergence in standard CKF) are decoupled from the gain calculations. The resulting Transformed Kalman Filter (TKF) and its steady-state version utilize only the convergent error covariance matrices of the observable states, thereby eliminating diverging terms and ensuring numerical stability.
Explicit Ensemble Mean Synchronization Algorithm:
Using the state estimates from the TKF, the authors propose an explicit ensemble mean synchronization algorithm. The control input is designed to synchronize the time deviations of individual clocks to an "unobservable weighted average phase state."
- Synchronization: A state-feedback control law is applied to drive the synchronization error to zero, ensuring that the ensemble converges to a common time scale defined by a user-specified weight vector $q$ .
- Optimization of Frequency Stability: To address the trade-off between short-term and long-term stability, the authors analyze the free-running dynamics of the unobservable state. They derive an optimal weight vector $q$ $q$ that minimizes the Hadamard variance (HVAR) of the generated time scale.
  - For short-term stability, the weight vector is derived by minimizing HVAR as the sampling interval $\tau \to 0$ , effectively prioritizing clocks with low white frequency noise (typically Hm-type).
  - For long-term stability, the weight vector is derived by minimizing HVAR as $\tau \to \infty$ , prioritizing clocks with low random walk frequency noise (typically Cs-type) and assigning zero weight to Hm-type clocks due to their frequency drift.

Key Contributions

Stable Filtering for Mixed Ensembles: The paper provides a rigorous derivation of a Kalman filtering algorithm based on observable canonical decomposition that resolves the divergence issue inherent in conventional filters for undetectable mixed clock ensembles.
Explicit Synchronization Control: Unlike previous "implicit" ensemble mean approaches, this work introduces an explicit algorithm where the user can define the weighting vector to synchronize clocks to a specific target state.
Optimized Time Scale Generation: The authors demonstrate how to select weight vectors to optimize the frequency stability of the generated time scale (GTS) for specific time intervals. They show that by choosing appropriate weights, the GTS can achieve superior short-term stability (leveraging Hm-type clocks) or superior long-term stability (leveraging Cs-type clocks) compared to individual clocks or simple averages.

Results
The efficacy of the proposed algorithm is validated through a numerical example involving an ensemble of 10 clocks (7 Cs-type and 3 Hm-type).

Short-Term Optimization: When the weight vector optimized for short-term stability ( $q^0_H$ ) was applied, the simulation showed that the controlled clocks' time deviations closely followed the virtual trajectory of the optimized free-running dynamics. The resulting GTS exhibited significantly reduced divergence from ideal time over a 1,000-second interval compared to free-running clocks.
Long-Term Optimization: When the weight vector optimized for long-term stability ( $q^\infty_H$ ) was applied, the GTS demonstrated excellent frequency stability over a $10^5$ -second interval. The HVAR curves confirmed that the generated time scale maintained stability superior to individual free-running clocks over long durations.
Hadamard Variance: The HVAR analysis confirmed that the proposed method successfully minimizes the variance for the targeted time intervals, validating the theoretical optimization of the weight vectors.

Significance and Claims
The paper claims that the proposed "Explicit Ensemble Mean Synchronization Algorithm" offers a robust solution for time scale generation in mixed atomic clock ensembles. By eliminating the divergence of the error covariance matrix, the method ensures numerical stability where conventional filters fail. Furthermore, by explicitly allowing the user to define the weighting vector, the algorithm provides a flexible mechanism to tailor the generated time scale's frequency stability to specific operational needs—either prioritizing short-term precision or long-term accuracy. The authors note that while the current work optimizes for either short-term or long-term stability separately, future work will focus on strategies to achieve a balanced stability profile for both intervals simultaneously.

The Problem: The "Ghost" in the Machine

The Solution: A New Way to Look at the Data

The "Explicit Ensemble Mean" Strategy

The Results: A Better Timekeeper

The Catch

More like this