Local Oscillator Holdover Using Trained Timing Correction Models
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Network nodes with local oscillators disciplined by external timing references face performance impairment when the external reference is unavailable, requiring costly and time-consuming human intervention to maintain timing accuracy, especially in stringent standards like WiMAX and 4G LTE systems.
Innovation Solution
A method and system that train mathematical models of the local oscillator using samples of the external reference signal to predict correction signals, allowing for autonomous operation by selecting the model with the smallest time error to discipline oscillator drift, thereby maintaining timing accuracy without external references.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If an external timing reference source is used to discipline the local oscillator, then timing accuracy is improved, but system reliability deteriorates when the external reference becomes unavailable
Solution Approach 1:
The system performs preliminary training of multiple oscillator models during normal operation when the external reference is available. Multiple mathematical models (e.g., polynomial models of different orders) are trained and stored in advance, so that when the reference becomes unavailable, the system can immediately switch to using one of the pre-trained models without requiring real-time computation or human intervention.
Solution Approach 2:
The system changes the operational parameter from relying on external reference signals to using internally stored mathematical models. By representing oscillator behavior through multiple mathematical models with different complexity levels, the system can maintain timing accuracy through parameter-based prediction rather than external signal dependency.
2Measurement precision
If human intervention is used to correct timing when external reference is unavailable, then timing accuracy is restored, but operational efficiency deteriorates due to time-consuming manual reset
Solution Approach 1:
The system implements self-service by automatically detecting when the external reference becomes unavailable and autonomously switching to use one of the pre-trained mathematical models. The system monitors the availability of the external reference and automatically selects the appropriate operation mode (external reference discipline or model-based holdover) without requiring human intervention, thereby maintaining both timing accuracy and operational efficiency.
Solution Approach 2:
The system incorporates feedback mechanisms to monitor the status of the external reference source and automatically adjust its operation. By continuously checking reference availability and switching between operation modes based on this feedback, the system maintains timing accuracy while eliminating the need for manual intervention.
3Measurement precision
If multiple mathematical models are trained and tested, then prediction accuracy is improved, but system complexity increases
Solution Approach 1:
The system segments the oscillator modeling function into multiple independent mathematical models of varying complexity (e.g., first-order polynomial, second-order polynomial, higher-order models). Each model is trained separately on historical data and stored independently, allowing the system to evaluate and select the most appropriate model without requiring complex integrated processing.
Solution Approach 2:
The system trains multiple models with varying degrees of complexity, using more models than strictly necessary. This excessive action ensures that among the trained models, there will be suitable candidates for different operating conditions and holdover durations, with the selection criterion being the model that produces the smallest predicted time error for the specific scenario.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
Embodiments of the invention include a method for use in a device having a local oscillator. The method includes - for the local oscillator that is disciplined by an external reference signal while locked to the external reference signal - training a mathematical model of the oscillator to determine a predicted correction signal based at least in part on a first set of samples of a correction signal that is a function of the external reference signal and which is used to discipline drift in the oscillator. The method further includes testing the mathematical model using a second set of samples of the correction signal to assess suitability of the trained mathematical model for generating a correction signal for controlling operation of the oscillator when the reference signal is not available.