System and method for built in self test for timing module holdover

• US 8,884,706 B2
• Filed: 03/25/2011
• Issued: 11/11/2014
• Est. Priority Date: 07/06/2009
• Status: Active Grant

• Current Assignee: Malikie Innovations Limited (Key Patent Innovations Limited)
• Original Assignee: Blackberry Limited
• Inventors: Nicholls, Charles, Wu, Philippe
• Primary Examiner(s): Chang, Joseph
• Assistant Examiner(s): SHIN, JEFFREY M
• Application Number:
  US13/072,285
• Publication Number:
  US 20110169577A1
• Time in Patent Office: 1,327 Days
• Field of Search: 331/2, 331/14, 331 16- 18, 331/44, 331/46, 331/47
• US Class Current: 331/17
• CPC Class Codes: H03L 1/026 by using a memory for digitally storing correction values H03L1/025 takes precedence
• CPC Class Codes: H03L 1/028 of generators comprising piezoelectric resonators H03L1/021, H03L1/022 take precedence; oscillation generators with a piezoelectric resonator H03B5/32
• CPC Class Codes: H03L 7/0805 the loop being adapted to provide an additional control signal for use outside the loop
• CPC Class Codes: H03L 7/146 by using digital means for generating the oscillator control signal H03L7/141, H03L7/143 take precedence
• CPC Class Codes: H04J 3/0688 Change of the master or reference, e.g. take-over or failure of the master

Abstract

Embodiments of the invention include a method for use in a device having a local oscillator. The method includes performing, for the local oscillator that is disciplined by an external reference signal, while locked to the external reference signal, training at least two mathematical models of the oscillator to determine a predicted correction signal for each mathematical model based at least in part on a correction signal that is a function of the external reference signal and which is used to discipline drift in the oscillator. The method also includes selecting a mathematical model of the at least two mathematical models that results in a smallest time error when disciplining the oscillator to use when the external reference signal is unavailable and an alternative correction signal is to be used to discipline drift in the oscillator. The method further includes testing the selected mathematical model using a sampled version of the correction signal such that the selected mathematical model can be used without the need for a testing duration that is in addition to a period of time used for the training.

Claims

1. A method for controlling an oscillator, comprising:

receiving a reference signal during a training period;

determining, based on the reference signal, a correction signal suitable for controlling operation of the oscillator by correcting oscillator frequency drift;

sampling the correction signal for the training period;

storing the correction signal samples;

training a mathematical model of the oscillator based on the correction signal not including the correction signal samples; and

testing the trained mathematical model of the oscillator against the correction signal samples 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.

2. The method as defined in claim 1, comprising:

training at least two mathematical models based on the correction signal;

testing the at least two trained mathematical models against the correction signal samples; and

responsive to the testing, selecting one of the at least two trained mathematical models for generating a correction signal for controlling operation of the oscillator when the reference signal is not available.

3. The method as defined in claim 2, comprising:

training the at least two mathematical models based on the correction signal;

testing the at least two trained mathematical models against the correction signal samples by;

using each trained mathematical model to generate a respective modelled correction signal;

determining a difference between each modelled correction signal and the correction signal determined based on the reference signal to determine a respective error signal for each trained mathematical model; and

integrating each error signal over time to derive respective error parameter; and

selecting, based on the respective error parameters, one of the at least two trained mathematical models for generating a correction signal for controlling operation of the oscillator when the reference signal is not available.

4. The method as defined in claim 1, wherein testing the mathematical comprises:

using the trained mathematical model to generate a modelled correction signal;

determining a difference between the modelled correction signal and the correction signal determined based on the reference signal to determine an error signal and

integrating the error signal over time to derive an error parameter.

5. A system for controlling an oscillator, comprising:

a receiver operable to receive a reference signal during a training period;

a correction signal generator operable to determine, based on the reference signal, a correction signal suitable for controlling operation of the oscillator by correcting oscillator frequency drift;

a sampler operable to sample the correction signal;

memory to store the correction signal samples;

a training element operable to train a mathematical model of the oscillator based on the correction signal not including the correction signal samples; and

a comparator operable to test the trained mathematical model of the oscillator against the correction signal samples 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.

6. The system as defined in claim 5, comprising:

at least two training elements, each training element operable to train a respective mathematical model based on the correction signal;

wherein the comparator is operable to test the at least two trained mathematical models against the correction signal samples; and

the system further comprises a selector operable responsive to the testing to select one of the at least two trained mathematical models for generating a correction signal for controlling operation of the oscillator when the reference signal is not available.

7. The system as defined in claim 6, wherein:

the at least two training elements are operable to train the at least two mathematical models based on the correction signal;

the comparator is operable to test the at least two trained mathematical models against the correction signal samples by determining a difference between a respective modelled correction signal generated by each mathematical model and the correction signal determined based on the reference signal to determine a respective error signal for each trained mathematical model; and

the system further comprises;

an integrator operable to integrate each error signal over time to derive respective error parameter; and

a selector operable to select, based on the respective error parameters, one of the at least two trained mathematical models for generating a correction signal for controlling operation of the oscillator when the reference signal is not available.

8. The system as defined in claim 5, wherein:

the trained mathematical model is operable to generate a modelled correction signal;

the comparator is operable to determine a difference between the modelled correction signal and the correction signal determined based on the reference signal to determine an error signal; and

the system further comprises an integrator operable to integrate the error signal over time to derive an error parameter.