Method for training a neural network, method for automatically characterizing a measurement signal, measurement apparatus and method for analyzing

Patent metadata

Field	Value
Publication number	US-11640523-B2
Application number	US-201916730279-A
Country	US
Kind code	B2
Filing date	Dec 30, 2019
Priority date	Jan 2, 2019
Publication date	May 2, 2023
Grant date	May 2, 2023

Abstract

The present invention relates to a method for training a signal characterization neural network. The method comprises the steps of: providing a measurement signal having at least one distortion; assigning at least one predefined signal integrity identifier to a corresponding distortion within the measurement signal; generating at least one input training vector based on the provided measurement signal and the corresponding assigned signal integrity identifier; and applying the generated input training vector on input terminals of the signal characterization neural network for training the signal characterization neural network. The present invention also relates to a method for automatically characterizing a measurement signal. The present invention further relates to a measurement apparatus and a corresponding method for analyzing a waveform signal.

First claim

What I claim is:

1. A method for training a signal characterization neural network, the method carried out by a processor of a measurement apparatus and comprising:
providing a measurement signal having at least one distortion;
assigning at least one predefined signal integrity identifier to a corresponding distortion within the measurement signal;
generating at least one input training vector based on the provided measurement signal and the corresponding assigned signal integrity identifier; and
applying the generated input training vector on input terminals of the signal characterization neural network for training the signal characterization neural network.

2. The method of claim 1 , wherein an assigned signal integrity identifier refers to at least one of the following distortions within a measurement signal:
a runt,
a glitch,
a duty cycle distortion,
a slew rate,
a crosstalk,
intersymbol interferences,
reflections,
ripples,
jitters,
noise.

3. The method of claim 1 , wherein the measurement signal is provided by a device under test.

4. The method of claim 1 , wherein the measurement signal is provided by a signal generator.

5. The method of claim 1 , wherein the measurement signal is provided by an oscilloscope or a spectrum analyzer.

6. The method of claim 1 , wherein the measurement signal is provided by a software application.

7. A method for training a signal characterization neural network, the method carried out by a processor of a measurement apparatus and comprising:
providing a measurement signal having at least one distortion;
assigning at least one predefined signal integrity identifier to a corresponding distortion within the measurement signal;
generating at least one input training vector based on the provided measurement signal and the corresponding assigned signal integrity identifier; and
applying the generated input training vector on input terminals of the signal characterization neural network for training the signal characterization neural network; and further comprising:
generating a histogram representation of the measurement signal, wherein the generated histogram representation is used for generating the input training vector.

8. The method of claim 7 , wherein generating a histogram representation further comprises:
applying at least one predefined threshold to the provided measurement signal;
iteratively slicing the provided measurement signal based on the predefined threshold;
overlaying the iteratively sliced measurement signal to form the histogram representation of the measurement signal.

9. The method of claim 7 , wherein the histogram representation is provided in the form of a pulse-width histogram.

10. The method of claim 7 , wherein the histogram representation is provided in the form of a time histogram.

11. The method of claim 7 , wherein the histogram representation is provided in the form of an eye pattern.

12. A method for automatically characterizing a measurement signal, the method carried out by a processor of a measurement apparatus and comprising:
providing a signal characterization neural network;
providing a measurement signal having at least one characteristic property;
generating a histogram representation of the provided measurement signal;
generating an input vector based on the generated histogram representation;
applying the generated input vector on input terminals of the signal characterization neural network for automatically identifying the at least one characteristic property within the provided measurement signal.

13. The method of claim 12 , wherein the characteristic property refers to at least one of the following distortions within the measurement signal:
a runt,
a glitch,
a duty cycle distortion,
a slew rate,
a crosstalk,
intersymbol interferences,
reflections,
ripples,
jitters,
noise.

14. The method of claim 12 , wherein the measurement signal is provided by at least one of:
a device under test;
a signal generator;
an oscilloscope;
a spectrum analyzer;
software application.

15. A method for automatically characterizing a measurement signal, the method carried out by a processor of a measurement apparatus and comprising:
providing a signal characterization neural network;
providing a measurement signal having at least one characteristic property;
generating a histogram representation of the provided measurement signal;
generating an input vector based on the generated histogram representation;
applying the generated input vector on input terminals of the signal characterization neural network for automatically identifying the at least one characteristic property within the provided measurement signal;
wherein the generated histogram representation is provided in the form of at least one of:
a pulse-width histogram;
a time histogram;
an eye pattern.

16. A measurement apparatus for analyzing a waveform of a signal, the apparatus comprising:
an acquisition device which is configured to acquire a waveform of a signal;
a segmenting device which is configured to identify a number of sections in the acquired waveform;
a processing device which is configured to assign a signal integrity identifier to each section of the number of sections in the acquired waveform; and
an output device which is configured to output, for each section of the number of sections, a representation of the respective assigned signal integrity identifier.

17. The measurement apparatus of claim 16 , wherein the representation of the signal integrity identifier comprises at least one of:
a graphical representation,
an alphanumeric element,
a color,
an audio output.

18. The measurement apparatus of claim 16 , wherein the output device is configured to display a representation of the waveform of the signal, and to output the representation of the signal integrity identifier in association with the displayed representation of the waveform of the signal.

19. The measurement apparatus of claim 18 , wherein the output device is configured to automatically zoom the displayed representation of the waveform of the signal relating to a predetermined signal integrity identifier.

20. The measurement apparatus of claim 16 , further comprising a selection device, wherein the selection device is configured to automatically select a section of the output representation of the respective assigned signal integrity identifier which relates to a predetermined signal integrity identifier.

21. The measurement apparatus of claim 16 , further comprising an analyzing device, wherein the analyzing device is configured to compute histogram data of a section of the acquired waveform of the signal.

22. The measurement apparatus of claim 21 , wherein the processing device is configured to compare the computed histogram data with a number of prestored reference data and to assign the respective signal integrity identifier based on a result of the comparison.

23. The measurement apparatus of claim 21 , wherein the analyzing device is configured to divide the waveform of the signal into a number of slices based on at least one threshold value, and to generate the histogram data based on the sliced waveform.

24. The measurement apparatus of claim 21 , wherein the computed histogram data comprises a pulse-width histogram.

25. The measurement apparatus of claim 21 , wherein the segmenting device is configured to identify

Assignees

• Rohde & Schwarz

Inventors

• Schaefer Andrew

Classifications

• G06N3/09

  Supervised learning · CPC title

• G06N3/0499

  Feedforward networks · CPC title

• G06N20/10

  using kernel methods, e.g. support vector machines [SVM] · CPC title

• G01R13/02

  for displaying measured electric variables in digital form · CPC title

• G01R13/029 Primary

  Software therefor · CPC title