Machine learning model training using de-noised data and model prediction with noise correction

Patent metadata

Field	Value
Publication number	US-12416662-B2
Application number	US-202318094947-A
Country	US
Kind code	B2
Filing date	Jan 9, 2023
Priority date	Jan 14, 2022
Publication date	Sep 16, 2025
Grant date	Sep 16, 2025

Abstract

A test and measurement system has one or more inputs connectable to a device under test (DUT), and one or more processors configured to execute code that causes the one or more processors to: gather a set of training waveforms by acquiring one or more waveforms from one or more DUTs or from simulated waveforms, remove noise from the set of training waveforms to produce a set of noiseless training waveforms, and use the set of noiseless training waveforms as a training set to train a neural network to predict a measurement value for a DUT, producing a trained neural network. A method of training a neural network having receiving one or more waveforms from one or more DUTs, or generating one or more waveforms from a waveform simulator, removing noise from a set of training waveforms gathered from the one or more waveforms to produce a set of noiseless training waveforms, and use the set of noiseless training waveforms as a training set to train a neural network to predict a measurement value for a DUT, producing a trained neural network.

First claim

The invention claimed is:

1. A test and measurement system, comprising:
one or more inputs connectable to a device under test (DUT); and
one or more processors configured to execute code that causes the one or more processors to:
gather a set of training waveforms by acquiring one or more waveforms from one or more DUTs or from simulated waveforms;
remove noise from the set of training waveforms to produce a set of noiseless training waveforms; and
use the set of noiseless training waveforms as a training set to train a neural network to predict a numeric measurement value for a DUT, producing a trained neural network.

2. The test and measurement system as claimed in claim 1 , wherein the code that causes the one or more processors to remove noise from the set of training waveforms comprises code that causes the one or more processors to capture a number of original waveforms and average the number of original waveforms to produce a noiseless waveform repetitively until the set of noiseless training waveforms is complete.

3. The test and measurement system as claimed in claim 1 , wherein the one or more processors are further configured to execute code to determine a correction factor from the noise removed from the set of waveforms.

4. The test and measurement system as claimed in claim 1 , wherein the numeric measurement value is a Transmitter Dispersion Eye Closure Quaternary (TDECQ) value.

5. The test and measurement system as claimed in claim 4 , wherein the code that causes the one or more processors to use the set of noiseless training waveforms to train a neural network to predict a TDECQ value comprises code that causes the one or more processors to use the set of noiseless training waveforms to train the neural network to predict tap values for a feed forward equalizer (FFE), and to determine the TDECQ value from the FFE tap values.

6. The test and measurement system as claimed in claim 1 , wherein the one or more processors are further configured to execute code to cause the one or more processors to normalize amplitudes of the set of noiseless waveforms.

7. The test and measurement system as claimed in claim 4 , wherein the one or more processors are further configured to execute code that causes the one or more processors to:
acquire one or more waveforms from a DUT in a production environment; and

apply the trained neural network to produce a predicted TDECQ value for the DUT based upon the one or more waveforms.

8. The test and measurement system as claimed in claim 7 , wherein the one or more processors are further configured to execute code that causes the one or more processors to:
determine a correction factor from the noise removed from the set of training waveforms; and
apply the correction factor to an accuracy level for the predicted TDECQ value.

9. The test and measurement system as claimed in claim 7 , wherein the one or more processors are further configured to execute code that causes the one or more processor to:
gather a set of operational waveforms;
determine a correction factor from noise removed from the set of operational waveforms; and
apply the correction factor to an accuracy for the predicted TDECQ value.

10. The test and measurement system as claimed in claim 7 , wherein the code to cause the one or more processors to apply the trained neural network causes the one or more processors to predict feed forward equalizer (FFE) tap values and determine the TDECQ value from the FFE tap values.

11. A method of training a neural network, comprising:
receiving one or more waveforms from one or more DUTs, or generating one or more waveforms from a waveform simulator;
removing noise from a set of training waveforms gathered from the one or more waveforms to produce a set of noiseless training waveforms; and
using the set of noiseless training waveforms as a training set to train a neural network to predict a numeric measurement value for a DUT, producing a trained neural network.

12. The method as claimed in claim 11 , wherein removing noise comprises capturing a number of original waveforms and averaging the number of original waveforms to produce a noiseless waveform, repetitively until the set of noiseless waveforms is complete.

13. The method as claimed in claim 11 , wherein the one or more processors are further configured to determine a correction factor from the noise removed from the set of waveforms.

14. The method as claimed in claim 11 , wherein the numeric measurement value is a Transmitter Dispersion Eye Closure Quaternary (TDECQ) value.

15. The method as claimed in claim 14 , wherein using the set of noiseless training waveforms to train a neural network to predict a TDECQ value comprises using the set of noiseless training waveforms to train the neural network to predict tap values for a feed forward equalizer (FFE) from which the TDECQ value can be determined.

16. The method as claimed in claim 14 , further comprising:
acquiring one or more waveforms from a DUT in a production environment; and
applying the trained neural network to produce a predicted TDECQ value for the DUT based upon the one or more waveforms.

17. The method as claimed in claim 16 , further comprising:
determining a correction factor from the noise removed from the set of training waveforms; and
applying noise correction to an accuracy level for the predicted TDECQ value to produce a final TDECQ value.

18. The method as claimed in claim 16 , further comprising:
gathering a set of operational waveforms;
determining a correction factor from the noise removed from the set of operational waveforms; and
applying the correction factor to an accuracy level for the predicted TDECQ value to produce a final TDECQ value.

19. The method as claimed in claim 16 , wherein applying the trained neural network comprises predicting feed forward equalizer (FFE) tap values and determining the TDECQ value from the FFE tap values.

Assignees

• Tektronix Inc

Inventors

• Sun Wenzheng
• Zivny Pavel R

Classifications

• G06N3/08 Primary

  Learning methods · CPC title

• G06N3/084

  Backpropagation, e.g. using gradient descent · CPC title

• G06N3/045

  Combinations of networks · CPC title

• G01R31/2603 Primary

  for curve tracing of semiconductor characteristics, e.g. on oscilloscope · CPC title