Machine learning boosts power waveform analytics in the cloud

Machine learning boosts power waveform analytics in the cloud

Technology News |
By Nick Flaherty

Keysight Technologies has used machine learning algrithms to detect anomalies in voltage and current signals and reduce data storage costs during pre-silicon validation. 

The patented algorithms are used in PathWave Waveform Analytics, an edge-to-cloud computing application with a a new data compression technology that enables long-duration waveform compression, high resolution playback and analysis exceeding several terabytes of data. The machine learning improves the discovery of voltage and current anomalies, as well as transient trends captured by the waveforms.

“Highly power-efficient semiconductors require robust, reliable and secure analytics during design qualification,” said Christopher Cain, vice president and general manager of Keysight’s Electronic Industrial Products. “Keysight’s innovative big-data waveform analytics solutions enable those semiconductor designers to automate design analysis, improving productivity of those tasks by up to 90 percent, thus accelerating their companies’ time-to-market opportunity.”

Keysight sees this being used in automotive, IoT and mobile devices markets that are growing rapidly and need speedy development of products that are robust, reliable and secure against malicious intrusions while reducing power consumption. During validation and characterization process, R&D labs are storing huge sets of continuous long duration waveforms from their multiple runs of their tests for monitoring or details analytics. The total sizes of the waveform data from all their multiple runs, for just one IC validation, is huge (in the 100’s of gigabit range), incurring huge cost of storage space and maintenance. Manual and tedious human eyeballing is needed to perform multi-correlation of multi-channel waveforms to discover anomalies such as spikes and time lags. 

The abilty to detect anomolies quickly and easily in gigabytes of data shortens analysis time in pre-silicon validation and reduces overall project costs by debugging in pre-silicon, rather than in the more expsive post-silicon validation phase. The waveform analytics also identifies outlier waveform shapes via high level view into clustering results, analyzes data in high resolution with hierarchical clustering for multi-level drill down and allow engineers to query and analyze any portion of the big vector data or waveform that was captured and stored. An unlimited number of channels on a single dashboard, with options to pin or move the waveform in the interested channel for comparison.

The tool itself if power eficient as the 60 percent reduction in data from the compression allows more to be stored locally for a faster response and less data to be sent to the cloud, while still making full use of the cloud-based machine learning algorithms.

If you enjoyed this article, you will like the following ones: don't miss them by subscribing to :    eeNews on Google News


Linked Articles