We live in a tech-heavy environment, with electronic devices found almost anywhere. Unfortunately, electronics placed nearby do not always get along. Radiation emitted from one device can disrupt the function of another in a phenomenon known as electromagnetic interference (EMI), which affects performance efficiency.
To this end, a collaborative team led by Richard Gao Xianke, a Senior Scientist at A*STAR’s Institute of High Performance Computing (IHPC), has been developing advanced methods for detecting, identifying and diagnosing EMI sources.
Traditionally, researchers rely on mathematical models of near-field radiation to gauge energy emanating at close range from a potential EMI source. These models enable the analysis of dipoles—pairs of opposite charges across which radiation flows—and making predictions based on the number, location and intensity of dipoles in a device.
However, these conventional methods can be impractical for analyzing EMI patterns as they can only be applied to a single output frequency. For example, global optimization algorithms can only predict a unique location for individual dipoles at a specific frequency. This approach fails due to the significant errors often incurred while modeling multiple dipole locations across a range of frequencies.
Seeking to overcome this challenge, Gao and colleagues explored novel ways of modeling EMI sources at a wider range of frequencies. The researchers selected three frequencies for scanning potential EMI sources and measured the magnitudes only of near-field radiation emitted at every frequency. To minimize the risk of errors while using the global optimization algorithm, they defined a spatial range within which dipoles should be located based on the size of the scanned region.
This data was then fed into a new computational framework capable of predicting the presence of dipoles at the same location at all three frequencies. Measuring dipole moments, however, still proved to be difficult, as the intensity of radiation would vary depending on the frequency.
“To prevent this issue, we used a method called interpolation, which uses dipole moments modeled at predefined frequencies to predict values for the unmeasured ones,” Gao explained.
Finally, the team validated their new modeling method in a real-world setting using devices with integrated circuits. They showed that their framework accurately reconstructed radiation patterns, even at new frequencies within the range of the predefined ones.
Ultimately, the study demonstrates how computational methods could leverage gathered data by measuring EMI at predefined frequencies to predict a source’s near-field radiation pattern across a wider band of interest. “Our method shows promise in being reliable and efficient for quickly diagnosing unknown EMI sources in a complicated environment,” Gao concluded.
The A*STAR-affiliated researchers contributing to this research are from the Institute of High Performance Computing (IHPC).
