PRECEDER: Prediction of the Electrical Behavior of Electronic Devices under Radiation

ABSTRACT

When cost and time is leading your project requirements, the use of COTS looks very attractive. However, there are many things to consider before deciding if this is the most appropriate solution. When is it wise to pursue a COTS-based system approach? When is it best to hold back? How could we minimize the risk of applying the COTS-based approach in our system?

Statistical process control (SPC) is commonly used in manufacturing to measure how consistently a product performs according to its specifications. By achieving consistent quality and performance, some of the benefits manufacturers can achieve are Reduced scrap, rework, and warranty claims. It is easy to see why larger datasets can be used to generate value. If we have a bigger data set to work with, we’re comfortable with the data quality, and we have defined our approach well, we can probably qualify for a Device Under Test (DUT), without the need for some expensive tests. But how could we achieve a similar goal with a reduced dataset?

This work presents how a new approach, based on Machine Learning, to analyze any available information on the performance of COTS could help to carry out this assessment. It evaluates how valuable information might be extracted from archival data despite the size limitation of the analyzed dataset in a few practical examples of Total Ionizing Dose (TID) and Displacement Damage (DD) tests. The impact of the well-known and dramatic lot-to-lot, or even part-to-part, variability for some technologies or the different test conditions, such as bias and dose rates, is analyzed.

Speaker: Aintzane Lujambio – ALTER TECHNOLOGY / CNA (Centro Nacional de Aceleradores)