Presentation
DVD Diagnostics - Debugging Dynamic IR problems using Advanced Analytics
TimeMonday, July 11th5pm - 6pm PDT
LocationLevel 2 Exhibit Hall
Event Type
Engineering Track Poster
Engineering Tracks
AI
Back-End Design
Cloud
Embedded Systems
Front-End Design
IP
DescriptionThe traditional approach to fixing dynamic voltage drop (DvD) issues in a design has been to find out the instances with a voltage drop higher than a certain ratio of the nominal supply voltage level. Whereas the power delivery network (PDN) analysis tools do a good job in reporting the violating instances, it has not been straight-forward to determine the root-causes leading to these violations. This paper presents a new methodology, which helps identifying the root-causes of the DvD hot-spots and hence enables the designers to focus on a reduced number of issues. It introduces a metric which performs DvD Diagnostics using big-data analytics and Python frame-work. It is shown that an average reduction of 4X in the number of violations to fix can be achieved, which shortens the ECO time significantly.