Behind the Scenes of Semiconductor Testing and Data Visualization

Semiconductor testing is a critical part of making sure the chips in everything from smartphones to cars work reliably. Every step of production needs precise testing to keep performance high and failures low. Data visualization plays a huge role in this process, helping engineers catch issues early and optimize production.

Our latest Plotly Hangout featured Mike Purtell, a principal test engineer with extensive experience in test program development, board design, and platforms like Advantest and Teradyne. He specializes in production verification testing (PVT) characterization and automated test equipment (ATE) yield correlation.

Mike is a Python enthusiast and a fan of Plotly to craft advanced data analysis and visualizations. He is also a key contributor to our Figure Friday series on the Plotly Forum.

Q&A with Mike Purtell, principal test engineer

In this Plotly Hangout session with Adam Schroeder, Mike answered questions about his career path in semiconductor testing and his transition to using Python, Polars, and Plotly for data analysis. Read through some of his responses, and make sure to watch the full recording for more information.

How did you get started in semiconductor testing?

I’ve been in the semiconductor industry ever since I moved to Silicon Valley after graduating from Stony Brook University in New York. My background is in electrical engineering, and I’ve always been a hardware guy, tinkering with meters and oscilloscopes. In the 1980s, when I started, I worked with companies heavily involved in semiconductor development and production testing, and I’ve been in the field ever since.

How has your work balanced between hardware and software?

It’s really been about 50/50. When I talk about semiconductor testing, I primarily mean production testing, where we might be producing tens or hundreds of thousands of units in a short period. The challenge is to write programs for sophisticated test equipment that can evaluate complex circuit designs in just a few seconds. It requires both programming and working with physical measurement tools to ensure accuracy.

What tools and programming languages have you used throughout your career?

When I started, I was using LTX test equipment and programming in an old version of BASIC. It was primitive by today’s standards, but at the time, it was cutting-edge. Over the years, I’ve worked with languages like Pascal, C, C++, and now primarily Java for test equipment. For data analysis, I eventually moved to Python, which has become my favorite language due to its flexibility and powerful libraries.

When did you start using Python, and what led you to Plotly?

I started using Python about ten years ago, primarily to control electronic measurement equipment using a library from National Instruments. Then, in 2016, I began learning Pandas to improve my data analysis capabilities. That’s when I discovered Plotly, where I could create interactive, dynamic visualizations instead of sending colleagues static PDFs filled with charts. This made it much easier to communicate insights from test data.

What led you to start using Polars instead of Pandas?

About a year ago, I started using Polars because I was dealing with massive datasets, sometimes with hundreds of thousands of measured devices. Processing these with Pandas was taking several minutes, and Polars significantly reduced that time. I initially came for the speed but stayed for the syntax — it’s more intuitive for me. While Pandas is more mature and has a wider range of methods, I prefer Polars for most of my data analysis tasks now.

What is Figure Friday, and why do you participate?

Figure Friday is a weekly initiative on the Plotly Forum where a dataset is released, and the community creates visualizations based on it. I started participating because, in my field, semiconductor companies keep their data proprietary. Figure Friday provided an opportunity to practice and share my visualization skills using public datasets. It has helped me improve my skills, explore different types of visualizations, and learn from the community.

How do you use Plotly in semiconductor testing?

A lot of my work involves analyzing test data and looking for patterns in performance metrics. For example, I create wafer maps to visualize semiconductor device failures across a wafer. Plotly allows me to generate interactive heat maps, histograms, and correlation plots that make it easier to analyze data and identify issues. One of the biggest advantages is being able to export visualizations as HTML files, making it simple to share interactive reports with colleagues who may not use Python.

What advice do you have for those looking to improve their data visualization skills?

Master the hover tool in Plotly. Interactive tooltips can make your visualizations much more informative and user-friendly.
Create self-contained visualizations. Export your figures as HTML so colleagues can explore the data without needing to install Python or Plotly.
Focus on storytelling with data. Use clear, effective visualizations that emphasize key insights. Avoid clutter and highlight what matters most.
Explore data outside your field. Working with different datasets broadens your skills and can introduce techniques that apply back to your primary work.