Practical Guide to AI working with Instruments for the RF Test Engineer
Speakers: Joel Dunsmore & Jordan Trout, Keysight Technologies
Abstract
AI tools can provide a step-function improvement in productivity for test engineers seeking to understand and automate test processes and programs, particularly during the design and validation stages of a project. This 2-hour short course presents an introductory guide to utilizing AI tools, using Claude Code and the PNA-X as detailed examples, organized into three sections.
The first section (40 minutes) is an AI tool boot camp covering how to get AI access and best practices for setting up your AI workflow, including how to incorporate your own domain knowledge into Claude sessions. We will walk through the practical steps of requesting and configuring AI access within an organization, along with fundamentals of writing effective prompts. A live demonstration will show how to build a domain-knowledge reference, such as notes on SCPI commands and instrument behavior, so that Claude’s responses are grounded in verified information rather than guesswork.
The second section (30 minutes) covers using Claude Chat to understand and optimize a test setup, and how to gain deeper insight into using instruments in your test setups, including a demonstration of Claude Chat and asking it to identify likely sources of measurement error or inefficiency from a block-diagram figure or description. Additional examples will show Claude Chat helping interpret SCPI documentation, walk through calibration procedures, and diagnose unexpected measurement results, illustrating how AI-assisted research can significantly reduce the time spent digging through manuals.
The third section (50 minutes) covers using Claude Code to create Python test scripts that accomplish tasks quickly and efficiently, including how to set up startup rules and use your domain knowledge to enhance your Claude Code experience. We will build a Python automation script live, connecting to a PNA-X over SCPI to run a swept measurement, and show the iterative write-test-debug workflow that Claude Code enables. The session will also cover best practices for organizing project-level rules and reference material, and demonstrate how Claude Code can help diagnose and fix issues when a script fails against real hardware.
We end with introduction of custom tools you can use for deep knowledge of several specific instrument classes.
