This repo is a demonstration of how to run LTspice simulations in a continuous integration pipeline. LTspice is a professional electronics circuit simulator from Analog Devices, which can be downloaded free of charge.

Why do we want to run LTspice simulations in a CI pipeline?

• Run long simulations on a build server
• Automatically generate documentation from simulation results
• Run multiple simulations in parallel

And probably many other good reasons too!

The circuit to be analyzed in this repo is a simple amplifier with a capacitive load. The purpose of this simulation is to determine the phase margin of the amplifier. The phase margin can be analyzed by inserting a small signal generator between the amplifier input and output (feedback):

The phase margin can be measured with a .meas statement, and the results of these measurements are output in the .log file produced by LTspice. We are also interested in the waveforms, which LTspice stores in the .raw file.

LTspice is a Windows program only, but it officially supports Wine (a compatibility layer allowing Windows programs to run on Linux). This means we can run LTspice in a Docker image on a Linux server. One issue with running GUI programs like LTspice in a terminal-only environment is that they tend to throw errors because there is no display available. To the rescue comes xvfb (X virtual framebuffer) which emulates a dumb framebuffer.

The Docker image with LTspice used here is made by Aanas Sayed. Dockerfile link.

LTspice produces two output files we want to process:

• .log-file (text file with measurements)
• .raw-file (binary file with waveforms)

Thankfully there is a Python package called PyLTSpice that can interact with LTspice designs. Using PyLTSpice makes it simple to extract measurements from the log and waveforms from the raw-file. What we want to make is a bode plot like this:

The pipeline consists of two jobs:

• LTspice simulation.
  • The two output files (.log and .raw) are made available as artifacts to the next job.
• Post processing.
  • A Python script utilizing PyLTSpice extracts the phase margin and bode plot data and outputs a markdown report.
  • The tool CML (Continuous Machine Learning) is used to publish the report as a comment to the commit on GitHub. This means you get an email with the report (including plot) too. Take a look at the CML report in one of the latest commits.

That's it, enjoy automation of LTspice simulations!