There are two honest ways to get hardware into your CI. You can build a rig yourself, or you can buy a BenchPod. We've done both — the BenchPod exists because we built our own first — so this is a fair comparison, not a sales pitch.
The DIY path is real, and worth taking
If you want to understand HIL deeply, build a rig. It's not exotic. The progression looks like this, and we wrote it up step by step:
- Power control first — a relay board or a protected supply to put the target in a known state.
- Booting from a known image so each run doesn't inherit the last one's state.
- Console access — a UART hub and stable device names.
- Signal routing and peripheral simulation — usually a small MCU between the Pi and the DUT for deterministic bus timing.
- Orchestrating it in CI — runners, flashing on every push, reporting.
For a single board, a fixed setup, or a team that wants full control of every wire, a DIY rig is the right call. You'll learn an enormous amount, and you'll own every part of it.
Where DIY gets expensive
The cost of rolling your own isn't the parts — it's the long tail:
- Each capability is a separate build. Power, flashing, UART, a logic analyzer, sensor emulation, and an analog source are five or six sub-projects, each with its own firmware and quirks. The BenchPod folds them into one board.
- Determinism is hard. A Raspberry Pi can't give you cycle-accurate bus timing; you end up adding an MCU (and its firmware) anyway. The pod puts that work in an FPGA from the start.
- Remote and multi-board access is its own project. Reaching a rig behind NAT, naming devices, and handing CI secret-free access is real infrastructure. The pod dials out to the cloud and a GitHub Action drives it by name.
- Scaling means rebuilding. Covering a test matrix usually means re-cabling one rig. Because pods are inexpensive, you give each board its own and stop rewiring.
- You maintain it forever. Every fix, every new sensor, every flaky cable is yours.
How to choose
A rough rule:
- Build your own if HIL itself is something you want to learn or customize down to the wire, you have one or two fixed targets, and maintaining the rig is an acceptable cost.
- Buy a pod if you want power, flashing, UART, logic analysis, sensor emulation, and a 16-bit analog front-end working together on day one, reachable from CI without secrets, and cheap enough to deploy one per board.
The two aren't enemies. The DIY guides above are the best way to understand what a HIL rig actually does — and that understanding is exactly what makes a BenchPod obvious once you'd rather spend your time on firmware than on the bench. When you're ready, your first test takes about 15 minutes.