A 1977 industrial pottery kiln, rebuilt into a smart, dual-core controller that never crashes.
A kiln is a box that holds over 1000°C for hours, unattended, in a workshop. Pico Kiln is my project to control one safely: a 1977 “Bretagne” pottery kiln, originally wired for 380V 3-phase factory power, that I converted into a 220V single-phase smart kiln driven by a $6 microcontroller.
The system rewires and controls a real, physically dangerous piece of equipment: three 1500W heating elements, solid-state relays, and a firing cycle that has to hold precise temperatures (cone 6, ~1222°C) for hours without supervision. On top of the hardware, it runs a physics-based PID control loop with gain scheduling to counter the kiln’s non-linear radiative heat loss, a multi-mode auto-tuner, crash-recovery that resumes an interrupted firing from its own log, and a cross-platform app (web, macOS, Android) built once in React via Tauri.
This isn’t a toy IoT project. Bugs in this codebase can mean a fire left unattended runs away, or a $600 kiln element burns out because a garbage collector paused the control loop at the wrong second. Embedded work with real physical stakes forces a different discipline than typical app development: every abstraction has to earn its place next to a 9kW heating element.
Sole creator, end to end: electrical rewiring of the 400kg unit, firmware, control theory, the app, and later the full rewrite.
The series runs five parts. Part 1 covers the physical rewiring: taming 3-phase industrial power down to a safe 220V single-phase circuit, sizing breakers, and solving ground-fault issues. Part 2 builds the firmware in MicroPython on the Pico’s dual-core RP2040, splitting real-time control from best-effort WiFi/HTTP handling. Part 3 ships a single React codebase as a web app, native macOS app, and Android APK via Tauri. Part 4 replaces eyeballed temperature curves with a physics-based thermal model and proper PID tuning from CSV firing logs.
Part 5 is the one I’m proudest of: after months of real firings validated the design in MicroPython, I tore out the interpreter and rewrote the firmware from scratch in bare-metal Rust on the RP2350. The rewrite is organized around one rule enforced at compile time, not by convention: the control brain never touches the world. Pure, host-testable logic; hardware access behind embedded-hal traits; golden-replay tests against real firing logs; and a linker-enforced safety boundary between the control core and everything else. It hasn’t crashed, frozen, or hiccuped since. I sleep through overnight firings now.
Replacing the MicroPython interpreter with a bare-metal, dual-core system that never crashes.
Migrating the pico-kiln firmware from MicroPython to bare-metal Rust on the RP2350: a compile-time safety boundary, golden-replay tests, hardware stack guards, and plug-and-play provisioning.
From CSV Logs to Optimal Control Theory
Analyzing kiln performance with physics-based phase detection, thermal modeling, and multi-method PID tuning. Because eyeballing temperature curves doesn't cut it.
One React App, Three Platforms, Zero Mixed Content Drama
Building a kiln controller UI that runs as a web app, macOS app, and Android APK: all from a single React codebase. Solving the HTTP/HTTPS mixed content nightmare without compromise.
Industrial Control on a $6 Chip
Deep dive into the dual-core control logic, PID implementation with Adaptive Gain Scheduling, and thread-safe communication on the Raspberry Pi Pico W.
The electrical rewiring journey: Taming a 400kg vintage beast, converting 3-phase power, and solving critical ground faults.
A deep dive into rewiring a 1977 'Bretagne' pottery kiln. We cover converting 3-phase to single-phase, handling massive loads, and solving dangerous electrical noise issues.
Curious about the technical details, or have a similar challenge? Reach out.
Book a call