Reflow Project Kicad

Reflow Oven Controller

A toaster oven converted into a PCB reflow oven, driven by a custom STM32F103 board with a K-type thermocouple, a solid state relay and a 7" Nextion touch panel.

The firmware runs a PI loop on the MCU. A host can steer it either from a small web app over SWD (working today) or from the Nextion panel (in progress).

System

flowchart LR
    MAINS["120 V AC mains"] --> PSU["PSU<br/>120 VAC to 24 VDC"]
    PSU -->|24 V| SSR["SSR-25DD<br/>solid state relay<br/>3-32 V in, 25 A"]
    SSR -->|switched 24 V| HEAT["Heating element<br/>3.6 ohm, 160 W"]
    HEAT -.->|heat| TC["K-type thermocouple"]
    TC --> MAX["MAX6675<br/>cold-junction comp<br/>+ 12-bit ADC"]
    MAX -->|SPI| MCU["STM32F103C8T6<br/>custom PCB"]
    MCU -->|PA8 PWM<br/>TIM1_CH1| SSR
    MCU <-->|USART1<br/>PA9/PA10| NEX["Nextion NX8048P070-011C<br/>7in 800x480 capacitive"]
    MCU <-->|SWD| HOST["Host PC<br/>OpenOCD + web app"]
    PSU5["5 V supply, 1 A"] --> NEX

Bill of the major pieces

Pin map

Two board quirks worth knowing before wiring anything:

• J4 (the 5 V serial header) does not work. Its BSS138 level shifters are rotated one pin position, so a signal arriving there reaches only a MOSFET gate - there is no DC path to the MCU. Go directly to PA9 / PA10 (TP6 / TP5). No level shifting is needed in either direction.
• SW1 selects BOOT0. In the +3.3 V position the part boots from SRAM and never runs your firmware. It belongs in the GND position.

Repository layout

Hardware/
  KiCad/            schematic, PCB, gerbers, BOM, 3D step
Software/
  webapp/           browser UI + Python bridge (bridge.py, index.html)
  run_webapp.ps1    build, flash, serve the web app
  run_profile.ps1   open-loop step tests, PI profiles, hardware probes
  run_monitor.ps1   plain thermocouple monitor
  Display/          Nextion HMI designs
  Embedded/
    Thermo_Monitor/ the live firmware (all current work)
    Reflow_Oven_STM32_SOURCE_V1_1/  original 2021 firmware, Nextion HMI
    Reflow_Oven_STM32_Source/       earlier revision
skill.md            project goal, architecture and the pitfalls found so far

Quick start

Everything is driven from Software/, with the ST-Link connected. The STM32CubeIDE toolchain is used in place, so nothing extra needs installing; make comes from Git Bash.

# live temperature only
.\run_monitor.ps1 -Seconds 30

# browser UI: setpoint + Kp/Ki in, temperature and duty out
.\run_webapp.ps1 -Kp 0.2 -Ki 0.002 -MaxDuty 1.5
# then open http://127.0.0.1:8770/

# open-loop step test, for characterising the plant
.\run_profile.ps1 -Mode step -PwmHz 1 -StepDuty 10 -BaseSeconds 180 -Seconds 500 -Csv step.csv

The firmware builds in several modes, selected at compile time:

Status

Working: thermocouple readings, SSR drive, PWM on TIM1_CH1, PI loop on the MCU, web app with live charts and a dead-man switch, all the safety cutoffs.

Working on the panel: wire it straight to PA9 / PA10 - not J4, whose level shifters are miswired. MODE=nextion is a full controller driven from the screen: Start and Stop arm and disarm the loop, Apply sets the setpoint and gains, and the firmware pushes temperature, setpoint, duty, error, state, the status LED, the P/I split and the trend waveform back to the panel. The object map is Software/Display/README.txt.

Remaining on the panel: a run on real hardware to confirm the arming path end to end. Note the trend waveform's add command takes a single byte, so the trace clips above about 91 °C - fine for bench work, not for a reflow profile.

Not started: reflow profiles proper (soak / ramp / peak / cool), enclosure, mains-side safety.

See skill.md for the detail, and especially for the things that cost real time to discover.