FirePick Delta Project Log: Building an Open‑Source MicroFactory with OpenPnP

Software Topology and Theory of Operation

OpenPnP serves as the high‑level control engine, orchestrating tasks such as picking and placing components. Its architecture integrates numerous libraries and subsystems, funneling data to motor controllers through dedicated modules. Created by Jason von Nieda before the FirePick Delta project, OpenPnP is still in alpha but has proven reliable on several production machines. We have now expanded its documentation to bring full transparency to the community.

The Java modules referenced by the JAR icons can be swapped or extended to tailor machine configurations. Each module—except the configuration models—can be specified in machine.xml. Even without Java experience, the declarative approach allows you to point to .java files, dictating class instantiation and behavior, which keeps the codebase modular and maintainable. OpenPnP ships with reference files for a vanilla pick‑and‑place setup using TinyG motion control, a single head, nozzle, actuator, drag‑tape feed, and a downward‑looking camera. We have adapted and expanded these files to support the advanced features required by FirePick Delta.

Motion driver: Our custom motion controller board, inspired by RepRap RAMPS 1.4 and Melzi, runs a forked Marlin firmware. An OpenPnP Marlin driver translates cartesian commands into G‑code for the board. While functional, offloading delta math and Z‑probe corrections to the Pi offers superior precision and speed. FireFUSE bridges the gap, converting cartesian coordinates into delta space, applying calibration offsets, and relaying the corrected G‑code to Marlin. The only requirement for OpenPnP to talk to FireFUSE is a compatible driver, which is currently under development.

Camera driver and Vision Framework: OpenPnP’s native camera support lagged behind the 5‑megapixel Raspberry Pi camera. Karl Lew addressed this with FireSight, an open‑source vision library built atop OpenCV. It exposes a high‑level pipeline API, simplifying complex image processing without deep C++ knowledge. FirePiCam, a customized raspistill wrapper, captures images directly to the FUSE filesystem, preserving SD card lifespan. The firerest-client Java library then feeds these images back to OpenPnP for real‑time vision.

GUI: The bundled OpenPnP GUI is limited and hard to integrate into modern workflows. We’re developing a web‑based interface using Node.js, Express, AngularJS, and Bootstrap, which can run on any HTML5‑capable browser. A node‑java bridge allows seamless invocation of OpenPnP’s core Java APIs from JavaScript, bypassing the legacy GUI while retaining full machine control.

Hardware Topology and Theory of Operation

Our core controller is a Raspberry Pi Model B+, chosen for its cost‑effectiveness and ample GPIO headers. While the Broadcom 2835 is modest, it reliably handles computer vision and web services within our budget constraints. We designed the system to be portable across other SBCs—BeagleBone Black, Allwinner A4, Intel Galileo, or Mini‑ITX—though the Pi remains the preferred baseline.

ERPIHAT01 HAT Board

Following the RPi HAT specification, we introduced the FirePick Delta HAT (ERPIHAT01), a mezzanine board that expands the Pi’s capabilities without clutter.

• Provides a 16×2 character LCD connector wired to the Raspberry Pi GPIO.
• Includes connectors for a rotary encoder, push‑button switch (with LED), and a 12 V→5 V step‑down supply to the Pi.
• Pass‑through connector for the Raspberry Pi camera module, allowing extension to the delta end‑effector.
• Equips a piezo buzzer driven from the Pi’s single PWM pin, cheaper than using the audio jack.
• Per the HAT spec, supplies the required EEPROM chip and +5 V back‑feed protection, enabling power input via GPIO instead of the micro‑USB port.

For more detail: A project log for FirePick Delta, the Open Source MicroFactory