Building a Custom 45-Module, 64-Flap Electromechanical Display: From Design to Calibration

Adam G Makes

Introduction [0:00]

The video's creator, Adam G, introduces his ambition to build a custom split-flap display after being inspired by commercial versions.

Split Flap Explanation [0:39]

Adam explains the basic mechanics of a split-flap display using a clear acrylic open-source model:

• Each module contains a hub with holes where half-flaps (split flaps) ride.
• When the hub rotates, gravity causes flaps to drop, displaying a character.
• A stepper motor precisely controls the hub's rotation.
• A Hall effect sensor detects a magnet on the hub, establishing a "home" position for accurate character display.

  

A Look at My First Attempt [2:12]

Adam reviews an earlier 8-character model he built based on a tutorial:

• Likes:
  • Compact module size with the motor integrated into the hub.
• Dislikes:
  • Limited to 8 modules, not scalable for larger displays.
  • Assembly was difficult due to threaded rods and messy wiring, making maintenance challenging.

    

Design Criteria [3:37]

Adam outlines five key criteria for his new, improved split-flap display:

1. More Characters: Expand from 37 to 64 flaps per module to include A-Z, 0-9, and special characters, including colored ones.
2. More Modules: Aim for 15 modules per row, across 3 rows, totaling 45 modules, a significant increase from 8.
3. Easier to Assemble: Implement a totally modular, snap-on/snap-off design without threaded rods, allowing for easy servicing of individual modules.
4. Simpler Electronics: Minimize soldering by using self-enclosed boards where motors and sensors can simply plug in.
5. Cost-Effective: Keep the total project cost under $1,000, competitive with commercially available displays of similar size.

   

Adding More Flaps [4:54]

The first challenge was increasing the number of flaps:

• The existing design had 37 flaps, the goal was 64.
• The stepper motors used (28BYJ-48 5-V) have limited resolution, so accommodating 64 flaps was a concern.
• Adam redesigned the hub in Onshape, removing old holes and extruding 64 new, smaller ones.
• Flaps also had to be thinned to fit.
• A successful test of a single module confirmed that the motors could handle the 64 flaps with sufficient resolution.

  

  

PCB Design (First Attempt) [6:02]

The electronics strategy was a major re-think:

• Previous Approach: A single ESP32 microcontroller controlling all 8 motors.
• New Approach: An individual control chip on each module for its motor and sensor, with a central Raspberry Pi sending high-level commands.
• Custom PCB Design: This involved designing a custom Printed Circuit Board (PCB) for each module.
  • Components included a microcontroller, motor controller, and connectors for the motor and Hall effect sensor.
• Initial Test: The first batch of PCBs was manufactured and tested, showing motors spinning as expected. A minor issue with a resistor was identified and corrected.

  

Module Mount and Data Bus [7:48]

A robust solution for mounting and electrical connectivity was needed:

• DIN Rail Mounting: DIN rails, a standard for mounting electrical components, were chosen for their modularity.
• 3D Printed Mounts: Custom 3D-printed mounts with a compliant mechanism allow modules to easily snap onto the DIN rails.
• Pogo Pin Interface: Four holes were incorporated into the mounts for pogo pins, which would connect directly to a data bus PCB.
• Data Bus PCB: A new PCB was designed with four long traces to carry power (5V), ground, and RS485 signal lines (A and B).
• Final Module PCB Design: The module PCBs were modified to include pads for these pogo pins. The final blue PCBs feature an ATtiny1616 microcontroller, an RS485 transceiver for communication with the Raspberry Pi, a motor controller chip, and power circuitry.

  

  

Assembling the First Module [10:25]

A speedrun was performed to estimate assembly time:

• All necessary components for one module were laid out.
• The assembly process involved attaching flaps, soldering pogo pins, inserting the motor and Hall effect sensor, connecting wires, and enclosing the module.
• The first module took 14 minutes to assemble.
• Estimating for 45 modules, the total assembly time was projected at 10 hours, which could be optimized by batching processes like soldering.
• Extensive 3D printing (hundreds of hours) was required for all 45 sets of flaps, enclosures, and DIN rail mounts.

  

  

  

Putting It All Together [12:32]

The final assembly of the complete display:

• After a month of intensive building, all 45 modules were complete.
• A backing board was prepared to mount the three DIN rails, along with a power supply and a Raspberry Pi.
• Initial Power-Up Challenges (Integration Hell):
  • Not all modules spun correctly, and some displayed incorrect characters (e.g., 'A' instead of blank space, or '0'/'2' instead of '1').
  • This was attributed to minor discrepancies in 3D printed parts and motors, leading to varying home offsets.
• Software Calibration Solution:
  • Adam developed a custom mobile application to perform "per-character tuning."
  • This app allowed him to individually adjust the "home offset" for each module and fine-tune the step count for every character on each module that was showing incorrectly.
  • The errors were "consistently wrong," making them solvable through software adjustments.
• Successful Tuning: After significant effort, all modules were accurately calibrated.
• Demonstration:
  • The display successfully showed all modules as blank, then all '1's.
  • It then displayed a custom message: "HEY YOUTUBE! IS THIS THING WORKING?? YES!"
  • The display also showed live-updating time (e.g., "11:34 PM").
  • The system demonstrated reliability, maintaining accuracy for days without needing re-homing.

    

    

    

    

Finale [19:30]

The completed display is shown in its final enclosure, demonstrating various capabilities:

• Smooth character transitions.
• Displaying colored characters.
• Showing dynamic data like NHL scores and stock prices.