Let the Calibration Begin
May 3, 2025
Bringing It to Life
So, the device is assembled — and at first glance, it’s alive.
But before diving into any serious software development, we need to verify that every component is correctly wired, functional, and responds as expected.
That includes a camera, 9 servo actuators, 2 continuous servo motors, 3 IR sensors, 2 fans, and 2 LED channels.
Not a huge list, but still too much to test manually by typing commands in the console every time. That would be exhausting — and frankly, a waste of time.
So we need a custom tool to help us test everything now and streamline future assembly and diagnostics.
Power, Light, and Vision
First, we test the ULN2003 driver — it controls several key components:
– the exhaust fan for the card feeder,
– the cooling fan for the power supplies,
– the main LED lighting for the lightbox,
– and the secondary LED lighting used for detecting foil-coated cards.
We also take this chance to calibrate the camera settings — adjusting exposure and white balance for both lighting modes.
The white balance still needs some fine-tuning, but that will come later, once we start developing the card recognition module.
As for foil detection, the extra lighting setup isn’t fully validated yet — but the potential is there, and we’ll know more as the project moves forward.
Servos
The device has three sorting platforms, each with three servo drives.
It’s crucial to set the end positions carefully — if a servo pushes against a hard stop and starts slipping, it can damage the gears or overheat the motor.
Slipping is usually easy to spot thanks to the lovely crunching sound of stressed gears :), but overheating is a slow and silent process.
So when calibrating limits, it’s best to be cautious… or rich. ;)
The video shows sliders for all nine servos, but I only recorded the calibration of two — watching all nine would be a bit much.
Can It Feel?
Each sorting platform has an IR sensor that allows the device to detect when a card is present — and eventually, decide what to do with it.
In the video, we check that all three sensors are working correctly.
You can catch a faint glow from the IR LEDs on camera, but in real life, they’re completely invisible to the human eye — their wavelength is beyond the visible spectrum, so they won’t interfere visually.
Let’s Make It Move
With calibration done and all components behaving as expected, it was time to try feeding a single card.
Manually controlling each servo with a finger isn’t exactly fast — but for testing, it’s bearable.
Off-camera, I noticed that if a card sits on a sorting platform for a few seconds, it doesn’t always want to slide down as expected.
At first I suspected static electricity, but the real issue was simpler: too little air between the card and the platform surface, leading to too much friction for gravity to overcome.
The fix was straightforward, but still required reprinting some parts and rebuilding all the sorting platforms.
And yes — this is the way. :)
What’s Next
Now that all components have been tested and are working properly, it’s time to think about stress-testing the system.
I’ll need to build a small demo app that cycles through the slots and distributes cards — no recognition yet, just pure mechanics. This will help reveal weak points early, while they’re still easy to fix.
And of course, it’ll start to look much more like an actual working card sorter.
So don’t forget to subscribe to the blog updates — you won’t want to miss what’s coming next. ;)
To be continued...
Want to support the project?
See how you can help →