TEKNOFEST 2026: Our Hardware Choices and Why They Matter

Honestly, we didn't know from the start. Our first parts list changed twice. The list we ended up with has a reason behind every part, but we only got there after months of testing. In this post I'm breaking down the three most critical choices: why we started with an F450 frame and dropped it, why we run two separate processors, and why we still 3D print in ABS alongside carbon fiber.

X-quadrotor, 520 mm: not math, visual inspection

In our first attempts we considered the F450 plastic frame. Cheap, easy to source. But when we mounted the Pi Camera underneath and hovered, propeller shadow crept into the edge of the frame. That breaks target detection immediately. After switching to the X-type 520 mm frame, the wide-angle scan stayed clean. Bonus: it gave enough clearance for 10-inch propellers while staying compact enough to fit in a car trunk.

Total weight

2.4 kg

Hover time

19.76 min

Max speed

59.4 km/h

Hover throttle

45.9%

We didn't cram it onto one board

Running both the PID loop and the OpenCV pipeline on a single SoC is possible in theory. In practice, the moment image processing gets heavy, flight stability suffers. We didn't want to accept that. The Pixhawk 6C handles flight control, with dual-IMU redundancy. The Raspberry Pi 5 (8 GB) runs all of the image processing. The bridge between the two boards is MAVLink over FTDI USB-TTL; latency stays well under a second.

• Pixhawk 6C: ArduPilot stack, sensor fusion + motor PWM output
• Raspberry Pi 5: OpenCV image processing, DroneKit/PyMAVLink bridge
• Active cooler: required to keep the Pi 5 from throttling, we found this out in the first test
• Holybro PM07 + Pololu D24V50F5 BEC: supplies the Pi 5's 5V/5A draw over an isolated line

We didn't pick the cheapest option, we picked what actually works in the field. Reliability on competition day matters more to us than sponsor cost.

Both carbon fiber and ABS, because they serve different jobs

The main airframe plates are 3K carbon fiber. Its strength-to-weight ratio leaves aluminum far behind, and it also helps with electromagnetic shielding and vibration damping. But making everything out of carbon would be both wildly expensive and impractical. Parts that are likely to need replacing during competition, like motor mounts, wing attachments, and the payload bay, are printed in ABS filament. We bring our 3D printer to the field; if a part breaks, we print a new one. Our field repair time is under 10 minutes.

Why this all matters

Looked at one by one, each part seems reasonable. What actually matters is that they work together in the field. Two clear goals sit behind these three choices: (1) the target detection pipeline won't drop out mid-mission, and (2) if a part breaks, we can repair it trackside and fly again. We cover the software side, especially the move from SITL to the field, in a separate post.