Nighthawk (SAFMC Competition)
3rd Place Competition Finish
50g
Build Weight
5-8 mph
Cruise Speed
5
Prototypes
The SAFMC (Singapore Amazing Flying Machine Challenge) competition brief was simple: design an indoor RC (Remote Control) glider, navigate through gates, perform figure-eights, and land accurately — no time limit.
That last detail should change everything; however, most teams still optimized for speed, twitchy control surfaces, high thrust-to-weight ratios, and aggressive flight profiles. We spent a few days watching other teams test and realized they were building the wrong aircraft. In a gymnasium with unpredictable air currents and zero margin for error, a fast plane overshoots every gate and overreacts to every input. What you actually need is a platform that flies slowly and responds predictably.
Demo 🍿
Starting from Zero
None of us had built an aircraft from scratch before. Most teams were using proven templates—trace the outline, laser-cut foam board, assemble in a weekend. That works fine if you want a guaranteed flyable result, but you learn and explore basically nothing! It ends up being a boring LEGO building competition instead of engineering.
Our vision was different; we wanted to experiment. So I opened Fusion 360, and we started working through the fundamentals:
- What makes a plane stable?
- How do you calculate the center of gravity?
- Why does dihedral matter?
- What's the relationship between wing area and stall speed?
The first five days were pure theory. We shared YouTube videos, worked through lift equations, and sketched force diagrams on whiteboards.
By day six, we had a shared understanding of the physics and could start making actual design decisions.
The Slow Flight Thesis
The competition scoring had no time bonus. You could take ten minutes to complete the course and still win if you hit every gate cleanly. That constraint completely changed the optimization problem.
Fast planes have problems indoors:
- Overshoot gates and targets
- Require aggressive corrections that destabilize the flight
- Less time to react to obstacles or air currents
- Higher kinetic energy makes crashes more destructive
Slow planes win on precision:
- More time to line up approaches
- Gentler control inputs for fine adjustments
- Better low-speed handling in confined spaces
- Crashes are survivable (foam doesn't shatter at 5 mph)
We optimized for controllability at 5-8 mph cruise speed instead of raw performance. That decision drove every subsequent design choice.
So what did we do?
Bi-Wing Configuration
To fly slow, you need lots of lift at low velocity.
At half the speed, you need four times the wing area to generate the same lift (velocity is squared). We couldn't just make the wings wider — that increases drag and makes turns harder in tight spaces.
Solution: Stack two wings vertically. Bi-plane configuration gives us double the lifting surface without extending the wingspan. The horizontal stabilizers at the rear effectively function as a second set of wings, creating biplane aerodynamics.
Benefits we actually used:
- Higher total lift at low speeds (larger total wing area)
- Neutral static stability (plane naturally returns to level flight after disturbances)
- Increased drag for slower cruise speeds (usually a disadvantage, exactly what we wanted)
- Better roll authority from differential lift on stacked surfaces
The tradeoff is reduced top-end speed and slightly worse efficiency. Neither mattered for our use case.
Control System: Why Strings
Standard RC planes use rigid pushrods—metal or carbon fiber rods connecting servos to control surfaces. At our weight budget (50g total), the metal linkages were too heavy.
Standard pushrod weight: ~12 grams for three control surfaces
Our string linkage weight: ~3 grams
We used a tensioned nylon string running from three micro servos to control the surfaces:
- Two elevons on the main wings (pitch and roll control)
- Two rudders on vertical stabilizers (yaw control)
The catch: Strings stretch slightly over time and need re-tensioning between flights. For a competition aircraft that only needs to fly five or six times, that's manageable. For a daily flyer, you'd want pushrods.
Magnetic Wings: Learning from Crashes
Prototype 3 crashed during outdoor testing. Crosswind caught it wrong, rolled inverted, and hit the pavement nose-first. Both wings shattered at the fuselage joints where we'd hot-glued them.
Hot glue is great in shear (holds well under sliding forces), but terrible in impact. Every crash meant disassembly, re-gluing, and waiting for the adhesive to cure. We were losing entire afternoons to repair work.
Better solution: Neodymium disc magnets embedded in wing roots and fuselage attachment points.
Wings snap on magnetically. The hold is strong enough for normal flight loads (several times the aircraft's weight) but releases cleanly on impact. The wings pop off instead of breaking.
We also 3D-printed mounting brackets—removing unnecessary material while maintaining strength around magnet pockets.
Dialing In Control Sensitivity
Early test flights were over-responsive. Small stick inputs caused large deflections. The plane would oscillate: pilot corrects left, overshoots, corrects right, overshoots, repeat until crash.
The problem wasn't the linkages—it was the ratio between stick movement and control surface throw. We'd set the servo travel too high.
Fix: Reduced servo throw in transmitter settings (from 100% to 60%) and physically reduced elevon surface area by about 15%.
It took four or five test sessions to find the right balance. Too much throw: twitchy and unstable. Too little: sluggish and can't make tight turns. There's a sweet spot where control response matches pilot reaction time.
What We'd Change
- Calculate CG before building. We wasted prototype 3 discovering our CG was aft of the aerodynamic center—something a spreadsheet would've caught in ten minutes.
- Log configuration changes systematically. We took photos during assembly, but didn't consistently document what we changed between test sessions. That made it harder to track which modifications actually improved performance.
- Start pilot training earlier. The plane's handling characteristics only matter if the pilot can fly it. We should've been practicing on a basic trainer aircraft weeks earlier instead of learning on our custom design during competition prep.
- Test string linkages under vibration. The strings performed well in smooth flight but loosened faster than expected under prolonged motor vibration. We should've run longer endurance tests to catch that.
Technical Specs
Component | Specification |
Total weight | 50g |
Motor | 1850KV brushless |
Propeller | 5030 (5" diameter, 3" pitch) |
Battery | 1S 450mAh LiPo (~15g) |
Servos | 3× micro servos (~3g each) |
Frame | Laser-cut 5mm foam board |
Wingspan | ~600mm |
Control surfaces | 2 elevons, 2 rudders |
Linkage system | Nylon string (3g total) |
Why Build From Scratch?
Some teams using template designs placed higher than us. Templates work. They're proven designs that have been refined through multiple iterations.
But copying doesn't teach you why something works. We spent weeks debugging CG issues, tuning control throws, and testing wing configurations. That iterative process is where the learning happened.
By the end, we could look at any part of the aircraft and explain exactly why it was designed that way: why bi-wing instead of monoplane, why dihedral angle, why string linkages, and why magnetic attachment. That depth of understanding doesn't come from following instructions.
For anyone considering this approach: Expect to crash more, iterate more, and spend more time troubleshooting than teams using standard kits. You'll place lower in the competition rankings. But you'll understand aircraft stability and control systems at a level that tracing a template doesn't provide.
The question is what you're optimizing for: competition results or building something uniquely epic. Both are valid. We chose the latter.
End Credits
