Grams, EMI and Crash Kits: Building Small-UAV Hardware on a Startup Clock

A mapping drone comes back from its second flight test with a sheared motor mount and a bent arm. The team that treats this as a failure loses a week. The team that designed for it swaps in a printed mount from the field kit, uploads a reinforced revision that evening, and has new parts in hand within days — before the flight-test window closes. Small-UAV hardware is aerospace engineering on a startup clock, and the manufacturing choices decide which team you are.

Every gram is flight time

For a hovering multirotor the physics is unforgiving: actuator-disk theory puts hover power proportional to the 3/2 power of mass, so adding 10% mass costs roughly 15% more hover power — and endurance falls faster than the weight went on. Fixed wings are gentler but not free: every gram becomes induced drag, climb rate and launch speed. On a platform whose entire empty weight is a few kilograms, a 40 g bracket is not a detail. It is minutes of mission.

So drone teams end up running the same discipline as airliner programs — a weight budget with a named owner — at one-millionth the budget.

What small airframes are actually built from

Selective laser sintering in PA12 nylon has quietly become the default for complex airframe parts. At about 1.01 g/cm³ and printed without support structures, it turns wing ribs, servo trays, gimbal mounts, antenna mounts and cooling ducts into parts that need no tooling and arrive in days. Because SLS parts nest freely in the build volume, a full shipset of ribs and mounts often prints as a single batch.

Where stiffness rules — motor mounts, landing gear, wing-spar joiners — carbon-filled nylon roughly doubles the modulus of neat PA12 for a small weight premium. The catch is environmental: nylons absorb moisture, and the part you measured the day it left the printer is not the part after a humid week in the field. Critical fits get designed with that drift in mind, and load-bearing prints get retested after conditioning, not just fresh out of the machine.

The design move that matters: modularity around the crash. The teams that iterate fastest design the airframe as printable modules on a carbon spar or plate core, with the parts most likely to break — mounts, legs, wingtips — as cheap, replaceable prints. A crash consumes a module, not an airframe, and every flight test feeds a revision that flies the same week.

Avionics boxes: CNC aluminium and the EMI trap

Flight controllers, radios and power electronics live in machined aluminium for three reasons: heat, stiffness and shielding. 6061-T6 covers most enclosures — around 310 MPa tensile strength, friendly to machining and anodizing, corrosion-resistant. 7075-T6 nearly doubles that strength at about 570 MPa for load-bearing housings and gimbal structures, at higher cost and with worse corrosion behavior. Walls of 1.5 mm machine cleanly in either alloy; below about 1.2 mm, chatter and distortion start eating your tolerances.

EMI is where enclosures quietly fail. An aluminium box shields only if it is electrically continuous, and standard anodize is an insulator — anodize the whole box including the lid land, and you have built a slot antenna with a gasket in it. The working recipe: chromate conversion coating (MIL-DTL-5541 chem film) on mating surfaces and grounding points to keep them conductive, anodize only where you want wear resistance, conductive gaskets seated in machined grooves, and honest attention to every cable feedthrough — because the openings, not the walls, set the shielding you actually get.

Aerospace supply-chain discipline at startup scale

The aerospace part of small-UAV work is not the altitude — it is the paperwork culture. When a wing rib cracks at flight 200, you need to know which powder lot built it and which other airframes share that lot. That is traceability. Big aerospace buys it with AS9100 systems; a five-person team gets most of the value with plain rigor: material certs filed for every load-bearing part, print lots recorded against part serials, drawing revisions tied to flight configurations.

• Material certs — mill certs, powder or filament lot numbers — on file for every load-bearing part before it flies
• Every printed or machined part marked with a serial that maps back to its build lot and drawing revision
• Fasteners called out by grade and torque, never as generic hardware
• One frozen drawing revision per flight-test configuration, so a crash points at one design instead of a guess
• A field kit of the crash-prone printed modules, so a broken mount costs minutes rather than the test window

Taiwan is a good place to run this loop. SLS and CNC vendors with day-scale turnaround sit within a courier day of any lab on the island, so a Friday crash can put revised ribs and a re-machined mount back on the bench before the next weekend's flight window. The cadence, more than any single part, is what compounds into a better aircraft.

Weigh everything, trace everything, and design the airframe to be rebuilt — because it will be. When the next revision is drawn, upload the whole shipset in one go and keep the iteration on a weekly cadence.