Confirmed torque as the failure mode before committing to any swap: the 3.6g micro servo (0.7 kg/cm) was running at its limit under cable tension during pan motion, the peak-load moment.

Mechanical fit — measured the 9g servo against the camera-arm drawing: width absorbable into the casing resize, but its 13.5mm height jump (35.5 vs 22mm) is the binding constraint.

Voltage — cross-referenced the servo's 4.8V max against the 10.8V pack — a BEC that may sag voltage under dynamic load, a failure that mimics an undersized servo but has a different root cause.

Cable routing — a PTFE liner cut friction, but the real fix was geometric: rerouting shortened the lever arm on the servo horn.

2.5× torque headroom (0.7 → 1.8 kg/cm) from the 9g swap, with metal gears for fatigue life.

Two distinct failure modes isolated — torque margin and BEC voltage sag — instead of a single assumed cause.

Fix bundled into one CAD revision with the planned 1% casing resize: one iteration, not two. Recommendation issued conditional on clearance verification rather than as a blanket swap.

Instrument BEC output voltage under dynamic load to confirm or rule out the sag hypothesis before committing the swap.

Verify 13.5mm vertical clearance for the 9g servo in the next CAD revision, and cut the mount-pocket update into the same casing resize.

Bench-validate the 9g servo under cable tension across the full pan range.

Architecture trade study. Weighed a twin-boom pod against a slender monocoque on drag, structural joints, and first-build complexity.

Tradeoff: a belly camera aperture instead of an unobstructed boom-mounted view. Hit our AUW targets.

Full CAD assembly — 15% gyroid wing infill visible

Internal layout. Mapped every avionics component's constraint back onto the fuselage envelope while holding CG, vibration isolation, and serviceability.

Every component constrained the fuselage geometry — the internal envelope had to fit the full avionics stack while holding CG placement, vibration isolation, and serviceability.

Fuselage internal cavity — designed around component stack

Component	Constraint imposed on fuselage
PM02D Power Module	Near CG — short high-current runs to PDB, placement fixed internal bay geometry
Dual GPS	Requires separation + clear sky exposure — constrained upper fuselage geometry
Pixhawk 6X	Positioned for IMU isolation from motor vibration — affects mount stiffness design
D3548 Motor	Tractor config — CG and nose geometry set by motor mass forward of wing
Raspberry Pi	Thermal management — requires airflow clearance, can't be adjacent to battery

Manufacturing & materials. Down-selected the airframe material class, then the filament, then split the print by load path — and redesigned for serviceability.

Ran a material decision matrix across eight classes — wood, plastic, fiberglass, carbon fiber, composite, aluminum, foam, and 3D printing — scored on strength-to-weight, manufacturability, fatigue, repairability, cost, and environmental resistance; 3D printing won on customizability, cost, and in-house manufacturability. Then down-selected filament from SolidWorks part volumes (wing/aileron 916 cm³, fuselage 574 cm³, tail 223 cm³): standard PLA 4.68 lb, LW-PLA mild 3.17 lb, LW-PLA strong 2.27 lb.

Chose a load-path hybrid over one material everywhere: standard PLA for the fuselage (high local stress from motor thrust and vibration, payload mounting, wing attachment, and landing loads — better impact resistance and screw retention, and only a small share of surface area), and ColorFabb LW-PLA in strong foaming for the wings, control surfaces, and skins (loads spread over large areas, stiffness geometry-dominated, lower wing loading and bending moments). Carbon spars carry the primary loads; 15% gyroid infill adds isotropic stiffness.

Added a modular removable nose for direct avionics access without full disassembly — a DFM call for field serviceability. Nozzle temperature tuned to 240°C through controlled test prints.

Modular removable nose — direct access to internal avionics

Aerodynamics. Sized and checked the wing in XFLR5 at the cruise Reynolds number (results below).

Slender-pod monocoque built and assembled — skin-carrying structure hit the AUW targets with less internal reinforcement than the twin-boom alternative.

16% static margin (neutral point at 38.5% MAC) for stable, responsive handling; S1223 wing (CL max ~2.2, cruise CL ~1.4, gentle stall at 13° AoA) with a symmetric NACA 0012 tail, lift confirmed at 16.98 lbf at 22.35 m/s.

Hybrid PLA + LW-PLA airframe — standard PLA fuselage where local loads are highest, ColorFabb LW-PLA (strong foaming) wings and skins at 240°C with 15% gyroid infill. Estimated 2.27 lb, ~52% lighter than an all-standard-PLA print (4.68 lb). Modular removable nose for avionics access.

Flight-test campaign: collect in-flight structural and thermal data and validate CG and static margin in the air.

Iterate nose and avionics-mount fit against learnings from the first assembly.

Requirements spec anchored to a real reference load — my own iPhone 16 Pro (3,582 mAh, 13.94 Wh) — with every requirement given a target, type, and validation method.

Energy-budget waterfall from cell nameplate down through usable-capacity fraction, boost-converter efficiency, board quiescent loss, and the load-bearing phone-side charging assumption — each number cited and defensible, run at conservative / nominal / optimistic.

Cell-architecture trade study comparing one 21700 against a Li-po pouch and two 18650s on energy, volume, cost, sourcing, and protection.

Benchmarking against in-class commercial banks to locate a defensible point of differentiation.

Feasibility resolved. A full 5→95% phone recharge isn't achievable from one 21700 (needs ~91% cell-to-port vs an ~86% ceiling). Reframing to a 10→90% window sets REQ-010 at 13.1 Wh delivered, requiring 0.73 cell-to-port efficiency — which the energy model shows is achievable.

Architecture selected: one protected 21700 (18 Wh nameplate, 34.7 cm³ cell envelope) over the pouch and dual-cell options — chosen on serviceability, not raw energy density.

Positioning: a 21700 can't beat flat-pouch banks on volume (in-class median 61.9 cm³ vs a 75 cm³ target), but lands ~36% smaller than the comparable same-architecture Anker PowerCore 5000 (116.5 cm³). The differentiator is the user-replaceable cell.

Provisional BOM closed at $73.98, within the project budget.

SolidWorks V1 enclosure next: package the cell, boost board, USB-C, and a 4-step state-of-charge indicator into a pocket-safe shell (≥1.0 mm edge radii).

Finish the remaining calculations — thermal model at 10 W, snap-fit for the replaceable-cell door, and tolerance stacks on the connector interface.

Print coupons and run metrology to lock print settings, then build and discharge-test V1 against REQ-010 (13.1 Wh) — targeting a working prototype at sprint close.