Airframe Design Protocol

Use this protocol for each change to the outer shape of the aircraft: wing, tip, keel, stilt, pod or fuselage. Do the cheap gates first. Stop at the first gate that fails, and record the result. Ryan decided the gate order on 2026-09-29: calculations and strength come before the tests. th-aeronautics owns this protocol.

Roles

Session	Gates
th-aeronautics	Owner. Idea card review, calculations, strength, screen, CFD, the report
th-blender	Geometry generator, the Blender review with Ryan, CFD geometry export
th-main	Sliced mass and print time
th-hangar	Fit on the pad, in the box and in the capture guides
jbsim-06	Flight modes from the linear model
th-simulator, th-testpilot	SITL model variant and the test battery
Ryan	Approval of the geometry, of each dell01 CFD run, and the final decision

Procedure

1. Write an idea card in docs/agent-threads/airframe-design/. Give the problem, the number that must change, its baseline value, and the part that changes.
2. Build the geometry as a generator script in tools/, not as a hand edit. Write the kit-derived parts to parts/ only. They must not go to the public site.
3. Show the geometry to Ryan in Blender. Discuss it with him on mcspeak. Continue only if he agrees.
4. Do the calculations. Compare these values with the baseline:
   • mass and the CG shift, with the nose ballast that holds the CG at 270 mm
   • reference area, span, mean aerodynamic chord and aspect ratio
   • the change in trim, from the change in lift and moment of each part
5. Do the strength check. Use the load cases in the table below. The minimum margin is 2 on the material’s yield strength (an estimate, until the first test coupons). A part that is also a charge contact must have a margin of 10, as the keel cores have.
6. Get the sliced mass and print time from th-main. Get the fit check from th-hangar.
7. Screen the shape in VSPAERO, and in XFOIL if a section changes. Use the baseline mesh settings (8 mm panels, Tess_W 49). Compare differences only, because the absolute zero-lift terms are not converged (jsbsim thread, 5cba52f).
8. Send the VSPAERO results to jbsim-06 for the flight modes.
9. Ask Ryan for dell01 time. If he agrees, run a coarse CFD pair at alpha 6.7 and 14 on the new shape and on the baseline, from one generator setting each. Run fine-mesh cases only for a shape that passes the coarse pair.
10. Give th-simulator the SITL model variant. th-testpilot flies the test battery on it.
11. Write the report in the idea card’s thread. Put the table of gate results first.
12. Ryan decides. Record the decision in the thread, and send it to th-testpilot for the CLAUDE.md decisions table.

Load cases

Case	Load	Source
Manoeuvre	3.5 g at 2.59 kg, on the wing and its joints	Estimate: ArduPlane roll and pitch limits, with margin
Gust	7.5 m/s vertical gust at 30 m/s	Estimate: the Dryden model in the test battery
Landing	0.5 m/s touchdown on one keel sole, 12.4 degrees nose down	Dock touchdown study
Hard landing	2 m/s on one keel sole	Estimate: a hover failure over the pad
Pad push	Push bar force on the toe and heel faces	th-hangar

Pass criteria

Each value is compared with the baseline in the same model. A value outside the limit stops the change unless Ryan accepts it.

Value	Limit	Baseline (keel A, drooped tip)
Static margin	5 to 15 % MAC	8.2 % (CFD), 11.4 % (VSPAERO)
Elevon trim at 18 to 30 m/s	Within 12 degrees of neutral	5.8 degrees TE up at 20 m/s (CFD model)
Cruise drag at 20 m/s	No increase above 2 %	CD 0.053 (fine CFD, alpha 6.7)
Stall speed at 2.59 kg	No increase	17.9 m/s (SITL stall probe)
Spiral, time to double at 19 m/s	No decrease	6.6 s
Dutch roll damping ratio	0.10 or more	0.141
Mass	No increase above 20 g without a reason	Parts list
Hangar fit	Fits, or th-hangar has a change	th-hangar

Baseline

The baseline is the flight configuration on 2026-09-29: the kit wing with the drooped kit tip, keel A as the foot, no Winglet Stabilizers, CG 270 mm, 2.59 kg. Its CFD geometry is the “14 Flight Config” scene, exported by tools/airframe_stl.py. Its SITL model is sim/terrahawk-keel-notips.json. When Ryan accepts a change, the accepted shape becomes the new baseline.

---

Source: docs/design-protocol.md in the repo. Edit it there; this page is generated.