Design-Build-FlyC0 lift + cruise VTOL

Lift + Cruise VTOL Design Review

0 Report info

Design-Build-Fly UC Berkeley capstoneDBF-RPT-001 Issue 6

Lift + Cruise VTOL Design Review

Hybrid VTOL with vertical lift rotors for takeoff and landing, and a pusher prop and fixed wing for cruise.

Engineering design review DBF-RPT-001, issue 6, 26 October 2026. Design revision 0.4.0, ready for build: drawing set DBF-DRW-001 revision E.

Revision record
IssueDateDescription
12026-10-13Design review after the budget redesign and the six-reviewer audit
22026-10-18Drawing set DBF-DRW-001 (5 sheets), wing saddle, interactive CAD model, accessibility fixes
32026-10-21Design 0.4.0: aerodynamics and CFD, resized wing, leg fairings, every fastener in the CAD, drawing rev B (9 sheets), build manual, test plan, flight software, trade study
42026-10-24One chapter at a time with Part tabs; animated CFD flow views (2-D section, 3-D aircraft); drawing rev C (layout, ballooning and callouts corrected, geometry unchanged)
52026-10-25Review platform: sidebar navigation, visual-first CAD, CFD and drawing workspaces, light theme, report view; wing sized to an 11.75 m/s design stall, a 4 % CL_max margin under R11 (sizing reads CL_max from the aero package); CAD rev D, drawing rev E; FEA and CFD re-run on the final wing; heat-dissipation analysis; airframe-specific SITL; code-review fixes
62026-10-26Benchmarks against 22 published aircraft and 37 literature values; cruise moved to 18 m/s, the R11 lower bound (22 % less cruise power); Sources section; requirement traceability, review checklist, margins at a glance, failure modes, scored risk register, findings log, start-here guide; build-manual step figures; companion pages one section at a time

Approval

Design-Build-Fly teamPrepared · 26 October 2026
Checked by, with date
Approved for the client, with date

0 Overview

C0 is ready to build within the $2,000 cap: every part, screw, assembly step, test and software parameter is specified. It hovers on W and cruises at m/s on W. One pack gives min of cruise after the hover allowance and 20 % reserve; the mission needs 10 min (R10).

analysis/sizing.mjs → analysis/results.json (C0); docs/DESIGN-BASIS.md v0.4.0 R2, R6, R7, R10; analysis/aero/aero_results.json (leg drag); parts from docs/research/budget-*.md and hardware-spec.md (prices checked 2026-10-13 and 2026-10-21)

Committed spend (R6)
cap $2,000 incl. tax + shipping
Crash reserve (R7)
target ≥ $100: met; also covers the unpriced items (§7)
Sized MTOW (R2)
kg
below the 8–15 kg working target
Next client action
2026-11-02
approve C0 so Order A (test motors, packs, charger) goes out; first hover 2027-03-13
The aircraft as modelled in SOLIDWORKS, every part and fastener included

A 4-lift twin-boom quadplane: a 2.4 m NACA 2415 wing ( m², chord mm), four SunnySky X4110S lift motors on 17-in props and a fifth on a 14 × 8.5 pusher, 6061 aluminium booms and spar on a taped XPS foam wing with printed fairings on the legs, and ArduPilot QuadPlane on a CoreWing F405 Wing board. Take-off mass kg. Full model: §2.

What we ask, and what happens next

Approve C0 and its spend by 2026-11-02. Then:

  1. 2026-11-02 Order A: test motors, ESCs, thrust-stand kit, flight packs and charger
  2. 2026-11-16 Thrust-stand tests T1–T5 confirm the lift motors and the R13 thrust margin
  3. 2026-12-04 Design freeze, then Order B (airframe, propulsion, avionics) on 2026-12-07
  4. 2027-01-04 Build begins from the build manual; flight-software rehearsal in simulation
  5. 2027-03-13 First hover, then transitions
  6. 2027-04-17 Full mission flight (R9, R10), with a buffer to 2027-04-30

Order-by dates and schedule: §15; tests: §13; build: §11

Start here

  1. Read this page first. The summary, Margins at a glance and requirement status show where the design stands.
  2. Then your part. Structures: Structures, Drawings, build manual WP3–WP7. Electrical and software: Propulsion, Controls, WP9–WP11. Testing: Test plan, Flight test. Everyone: Requirements, Budget, Schedule.
  3. Search. Ctrl K (⌘ K on a Mac) searches every heading. Report view shows the whole review as one printable document.
  4. Trace any number. The grey line under each table or figure names its source file; Sources lists every cited link.

Status labels: met shown by analysis or inspection · risk met with a small margin, or waiting on a decision · open needs a test · fail not met · est engineering estimate · "—" not known yet.

Status

Margins at a glance

Each design limit, today's predicted value and the room left. Green: ≥ 10 % margin. Amber: met with less. Red: not met.

What is limitedPredictedLimitMarginStatusDetails

analysis/results.json (C0 sized, BOM, CG); analysis/aero/aero_results.json (stall, mission); analysis/thermal/thermal_results.json (the two tightest mission cases at the 25 °C design ambient, R21); limits from docs/DESIGN-BASIS.md and the test plan abort table

Workstream status

Every workstream has a written output except the team sign-off and the flying-site approval.

The ten workstreams: what exists and what does not
WorkstreamStatusWhat existsWhat does not
Requirementsdraftdocs/DESIGN-BASIS.md v0.4.0, R1–R20Team sign-off; car measured (R4)
Configurationdone29-variant trade study with literature survey (§2)—
Sizing and aerov0.4Sizing, XFOIL, NeuralFoil, AVL, drag build-up, 2-D RANS CFD (§4)Any measured input (M-items)
CAD and drawingsrev D— components with every fastener; drawing set DBF-DRW-001 rev E, 9 sheets (§2)As-built check against the first article
SimulationrunFEA boom + spar, modal, thermal, vortex lattice (§9)FEA re-run 2026-10-25 on the CAD rev D inputs; every result within 15 % of its hand calc
BuildwrittenBuild manual DBF-MAN-001, cut, print and wiring lists, templates (§11)Parts on hand
Flight softwareconfiguredParameter file, setup guide, checklists, SITL A–G (§12)Bench setup on the real board
TestingplannedTest plan DBF-TP-001, cards, data scripts (§13)Tests run
BudgetfitsItemised BOM and purchasing CSV (§10)Cart quotes; orders placed
Compliance / siteopenRules located; Remote ID module on FAA listSite approval; TRUST

What changed in v0.4

v0.4 made the concept buildable: new aero data resized the wing and added leg fairings, the CAD carries every fastener at real size, and the build manual, test plan and flight software are written.

The v0.4 design changes
  • wing area up to m² (chord mm) on the trimmed CL_max of
  • printed teardrop fairings on the four landing legs (bare legs would fail R10)
  • neutral point moved forward by the pod ( % MAC): fly with 0.5 kg of ballast when there is no payload
  • split motor mounts, butt-spliced booms, 46 mm split clamps, M5 × 20 wing bolts, rudders on both fins
  • wing joiner made a slip fit with printed stops, so the pod fits the car envelope
  • firmware output map rebuilt around the board's timer groups; ArduPlane 4.7.2 stable
  • hinge tape, spare servo and spare pushrods dropped to hold the $100 crash reserve
Earlier history: v0.2 cost and failed the cap; v0.3 replaced every high-cost part
Areav0.2 goodsv0.3 goodsSaving

analysis/results_v0.2_baseline.json vs analysis/results.json, category totals of the C0 BOM (goods, pre-tax)

v0.3.1 (same day): a six-reviewer audit (§12; findings log N01) confirmed 79 findings. Changes: booms moved outboard so the pusher disc clears the rear lift discs (it overlapped by 30 mm); joiner 0.76 m for 0.30 m engagement per side; NACA 2415 with the spar at 30 % chord, for foam cover over the socket; a fifth servo for twin rudders; PA6-CF motor mounts; missing wiring, glue, battery straps and hot-wire allowance added; transition speed 16 m/s (R11).

0 Requirements

Requirements

Requirement status

IDRequirement (short)StatusEvidenceVerified byClosing tests
R1Lift rotors + pusher + wing; lift stopped in cruisemetC0 (§3), CAD (§2)
R2MTOW 8–15 kg target, < 25 kg caprisk kg: cap met; target deviation pending sign-off
R3Span 2–3 mmet2.4 m
R4Largest piece fits a car (≤ 1.5 × 0.5 × 0.4 m est.)riskLongest piece m; joiner is now a slip fit so the pod stays inside the envelope; car not measured (§2 M2)
R5Two-person carry; assemble ≤ 15 minopenMass fine; assembly untested
R6≤ $2,000 incl. tax + shippingmet committed (§7)
R72nd pack, spares, crash reserve ≥ $100metReserve ; three optional spares moved into reserve cash
R8Flight-tested by 2027-04-30risk12-day buffer, 2027-04-18 to 04-30 (§11)
R9Full mission profileopenMission flown in SITL (scenario A); flight test cards written (§8)
R1010 min cruise + 2 min hover, 20 % reservemet min sizing, min aero build-up, with leg fairings (§4)
R11Cruise 18–22 m/s; stall ≤ 12 m/s; transition ≥ 1.3 × stallriskDesign stall m/s, % under the 12 m/s limit; cruise m/s; AIRSPEED_MIN 16 m/s; stall test decides (§4)
R120.5 kg payload allowancemetIn mass roll-up
R13Hover T/W ≥ 1.6open on nominal-voltage maker data; thrust stand must show ≥ kgf per motor at storage voltage (§5 M1)
R14Motor-out: cruise hard, hover desiredriskC0 fails the hover part (§8)
R15Pusher-out → hover recoveryopenQ_ASSIST plan (§10)
R16Wind ≤ 5 m/s, gust ≤ 8 m/sopenAbout 9° of hover tilt holds an 8 m/s gust (docs/research/configurations.md); flight demonstration limited to steady 3–5 m/s wind
R17Maintained VTOL autopilotmetArduPlane 4.7.2 stable includes the board; parameter file checked; SITL scenarios A–G pass (§10)
R18FAA path, TRUST, registration, Remote IDopenBeeID on FAA DoC list (RID000001995)
R19Campus policy and siteopenCapstone covered (senior projects); RFS only campus transition site
R20Arming switch, failsafe, geofenceopenAmass AS150U disconnect + RC arm switch; failsafes and 110 m fence pass in SITL; hardware test pending (§10)

docs/DESIGN-BASIS.md R1–R20; analysis/results.json; "Verified by" and "Closing tests" from docs/test/test-matrix.csv (A analysis, I inspection, D demonstration, T test)

Ready to build? Review checklist

What a critical design review asks before building (NASA NPR 7123.1D, App. G, Table G-7, cut to what applies to a student aircraft), and where each answer is.

Question a reviewer asksStatusWhere to look
Is the design complete and does it meet every requirement with margin?riskYes except R14 hover motor-out (accepted with limits) and the open tests; Margins at a glance, Requirement status
Can every requirement be verified, and is it clear how?openRequirement status, columns "Verified by" and "Closing tests"
Are the interfaces defined (joints, fasteners, wiring)?metDrawings sheets 4–9, fastener schedule; wiring list in Downloads
Are the risks known, scored and owned?metRisk register (owners are roles until names are assigned)
Have the hazards been analysed and controlled?metFailure modes and controls; LiPo procedure in Flight test
Is there a manufacturing plan?metBuild; build manual DBF-MAN-001 with step figures
Is the flight software ready to test?metControls: parameter file checked, simulator scenarios pass
Do cost and schedule fit?riskBudget (fits the cap); Schedule (12-day buffer before 2027-04-30)
Are the earlier review findings closed?openReview findings log

NASA NPR 7123.1D Appendix G (explorers.larc.nasa.gov/APSMEX25/SMEX/pdf_files/NASA02_NPR_7123_1D.pdf, checked 2026-10-26); docs/research/review-formats.md; closing tests from docs/test/test-matrix.csv

Open items

Decisions needed from the client

Approve design C0 and its spend, so Order A goes out by 2026-11-02.
Order A is the test hardware: two lift motors with props, an ESC pack, the thrust-stand kit, both flight packs and the charger. Ordered by 2026-11-02, the thrust-stand tests start 2026-11-16; the motor and battery choices depend on them. Order B (airframe, propulsion, avionics) follows on 2026-12-07.
Recommended: approve C0 as specified, holding the reserve for crash repair and unpriced items. Alternative: approve Order A only and hold Order B until the thrust-stand results (no cost; 12-day schedule buffer).

analysis/results.json C0.bom; order-by dates §15; thrust-stand tests §13

Also for sign-off: five acceptances, each recommended
Which legal path and which flying field? Flight tests can't be scheduled until this is settled.
Campus policy allows flying for UC courses, including senior projects, so the capstone qualifies. Site limits:
  • Richmond Field Station: 400 ft AGL
  • Clark Kerr track: 200 ft AGL
  • La Loma field: below 20 ft, inside the fences
  • All sites: 6 a.m. to sunset; no flying on game or graduation days
Each flight needs a request filed ≥ 7 business days ahead and ≥ $1M liability cover per pilot. A transition needs ≥ 30 m AGL and 50–100 m of straight flight, so La Loma is out, Clark Kerr allows only hover and low passes, and Richmond Field Station is the only campus transition site.
Recommended: by 2026-11-06, request Richmond Field Station from UCPD/Risk Services and confirm the capstone's institutional sponsorship (implied by the § 350 educational path). Each pilot holds TRUST and AMA Full ($89, outside the cap). Alternative: one member passes Part 107 ($175) and the team flies off campus.

R18, R19; oercs.berkeley.edu/policies/campus-policy-library/unmanned-aircraft-system-drone-policy (site limits and senior-project clause, checked 2026-10-13) · §11

The wing is sized to an 11.75 m/s design stall, % under R11's 12 m/s. Accept this margin and confirm stall in flight?
Sizing uses the trimmed CL_max (maximum lift coefficient) from the aero package, now , and designs to 11.75 m/s. A CL_max up to 4 % low, or about 0.28 kg of mass growth, still meets R11. The margin cost 88 g and 0.6 min of cruise, no money and no pack change (docs/research/stall-margin.md compares 11.25–12.0 m/s). Transition completes at 16 m/s, 1.36 × design stall.
Recommended: accept, and measure stall in fixed-wing mode early in the flight build-up. Then set AIRSPEED_STALL to the measured value and AIRSPEED_MIN to the larger of 16 and 1.3 × measured stall. A stall above 13.08 m/s is an R11 failure.

R11; analysis/aero/aero_results.json stall; docs/research/stall-margin.md; docs/research/aerodynamics.md · §4

Accept no hover motor-out tolerance (R14)?
C0 cannot hold attitude after losing a lift motor in hover. The 8-lift (C1) keeps T/W after a motor-out but costs , over the cap.
Recommended: accept for C0. Keep hovers under 5 m AGL and 60 s, and qualify every motor on the thrust stand first. Alternative: none fits under $2,000.

R14; analysis/results.json · §3

How do the stopped lift props sit in cruise?
BLHeli_S ESCs cannot stop a prop at a set angle, so a 2-blade prop stops wherever it lands. The four stopped motors, mounts and props are about a fifth of zero-lift drag (estimated coefficients). C0 still meets R10 at twice that drag, but only with the 8000 mAh pack, which uses most of the crash reserve.
Recommended: run the car-roof test (§3 M1) on one stopped unit at m/s. If the prop windmills or sits crosswise, print a stop pawl. Alternative: Eolo folding 17×6.2 at $49.99 a pair; two pairs are $99.98 (about $110 with tax), more than the reserve.

docs/research/budget-lift.md caveat 4; Props table (folding) · §3

MTOW lands at kg, under the 8–15 kg working target.
8–15 kg was a working target; the hard cap is 25 kg. The cost cap and short mission drive mass down, and a smaller aircraft is cheaper, easier to carry (R5) and safer to test.
Recommended: accept, and record it in the design basis pending team sign-off. Alternative: add payload or ballast to reach 8 kg: more battery, motor and cost for no benefit.

R2; docs/DESIGN-BASIS.md v0.4.0 · §4

What the team still has to measure

Every open measurement, by chapter

    0 CAD and drawings

    The SOLIDWORKS model matches the aircraft down to the screw: — part files in a —-component assembly, every screw, nut, washer and heat-set insert at real size. cad/build.ps1 regenerates it from the sizing outputs and cad/hardware.json, so CAD and numbers can't drift apart. Holes, counterbores and nut pockets use the hardware-spec clearance and insert sizes, with printed-hole compensation.

    What the model includes
    • NACA 2415 wing panels (chord mm) with 26 % chord ailerons, spar sockets at 30 % chord and a slip-fit joiner held by printed stops
    • a foam wing saddle with printed root plates, a bolt pillar and root tongue for the M5 × 20 wing bolts
    • split PA6-CF motor mounts with a 4.5 mm floor, and the boom butt-spliced inside the rear mount on a printed plug
    • 46 mm split boom, gear and tail clamps; fin shoes; rudders on both fins
    • printed teardrop fairings over the four landing legs
    • servo trays, horns with hardpoints, pushrods, flight-controller and battery trays, pitot clamp, pusher cup

    Checks. Pusher-to-lift-disc clearance is mm (minimum 20 mm, checked by params.mjs). SOLIDWORKS interference detection on the full assembly found 286 contacts: 201 intended heat-set insert press fits, and 85 slivers totalling 0.42 cm³, mostly flat prints on curved foam (largest 0.10 cm³, the rudder clevis around its horn upright). It also found two real clashes, both hardware-spec errors, now fixed: the landing-leg screw is M3 × 30 so nut and head clear the tube, and the motor counterbore is Ø7.5 to take the Ø7.0 washer.

    cad/BuildVtol.cs, cad/params.mjs ← analysis/results.json and cad/hardware.json; checks in cad/verification.txt; outputs cad/model/VTOL_Concept.SLDASM, VTOL_Exploded.SLDASM, .STEP, .EASM (open the EASM in the free eDrawings viewer, or the STEP in any CAD tool)

    3-D model

    Drag to rotate, right-drag to pan, scroll to zoom (touch: tap once, then two fingers to pan or pinch). Point at or tap a part to name it. X-ray shows the booms, spar sockets, joiner and battery inside the skin; Explode pulls the parts apart.

    Loading the 3D model…

    View
    Display
    SOLIDWORKS model · C0 rev D

    cad/model/*.SLDPRT exported by cad/ExportMesh.exe (binary STL, 0.4 mm chord deviation, 12° angle tolerance) → review/mesh.mjs; placed per cad/params.txt ( triangles, components). Colours are the CAD display colours, not finishes.

    Drawings

    Drawings

    Engineering drawings

    DBF-DRW-001 rev E: nine ANSI B sheets generated from the SOLIDWORKS model and dimensioned from cad/params.txt (no hand-entered numbers). Each sheet has a general tolerance block (printed parts ±0.5 % with a ±0.5 mm floor, cut tubes ISO 2768-m, hole positions Ø0.4, motor pattern Ø0.2); critical bores carry their fits. Select a sheet to enlarge it, or save the vector PDF.

    SheetTitleContents
    1General arrangementThree views and isometric, overall dimensions, revision table, breakdown and ballast notes
    2Exploded assemblyExploded view with 24 balloons, one per major part type, numbered as the sheet 3 parts list
    3Parts listThe SOLIDWORKS bill of materials, every fastener included
    4Wing panel, saddle and bolt pillarSpar socket, joiner bore with its slip fit, M5 hole and nut pocket
    5Motor mount and spliceBore 25.6 ± 0.2, floor 4.50 ± 0.15, motor pattern position Ø0.2, counterbores, insert holes, splice bolts
    6ClampsBoom, gear and tail clamp halves with hole callouts and the leg-socket fit
    7Tube cut list, leg and fairingCut lengths with ISO 2768-m tolerances and cross-hole positions; landing leg; leg fairing
    8Fastener scheduleID, standard, size, quantity, joints, torque and threadlocker for every fastener
    9Printed hardpoints and traysFin shoe, spar collar, rear hardpoint, pusher cup, servo tray, root tongue

    cad/Drawings.cs → cad/drawings/DBF-DRW-001.SLDDRW and DBF-DRW-001.PDF (vector, all sheets); dimensions from cad/params.txt ← analysis/results.json

    Parts and fits

    Fasteners and fits

    Every fastener, with its joints and torque
    IDSizeStandardQtyTorque, N·mJoints
    Fits: designed clearance or interference at each critical bore
    IDJointMin, mmMax, mm

    cad/hardware.json and docs/research/hardware-spec.md: SunnySky X4110S drawing (motor pattern and thread depth), ISO 273 clearance holes, ruthex insert datasheet, Fastenal stainless torque chart, Loctite 242 data sheet; vendor pages checked 2026-10-21. Insert torques in printed parts are estimates until the coupon test (test plan C02).

    Parts and mass (C0, sized)

    ItemMass kgx from LE, mSource

    analysis/results.json concepts.C0.cg.items; EST rows are estimates in analysis/sizing.mjs BASE

    Breakdown and transport (R4)

    PieceQtySize, mFits 1.5 × 0.5 × 0.4 m?
    Wing half with aileron and spar socket2 × × yes
    Front boom with motor mounts, motors and props (props off)2 × 0.05 × 0.08yes
    Tail assembly: stabiliser, elevator, fins, aft boom stubs1 × × width over 0.5 m; lies flat on a seat
    Fuselage pod with joiner, avionics, pusher10.85 × 0.16 × 0.16 estyes
    Gear legs (4), flight packs (2) in the ammo can——yes

    analysis/results.json geo.segs, geo.hSpan; R4 envelope is an estimate

    What the team must measure

    1. M1
      Built component masses. Kitchen scale (5 kg, 1 g) on each piece as it comes off the bench, logged against the parts table. Empty mass (no battery or payload) over kg + 15 %: re-run the sizing, as T/W falls toward R13 and cruise time below the 25 % margin.
    2. M2
      Car load space. Tape-measure the boot of the team's actual car. A clear length under m fails R4.
    3. M3
      Assembly time. Stopwatch two people from car to armed pre-flight, three runs. A median over 15 min fails R5.

    Manufacturing readiness

    Is each subsystem specified well enough to build (dimensions, tolerances, fits, fasteners, adhesives, assembly order)? Checked against the CAD, drawings and hardware spec.

    SubsystemLevelWhat blocks it

    Tolerance stacks

    StackWorst caseRSSRequirementVerdict

    analysis/validation/manufacturing_readiness.json (check_struct.py), from cad/params.txt, cad/hardware.json, docs/research/hardware-spec.md, cad/drawings/ and docs/build/

    0 Aerodynamics

    analysis/aero/run_all.py → analysis/aero/aero_results.json; method, tools and validation: docs/research/aerodynamics.md. Tools: XFOIL 6.99, NeuralFoil 0.3.3, AVL 3.52, AeroSandbox 4.2.10, SU2 8.5.0, gmsh 4.15.2 (download URLs in the method note, checked 2026-10-21)

    Drag and stall

    Where the drag comes from

    Zero-lift drag by component group, as a share of the aircraft's CD0

    analysis/aero/drag_buildup.py: XFOIL section drag for the wing and tail; Raymer form factors (eq. 12.27–12.31) for pod, fins and booms; Hoerner cylinder and strut data for legs, stopped motors and props

    Wing section and stall

    NACA 2415 lift curve at the cruise Reynolds number: XFOIL, NeuralFoil and SU2 RANS, with the wind-tunnel stall point

    analysis/aero/out/xfoil_polars.json and neuralfoil_polars.json (Re as labelled, Ncrit 9); SU2 2-D RANS points: analysis/aero/su2_2d.py; experiment: Genç, Lock and Kaynak, AIAA ATIO 2008 (Re 2×10⁵)

    2-D CFD

    CFD: mesh study and validation

    MeshCellsCL at 10°CD at 10°
    SU2 against XFOIL and NeuralFoil, angle by angle
    αXFOIL CLSU2 SA CLSU2 SST-LM CLXFOIL CDSU2 SA CDSU2 SST-LM CD

    How the air moves over the wing section

    White particles move with the SU2 velocity field around the NACA 2415, over a background coloured by speed or pressure. The whole-aircraft flow is in the 3-D CFD tab.

    2-D: analysis/aero/flowfield_2d.py → out/flowviz/flow2d.json (SU2 8.5.0 incompressible RANS, Spalart–Allmaras fully turbulent, fine mesh unless labelled; field resampled to a 220 × 128 grid; motion slowed for viewing). Static renders: analysis/aero/plots/flow_*.png. Method, convergence and limits: docs/research/aerodynamics.md §11

    3-D CFD

    SU2 solved the inviscid (friction-free) flow around the whole aircraft on a gmsh tetrahedral mesh; everything shown is read from the solver files. Drag the slice (or press Sweep) to see pressure, speed or vorticity anywhere; the vortex volume shows tip vortices and wakes.

    Full-aircraft flow at cruise. Select a simulated condition; nothing is interpolated between cases.
    Flow Mesh

    Open this tab to load the CFD data.

    Drag to rotate · wheel or pinch to zoom · shift-drag or right-drag to pan · double-click to reset · hover the surface for Cp and speed. The run is a half model with a symmetry plane, mirrored to port for display.

    Solver history for the selected case: pressure residual and lift coefficient

    Results, convergence and checks

    CaseαTetsCLCMyCDi TrefftzOwn CL integrationIterationsrms p drop

    CheckStatusResult

    Computational mesh

    LevelTetrahedraNodesWing trianglesWall size mmTE size mmmin SICN1 % SICNmean SICN

    Is the mesh good enough? Scored against published standards

    CriterionStandardL1L2L3Note
    What a passing mesh needs
      The meshing strategy, rule by rule, with the code that sets each rule

      Mesh strategy, domain and solver settings
      • Mesher: gmsh 4.15.2 (OpenCASCADE geometry, Frontal-Delaunay surfaces, HXT tetrahedra, Netgen optimiser). Linear tetrahedra, linear wall triangles, no prism layer (inviscid run).
      • Sizing (mesh rev 2): wing, tail and fins 2.2 mm × f on the surface; curvature 24/f elements per 2π radius; distance-to-wall threshold from 8 mm × f to 4 m × f at 30 m; trailing-edge lines 2.5 mm × f and their end points 0.6 mm × f; boxes around the aircraft, its wake and the starboard tip-vortex path. Every rule scales with one length factor f (1.5 / 1.0 / 0.72 / 0.52), so the four levels are one grid family (analysis/cfd3d/build_mesh.py).
      • Domain: half model (starboard), symmetry plane at the centreline; box about 34 m ahead of the nose, 44 m behind, 35 m above and below and 34 m outboard of the tip, at least 105 reference chords everywhere (build_mesh.py DOMAIN). Far field: SU2 MARKER_FAR; walls: slip (MARKER_EULER).
      • Solver: SU2 8.5.0 (win64, OpenMP, 20 threads), INC_EULER, constant density 1.225 kg/m³, FDS upwind with MUSCL and Venkatakrishnan limiter, implicit Euler with adaptive CFL 5–30, FGMRES/ILU. Stop on a Cauchy criterion (CL, CD, CMy, 200 iterations, 1e-6) or 3000 iterations.
      • References: S 0.7727 m² (CAD wing 0.3219 × 2.4 m), c 0.3219 m, moment about the design CG (0.0901 m aft of the LE, 28 % chord, on the chord line; analysis/results.json cg.target; the CG height is not tracked by the sizing).
      • Not modelled: viscosity (no skin friction, no physical separation, no stall), propellers (power off), control deflections (0°), sideslip (β 0°). Lift, pitching moment, pressure distribution, induced drag, downwash and tip vortices are what this run resolves.
      CFD geometry against the SOLIDWORKS model: what was simplified and why

      The outer surface was rebuilt in gmsh from the same cad/params.txt parameters and placements as the SOLIDWORKS parts (analysis/cfd3d/build_geometry.py). The raw STEP has 0.1 mm bore gaps, 1.2 mm clamp splits, open hinge gaps and fastener holes, each of which a volume mesh would have to resolve. Each rebuilt part is compared with its STEP solid below.

      CAD partsCFD representationDifferences
      PartSTEP solidsLargest bounding-box difference, mmNote
      Hover and transition CFD are not run. Both need the four 17 in lift rotors modelled: SU2's actuator disk works only in the compressible solver, and resolved rotating blades need a sliding or overset mesh and hours per revolution on this workstation. The view shows cruise only rather than invented fields. The cruise runs are power-off (no pusher slipstream), at 0° elevator (not trimmed) and inviscid, so they do not predict drag.

      analysis/cfd3d/build_geometry.py → out/oml_half.brep, oml_fidelity.json; build_mesh.py → out/L1–L4/mesh.su2, mesh_stats.json; run_cases.py → out/cases/*/ (case.cfg, su2.log, history.csv, restart_flow.csv, forces_breakdown.dat) and out/cfd3d_results.json; postprocess.py → out/post/*.json, out/viz/cfd3d_viz.json; verify_cfd.py → analysis/validation/claims_cfd.json. Run on 2026-10-29; L4 added 2026-11-03.

      Stability

      Stability

      AVL stability derivatives at cruise trim
      DerivativeValueMeaning

      What the team must measure

      1. M1
        Gear-fairing benefit. Watt-meter at m/s in level flight, the same pack state, with and without the leg fairings. Fairings saving less than half the predicted power leave no cruise margin (R10): shorten the legs or fly the lean pack plan.
      2. M2
        Stall speed. Slow to the stall warning at 30 m AGL in fixed-wing mode; log calibrated airspeed (§12 M1). Above 12 m/s fails R11: raise AIRSPEED_MIN and Q_ASSIST_SPEED by the difference.

      CG and stability

      The CG target is 28 % MAC (mean aerodynamic chord), and the lift thrust centroid sits on it by construction. The sizing puts the neutral point at % MAC: the tail-volume value for wing and tail ( %; vortex lattice gives %) less the pod's destabilising shift from AVL. Static margin (CG to neutral point) is % with payload and % without. Full AVL with the pod is slightly more stable, at % (§5); the design uses the more cautious sizing value. Without payload the CG moves aft and the rear rotors carry % more thrust, so the bay carries 0.5 kg of ballast whenever it flies empty.

      C0 CG envelope against neutral point and thrust centroid (% MAC from leading edge)

      analysis/sizing.mjs cg(), geometry() (hn = tail-volume neutral point + AVL pod shift); analysis/vlm.mjs → aero.json (wing and tail only); analysis/aero/aero_results.json (AVL with pod)

      CG build-up

      Itemkgx, mkg·m

      Stability numbers

      QuantityC0C1C2Formula

      What the team must measure

      1. M1
        Actual CG. Two kitchen scales under the wing at known stations. Aft of % MAC (SM under 8 % est) fails cruise stability. Without payload, fly only with ballast; move the battery to bring the CG within 2 cm of the thrust centroid.
      2. M2
        Hover pitch trim. Front vs rear motor outputs in a 30 s hover log. A split over 15 % means the CG is more than about cm off the thrust centroid, eating hover control margin (R13).

      0 Structures and heat

      The 6061-T6 wing joiner is the critical member. At ultimate load (4.5 g) the hand calc on the true 0.875 in joiner gives MPa, a reserve factor of on yield; this figure governs. The FEA, with the joiner bonded in its socket (OD = socket ID), gives MPa on a 2 mm mesh and confirms the load path under that idealisation. Booms have strength to spare: MPa on a 1.5 mm mesh, mm tip deflection at full thrust. Spring-like landing gear needs about 100 mm of travel.

      analysis/sizing.mjs structures(); cad/Sim.cs → cad/sim/results.txt; allowables: SOLIDWORKS library 6061-T6 (SS), E 69 GPa, yield 275 MPa

      Loads

      Load cases

      CaseLoadValueBasis
      Cruise manoeuvren_ult = 3.0 × 1.5 gLimit 3.0 est; FS 1.5
      Hover, gust/control2 × (W/2) at boom station N·mFactor 2 est
      Lift motor full thrustT_max at front rotor NX4110S 3,130 gf
      Hard landing0.3 m drop2.43 m/sv = √(2gh); h est
      Boom root torsionone front motor at T_max, rear at idle N·mSocket shear MPa (τ = T r / J); clamp slip checked by proof load M5

      Landing loads

      Gear stroke, mmDamped n = 1 + h/sSpring n = 2(1 + h/s)Verdict (spring bound)

      Energy balance, drop h = 0.30 m est. Bare aluminium legs act as springs, so read the right-hand column. A foam or TPU foot with about 100 mm of travel keeps n ≤ 8.

      Spar, booms, joints

      Wing joiner and spar

      Weight
      N
      m g
      Root moment, ultimate
      N·m
      n_ult W b / (3π)
      Joiner I (0.875 × 0.058 in)
      mm⁴
      π/64 (22.23⁴ − 19.28⁴)
      Joiner stress, hand
      MPa
      M (D/2) / I
      FEA von Mises
      MPa
      SOLIDWORKS, 3 mm tets

      The joiner carries the full root moment (foam neglected, conservative). It slides into the 1.000 × 0.058 in sockets with 0.23 mm clearance; its 0.76 m length gives m of engagement per side, about diameters. The sockets sit at 30 % chord, leaving about 4.6 mm of foam each side in the 15 % section.

      Booms

      Arm, spar to front rotor
      m
      x_front + 0.25 c
      Moment at full thrust
      N·m
      T_max L
      Stress hand / FEA
      / MPa
      M c / I; FEA
      Tip deflection hand / FEA
      / mm
      P L³ / 3EI
      First mode hand / FEA
      / Hz
      tip mass = lift unit (motor + ESC + prop + mount) + 0.236 m_boom; rigid root

      X4110S no-load speed is 400 KV × 22.2 V = 8,880 rpm (148 Hz); hover speed is not published (—). If it lands near a multiple of Hz the booms resonate; the tap test (M2) settles it.

      Joints and materials

      • Wing–fuselage: 0.76 m 6061 joiner, slip fit in the saddle, held by printed stops 300 mm inside the spar sockets. One M5 × 20 button head per half, into a nylock in the printed bolt pillar, sets incidence and stops rotation (hardware spec J10).
      • Boom–wing: two 46 mm split PETG clamps per boom, closed by M3 socket heads into heat-set inserts, grip the tube by friction; front clamps screw to a spar collar, rear clamps to a printed hardpoint (J05). Typical PETG HDT (heat-deflection temperature) is 68 °C (Prusament datasheet); the ELEGOO PETG bought has none published (—).
      • Motor mounts: printed PA6-CF20 (HDT 173 °C, Polymaker datasheet). The lumped thermal model puts the motor near °C after a 90 s hover, too close to PETG's limit.
      • Wing: hot-wire FOAMULAR XPS core, packing-tape skin. σ_wing = kg/m² by F11; tape mass per area is unpublished, so estimated est (M4).

      docs/research/budget-airframe.md §1–§5

      What the team must measure

      1. M1
        Wing proof load. Invert the wing on pod-width supports and sandbag it to limit load (3.0 × W), hold 30 s. Any permanent set in the joiner, or tip deflection over 10 % of semi-span, fails the wing for flight.
      2. M2
        Boom natural frequency. Tap the motor and record with a phone spectrum app. Within ±15 % of the hover rotor frequency (or twice it), add a strut before flight.
      3. M3
        Gear stroke. Drop the empty airframe 0.3 m onto grass; read peak g from the flight-controller log. Over 8 g, fit softer feet before the hover test.
      4. M4
        Tape skin mass. Weigh a 0.5 m² taped foam sample against bare foam. Over 0.1 kg/m² per side, update F11 and re-run the sizing (R2, R10).
      5. M5
        Boom clamp proof load. Hang 1.5 × T_max ( N) from the front motor station with the rear free, and check for slip or rotation at the clamps. Any rotation over 1° or any crack fails the boom joint (R9).

      FEA and heat

      Simulation and verification

      Every SOLIDWORKS Simulation result is within 15 % of its hand calculation, and every margin passes on the loads the studies were given. But the 2026-10-29 audit found those loads understate the lift-unit mass by 37 %, and no landing case exists (Validation matrix STR-05, STR-11). Each FEA model uses the same idealisation as its hand calc (rigid root, joiner continuous with its sockets), so it confirms the hand calcs rather than validating them independently; an independent gmsh/C3D10 re-solve of the boom, with no SOLIDWORKS code, is in the validation matrix. The rev B slip-fit joiner carries bending the same way once seated, but the studies were not re-run on the rev B clamps; clamp compliance would lower the boom frequency.

      What was simulated, and with which tools

      FEA of the boom and the spar/joiner, a boom modal study and a lift-motor thermal study. A vortex-lattice model confirms the wing-and-tail neutral point and lift slope; AVL adds the pod, which moves the neutral point forward (§4). An independent Python check re-derives every sizing formula and the BOM. SOLIDWORKS Flow Simulation would not load on this licence, so CFD runs in SU2: 2-D sections and, since 2026-10-29, the whole aircraft in 3-D (§4).

      cad/Sim.cs → cad/sim/results.txt and *.SLDPRT (studies inside); analysis/vlm.mjs → aero.json; analysis/verify.py

      Results and independent checks

      Against the hand calculations

      StudyQuantityFEA / modelHand / sizingDiffMargin

      SOLIDWORKS 2025 Simulation (Educational licence, watermark on plots). Linear static, 2nd-order tets ≈ 3 mm; 6061-T6 (SS) library material. Simplified load-path models: boom arm cantilevered at the spar clamp with a tip block sized to the lift-unit mass (motor + ESC + prop + mount, kg); spar + joiner half-span fixed at the centreline, carrying a uniform load scaled to the elliptic root moment.

      Aerodynamics: vortex lattice

      QuantityVLMSizing estimateComment

      analysis/vlm.mjs: horseshoe vortices, wing 48 × 4 + tail 16 × 2 panels, cosine spanwise spacing (Katz & Plotkin, Low-Speed Aerodynamics, ch. 12). Inviscid; stopped rotors, booms and fuselage excluded.

      Stopped-rotor and boom drag is still an estimate.
      The whole-aircraft CFD (§4) is inviscid: it gives lift, moment and induced drag, but not the skin friction and base drag of the four stopped motors, mounts and booms. Those come from handbook coefficients (Hoerner) and are the largest cruise-drag term after the wing. A viscous run needs prism layers on every part; it is the next CFD step.
      Recommended: the car-roof drag test (§4 M3) on one stopped motor and prop at m/s measures it directly, for the cost of a luggage scale. Alternative: ask whether the department's Education licence includes Flow Simulation, and run an external-flow study on cad/model/VTOL_Concept.SLDASM.

      Independent checks (analysis/verify.py)

      Check groupWhat is re-derivedResult
      Sizing F1–F12Every formula for C0–C3, from the stored inputs, without model.jspass
      BOMLine sums, tax, committed total, reserve, cappass (C1/C2 flagged over cap)
      StructuresRoot moment, joiner stress, boom deflectionpass
      FEA vs handBoom stress, deflection, mode; spar stress; motor steady temperature (±15 %)pass
      VLM vs sizingNeutral point and lift slope (±10 %); e assumption below inviscid epass
      Project audit (workflow)Six reviewers (citations, engineering, budget, requirements, CAD/sim, page), an adversarial verifier and a completeness critic79 confirmed findings fixed in v0.3.1 or listed as open items; 2 rejected

      What the team must measure

      1. M1
        FEA boundary check. Dial gauge or ruler on the boom tip with the measured clamp, under a known hanging weight. Over 1.5 × the FEA deflection means clamp compliance dominates: re-run the boom modal study with the measured stiffness before transition tests (R9).
      2. M2
        Motor temperature curve. IR thermometer every 15 s during T3. A measured time constant under 100 s, or a 90 s value over 80 °C, invalidates the lumped model and the hover limits (R9).

      Explore the results in 3-D

      Each SOLIDWORKS Simulation result drawn in place on the aircraft. Coloured: study meshes (stress or temperature). Grey: the rest of the CAD; parts around a result are drawn faint. Each study is a separate simplified model of one load path or heat source, not a whole-aircraft analysis. The boom result appears on the two front arms (the 0.462 m arm it models); the shorter rear arms stay grey. Hover to read a value; "Go to the peak" jumps to the maximum.

      SOLIDWORKS Simulation results placed on the whole aircraft (grey: outside the study shown): choose a result

      Drag to rotate · scroll or pinch to zoom · shift-drag or right-drag to pan · double-click to reset · hover to read a value · keyboard: arrows, + and −, h

      cad/SimFields.cs (nodal von Mises and displacement from the fine-mesh static studies saved by cad/Sim.cs; the von Mises column is matched to SOLIDWORKS' own maximum) → analysis/fea_fields.py → analysis/fea_fields.json; thermal fields from cad/Thermal.cs → analysis/thermal/thermal_results.json. Meshes reduced by vertex clustering for the web: each cluster keeps its peak stress, so the hot spots survive. Aircraft: the CAD assembly mesh of §2 (cad/mesh, cad/params.txt comp lines). Each study is placed from the CAD positions and cad/params.txt, not by eye (boom root at the wing quarter chord, x = −c/4, since its arm feaBoomL = xFront + c/4; spar at sparX, sparY, mirrored; lift unit on each Lift_Motor; pod by the battery and flight-controller positions; pusher by the cup floor and motor can); the derivation is in review/src/fieldviz.js place(). Where a model and the CAD differ (pod cylinder position, pusher cup, front ESC side), the note under the view says so.

      Heat dissipation

      Across the aircraft

      SOLIDWORKS thermal studies of the mission's heat sources: each lift unit (motor, PA6-CF mount, boom, ESC), the pod (battery and flight controller in foam bays) and the pusher. Each is a chained transient through the mission (60 s hover, 20 s transition, 600 s cruise, 20 s transition, 60 s hover) at the 25 °C design ambient (R21: no full mission above 25 °C, measured at the field), cooled by prop wash in hover and the free stream in cruise. Heat loads are loss estimates est; the tests that replace them are listed below.

      Temperature fields from SOLIDWORKS Simulation, shown as a thermal camera would see them

      Component temperatures through the mission at 25 °C ambient, against their limits
      Every component against its limit, both days (table)
      ComponentCasePeak, °C (25 °C day)Peak, °C (20 °C day)Limit, °CMargin, KStatusLimit source and note
      Every flight case passes at the 25 °C design ambient, but three margins are under 10 K and rest on estimated heat loads.
      Mission peaks against limits: lift motor 75 / 80 °C (+5 K), pusher motor 76 / 80 °C after 600 s cruise (+4 K), battery 53 / 60 °C (+7 K, packs at the 6.0 mΩ per cell acceptance limit), lift ESC 62 / 70 °C (+8 K), flight controller 60 / 85 °C (STM32F405 rating). The 25 °C envelope (R21) comes from these results; Berkeley's April average high is 20 °C. The motor losses are estimates with no published value (lift: 20 % of hover input; pusher: 33.2 W, 15 % × 75 % of cruise input) est, and the pod model has no bay vents, which could only lower the battery and FC temperatures.
      Recommended: accept packs at ≤ 6.0 mΩ per cell (C11); fly no full mission above 25 °C measured at the field (R21); keep the 60 s hover rule. Measure before changing the design: the lift-motor rise on the thrust stand (C08, 90 s) and the pusher on the ground (G03, 3 min at cruise throttle). If either runs more than 4 K hotter than the model at the same time and ambient, vent the bays (docs/research/thermal.md §6) or shorten the hovers.

      analysis/thermal/thermal_results.json findings (SOLIDWORKS run 2026-11-02, cad/sim/thermal/results.txt); analysis/thermal/thermal_inputs.py (loads, limits, ambient)

      Static renders and SOLIDWORKS plots (for print)

      cad/Thermal.cs (SOLIDWORKS Simulation 2025 thermal, steady and chained transient; the transient API now saves every step) ← analysis/thermal/thermal_inputs.py (loads and film coefficients with sources) → cad/sim/thermal/results.txt → analysis/thermal/thermal_report.py → analysis/thermal/thermal_results.json and img/. Validation: motor-only steady 143.9 °C = Sim.cs; 90 s transient 77.2 °C vs lumped 77.0 °C. Correlations, materials and limitations: docs/research/thermal.md

      One lift motor: steady FEA and 90 s hover

      A 90 s hover takes a lift motor to about °C (lumped estimate; band °C for c = 385–500 J/kgK; time constant s). Inputs: motor heat 20 % of hover input ( W) est, prop-wash convection 50 W/m²K est, 35 °C ambient (the motor-only solver-validation case; the mission studies above use 25 °C). Steady FEA gives °C, the bulk temperature for unlimited hover; real hovers last 60–90 s, so the time response decides. Three back-to-back 90 s runs with 30 s rests would reach about °C, so T3 cools the motor to 45 °C between runs.

      h·A from steady FEA
      W/K
      Q / (T_ss − T_amb)
      Heat capacity
      J/K
      m c = 0.165 kg × 470 J/kgK (est, Cu/steel/Al mix)
      Time constant
      s
      τ = m c / (h A)
      T after 90 s
      °C
      T_amb + ΔT_ss (1 − e^(−t/τ))

      The 90 s value here is lumped capacitance; the SOLIDWORKS transient study in "Across the aircraft" above (cad/Thermal.cs) gives 77.2 °C for the same case. Motor heat capacity uses a mass-weighted c of 470 J/kgK est. Copper (385) and steel (460) have lower c than aluminium (896), so an all-aluminium assumption would understate the temperature rise.

      0 Mass and power

      MTOW is kg on the ( Wh), the cheapest real pack that keeps ≥ 11.5 min of cruise: a 15 % design margin est over R10, on top of the 20 % energy reserve. Mass comes out of a closed loop: mission energy sets battery mass, mass sets wing area and hover power, and those feed back into the energy. With an ideal, continuously sized battery the loop closes at kg and Wh.

      analysis/model.js F1–F12; analysis/sizing.mjs CRUISE_MARGIN_MIN

      Mass

      Mass roll-up

      GroupC0 design (cont. battery)C0 real packC1 real packC2 real packBasis

      Live back-calculation chain

      The ideal loop with a continuously sized battery, so its numbers sit slightly under the built aircraft's (§1). Move a slider to see how each input drives the mass.

      The chain runs analysis/model.js size() in the page with a continuous battery. Defaults reproduce results.json "design". Sliders marked est are estimates.

      Sensitivity

      MTOW change for ±20 % on each input (current concept, continuous battery)
      Real packs on C0: cruise minutes available and take-off mass (chosen pack highlighted)

      analysis/results.json sweepC0; packs from docs/research/budget-energy-test.md §1

      Propulsion

      Propulsion

      Lift: SunnySky X4110S 400KV on 17-in props with LANRC 45 A BLHeli_S ESCs, $63.79 per station ($46.99 motor + $18.60/2 prop + $30.00/4 ESC, analysis/results.json BOM). Max thrust 3.13 kgf per motor gives T/W ; each motor draws W in hover. The pusher is a fifth X4110S on an APC 14×8.5EP, so one spare motor covers every position. Power: two Ovonic 6S 6500 mAh LiPos ($138.99 for the 2-pack) and an HTRC C150 charger. Re-terminate the packs' EC5 leads to Amass AS150U connectors (70 A continuous), which are also the R20 physical disconnect.

      docs/research/budget-lift.md A3, ESCs; budget-energy-test.md §1–2; pusher prop PB 1c

      Lift units

      PartLayoutMax gfg/W @ 1.5 kgMotor gStation $Notes
      SunnySky X4110S 400KV + 17×5.5 + LANRC 45A4-lift3,1308.3416563.79Table on 17×5.8; some KV340 rows look wrong (KV400 rows used)
      SunnySky X3515S 400KV + APC 16×5.5MR + SoloGood 45A (price at the time of the trade)4-lift (C3)3,3607.5117063.72Watts computed as V × A
      SunnySky V3508 KV380 + Eolo CN15×5.5 + LANRC 35A8-lift1,84010.30 @ 0.8 kg9759.128 stations; over cap as a layout
      Hobbywing X6 SE 280KV (v0.2)4-lift6,0588.5 @ 3.1 kg373134.98Oversized for a 6 kg aircraft
      Tarot 4114 320KV4-lift≈2,650———Below the 3 kgf floor; no table

      docs/research/budget-lift.md tables A/B and flagged-table notes; station $ = motor + ESC + one prop

      Battery

      PackChemistryWhkgWh/kgContinuous A$C0 MTOWC0 cruise min

      docs/research/budget-energy-test.md §1 (Wh = mAh × V_nom). LiPo C ratings (480–1,200 A) are not credible, so packs are accepted by measured resistance instead: at ≤ 6.0 mΩ per cell (C11) a pack holds 3.31 V per cell at the 82 A transition step (AERO-19).

      Current and rating check (C0)

      PathHoverPeakRatingVerdict
      Pack, hover A ( A at 21.0 V)—60 A assumedFine
      Pack, transition (hover + pusher at 2 × cruise power, est) A at 21.0 V111 A (4 × 27.7 A full lift)60 A assumedOver the assumed rating; burst test M4, and set ArduPilot battery current limits
      Each lift ESC A27.7 A45 A headlineMargin 1.6× on peak
      FC current sensor A111 A90 A cont / 215 A peakPeaks are brief
      Pusher ESC A cruise—45 A14×8.5 current unknown (M3)

      docs/research/budget-lift.md X4110S table (27.7 A at max); budget-avionics.md §1.2 (CoreWing sensor); I = P / 22.2 V

      What the team must measure

      1. M1
        Lift thrust curve on our pack. DIY stand (20 kg load cell + HX711 + Nano, calibrated with water jugs), watt meter in line, five throttle steps. A maximum under kgf per motor at storage voltage fails R13 (4 × thrust ≥ 1.6 × kg).
      2. M2
        Motor heat in a mission-length hover. Hold g for 90 s, read the can with an IR thermometer every 15 s, and cool below 45 °C before the next run. (A mission has one hover of about 80 s at each end, with about 10 min of cruise cooling between.) A can over 80 °C est after one 90 s run fails R9 hover duty (model: °C). Never fly back-to-back hovers: three 90 s runs with 30 s rests reach about °C in the model. PA6-CF mounts tolerate the base temperature; PETG (HDT about 68 °C) would not.
      3. M3
        Pusher current and static thrust on 14×8.5EP. Thrust stand + watt meter. Static thrust under 1.5 kgf at 60 % throttle, or a peak over 40 A, means dropping to a 13×8 prop.
      4. M4
        Pack sag at hover current. Watt meter on all four lift motors at hover throttle, tied down; then a 3 s full-throttle burst on all four. Below 3.4 V/cell est at hover current, or 3.2 V/cell in the burst, fails R13 margin: move to the 8000 mAh pack (+$61, more than the reserve).

      Hover efficiency calibration (FM·η from maker data)

      Sizing formula F3 needs FM·η: rotor figure of merit (how close to ideal the prop is) × motor-and-ESC efficiency. Each maker thrust-table point gives it as ideal power / measured electrical power. At C0's hover thrust, FM·η = .

      UnitThrust gP_elec WP_ideal W (F2)FM·η

      docs/research/budget-lift.md thrust-table excerpts (sunnyskyusa.com), checked 2026-10-13; ρ = 1.225 kg/m³

      What the team must measure

      1. M1
        Hover electrical power at full weight. Watt meter in a 30 s tethered hover, or one motor on the thrust stand at g. Above W, R10 fails on the selected pack. The maker-data estimate is W.
      2. M2
        Parasite drag. Log cruise current at m/s in still air and back out drag with F7. Above W cruise power (total CD0 above , against estimated), R10 fails on the selected pack.
      3. M3
        Stopped-rotor drag. Car-roof mast with a luggage scale on a pivot, m/s. Model: N per stopped unit at m/s. Above about N per unit, R10 fails on the selected pack.
      4. M4
        Pack usable energy. Discharge at about 16 A (cruise current) on the charger, logging Wh. Under Wh delivered, the 20 % reserve of R10 fails.

      0 Proof and validation

      How a claim is judged

      Verified: the calculation re-runs from its cited inputs and an independent method agrees within its stated tolerance. Partially Verified: some evidence, but an important check is missing. Unverified: not enough evidence. Failed: a calculation, simulation or independent check contradicts the claim. A picture of a simulation does not count as evidence, and two checks that share inputs are not counted as independent.

      analysis/validation/build_matrix.py (status), proofs.py (arithmetic re-run and links), source_audit.json (sources), link_check.py (every URL in the docs)

      Claims map

      Parts are coloured by the worst status of the claims about them. The labels point at where the key equations act; click a label or a part to open its proof.

      Hover a part to see its status; click to list its claims.

      Proof chains

      Each box is a claim; arrows show where one claim's result is another's input, and the left column is the external sources (datasheets, papers) the numbers start from. Grey arrows carry the same number on both ends; red means the downstream claim still uses an old value. Click any box to open its proof.

      All 78 claims
      IDClaimStatusRe-runWhy

      Sources

      Every external number the claims use, checked against the source itself: the value in the repo, the value on the page, and where on the page it is.

      SourceStatusValues checkedUsed by

      Benchmarks

      Benchmarks

      analysis/benchmarks/aircraft.csv (22 aircraft, each with its source); analysis/benchmarks/literature.csv (37 values); docs/research/benchmarks.md

      Cruise speed

      The wing is sized by the stall speed, so cruise speed is free to choose inside R11's 18–22 m/s. Flying slower brings the wing closer to its best lift-to-drag speed, so the same pack lasts longer.

      Cruise power and cruise time against speed, with the fixed pack
      Cruise speed, m/sInside R11L/DCruise power, WChangeCruise time on the pack, min

      analysis/cruise_speed_trade.mjs → analysis/cruise_speed_trade.json (analysis/model.js with C0's sized parameters; propeller efficiency held at 0.60)

      C0 among published aircraft

      Where C0 sits against 16 lift + cruise aircraft and six tilt-rotor or tailsitter aircraft of similar size. The clearest difference is how much battery each carries for its mass, which follows from how long it is meant to fly.

      Battery energy per kilogram of aircraft against take-off mass, for the aircraft that publish both; hover a point for its name. Most makers do not publish watt-hours, so many values assume 3.7 V per cell
      All 22 aircraft with their sources
      AircraftLayoutMTOW, kgSpan, mWing loading, kg/m²Disk loading, N/m²Cruise, m/sEndurance, minBattery, Wh per kg

      What the comparison shows: C0's wing and disk loadings are at the low end, which is what a 12 m/s stall limit and large 17 in props give. Its battery is a much smaller share of its mass (0.13 against 0.20–0.40) because R10 asks for 12 minutes, not the one to four hours these aircraft are sold for. Its weight × speed ÷ power, the one efficiency figure that can be worked out for all of them, is below the better lift + cruise aircraft; slowing the cruise closes part of that gap.

      analysis/benchmarks/aircraft.csv; derived columns computed by analysis/benchmarks/build_bench.py; endurance figures are makers' claims

      Our assumptions against the literature

      AssumptionC0 usesPublished valuesVerdict
      Drag added by the lift rotors and boomsCD0 0.0266 → 0.0445Retracting rotors removes 36–38 % of parasite drag; a sizing model adds 35 %typical
      Drag of one stopped lift unit0.0022 m²0.0015–0.0022 m² for the prop alone (0.1 σ A)low end
      Cruise propeller efficiency0.60about 0.65 measured on small props; 0.74–0.76 predicted for our pointconservative
      Hover efficiency FM·η0.516, from our motor's data0.567 from another maker's data at the same thrustconservative
      Transition energy64 kJ (40 s)about 29 kJ scaled from a 25 kg quadplane estconservative
      Battery specific energy176 Wh/kg (pack label)154–159 Wh/kg for 6S 10 Ah packslabel value
      Hover thrust-to-weightabout 2.0typical
      Static margin11.5 % with payload or ballast5–15 %typical
      Tail volumesVh 0.35, Vv 0.03Vh 0.30–0.60, Vv 0.02–0.05 for small UAVsin range
      Cruise speedwas 20 m/sbest lift-to-drag just above stallchanged

      analysis/benchmarks/literature.csv: Pollet et al. ICAS 2022-0383; Brandt and Selig AIAA 2011-1255; Russell et al. NASA 2016; Mathur and Atkins arXiv 2301.12316 and 2411.08661; Drela, MIT design rules (all checked 2026-10-26)

      What changed because of it

      • Adopted: cruise at 18 m/s instead of 20 m/s. Same pack, same wing, 22 % less cruise power and about three more minutes of cruise.
      • Kept: the conservative propeller, hover and transition figures, so the predicted margins are, if anything, understated.
      • Not changed: battery size (the mission does not need more), and the tail (inside the published range; the flight tests check handling).
      • Added to testing: the car-roof test records whether the stopped props park along or across the flow, a 25–30 % drag difference in the literature.

      0 Testing

      Open the test and evaluation plan, DBF-TP-001

      docs/test/TEST-PLAN.md, docs/test/test-matrix.csv, flight test cards in docs/test/cards/, data-reduction scripts in docs/test/analysis/

      SeriesTestsWhat it proves

      Weakest link: a lift-motor or ESC failure in hover, which C0 cannot survive (full ranking under Risks). The $69 test and safety kit qualifies every motor and ESC before flight.

      Ground tests

      Thrust-stand test plan

      #TestSetupPassRequirement
      T1Stand calibrationKnown masses 0–10 kg on the 20 kg load cellError ≤ 2 % of reading from 0.5 to 5 kg (2 % of full scale would be 400 g, a quarter of the hover point)—
      T2Lift thrust curve, each motorX4110S + 17 in, 5 steps, watt meterMax ≥ kgf at storage voltageR13
      T3Hover-duty heat g × 90 s, cool to 45 °C, repeat ×3Can ≤ 80 °C est after each run, no ESC cut-outR9, R10
      T4Pusher curve, 14×8.5EPX4110S, watt meterStatic ≥ 1.5 kgf @ 60 %R11
      T5Stopped-prop behaviourCar-roof mast at m/sNo windmillingR10
      T6System bench run4 lift + pusher, airframe tied downPack ≥ 3.4 V/cell at hover drawR10

      Mission duty cycle (C0)

      Electrical power through one mission

      R10 phase durations; P_hover and P_cruise from analysis/results.json; transitions at hover power est

      Motor-out behaviour

      EventC0 4-liftC1 8-liftC2 coax X8
      Lift motor out, hoverLoses roll/yaw authorityT/W on 6 balancedT/W
      Lift motor out, cruiseNo effect (stopped)SameSame
      Pusher out, cruiseQ_ASSIST below 12 m/s → QLANDSameSame
      Link lossRTL then QLAND (R20), §10

      Flight tests

      Flight test and operations

      Fly a strict build-up, F0 to F7, from simulator to full mission. Each step has entry, abort and success criteria. C0 has no hover motor-out tolerance, so early steps stay low and short.

      Build-up test matrix

      #StepEntry criteriaAbort ifSuccess
      F0ArduPilot SITL (QuadPlane, C0 parameter file)firmware/vtol_c0.param; scenarios A–G already pass—Every pilot flies scenarios A–G and recovers a failed transition
      F1Bench and tie-down runT1–T6 passed; CG measured (§9 M1)Any ESC desync or vibration warningAll outputs correct; motor directions; failsafe on link loss
      F2Tethered hoverF1; site request approvedOscillation, or attitude error > 15°30 s stable at 1 m; logs show < 60 % lift output
      F3Free hover, QLOITERF2; vibration logs cleanMotor > 80 °C after landing; any warning3 × 60 s at ≤ 5 m AGL, motors cooled below 45 °C between hovers
      F4Q-mode handling, low forward speedF3; tuning doneQ_ASSIST triggered unexpectedlyBox pattern at 5–8 m/s
      F5Partial transitionsF4; ≥ 30 m AGL; wind ≤ 5 m/s; start at the upwind edge, turn at once (SITL fence finding)Airspeed not rising within 6 sAccelerate to 12 m/s, back to hover
      F6Full transition and back-transitionF5 three timesQ_TRANS_FAIL (15 s)Fixed-wing circuit, back-transition, QLAND
      F7Full mission (R9, R10)F6 twicePack below the 20 % reserve10 min cruise, landing ≥ 20 % pack energy

      Crew, field procedure and charging

      • Crew roles: pilot in command (TRUST holder), visual observer, ground-station operator, and safety officer with the fire bag and extinguisher. Names: —.
      • Before each flight: props and CG checked; Remote ID broadcasting; geofence loaded; area clear; battery plugged in last through the AS150U.
      • LiPo: charge only in the LiPo bag inside the ammo can, never unattended. Check packs for swelling; store at storage voltage. Isolate a crashed pack outdoors for 30 min.
      • Field charging: the HTRC C150 gives about 6 A on 6S, roughly 1.1 h per 6500 mAh pack est, so two packs support about 3–4 flights in a 2–3 h session. Field power source: —.

      Risks

      Weakest-link ranking

      RankFailureLikelihoodSeverityMargin nowMitigation
      1Lift motor/ESC failure in hover (C0)medloss of aircraftnoneT2/T3 on every unit; hover ≤ 5 m AGL, ≤ 60 s
      2Hard landing, stiff gearhighmount/boom damage32 g at 20 mm (spring bound)Compliant feet, about 100 mm travel
      3BLHeli_S ESC overheatingmedpower loss A per ESC in hover vs 45 A ratedT3 with ESC temperature
      4Transition stallmedaltitude lossAIRSPEED_MIN 16 m/s vs stall m/s predictedAirspeed sensor; ≥ 30 m AGL; into wind
      5Flight-software configuration errorlowmode errorsParameter checker: 0 errorsSITL scenarios A–G before every parameter change (§12)
      6Wing joinerlowloss of aircraftRF on yield (hand calc on the true joiner, which governs; the idealised FEA gives )Proof load §6 M1

      Failure modes and controls

      What can go wrong in the shop or air, its effect, what prevents it, and the test that proves the prevention. The weakest-link table covers parts breaking; this one covers system failures.

      Failure modeEffectSeverityLikelihoodControlVerified by
      LiPo pack fire while charging or after a crashInjury, loss of the aircraft or the shopcriticallowCharge in the LiPo bag inside the ammo can, never unattended; crashed pack isolated outdoors for 30 min; AS150U disconnectG08 field procedure, I02 compliance
      Prop strike on a personInjurycriticallowBattery plugged in last; arming switch; props off for bench work; crew roles and keep-out distancesE02, F05, G08
      Radio link lostFly-awaycriticallowRC-loss failsafe: RTL, then QRTL inside 100 m; 110 m geofenceE04 range walk, E05, F10
      Transition fails to reach AIRSPEED_MIN, or the wing stallsLoss of height, possible crashcriticalmedQ_ASSIST below 13 m/s; transition cancelled after 15 s; pilots rehearse in the simulator firstE06, F06, F07
      Lift motor or ESC fails in hover (R14)Crash from low heightcriticallowHovers under 5 m and 60 s; every motor qualified on the thrust stand; ESC thermal soakC07, C08, G03
      Pusher motor or ESC fails in cruise (R15)Forced landingmarginallowQ_ASSIST brings the lift motors in and the aircraft lands verticallyE06, F10
      Battery voltage sags under the transition loadBrown-out, loss of controlcriticallowPack acceptance on internal resistance; battery failsafe on voltage and capacityC11, E05
      Wing joiner or spar fails in a pull-upCrashcriticallowDesigned to 4.5 g ultimate; proof load to the 3 g limit before flightS01, S07
      Centre of gravity out of range (no payload, no ballast)Pitch instabilitycriticallow0.5 kg ballast rule; weight-and-balance form signed before each sessionG01, G09
      Compass or GPS errorDrift in hover, poor position holdmarginalmedCompass calibration and compassmot; GPS mast away from the power wiringE07, G04

      firmware/vtol_c0.param and firmware/README.md (failsafes, transition limits); docs/test/test-matrix.csv (test IDs); docs/build/BUILD-MANUAL.md §1 (safety); severity and likelihood are engineering estimates est

      What the team must measure

      1. M1
        Temperatures after each flight. IR thermometer within 30 s of landing on every motor, ESC and the pack. Pack over 60 °C est, or any unit 15 °C hotter than its neighbours, grounds the aircraft (R9).
      2. M2
        Hover throttle in calm air. Mean lift output from a 30 s hover log. Above 60 % est, thrust is short of the maker curve and R13 is at risk: re-run T2 before transition tests.

      Risk register

      Likelihood and impact are scored 1 (low) to 3 (high). Score = likelihood × impact; residual = score after the response. Owners are build-manual roles (§2.4) until names are assigned.

      IDRiskLikelihoodImpactScoreTriggerResponseOwnerResidual
      K01Site approval delayed236No RFS approval by 2026-12-01Off-campus AMA field under its rulesIntegration lead3
      K02Crash during transition tests236Any airframe damageSimulator rehearsal first; spares on hand; re-order motor and props the same day; 12-day bufferIntegration lead4
      K03Lift motor lost in hover (R14 not met)133Any motor or ESC fault on the thrust standHovers under 5 m and 60 s; every motor qualified on the stand (C07)Electrical / avionics lead2
      K04Battery over 60 °C224Pack resistance over 6.0 mΩ per cell (C11), a pack warm before take-off, or a full mission above 25 °C6.0 mΩ acceptance (C11); no full mission above 25 °C (R21); shade the packsIntegration lead2
      K05Motor heat in hover224T3 over 80 °C60 s hover rule; ESCs in the prop wash; MN-class motor (over budget)Electrical / avionics lead2
      K06Stall above 12 m/s (R11)122F09 measures stall over 12 m/sRaise AIRSPEED_MIN and Q_ASSIST_SPEED by the difference; above 13.08 m/s treat as an R11 failureIntegration lead1
      K07Cart totals exceed allowance224Reserve under $100Borrow the hot-wire; drop the spare motor; then the airspeed sensorIntegration lead2
      K08Stopped props park across the flow212C10 car-roof test shows the prop windmilling or crosswisePrinted stop pawl; the 18 m/s cruise leaves energy margin for the extra dragStructures lead1
      K09Long-lead item late122Ship date after 2027-02-15Orders A and B early; US-stock alternativesIntegration lead1
      K10Flight software not supported (closed)122—ArduPlane 4.7.2 stable supports the board (R17; inspection E11)Electrical / avionics lead1

      What the team must measure

      1. M1
        SITL rehearsal pass. Each pilot flies F0 scenarios in ArduPilot SITL with the C0 parameter file. Any pilot who cannot recover a failed transition in SITL does not fly F5–F7 (R9).
      2. M2
        Field charge time. Time one full charge at the field supply. Over 1.5 h per pack est means fewer than 3 flights per session: plan more sessions before 2027-04-17 (R8).

      0 Budget and schedule

      C0 fits the fixed $2,000 cap; the 8-lift and coaxial layouts do not. Goods + tax (10.25 %) + shipping allowance = committed, leaving for crash repair. To keep that reserve above $100, v0.4 itemised every fastener (a $22 Bolt Depot order replaces the generic screw set) and dropped the hinge tape, spare servo and spare pushrods. FAA registration and AMA membership ($94) are fees, not hardware, so they are outside the cap.

      C0 committed
      goods
      C0 reserve
      target ≥ $100 (R7): met
      C1 8-lift
      over the cap
      C2 coax X8
      over the cap

      Budget

      Spend against the cap

      C0 spend by category, tax and shipping, and the reserve left under $2,000

      Side by side

      CategoryC0 4-liftC1 8-liftC2 coax X8

      analysis/results.json concepts.*.bom; tax 10.25 % (CDTFA, Berkeley, effective 2026-10-01) on all goods; shipping allowance est ($100 plus $10 for the Bolt Depot order, shown only at checkout): Amazon free, JeeFly $4.99; OnlineMetals UPS, SunnySky USA and Holybro (CN) shipping and duty are not shown until checkout

      Still unpriced (—), covered by the reserve: charger output adapter, 3.5 mm motor bullets if the X4110S ships without them, JeeFly and APC prop adapters. The packs' EC5 leads are re-terminated to the AS150U connectors in the BOM.

      Itemised BOM (C0)

      Purchasing copy by vendor: docs/purchasing/C0_BOM.csv.

      CategoryItemQtyUnit $Ext $Source / note

      analysis/sizing.mjs bom(); each line cites its research file and section

      Cost risks

      • Shipping over the allowance: OnlineMetals ships 72–96 in tube by UPS, priced only at checkout (M2).
      • A crash: a wing half plus a motor and props is about $120 est, more than the reserve. Spare foam, tube offcuts and a spare motor and props are already in the BOM; reserve cash covers a servo or pushrods.
      • China-shipped items (Holybro airspeed sensor, maybe the ESCs): allow for duty and longer lead time.
      • Prices are snapshots from 2026-10-13 (fasteners 2026-10-21): re-check before ordering.

      What the team must confirm

      1. M1
        Cart totals before ordering. Build each vendor cart to the shipping page without paying. If any cart pushes the reserve under $100 (R7), cut in order: borrow a hot-wire cutter ($44), drop the spare motor ($52), then the airspeed sensor ($66).
      2. M2
        OnlineMetals UPS quote for 2 × 72 in + 96 in + 36 in tube. Over $40 means buying the joiner tube locally.

      Schedule

      Schedule and compliance

      Plan: design freeze 2026-12-04, SITL rehearsal January–February, first hover 2027-03-13, full mission 2027-04-17, then a 12-day buffer. The site decision (by 2026-11-06) is the critical path. File flight requests for primary and backup dates together; each needs 7 business days. Academic calendar and winter-break availability are not yet reconciled (—).

      Milestones, 2026-10 to 2027-04

      Dates are this review's plan est; lead times from docs/research/budget-*.md

      Order-by dates

      OrderItemsLead time (source)Order byNeeded
      A, early test2 × X4110S + props, 1 ESC 4-pack, test kit, packs + chargerSunnySky USA (Ohio): —; JeeFly props 10–20 days (likely CN); Amazon Prime: days2026-11-02T1–T5 from 2026-11-16
      B, airframeAluminium tube, foam, tape, hardware kit, PETGOnlineMetals UPS: —; Home Depot El Cerrito in stock; Amazon Prime2026-12-07Build from 2027-01-04
      B, propulsion and avionics4 × X4110S, 2nd ESC pack, servos, FC, GPS, BeeID, radio, airspeedReadyMadeRC/RaceDayQuads US; Holybro ships CN: —2026-12-07Integration 2027-02-22

      Compliance checklist

      ItemRuleCostDueStatus
      Legal path§ 44809 (§ 350 higher-education note) or Part 107$0 / $1752026-11-06open
      TRUST, each pilot44809(a)(7)free2027-02-28open
      FAA registration + marking44809(a)(8)$5 (outside cap)2027-02-28open
      Remote ID broadcast module14 CFR 89.115(a); BeeID on FAA DoC list (RID000001995)$36.99 (in BOM)verify serial on receiptselected
      VLOS; ≤ 400 ft AGL in Class G44809(a)(3),(6)—every flightplanned
      Liability ≥ $1M per pilot (AMA Full: $2.5M)UC Berkeley UAS policy$89/yr per pilot (outside cap)2027-03-01open
      Flight request, each on-campus flight≥ 7 business days ahead—first by 2027-03-03open
      Post-flight reportwithin 3 business days—after each flightopen

      law.cornell.edu (49 U.S.C. § 44809; 14 CFR 89.115, 107.51); oercs.berkeley.edu UAS policy; FAA DoC list via docs/research/budget-avionics.md §4. All checked 2026-10-13.

      What the team must measure / confirm

      1. M1
        Field length. GPS-walk the usable straight at the approved site. Under 200 m est, or a ceiling under 30 m AGL, blocks R9 at that site.
      2. M2
        Actual lead times. Ship-date emails at order. Any item arriving after 2027-02-15 pushes first hover past 2027-03-13.
      3. M3
        Remote ID serial. Look up the received BeeID serial at uasdoc.faa.gov. If it isn't listed, flying outside a FRIA (FAA-recognised identification area) breaks 14 CFR 89.115.

      Design history

      Design iterations

      analysis/tradestudy.mjs → analysis/tradestudy.json (physics: analysis/model.js, unchanged); literature survey and references: docs/research/configurations.md

      Weighted score of every configuration studied, out of 10; the C0 baseline is marked, and the C0 variants above it spend margins the baseline keeps

      Review findings log

      Every finding from reviews, model checks and the benchmark study, and what was done. Reviewers: start here to check earlier comments were acted on.

      IDIssueRaised byFindingWhat was doneStatus
      N011Six-reviewer project audit79 confirmed findings, among them the pusher disc overlapping the rear lift discs by 30 mmv0.3.1: booms outboard, joiner 0.76 m, NACA 2415 with the spar at 30 % chord, fifth servo, PA6-CF motor mounts, missing items addedclosed
      N023Aerodynamic drag build-upBare landing legs would make the aircraft fail R10Printed teardrop fairings on all four legsclosed
      N034Drawing checkText overlapping the views and too many balloonsDrawing rev C, then rev E on the final wingclosed
      N045Model consistency checkHand-copied sizing inputs had drifted; stall came out at 12.05 m/sSizing now reads the aerodynamic package; wing sized to an 11.75 m/s design stallclosed
      N055Thermal analysis reviewThe pusher was modelled as the wrong motorRe-run with the X4110S; unsupported claims removedclosed
      N065Independent code and accessibility reviewFocus, history, search and theme problems on the review pageAll fixed and re-checkedclosed
      N075Automatic check (CI)Committed review page did not match its sources on a clean buildRender images committed; the build no longer depends on file timesclosed
      N086Benchmark study20 m/s cruise is far above the best lift-to-drag speedCruise moved to 18 m/s, the R11 lower bound: 22 % less cruise powerclosed
      N096Document checkPrint list still asked for pusher-cup vent slots after the thermal decisionRemoved; the build measures the pusher firstclosed
      N105Thermal analysisBattery could pass 60 °C at the end of a mission on a 35 °C dayResolved by requirement: packs accepted at ≤ 6.0 mΩ per cell (C11) and no full mission above 25 °C (R21) give 53 °C at mission end; confirm by measurementopen
      N111Requirement R14C0 cannot hold attitude after losing a lift motor in hoverAccepted with operating limits; the 8-lift layout that meets it is over the capaccepted
      N123Requirement R4The car has not been measured against the largest pieceTest S08open

      Revision record (Report info); docs/research/benchmarks.md; analysis/verify.py; git history

      How the score is built

      Each variant is sized with C0's physics, priced from real parts and scored 0–10 on eight criteria. Variants over the $2,000 cap or short of a hard requirement are shown but not ranked. Tailsitters and tilt-rotors break R1 (lift rotors stopped in cruise), so they are ranked separately.

      CriterionWeight10 points means

      Every variant with its eight criterion scores
      VariantTotalMTOW kgCruise minReserve $Status

      analysis/results.json concepts.*; docs/reference/lift-cruise-reference.png (Alia = 4-lift reference, "Lift Plus Cruise" = 8-lift reference)

      Considered and dropped

      • 6-lift (hex): on two booms the middle rotor lands on the wing; costs like C1.
      • Tilt-rotor, tailsitter: outside R1, which specifies stopped lift rotors.

      Morphological chart

      FunctionOption A (C0)Option BOption C
      Lift layout4 rotors, 2 booms8 rotors, 4 booms4 coaxial pairs, 2 booms
      Lift motorSunnySky X4110S 17 inSunnySky X3515S 16 inSunnySky V3508 15 in (8-lift)
      ESCBLHeli_S 45 A (no prop stop)Hobbywing X-series integratedLocking ESC (VL5009, over budget)
      Pack6S LiPo 6500 mAh6S LiPo 8000 mAh6S Li-ion 21700
      Booms / spar6061-T6 aluminium tubeRoll-wrapped carbonMixed (Al booms, carbon sockets)
      Wing skinPacking tape over XPSGlass + epoxyLaminating film
      TailH-tail on twin boomsInverted VConventional (pusher conflict)
      Flight controllerCoreWing F405 Wing V2Matek F405-WMOPixhawk 6C (over budget)
      Prop parkingFixed blades + stop pawl if neededFolding propsESC position lock

      docs/research/*.md; Option A (bold, highlighted) is the C0 choice

      Pugh matrix (datum C0 = 0)

      CriterionEvidence (C0 / C1 / C2)C0 4-liftC1 8-liftC2 coax X8

      + better than datum, − worse, 0 same; values from analysis/results.json

      What the team must measure

      1. M1
        Stopped-prop attitude. Car-roof mast, one motor and prop unpowered, m/s on a closed lot; film the prop. Windmilling, or a stop more than 30° off the airflow, adds drag beyond the R10 estimate: fit a stop pawl, or folding props ($110 with tax).

      0 Build and software

      Open the build manual, DBF-MAN-001

      docs/build/BUILD-MANUAL.md (work packages, steps, torques, checks, vendor links); docs/research/hardware-spec.md and cad/hardware.json (fasteners, holes, fits)

      Build

      Work packages

      WPWork packageLabour, person-hMachine time

      The cut, print and wiring lists, templates, fastener data and the drawing PDF are under Downloads.

      Fabrication notes

      • Wing cores: hot-wire cut FOAMULAR on printed NACA 2415 root and tip templates; socket channel cut with a heated tube; packing-tape skins.
      • Prints: 4 motor mounts (PA6-CF, needs a hardened nozzle — check the printer), 4 boom clamps, 4 gear clamps, pod frames, servo mounts (PETG). About h of printer and oven time (print list, build manual WP2).
      • Build order: wing cores and sockets → booms and clamps → tail → pod and avionics → wiring harness from the CAD positions → CG and proof loads (§6).
      • Ownership: assign an owner to each workstream (§1) and each M-item. Course deliverable dates still need adding to the §11 schedule (—).

      Flight software

      Controls and flight software

      The flight software is written, checked and rehearsed: ArduPlane 4.7.2 stable QuadPlane on the CoreWing F405 Wing APP V2, with a 129-parameter file that passes the checker and seven simulated failure scenarios. The board has been in a stable release since 2026-10-11, closing the earlier beta risk. Lift and pusher motors run DShot150; the board has no Lua scripting. Transition completes at 16 m/s, Q_ASSIST (lift motors help below this airspeed) is 13 m/s, and the geofence is 110 m, 12 m under the 400 ft limit. Telemetry needs no extra radio: CRSF to the handset, or ELRS MAVLink to a laptop.

      Download the parameter file Open the setup guide, DBF-SW-001

      The file saves as vtol_c0.param.txt; Mission Planner loads it as is (choose "All files"). firmware/README.md (flash, calibrate, first flight), firmware/vtol_c0.param + overlays, firmware/tools/check_params.py, firmware/sitl/; decisions and sources: docs/research/firmware.md (checked 2026-10-21)

      Flight controller choice

      BoardOutputsCurrent sensorArduPilot target$Verdict
      CoreWing F405 Wing APP V21290 A cont / 215 A peak4.7.2 stable67.26Chosen
      Matek F405-WMO9–10"132 A" vs "13.2 A" on pagestable69.99Fallback
      Holybro Kakute H743-Wing14120 Astable, full features105.00Needed only for C2 (OctaQuad)
      Holybro Pixhawk 6C Mini (v0.2)14external PMstable179.00Over budget with PM08

      Flight-mode state diagram

      QuadPlane modes and transitions (C0)

      Transition and threshold table

      ParameterValueBasis
      Q_FRAME_CLASS / Q_FRAME_TYPE1 Quad / 1 XIncluded in F405 builds
      AIRSPEED_STALL12 m/sR11
      AIRSPEED_MIN (transition complete)16 m/sR11: ≥ 1.3 × 12 = 15.6 m/s (docs suggest ≥ 20 % above stall)
      AIRSPEED_CRUISE m/sR11
      Q_ASSIST_SPEED13 m/sDocs: 3 m/s below AIRSPEED_MIN with a sensor (5 m/s without)
      Q_TRANSITION_MS5000Default
      Q_TRANS_FAIL / ACT15 s / QRTLDefault 0 = no limit; 15 s est; SITL scenario D gave up at 15 s and landed at home
      Minimum transition altitude30 m AGLPX4 guidance ≥ 20 m + margin
      Geofence ceiling110 m AGL12 m under the 400 ft (121.9 m) limit, § 44809(a), R18; SITL climb peaked at 113.1 m
      Low / critical batteryRTL / QLAND21.0 V or 1950 mAh used → RTL ending in a vertical landing; 19.8 V or 1300 mAh left (the R10 reserve) → land in place. Per-cell values est
      RC link lossRTL / QLANDFixed-wing modes return and land vertically at home; hover modes land in place (SITL B, C)
      Hover limits (test phase)≤ 5 m AGL, ≤ 60 sR14 mitigation; motor heat (§12)
      Arming / disconnect (R20)RC arm switch + Amass AS150UBattery plugged in last, props off for bench work; ArduPilot arming checks on
      Lift-motor current limitQ_M_BAT_CURR_MAX 85 AThrottles only the lift motors, so a lower cap would cut lift in transition; take-off throttle capped at 75 % (TKOFF_THR_MAX)

      Outputs and wiring (C0)

      Output / portConnectsNotes
      S1, S2Aileron L, aileron RES3005HV at 6 V
      S3ElevatorES3005HV
      S4Rudder LES3005HV
      S5, S6Lift motor 1 front-right (CCW), motor 2 rear-left (CCW)DShot150; S5–S8 share timers, so lift motors sit together
      S7, S8Lift motor 3 front-left (CW), motor 4 rear-right (CW)DShot150
      S9Pusher ESCDShot150
      S10Unused—
      S11Rudder RPWM; away from the DShot timer groups
      S12Spare—
      UART + I²CRMRC SAM-M10Q-5883 GPS + compassu-blox M10
      UARTRP3 V2 ELRS receiverCRSF; or MAVLink with the overlay file (one or the other, not both)
      I²CRMRC MS4525DO digital airspeed + pitotI²C address 0x28, no clash with the board barometer at 0x76 (ARSPD_TYPE 1, firmware/vtol_c0.param)
      5 V padBeeID Remote ID broadcast moduleFAA DoC RID000001995

      firmware/vtol_c0.param SERVOn_FUNCTION and SERIALn_PROTOCOL; output timer groups from the board's ArduPilot hwdef (firmware/README.md §2); wiring: docs/build/wiring.csv

      Software-in-the-loop rehearsal

      SITL proves modes, thresholds and failsafes, not gains or performance: it flies ArduPilot's generic QuadPlane, not C0. ArduPlane 4.7.2 SITL loaded the C0 parameter file (133 parameters read back unchanged, except three ESC-passthrough ones SITL lacks).

      #ScenarioResult
      AHover take-off, AUTO loop, vertical landingTransition done 10 s after entering AUTO; landed at home
      BRC loss in fixed-wing flightRTL, then vertical landing at home
      CRC loss in hoverQLAND in place
      DForward transition with no pusher thrustGave up at 15 s, returned and landed (QRTL)
      ELow battery in fixed-wing flightRTL, then vertical landing at home
      FClimb through the altitude fenceBreach at 110 m, peak 113.1 m
      GSticks over MAVLink only, link lostQLAND 1–2 s after the overrides stopped

      firmware/sitl/README.md, firmware/sitl/mission_test.py, firmware/sitl/results/ (ArduPlane V4.7.2 Windows SITL binary)

      A straight transition in SITL hit the 300 m fence radius about 6 s after transition, so real transitions start at the field's upwind edge with an immediate turn.

      What the team must measure

      1. M1
        Airspeed calibration. Constant-throttle circuit, compare against GPS on reciprocal legs (or ARSPD autocal). An error over 10 % puts the true transition speed at 14.4 m/s, below R11's 15.6 m/s.
      2. M2
        Transition distance and time. Log groundspeed from Q-mode exit to AIRSPEED_MIN at 30 m AGL, into wind. Over 15 s or 100 m means the pusher is short (§5 M3) and R9 is at risk.
      3. M3
        Servo load. Spring-scale hinge moment on a built aileron. Over 4 kg·cm (half the ES3005HV's 8 kg·cm at 6 V): fit a stronger servo (R9 control authority).

      0 Sources and files

      Where the numbers come from. Key sources are the ones the design depends on most; the full register lists every link cited in the project files. The search box filters both.

      Sources

      Key sources

      SourceWhat it supports in the designChecked

      Full register

      — sources by topic. Open a group for its links, check dates and citing files. A date marked * is stated once by the citing file for all its links.

      analysis/sources.py scans the repository for cited URLs → analysis/sources.json; key sources: analysis/sources_key.json. Regenerated by CI on every push.

      Files

      Downloads

      Companion documents generated from the repo, and the shop files the build team works from.

      Build manual, DBF-MAN-001 Test procedures, DBF-TP-001 Flight software Hot-wire templates

      Shop files

      Glossary

      MTOW
      Maximum take-off mass, with payload and battery.
      T/W
      Thrust-to-weight ratio of the lift motors at full throttle.
      FM·η
      Figure of merit times motor and ESC efficiency: how well electrical power becomes hover thrust.
      CdA, CD0
      Drag area of a part (m²), and the aircraft's drag coefficient at zero lift.
      CL_max, stall
      Highest lift coefficient the wing reaches; it sets the slowest speed the wing can fly.
      MAC, static margin
      Mean aerodynamic chord; the distance from the centre of gravity back to the neutral point, as a share of it. Positive means stable in pitch.
      QuadPlane, Q modes
      ArduPilot's lift + cruise vehicle type and its hover modes (QHOVER, QLOITER, QLAND, QRTL).
      SITL
      Software in the loop: the real autopilot code flying a simulated aircraft on a computer.
      AGL
      Height above ground level.
      § 44809
      The US law covering recreational and educational drone flying (49 U.S.C. § 44809).
      R, M, EST
      R1–R20 are the numbered requirements; M items are measurements the team still has to make; EST marks an engineering estimate.
      Esc