← Design review DBF-RPT-001DBF-SW-001

C0 flight software package: from a blank flight controller to the first transition

ArduPlane QuadPlane on the CoreWing F405 Wing APP V2. This package holds everything the team needs to program and configure each programmable part of the C0 lift + cruise VTOL, in the order the work is done. Every number has a source key; the full source list with check dates is in docs/research/firmware.md. Values marked (est.) are engineering estimates. "—" means the value is unknown.

Status, 2026-10-21. The parameter file passes the checker against the current ArduPlane metadata and the 4.7.1 stable metadata. All seven SITL scenarios passed on ArduPlane 4.7.2 (§11). Nothing here has run on the real board yet. Treat every gain as a starting value.

FileWhat it is
vtol_c0.paramFull parameter set for the aircraft, with a comment and source on every line
vtol_c0_hover_phase.paramOverlay for hover-only sessions: 10 m fence, no fixed-wing modes on the switch
vtol_c0_elrs_mavlink.paramOverlay: switch the ELRS receiver from CRSF to MAVLink mode for a $0 Mission Planner link
tools/check_params.pyValidator: syntax, names, ranges and the CoreWing DShot timer groups
sitl/SITL overrides, the scripted mission mission_test.py, results from 2026-10-21 (generic QuadPlane model) and 2026-10-25 (results_c0/, C0 airframe model)
sitl/c0_physics.py6-DOF physics model of C0 for ArduPilot SITL's JSON interface (reads mass, geometry and aero from analysis/ and cad/); --validate prints the trim checks
sitl/run_c0_sitl.ps1Starts the C0 model + SITL for hand-flying from Mission Planner (TCP 127.0.0.1:5760)
peripherals.mdESCs (BLHeli_S / Bluejay), ELRS receiver and transmitter, GPS/compass, airspeed, Remote ID, telemetry
checklists.mdBench, pre-flight, failsafe-test and post-flight log-review checklists
scripts/README.mdWhy there are no Lua scripts: this firmware build does not include scripting

0. Verified facts this package relies on

FactValueSource (checked 2026-10-21)
BoardCoreWing F405 Wing APP V2, STM32F405 with 1 MB flashBOM; hwdef FLASH_SIZE_KB 1024
ArduPilot targetCoreWingF405WingV2 (hwdef folder exists on ArduPilot master)github.com/ArduPilot/ardupilot libraries/AP_HAL_ChibiOS/hwdef/CoreWingF405WingV2
Stable firmwareArduPlane 4.7.2 stable now includes this target. It was beta-only in the review; that risk is closedfirmware.ardupilot.org/Plane/stable/CoreWingF405WingV2/firmware-version.txt and git-version.txt ("Plane: version to 4.7.2", 2026-10-11)
QuadPlane in the buildyes (HAL_QUADPLANE_ENABLED); Quad, Hexa and Octa framesfeatures.txt (stable); hwdef include minimize_fpv_osd.inc
Lua scriptingnot in the build (!AP_SCRIPTING_ENABLED)features.txt (stable)
DShotoutputs 1–10. Bi-directional DShot only on 1–4. Outputs 11–12 are PWM onlyhwdef PWM lines; board README; ardupilot.org board page
Output groupsS1-2, S3-4, S5-7, S8-10, S11-12. "If any DShot-capable output in a group uses DShot then all DShot-capable outputs in that group need to use DShot"hwdef README
Battery-power limiter BATT_WATT_MAXnot in the build (!AP_BATTERY_WATT_MAX_ENABLED)features.txt (stable)
MAVLink-radio RC input (RADIO_RC_CHANNELS)not in the build (!AP_RCPROTOCOL_MAVLINK_RADIO_ENABLED)features.txt (stable)
Gyro FFTnot in the build (!HAL_GYROFFT_ENABLED). Use the throttle-based notchfeatures.txt
QAUTOTUNE / QuickTuneboth in the buildhwdef (QAUTOTUNE_ENABLED 1); features.txt AP_QUICKTUNE_ENABLED
BarometerSPL06 at I²C bus 0, address 0x76 — the only I²C bus, shared with the external I²C/airspeed connectorhwdef BARO SPL06 I2C:0:0x76, I2C_ORDER I2C1
Lift/pusher ESCsSoloGood 45A BLHeli_S, "Support D-Shot 150/300/600". ESC MCU: —BOM; amazon.com/dp/B0DJQVGSBC listing title
RC receiverRadioMaster RP3 V2, ESP8285, CRSF output, ships with ELRS 3.0BOM; radiomasterrc.com product page

1. Install the tools (Windows)

  1. Mission Planner (primary ground station; parameter files, motor test, logs): installer firmware.ardupilot.org/Tools/MissionPlanner/MissionPlanner-latest.msi, guide ardupilot.org/planner/docs/mission-planner-installation.html.
  2. QGroundControl (optional second ground station): docs.qgroundcontrol.com/master/en/qgc-user-guide/getting_started/download_and_install.html. This package's parameter files use the Mission Planner/MAVProxy NAME,VALUE format. Load them with Mission Planner.
  3. STM32CubeProgrammer (needed once, for the first DFU flash): st.com/en/development-tools/stm32cubeprog.html. The ArduPilot page says it installs the DFU driver and may need Java (ardupilot.org/plane/docs/common-loading-firmware-onto-chibios-only-boards.html).
  4. ESC Configurator (web, Chrome/Edge): esc-configurator.com — BLHeli_S/Bluejay settings through the FC.
  5. ExpressLRS Configurator: github.com/ExpressLRS/ExpressLRS-Configurator/releases — receiver firmware and bind phrase.
  6. Python 3 + pymavlink for the tools in this folder: py -m pip install pymavlink. The checker uses only the standard library.

2. Firmware target and files

  • Target folder: firmware.ardupilot.org/Plane/stable/CoreWingF405WingV2
    • arduplane_with_bl.hex — bootloader plus firmware. First flash only, over DFU.
    • arduplane.apj — later updates from Mission Planner ("Install Firmware" > "Load custom firmware").
  • Do not use CoreWingF405WMiniV2. That target is for the Wing Mini V2, which has a different current-sensor scale (BATT_AMP_PERVLT 51 vs 64, board page).
  • Pin the firmware version for the season. Record ArduPlane V4.7.2 (git hash) from the Mission Planner Messages tab in the build log. Retest in SITL before any upgrade.

3. First flash (DFU)

The board ships without an ArduPilot bootloader (hwdef README "Loading Firmware").

  1. Remove the props. Unplug the flight battery. Leave only USB connected.
  2. Enter DFU mode. The board has no boot button. The hwdef README says to "pull the BOOT0 pin on the USB extender connector low while powering the board over USB, or hold the BOOT button on the wireless USB extender module if one is fitted". The included wireless module has that BOOT button, so use it. (On STM32 parts, BOOT0 is normally pulled high for DFU. The extender may invert it. Check the CoreWing Wing V2 user manual linked from the ArduPilot board page before bridging any pin.)
  3. In Device Manager, check for "STM32 BOOTLOADER" under Universal Serial Bus devices.
  4. STM32CubeProgrammer: connection USB > Connect > Open file arduplane_with_bl.hex > Download.
  5. Power-cycle. In Mission Planner, connect over USB (COM port, 115200, or auto). Messages should show ArduPlane V4.7.2 and the board name.
  6. Later updates: Mission Planner > Setup > Install Firmware > "Load custom firmware" > arduplane.apj (or pick Plane from the list once the board is detected).

4. Wiring (FC pad → device)

Pad names come from the hwdef README and the ArduPilot board page. Before soldering, check every pad on the board's own pinout images: the GPS pads and the GPS connector use different pin orders (hwdef README).

4.1 Signal and power

FC pad / connectorDeviceNotes
BAT+/BAT− (board power input, through the 90 A sensor)6S pack via XT90-S anti-sparkBoard input 3–6S, 10–28 V. Fit the included 35 V 470 µF capacitor on the input (board page)
S1 headerAileron servo, leftServo rail powered by the onboard servo BEC (8 A cont / 14 A peak). Set the rail to 6 V: the ES3005HV has no 5 V rating (budget-avionics §6)
S2 headerAileron servo, right
S3 headerElevator servo
S4 headerRudder servo, left boom
S5 headerLift ESC 1 — front-right, CCW (signal + GND only)DShot150. Do not connect an ESC BEC red wire to the servo rail (board page)
S6 headerLift ESC 2 — rear-left, CCW
S7 headerLift ESC 3 — front-left, CW
S8 headerLift ESC 4 — rear-right, CW
S9 headerPusher ESC (SoloGood 45A)DShot150 via SERVO_BLH_MASK/OTYPE
S10 header— leave emptyIn the DShot group with S8–S9
S11 headerRudder servo, right boomPWM-only output
LED pad (PWM12)sparePWM only; no serial LED (no DMA, hwdef README)
TX6 / RX6 (+5 V, GND)RP3 V2 ELRS receiver: RX6 ← receiver TX, TX6 → receiver RXSERIAL6 = CRSF (default)
GPS connector (UART5 + I²C + 4V5)HGLRC M100-5883: GPS TX → R5, GPS RX → T5, SDA/SCL → I²C4V5 is also live on USB, which makes bench work easier
I²C airspeed connectorAirspeed sensor (see §4.3)Same bus as the barometer
5 V pad (FC BEC, 5.2 V / 4 A)NewBeeDrone BeeID Remote ID modulePower only; BeeID uses its own GPS. Do not use the 4V5 USB rail
TELEM2/4 (GH1.25 6-pin: GND RX2 TX2 RX4 TX4 4V5) or pads T4/R4spare, reserved for a SiK telemetry radioSERIAL4 = MAVLink2
USB extender connectorCoreWing BLE/WiFi moduleSERIAL1 = MAVLink2 at 115200 (bench and field laptop link)

4.2 UART assignments

SERIALnUARTPadsProtocol in vtol_c0.paramDevice
0USBUSB-CMAVLink2 (fixed)Laptop
1USART1extender connector / T1, R12 = MAVLink2, 115200CoreWing WiFi/BLE module
2USART2SBUS pad (inverted RX)— (unused)—
3USART3VTX connectors−1 = none—
4UART4TELEM2/4, T4/R42 = MAVLink2spare (SiK radio)
5UART5GPS connector5 = GPSHGLRC M100-5883
6USART6TX6/RX623 = RCINRP3 V2 (CRSF)

4.3 Airspeed sensor: RMRC MS4525DO on the shared I²C bus

The BOM buys the RMRC digital airspeed sensor MS4525DO + pitot (rev 2026-11-01, docs/research/bom-savings.md; it replaced the Holybro MS5525DSO, whose 0x76 address clashed with this board's SPL06 barometer). The board has one I²C bus (hwdef BARO SPL06 I2C:0:0x76, I2C_ORDER I2C1), shared by the barometer and the airspeed connector.

The MS4525DO comes in address variants. ArduPilot's driver probes all three: MS4525D0_I2C_ADDR1 0x28, ADDR2 0x36, ADDR3 0x46 (libraries/AP_Airspeed/AP_Airspeed_MS4525.cpp, master, checked 2026-11-02). None is the barometer's 0x76, so no address strap is needed whichever variant RMRC ships (which one: — until the part label is read).

Setting used: ARSPD_TYPE,1 (I2C-MS4525D0), ARSPD_BUS,0. Bench check before flight:

  1. Power the FC on USB with the airspeed sensor plugged in. Mission Planner must show an airspeed reading, and the barometer must still read altitude (HUD). If either is missing, do not fly.
  2. Blow gently into the pitot: airspeed must rise. Then do the offset calibration (§6.8).

4.4 Output map (SERVOn_FUNCTION) and why it differs from the review

OutputGroupFunctionValue
S1, S21 (TIM4)Aileron L, R4
S32 (TIM3)Elevator19
S42Rudder L21
S5, S6, S73 (TIM8)Motor1, Motor2, Motor3 (DShot150)33, 34, 35
S84 (TIM2)Motor4 (DShot150)36
S94Throttle / pusher (DShot150)70
S104unused0
S115 (TIM1)Rudder R21
S12 (LED)5unused0

The review (§[controls] "Outputs and wiring") puts the lift motors on S7–S10 and the rudders on S5–S6. S7 shares timer group S5–S7 with S5 and S6, so DShot on S7 would force both rudder outputs into DShot. The map above keeps every servo out of a DShot group. It also matches ArduPilot's QuadPlane convention of lift motors on outputs 5–8 (quadplane-frame-setup.html). tools/check_params.py --corewing checks this automatically.

5. Load the parameters

  1. Validate first: py firmware/tools/check_params.py firmware/vtol_c0.param --corewing --require-every-line-commented.
  2. Mission Planner > Config > Full Parameter List > Load from file > vtol_c0.param > Write Params.
  3. Reboot. Q_ENABLE and the DShot settings (Q_M_PWM_TYPE, SERVO_BLH_*) take effect only after a reboot, and the Q_ parameters do not exist until then.
  4. Load the same file a second time, Write Params, and reboot again. Then use "Compare Params" against the file: every row should match.
  5. Hover-only sessions: also load vtol_c0_hover_phase.param. Reload vtol_c0.param before the first transition session.
  6. Save the full parameter list (Save to file) after each session as params_<date>.param in the team log.

6. Calibrations (props off, battery on only where stated)

#StepHowPass condition
6.1Board orientationFC arrow forward, top up → AHRS_ORIENTATION,0HUD horizon matches the airframe
6.2AccelerometerSetup > Mandatory Hardware > Accel Calibration (6 positions), then "Calibrate Level" with the wing at cruise attitude (common-accelerometer-calibration.html)"Calibration successful"
6.3CompassSetup > Compass > Start, rotate the assembled airframe on all axes. On a 2.4 m aircraft, use "Large Vehicle MagCal" with GPS lock and the nose pointed at a known heading (common-compass-calibration-in-mission-planner.html). COMPASS_AUTO_ROT,2 detects the module orientationFitness passes; the heading matches a phone compass within about 10° (est.)
6.3aCompass vs power currentThe pack trunk passes about 65 mm (EST) under the canopy-top GPS: about 157 µT at hover current for a single untwisted wire (docs/research/team-crosscheck.md X5). Twist the pack leads. Then, tethered with props on, log the MAG field magnitude with the lift motors off and at hover current. Use CompassMot if Mission Planner offers it for this firmware (Initial Setup > Optional Hardware > Compass/Motor Calib; procedure on ardupilot.org/copter common-compass-setup-advanced.html, checked 2026-11-03)Interference < 30 % ("acceptable"); 31–60 % is ArduPilot's "grey zone"; > 60 %: move the GPS away from the trunk
6.4RadioSetup > Radio Calibration: move all sticks and switches to their ends (common-radio-control-calibration.html). Check CH5 arm, CH6 six modes, CH7 E-stop, CH8 QAssist, CH9 RTLModes 1–6 show QSTABILIZE, QHOVER, QLOITER, FBWA, CRUISE, QLAND
6.5ESCsNo throttle calibration under DShot. DShot always uses 1000–2000 (common-dshot-escs.html). Configure ESC settings via passthrough, peripherals.md §1All 5 ESCs read and save in ESC Configurator
6.6Battery voltageMeasure the pack with a multimeter and adjust BATT_VOLT_MULT until the GCS matches within 0.05 V (est.)—
6.7Battery currentWatt meter in series. Hold one lift motor at a steady load (props on, aircraft tethered) and adjust BATT_AMP_PERVLTGCS current within 5 % of the watt meter (est.)
6.8Airspeed offsetCover the pitot loosely (foam) during boot or press "Calibrate Airspeed" (calibrating-an-airspeed-sensor.html)Reads 0 ± 1 m/s (est.) still air

7. Motor order, direction and servo checks

Motor test (props OFF, battery ON, aircraft restrained). Mission Planner > Setup > Optional Hardware > Motor Test, 5–10 % throttle, 2 s. Buttons go clockwise from front-right (copter/docs/connect-escs-and-motors.html):

ButtonMust spinOutputDirection (seen from above)
AFront-rightS5 (Motor1)CCW
BRear-rightS8 (Motor4)CW
CRear-leftS6 (Motor2)CCW
DFront-leftS7 (Motor3)CW
  • Wrong motor spins: fix the signal wire at the header, not the parameters.
  • Wrong direction: swap two of the three motor phase wires, or reverse it in ESC Configurator (peripherals.md §1).
  • With the props on and the aircraft tethered, find the lowest test % that spins all four cleanly → Q_M_SPIN_ARM. The lowest that gives a little thrust → Q_M_SPIN_MIN (quadplane-vtol-tuning-process.html step 4).
  • Props: check CW/CCW prop handedness against motor direction. Air must blow down.

Servo direction checks (FBWA or MANUAL, disarmed, rail at 6 V):

InputCorrect surface motionIf wrong
Right roll stickright aileron up, left aileron downSERVO1_REVERSED / SERVO2_REVERSED
Pull back (pitch up)elevator trailing edge upSERVO3_REVERSED
Right yaw stickboth rudders trailing edge rightSERVO4_REVERSED / SERVO11_REVERSED
Stabilization (FBWA, sticks centered): roll the airframe rightailerons command a roll leftFix the reversals above, never RLL2SRV signs
Pitch the nose downelevator moves upas above

Then set throws: SERVOn_MIN/MAX/TRIM so that no linkage binds at full stick. Check that the 5 servos together stay under the 8 A servo BEC (budget-avionics M3).

Pusher check: QSTABILIZE, armed, props on, tethered. The pusher must stay stopped in Q modes. In FBWA the throttle stick drives it. Air must blow aft.

8. Failsafe configuration and bench tests

EventMode on eventParametersBench test (props off)
RC loss in a Q modeQLAND in place (ArduPilot forces this in QSTABILIZE, QHOVER, QLOITER)THR_FAILSAFE,1; Q_OPTIONS bit 5 clear (apms-failsafe-function.html)Arm in QLOITER, switch the TX off → GCS shows "RC Short Failsafe: switched to QLand"
RC loss in fixed-wing modesCIRCLE (short), then RTL at 5 s, then QRTL inside RTL_RADIUS 100 m, then vertical landFS_SHORT_ACTN,0; FS_LONG_ACTN,1; FS_LONG_TIMEOUT,5; Q_RTL_MODE,1SITL scenario B
Battery low (25 %, or 21.0 V sag-compensated)RTL → QRTLBATT_LOW_MAH,1625; BATT_LOW_VOLT,21.0; BATT_FS_LOW_ACT,1Set BATT_LOW_VOLT above pack voltage → warning + mode change; restore
Battery critical (20 %, or 19.8 V)QLANDBATT_CRT_MAH,1300; BATT_FS_CRT_ACT,4as above
Forward transition not done in 15 sQRTLQ_TRANS_FAIL,15; Q_TRANS_FAIL_ACT,1SITL scenario D
Geofence: 110 m above home or 300 m radiusRTL (→ QRTL near home)FENCE_*SITL scenario F; on the bench, FENCE_RADIUS shows on the map
Motor emergency stopLift motors and pusher stop at onceRC7_OPTION,31Arm, flip CH7 → all motors stop
ArmingOnly from the CH5 switch, only in VTOL modes, only with a charged pack (≥ 23.4 V)RC5_OPTION,153; ARMING_RUDDER,0; Q_OPTIONS,262144; BATT_ARM_VOLT,23.4Try to arm in FBWA → refused

Receiver failsafe setting: set the receiver to output no pulses on link loss. ArduPilot recommends this over a preset failsafe throttle (apms-failsafe-function.html). Check the RP3 V2 failsafe option in its WiFi/Lua settings (peripherals.md §2). Do not set a failsafe position on the handset.

9. First-flight build-up

Fly at a site that allows the test (review §[schedule]). Hover tests: ≤ 5 m AGL, ≤ 60 s per hover (review §[controls], R14). Transitions: ≥ 30 m AGL, into wind. Wind ≤ 5 m/s, gusts ≤ 8 m/s (R16). The full checklists are in checklists.md.

9.1 SITL rehearsal (every pilot, before any flight)

Use the C0 airframe model, not the generic QuadPlane: sitl\run_c0_sitl.ps1 -SitlExe <ArduPlane.exe> starts sitl/c0_physics.py (a 6-DOF model built from the repo's mass, geometry, AVL aero, motor and prop data) and ArduPlane 4.7.2 SITL on its JSON physics interface; connect Mission Planner to TCP 127.0.0.1:5760 and fly with the Pocket Crush as a USB joystick: QSTABILIZE take-off, QLOITER, FBWA transition, CRUISE, QLOITER back-transition, QLAND. Practise the E-stop and RTL switches. Expect what the model shows (sitl/README.md, C0 section): hover at about 0.62–0.66 output, 37–38 A; forward transition 13–15 s and 140–170 m; a 30° nose-up, stalled-wing "airbrake" phase at the start of every back-transition; cruise at 18 m/s (AIRSPEED_CRUISE; SITL re-run 2026-10-27) with the pusher at about 0.6. sitl/mission_test.py --model JSON:127.0.0.1 --physics sitl/c0_physics.py ... runs the seven scripted scenarios (§11). Add -Wind "5,270" (5 m/s from the west) to rehearse the R16 wind limit.

9.2 Tethered bench run

Props on, aircraft tied down to a heavy base, battery on, vtol_c0_hover_phase.param loaded. QSTABILIZE, arm, raise throttle to light the skids. Check for oscillation, vibration and current (≈ 41.2 A hover at full weight; results.json C0.sized.I_hover).

9.3 First lift-off (QSTABILIZE): tuning-process steps 5–6

  1. Arm, disarm at once (proves the disarm path), arm again.
  2. Raise the throttle slowly. At lift-off, set it straight back down. Disarm and look for oscillation.
  3. If it oscillated: reduce Q_A_RAT_RLL_P/I/D and Q_A_RAT_PIT_P/I/D by 50 % and repeat (tuning-process step 6).
  4. Hover at about 1 m and give 5° roll/pitch inputs. Land at any oscillation.

9.4 QHOVER, hover throttle, notch

  1. Take off in QSTABILIZE to 5 m, switch to QHOVER for ≥ 30 s with no inputs. Q_M_HOVER_LEARN,2 saves the hover throttle. Expect about 0.52 (est.) from 1637.9 g needed / 3130 g max per motor (results.json C0.sized.Tunit_g, params.Tmax_g). Above 0.6 means the thrust margin is short (tuning-process step 7, "should be ~0.3–0.6").
  2. Then set Q_P_D_ACC_P = 0.1 × Q_M_THST_HOVER and Q_P_D_ACC_I = 0.2 × Q_M_THST_HOVER (tuning-process step 7).
  3. Notch setup within the 60 s hover limit. The tuning-process page asks for a 2-minute no-input hover. The review limits test hovers to 60 s (R14 mitigation). Two other ArduPilot sources cover the shorter hover:
    • The batch-sampler page asks for "a regular flight (not just a gentle hover) of at least 1 minute" (common-imu-batchsampling.html). One 60 s QHOVER with INS_LOG_BAT_MASK,1 and INS_LOG_BAT_OPT,4 (pre + post filter, both already set) meets that.
    • Throttle-based notch tracking needs only the hover noise frequency and the hover thrust. Hover-thrust learning needs "no pilot stick input for at least 10 seconds" (same page; Q_M_HOVER_LEARN,2).
    • Procedure: fly two separate 60 s no-input QHOVER flights on a calm day. Open each .BIN in firmware.ardupilot.org/Tools/WebTools/FilterReview (or Mission Planner > Setup > Advanced > FFT > "IMU Batch Sample"). Accept the peak only if both flights agree within about 10 % (est.); the average is hover_freq. More batches per flight are available by lowering INS_LOG_BAT_LGIN (batch page: "can be lowered to increase accuracy").
    • Set the throttle-based notch (common-throttle-based-notch.html): INS_HNTCH_ENABLE,1, reboot, INS_HNTCH_MODE,1, INS_HNTCH_FREQ = hover_freq, INS_HNTCH_BW = hover_freq/2, INS_HNTCH_REF = learned Q_M_THST_HOVER. Leave INS_HNTCH_HMNCS at its default 3 (1st + 2nd harmonic). The in-flight FFT auto-setup is not available (!HAL_GYROFFT_ENABLED).
    • Fly one more 60 s hover and check that the post-filter spectrum shows the peak attenuated. Then set INS_LOG_BAT_MASK,0 to free RAM (batch page note).
  4. Then QLOITER: position hold within about 2 m in light wind (est.).
  5. Manual roll/pitch tune, then QAUTOTUNE (in the build) per tuning-process steps 10+ and qautotune-mode.html.

9.5 Fixed-wing preparation

  • Airspeed ratio: ARSPD_AUTOCAL,1 for a 5-minute loiter, then back to 0 (calibrating-an-airspeed-sensor.html). Review M[controls].1 failure threshold: > 10 % error puts the real transition speed under R11's 15.6 m/s.
  • Fixed-wing gains start at firmware defaults. Run AUTOTUNE in FBWA after the first transitions (automatic-tuning-with-autotune.html).

9.6 First transitions

  1. Load vtol_c0.param again (removes the hover-phase overlay). Check Q_ASSIST_SPEED = 13.
  2. QLOITER to ≥ 30 m, nose into wind, with the longest clear line ahead.
  3. Switch to FBWA and push the throttle to about 70 %. The lift motors hold height. At AIRSPEED_MIN 16 m/s they taper out over 5 s. In SITL this took 10–11 s from mode change to "Transition done" on the generic model (sitl/results) and 13–15 s, 140–170 m on the C0 airframe model (sitl/results_c0.json).
  4. Start the turn as soon as "Transition done" appears. In SITL, a straight FBWA transition reached the 300 m fence circle about 6 s after transition end (first SITL run, scenario B, 2026-10-21). Start transitions from the upwind edge of the field.
  5. Back-transition: select QLOITER (or QHOVER). The pusher stops at once and the lift motors brake the aircraft (quadplane-transitions.html). Allow about 81 m stopping distance at the 18 m/s cruise (18² / (2 × 2 m/s² decel)). In the C0 airframe model the airbrake phase pitches to the 30° VTOL limit and briefly stalls the wing (−17°/+33° swing, ≤ 2 m altitude loss, sitl/README.md Gains item 1); if the aircraft does this, lower Q_TRANS_DECEL first (gentler, longer stop), then Q_ANGLE_MAX, before any attitude gain. The test cards F06–F13 carry this as a watch item.
  6. Check in the log: the transition current peak against the 75.7 A estimate (results.json C0.sized.I_trans), battery sag, and time to AIRSPEED_MIN. Review M[controls].2 failure threshold: > 15 s or > 100 m.
  7. Only then: CRUISE circuits, RTL test at altitude, and the full R9 mission in AUTO (sitl/mission_test.py scenario A is the template: NAV_WAYPOINTs then NAV_VTOL_LAND).

10. Decisions that changed the review's controls chapter

Details and rationale are in docs/research/firmware.md. They should be folded into the review:

  1. Output map: lift on S5–S8, pusher S9, rudder 2 on S11 (DShot timer-group conflict in the review map).
  2. Airspeed sensor (rev 2026-11-01): the BOM's RMRC MS4525DO is probed at 0x28/0x36/0x46, clear of the baro at 0x76; ARSPD_TYPE,1 (§4.3). The earlier MS5525DSO needed a CSB re-strap to 0x77 and was dropped.
  3. Telemetry (no radio in the BOM, $0): one RP3 V2 runs either CRSF or ELRS-MAVLink, not both (review row "CRSF + MAVLink"). Baseline for hover phases: CRSF (SERIAL6_PROTOCOL,23), with telemetry on the Pocket Crush. GCS link for transition flights: ELRS MAVLink mode (SERIAL6_PROTOCOL,2, SERIAL6_BAUD,460, RSSI_TYPE,5, ELRS ≥ 3.5 on TX and RX, Mission Planner on UDP 14550 via the Pocket Crush backpack WiFi). The RX then sends RC as RC_CHANNELS_OVERRIDE; SITL scenario G showed RC-loss failsafe 1 s after the overrides stop. Details and limits: peripherals.md §2.3–2.4.
  4. Transition threshold: the state diagram labels say "≥ 15 m/s" and "< 12 m/s". The table says AIRSPEED_MIN 16 and Q_ASSIST_SPEED 13. The parameters use 16 and 13.
  5. Battery current limit: BATT_WATT_MAX is not in this build. Q_M_BAT_CURR_MAX limits only the lift motors, so it is set to 85 A as a fault guard, not 60 A. Transition current is managed with TKOFF_THR_MAX 75 % and the pilot's throttle. Bench test M4 decides. THR_MAX 85 caps the pusher in every fixed-wing mode: the C0 SITL model gives about 890 W at 100 % throttle on a full 25.2 V pack (sitl/validation_c0.json, estimated motor model) against the X4110S 615 W rating; power scales roughly as throttle³, so 85 % is about 0.85³ × 890 = 547 W (est.). Cruise needs 233–302 W (SITL model / aero build-up). Bench test M1 (pusher power at full pack, TEST-PLAN C09) re-sets it; if THR_MAX is lowered, re-check that level speed at full throttle still exceeds AIRSPEED_MAX 25 m/s.
  6. Low-battery action, two stages: low (25 % / 1625 mAh, or 21.0 V sag-compensated; rev 2026-11-02, was 30 %, which fired before the planned 27.4 % landing) → RTL, which ends in QRTL and a vertical landing at home (Q_RTL_MODE,1). Critical (20 % / 1300 mAh, or 19.8 V) → QLAND in place. ArduPilot's failsafe page calls "BATT_FS_LOW_ACT = 1 (RTL) and BATT_FS_CRT_ACT = 2 (Land)" a reasonable setup. 4 = QLand is used for the explicit VTOL landing. The review had low = QLAND. The hover-phase overlay sets low → QLAND, because RTL would start a fixed-wing return. (Note: the parameter is BATT_FS_LOW_ACT; BATT_FS_LLOW_ACT does not exist.)
  7. Fence ceiling: 110 m above home, 12 m below the 400 ft (121.9 m) limit, so the climb overshoot after a breach stays legal. SITL peaked at 113.1 m, 3.1 m over the fence. Review wording: "fence 110 m, 12 m below the 400 ft limit".
  8. RC loss in hover: QLAND in place (ArduPilot behaviour in Q modes). The review's "RC loss → RTL then QLAND" applies to fixed-wing modes only.
  9. Firmware: stable 4.7.2 now covers the CoreWing. The 2027-02-22 go/no-go can close early.

11. SITL validation summary (2026-10-21)

ArduPlane V4.7.2 SITL (the Windows build Mission Planner uses, from firmware.ardupilot.org/Tools/MissionPlanner/sitl/PlaneStable), model quadplane, vtol_c0.param + sitl/sitl_overrides.param. 133 parameters read back. All match. Only the 3 BLHeli passthrough parameters (SERVO_BLH_*) are absent, because SITL has no BLHeli support. 7/7 scenarios passed. Results and limits are in sitl/README.md. The SITL airframe is ArduPilot's generic QuadPlane model, not C0. Its learned hover throttle was 0.15, against about 0.52 (est.) for C0. So SITL validates modes, thresholds, failsafes and the fence. It does not validate gains or performance.

Update 2026-10-25 — C0 airframe model. The same seven scenarios were re-run on sitl/c0_physics.py, a 6-DOF model of C0 itself (SITL --model JSON; mass, geometry, AVL derivatives, CL_max, drag polar, motor thrust/power table, APC 14x8.5 prop data and the output map all read from the repo files). 7/7 passed with vtol_c0.param unchanged; no vtol_c0_sitl_tuned.param was needed. Trims match the analysis: hover 0.523 thrust fraction at 22.2 V (TEST-PLAN 0.52), cruise α 2.84° / elevator −3.45° / 7.69 N (AVL 2.86° / −3.45° / 7.70 N), stall 11.75 m/s (target 11.75). SITL learned Q_M_THST_HOVER 0.44 on a full pack (the model predicts 0.45 at 25.2 V, because ArduPilot references thrust to Q_M_BAT_VOLT_MAX; 0.52 is the 22.2 V figure). Forward transition 13–15 s and 140–170 m with ≤ 2 m altitude loss, peak 71–75 A; back-transition ≤ 2 m loss with a −17°/+33° pitch swing during ArduPilot's airbrake stop (the C0 wing stalls at the 30° VTOL pitch limit). Model limitations (no prop wash, no rotor inflow, estimated inertia, reversed-flow aero) are listed in sitl/README.md. Re-run 2026-10-25 on the final data (results.json MTOW 6.552 kg, chord 0.2842 m; corrected AVL model with rudder derivatives Cnδr +0.000736/deg; CAD rev D): 7/7 again, same transition times (13–14 s), losses ≤ 2 m, peak 69–75 A; the model now takes all control derivatives from the AVL st file.