Solver Changelog

v1 — Model-accuracy release (June 2026)

This release closes a systemic thrust-over-prediction / current-under-prediction bias root-caused against wind-tunnel and manufacturer bench data (98 motors, 7,600+ bench rows). Models changed: propeller, motor, and system solver.

Propeller: calibration envelope guard + geometry-derived pitch

The propeller calibration multipliers now fade smoothly to neutral outside the validated wind-tunnel envelope instead of extrapolating beyond the data they were fit on. This removes thrust over-prediction at high tip speeds. In-envelope results are unchanged. Pitch-to-diameter is now derived from the actual blade twist at 75% radius instead of parsing the propeller name.

System: ESC losses now drawn on the DC bus

ESC inverter, PCB, AC-excess, ripple, freewheel, and six-step-harmonic losses are now sourced as current on the DC bus inside the solver, closing an energy-balance leak. Displayed bus current, voltage sag, SOC depletion, flight time, and pack heating rise by the ~3–7% the bus previously under-read.

Motor: resistance convention + bench-fit series resistance

Catalog motor resistance is now interpreted as the line-to-line value (per-phase R = Rm/2), matching how manufacturers publish it. This pairs with a bench-fit ESC + wire series-resistance default (fleet-median ≈ 18 mΩ) replacing the previous zero-resistance assumption, and a wire-gauge preset. Full-throttle RPM median error vs bench: +0.4%.

Motor: size-scaled, airflow-cooled thermal resistance + hard caps

The flat motor thermal resistance (0.5 K/W regardless of motor size — the cause of unrealistic 193–317 °C predictions on large motors) is replaced with a size-scaled, bench-fit law plus a forced-convection cooling model and a per-rotor cooling selector: cowled (still air), propeller exit airflow (default), or custom velocity / direct thermal resistance. Temperatures are hard-capped at 155 °C (winding) / 120 °C (magnet); the results page flags any rotor group that hits the caps.

Motor: part-throttle six-step throttle→duty map

A matched-RPM study against bench data showed the motor model is essentially exact at the bench operating point; the remaining part-throttle gap was throttle→duty endpoint semantics, now absorbed by a global bench-calibrated map (40–70% throttle band current accuracy improved from 0.94 to 0.99 of bench). Full-throttle results are unchanged. See the throttle-semantics note below.

Results

Results now carry the solver version stamp, the per-rotor applied motor cooling air velocity, and thermal-limit flags when the 155 °C / 120 °C caps are hit.

Throttle semantics

"Throttle %" in ThrustLab means inverter duty %, not transmitter stick %. The throttle you enter is the modulation duty command the ESC applies to the DC bus — it is not a radio/transmitter percentage, which real ESCs remap through their own endpoint and minimum-throttle calibration.

For six-step (BLDC) ESCs, ThrustLab additionally applies a global bench-calibrated throttle→duty map, duty = clip(−0.0040 + 1.0261 × throttle, 0, 1), fit against 7,079 part-throttle bench rows across 98 T-Motor motors. Real hobby ESC "throttle %" corresponds to a slightly higher effective duty than the commanded fraction; the map absorbs that endpoint convention so part-throttle current and RPM match published bench data. At 100% throttle the map clips to exactly 1.0 — full-throttle results are unchanged. FOC ESCs are untouched (duty = throttle exactly).

ESC type: API default vs app default

The web app defaults to esc_type="six_step" (the hobby BLDC ESC model), while the public API defaults to esc_type="foc". The two models differ by roughly +4.7% RPM / +9.6% thrust (FOC higher). API and SDK users who want results matching the app should set esc_type="six_step" explicitly.