BrotherHobby 1404 KV4600 — Semi-Empirical Motor Model for SiL Simulation

About This Project

Methodology:

• Built using the LIGHT-MBSE-PIPELINE-SKELETON to ensure full requirements traceability and automated validation.
• A data-driven C++ motor model for drone flight simulation, derived from independent stand test data.
• The model is semi-empirical: it combines the physical aerodynamic thrust law with data-driven gray-box polynomials.

Setup:

• Motor: BrotherHobby 1404 KV4600
• Propeller: iFlight Nazgul T4030 (2-blade) on 2S LiPo (7.4V)

Quick Start & Building

The pipeline is fully automated. You can build the firmware, run Renode simulations, export logs, and generate validation reports with a single command.

git clone [https://github.com/meugenom/light-mbse-pipeline-skeleton.git](https://github.com/meugenom/light-mbse-pipeline-skeleton.git)
cd light-mbse-pipeline-skeleton
./start.sh

Note: Test reports will be generated in ./logs/sensor_test_posix.log and ./logs/sensor_test_stm32.log.

Robust Orchestration

The pipeline is driven by a fail-safe bash orchestrator (start.sh with strict set -euo pipefail). It automatically handles:

• Toolchain dependency checks.
• Generation and verification of intermediate artifacts (LUT headers).
• Cross-platform builds (POSIX vs STM32) using CMake.
• Asynchronous execution and UART log extraction via Renode.
• On-the-fly parsing of validation traces into CSV formats.

Table of Contents

• Quick Start & Building
• Table of Contents
• Iteration Roadmap
• Motivation
• Environment & Toolchain (Reproducibility)
• Project Documentation
• Project Directory Structure
• Workflows
• Validation
• Known Problems and Limitations
• References
• License

Iteration Roadmap

Iteration Roadmap

Version	Status	Pipeline Requirements	Reference
v1.3.0 STATIC	Current	Octave -> C++ (LUT) -> Tests -> Report	Independent Datasheets
v2.0.0 DYNAMIC	Planned	Rotor inertia + back-EMF dynamics	Oscilloscope + test bench
v3.0.0 HIL	Future	Eddy currents + thermal drift + commutation noise	Hardware-in-the-Loop

Motivation

This project models the thrust and current of the BrotherHobby 1404 motor. Unlike manufacturer data, which is often "marketing-grade" (optimized for peak freestyle thrust), this model is designed for high accuracy in the hover and cruise ranges (20–60% throttle), making it suitable for control system engineering.

Project Documentation

File	Description
FULL SPECIFICATION	Component specifications, raw stand test data
CALCULATION DETAILS	Model derivation: math, pipeline, voltage scaling
VALIDATION REPORT	Validation results and performance metrics

Environment & Toolchain

Host: macOS (Apple Silicon) The following tools are required for reproducibility:

Tool	Version	Purpose
GNU Octave	11.1.0	Math modeling, LUT generation, Automated Reporting
Clang	21.0.0	Runtime model implementation (POSIX)
arm-none-eabi-gcc	15.2.rel1	Bare-metal target compilation (STM32)
Renode	1.16.1	Instruction-accurate hardware emulation
CMake	4.3.1	Build system management

Pipeline Overview

              new Iteration
                   |
                   ▼
┌─────────────────────────────────────────┐
│               Preparing:                │
│1. Searching Raw Data                    │     
│2. Convert Raw Data to Specifications    │
│3. Methodology of Calculation            │
└───────────────────┬─────────────────────┘
                    │                       
                    ▼                       
┌─────────────────────────────────────────┐
│        Octave Mathematical Model.       │
│1. Approximation of Curves               │
│2. Algorithm Logic and model design      │
│3. Export LUT Generation                 │
└───────────────────┬─────────────────────┘
                    │
                    ▼
┌─────────────────────────────────────────┐
│         C/C++ Model Design              │
│1. Import LUT.                           │
│2. Implement Algorithm Logic             │
│3. Optimize for Bare-Metal               │    
│4. Prepare Test Cases                    │
└───────────────────┬─────────────────────┘
                    │
                    ▼
┌─────────────────────────────────────────┐
│         Renode Model Testing            │
│1. Set up Test's Environment             │
│2. Run Simulations                       │
│3. Export Logs to CSV Data               │
└───────────────────┬─────────────────────┘
                    │
                    ▼
┌─────────────────────────────────────────┐
│         Results Analysis                │
│1. Compare with Datasheet and Stand Test │
│2. Errors Analyse  and Parameter Tracing │
│3. Validation Report                     │
└───────────────────┬─────────────────────┘
                    │
                    ▼
            End of Iteration

Project Directory Structure

├── README.md 
├── CALC.md
├── SPEC.md
├── VALIDATION.md
├── METRICS.md
├── LICENSE
├── ltspice/ /* LTSpice simulations for back-EMF and transient analysis */
├── octave/ /* Octave scripts for calculations */
├── references/ /* Reference materials, datasheets, papers */
├── src/ /* Source code */  
├── tests/ /* Test cases and validation scripts */
├── build(build_stm32)/ /* Build artifacts */
├── renode/ /* Renode simulation environment */
├── docs/ /* Documentation */
├── logs/ /* Logs */

Workflows

Octave Workflow

1. Load CSV from references/Brother-Hobby-1404-4600KV_Propeller-T4030.csv
2. Filter noisy data: exclude points with RPM < 2000 (idle and near-stall)
3. Normalize throttle: PWM 1000–2000 -> 0.0 - 1.0
4. Normalize RPM to $V_{nominal}$: $RPM_{norm} = RPM_{actual} \cdot V_{nom}/V_{actual}$
5. Normalize current to $V_{nominal}$: load component scaled by $(V_{nom}/V_{actual})^2$, idle current kept constant
6. Physical thrust model: $F = k \cdot n^2$ with coefficient $k$ from Least Squares on raw (unnormalized) data
7. pchip interpolation (Gas -> RPM): interp1(throttle_norm, rpm_norm, lut_throttle, 'pchip') — shape-preserving interpolation on all normalized points
8. pchip interpolation (Gas -> Current): interp1(throttle_norm, current_norm, lut_throttle, 'pchip') — shape-preserving interpolation on all normalized points
9. Evaluate all models on a uniform 101-point grid (0.00 - 1.00, step 0.01)
10. Enforce boundary conditions: zero thrust/RPM at 0% throttle, idle current floor
11. Export everything to src/includes/motor_lut.h (including MOTOR_R_INTERNAL)

Full algorithm description: CALC.md

Code Workflow

1. Import LUT: Include src/includes/motor_lut.h in src/core/motor.cpp
2. Implement Algorithm: in src/core/motor.cpp both for POSIX and STM32 platforms

Renode Workflow

1. Set up Environment: renode/test_run.resc — define the test bench, load the model firmware, set up peripherals
2. Logs: ./start.sh — execute the renode command with logs parameters
3. Logs Output: ./logs

Validation

Thrust Tests

  - PASS zero throttle -> zero thrust
  - PASS thrust is monotonically increasing (10%–90%)
  - WARN thrust drops at 100% throttle: 0.9148N -> 0.8389N (propeller saturation / Hall sensor RPM under read at > 20k RPM)
  - PASS higher voltage -> higher thrust

Current Tests

  - PASS current is monotonically increasing (10%–90%)
  - WARN current drops at 100% throttle: 8.7945A -> 8.1069A (consistent with propeller saturation at >20k RPM)
  - PASS current V_eff-scaling: I(8.4V)/I(7.0V) = 1.385

Model Accuracy

Max Thrust Error: 17.33% Max Current Error: 17.78%

WHY such errors?

1. Anomaly at 10% throttle (Physical Opportunities):

See the Test Report from tytorobotics.com:

• 10%: 255 RPM -> THRUST: 5.88 Gramms/N but should be 0,13 Gramms/N (idle current, propeller not spinning)
• 20%: 2825 RPM -> THRUST: 15.8 Gramms/N This is pure load cell drift. The sensor on the test bench was simply vibrating and it recorded this “noise” in the file. Dataset has 11 Points and it's too early to do an accurate analysis.

2. No-gas zone (0% gas):

What we see in the graph: The blue line (thrust error) drops to -100%. In the C++ model, we have strictly defined a boundary condition: at 0% PWM signal, the thrust is exactly 0.0000 N. This is a mathematical absolute. However, in the physical world, the measuring system is subject to gravity, residual mechanical stresses from previous tests (metal hysteresis), and micro-vibrations in the room’s air.

3. Transient Mode (10% Throttle): Aerodynamic Stall and Stall Torque

What we see in the graph: A spike in discrepancies, where the raw data indicates the presence of thrust, while the model (calibrated to the aerodynamic law $F = k \cdot n^2$) filters out these readings. Physical explanation: When the motor starts (low RPM), two physical barriers come into play:

• Magnetic resistance (Cogging torque)
• Low Reynolds numbers

4. Nominal Operating Mode (20%–70% Throttle): Perfect Convergence

What we see in the graph: Starting at 20% throttle, both curves (thrust and current) fall within a narrow range of 2% from the reference value, well within the 5% tolerance zone. Physical explanation: The motor reaches a stable RPM. The propeller transitions to the calculated aerodynamic regime (developed turbulent flow). The influence of static friction and mechanical hysteresis becomes negligible compared to the power being generated.

Known Problems and Limitations:

• Rotor Inertia
• Back-EMF Dynamics
• Commutation Noise
• System-level effects:
  • Magnetic field interference with IMU magnetometer
  • Power supply ripple on flight controller
  • EMC behavior of the complete drone system
• Temperature Effects

References

1. BrotherHobby 1404 KV4600 — Stand Test Data tytorobotics.com. Primary data source for this model

2. BrotherHobby 1404 KV4600 — Motor Specifications brotherhobbystore.com

3. iFlight Nazgul T4030 — Propeller Specifications shop.iflight.com

4. "A Comparative Study on Thrust Map Estimation for Multirotor Aerial Vehicles", Francisco J. Anguita, Rafael Perez-Segui, Carmen DR.Pita-Romero, Miguel Fernandez-Cortizas, Javier Melero-Deza, http://www.imavs.org

5. "Modelling and Control of a Large Quadrotor Robot", P.Pounds, R.Mahony, P.Corke, 2010

6. Propeller Performance Data at Low Reynolds Numbers, John B. Brandt and Michael S. Selig 2011, pages 1-18.

License

This project is licensed under the MIT License. See LICENSE