This repository implements Field Oriented Control (FOC) for stock hoverboards. Compared to the commutation method, this new FOC control methods offers superior performance featuring:
- reduced noise and vibrations
- smooth torque output
- improved motor efficiency. Thus, lower energy consumption
- field weakening to increase maximum speed range
## Firmware architecture
The main firmware architecture including:
- Estimations: *estimates the rotor position, angle and motor speed based on Hall sensors signal*
- Diagnostics: *implements error detection such as unconnected Hall sensor, motor blocked, MOSFET defective*
- Control Manager: *manages the transitions between control modes (Voltage, Speed, Torque)*
- FOC Algorithm: *implements the FOC strategy*
- Control Type Manager: *Manages the transition between Commutation and FOC Algorithm*
- The C code for the controller was auto-code generated using [Matlab/Simulink](https://nl.mathworks.com/solutions/embedded-code-generation.html) from a model which I developed from scratch specifically for hoverboard control. For more details regarding the working principle of the controller please consult the [Matlab/Simulink model](https://github.com/EmanuelFeru/hoverboard-firmware-hack-FOC/tree/master/01_Matlab).
- A [webview](https://github.com/EmanuelFeru/hoverboard-firmware-hack-FOC/tree/master/01_Matlab/BLDC_controller_ert_rtw/html/webview) was created, so Matlab/Simulink installation is not needed, unless you want to regenerate the code
- All the calibratable motor parameters can be found in the 'BLDC_controller_data.c'. I provided you with an already calibrated controller, but if you feel like fine tuning it feel free to do so
- The parameters are represented in Fixed-point data type for a more efficient code execution
- For calibrating the fixed-point parameters use the [Fixed-Point Viewer](https://github.com/EmanuelFeru/FixedPointViewer) tool
- conversion of the remaining filters (for the battery voltage, current, and temperature) from floating point to fixed-point. This will reduce further the SMT32 computational load
---
## Building
For building (and flashing) I recommend platform.io, plaformio.ini file included. Simply open the folder in the IDE of choice (vscode or Atom), and press the 'PlatformIO:Build' or the 'PlatformIO:Upload' button (bottom left in vscode).
Additionally, you can also flash using the method described below in the Flashing Section.
The original Hardware supports two 4-pin cables that originally were connected to the two sensor boards. They break out GND, 12/15V and USART2&3 of the Hoverboard mainboard.
Both USART2 & 3 can be used for UART and I2C, PA2&3 can be used as 12bit ADCs.
The reverse-engineered schematics of the mainboard can be found here:
To build the firmware, just type "make". Make sure you have specified your gcc-arm-none-eabi binary location in the Makefile ("PREFIX = ...") (version 7 works, there is a version that does not!) (if the ons in linux repos do not work, use the official version: https://developer.arm.com/open-source/gnu-toolchain/gnu-rm/downloads). Right to the STM32, there is a debugging header with GND, 3V3, SWDIO and SWCLK. Connect GND, SWDIO and SWCLK to your SWD programmer, like the ST-Link found on many STM devboards.
Do not power the mainboard from the 3.3V of your programmer! This has already killed multiple mainboards.
Make sure you hold the powerbutton or connect a jumper to the power button pins while flashing the firmware, as the STM might release the power latch and switches itself off during flashing. Battery > 36V have to be connected while flashing.
To flash the STM32, use the ST-Flash utility (https://github.com/texane/stlink).
If you never flashed your mainboard before, the STM is probably locked. To unlock the flash, use the following OpenOCD command:
First, check that power is connected and voltage is >36V while flashing.
If the board draws more than 100mA in idle, it's probably broken.
If the motors do something, but don't rotate smooth and quietly, try to use an alternative phase mapping. Usually, color-correct mapping (blue to blue, green to green, yellow to yellow) works fine. However, some hoverboards have a different layout then others, and this might be the reason your motor isn't spinning.
Nunchuck not working: Use the right one of the 2 types of nunchucks. Use i2c pullups.
Nunchuck or PPM working bad: The i2c bus and PPM signal are very sensitive to emv distortions of the motor controller. They get stronger the faster you are. Keep cables short, use shielded cable, use ferrits, stabilize voltage in nunchuck or reviever, add i2c pullups. To many errors leads to very high accelerations which triggers the protection board within the battery to shut everything down.
Most robust way for input is to use the ADC and potis. It works well even on 1m unshielded cable. Solder ~100k Ohm resistors between ADC-inputs and gnd directly on the mainboard. Use potis as pullups to 3.3V.
---
## Examples
Have a look at the config.h in the Inc directory. That's where you configure to firmware to match your project.
Currently supported: Wii Nunchuck, analog potentiometer and PPM-Sum signal from a RC remote.
A good example of control via UART, eg. from an Arduino or raspberryPi, can be found here:
ST Employee: [cedric H](https://community.st.com/s/question/0D50X0000B28qTDSQY/custom-foc-control-current-measurement-dma-timer-interrupt-needs-review)
for the very useful discussions, code snippets, and good suggestions to make this work possbile.
