Control of DC motor using the encoder and STM32F103 using Chibios
The STM32F103(Bluepill) has been used to achieve this project. It can either be programmed using JTAG or can be done using UART.
The pinout of the cheap JTAG connector can be seen here
and the pinout of the SWD pins of the Bluepull can be seen here
. Connecting each pin to their respective other you should be able to program the board.
| ST-LinkV2 Bluepill |
|---|
| 3V3 -> 3V3 |
| T_SWCLK -> SWCLK |
| T_SWDIO -> SWDIO |
| GND -> GND |
NOTE
It is important to note that the Chinese clones of Bluepill has a bootloader on them that prevents them to be programmed using SWD. But it can be easily overcome by connecting the
T_JRST to Reset pin on Bluepill.
The main IDE that I have used for the project is VSCode as it has support for debugging cortex microcontrollers using the Cortex-Debug extension and also it is fast. The addition of CMake support can help using other IDEs such as QtCreator. A good tutorial about setting up QtCreator can be found at Debugging STM32 from QtCreator.
OpenOCD has been used for debugging. If you run into problems with OpenOCD saying that Cannot halt device .. then you might be using the cheap chinese STM32 microcontroller which have a different ID. If you look into bluepill_openocd.cfg config file then you can see that I am also using a custom config file which contains the modified device ID.
You can copy and paste the contents of target/stm32f1x.cfg into a new file and modify the line
# this is the SW-DP tap id not the jtag tap id
set _CPUTAPID 0x1ba01477
to
# this is the SW-DP tap id not the jtag tap id
set _CPUTAPID 0x2ba01477
The results doesn't seem promising due to the limited availablity os speed measurement using the Encoders as the MCU doesn't get time to process other things because of too many interrupts. If the interrupts are removed and the hardware timers of STM32 are used in Quadrature encoder mode, then this problem can supposedly be solved. Another problem is that the RPM value calculated seems to be too noisy and the addition of low pass filter lead to overshooting as can be seen below.
Here is the result without any filtering of the RPM values
Here is the result by using a simple low pass filter.
The results without filtering seem to correlate with the simulation done in Matlab. It would be interesting to see the result and the tuned values after adding a low pass filter block.
- Use low pass filter block and find the PID values
- Parse line instead of getting individual chars and converting to number
- Implement a GUI for tuning.