#STM32H7 NUCLEO CODE#
#STM32H7 NUCLEO SERIES#

These signals look similar but like shifted or having different duty cycles at each period. Quickly single triggering also shows two main signal patterns, which is overlayed also in the above image:

#STM32H7 NUCLEO CODE#

So I enabled Instruction Cache (CPU ICache) in STM32CubeMX (which calls SCB_EnableICache() first thing in main) and re-run the code above. Cortex-M4 does not have a cache but M7 does, so does STM32H7. So the last thing I wonder is if processor cache changes anything (I expect it should). Since Flash memory is slower, there are wait states involved when fetching instructions. HAL_GPIO_TogglePin with L1 Instruction Cache You might ask why it is not 50%, it is probably because there is an asymmetric processing done somewhere, I did not check this in detail but we will see an example of this later. HAL_GPIO_TogglePin without SysTickДЄs you see, there is only a single signal, since there is no interrupt, there is nothing causing variation in the behavior of this code.ДЄlso the duty cycle is around 43%, which is close to 50%. All experiments below are with SysTick disabled, so you will not see its effect anymore. SysTick is used by HAL, so normally it should not be disabled, however, for the purpose of this post, it is harmless to disable it and it can simply be disabled by SysTick->CTRL = 0. This also causes the variation in the signal generated by toggling. Depending on where this handler is called, during high or low value of GPIO, that section naturally takes longer than other so duty cycle changes. The corresponding interrupt handler is called SysTick_Handler and it calls HAL_IncTick which increments a global variable, so it is a short quick running interrupt handler. Systick timer is a processor feature which generates an interrupt regularly at programmed intervals. This project, which is just created, has only one thing running concurrently with my TogglePin single line code, which is Systick.

Using history, I could also catch this one:

This picture does not show everything actually, time to time I also saw other waveforms. The color is because I selected temperature as waveform color, which shows blue for rare cases and white for frequent ones. Actually, the oscilloscope shows in continuous mode something like this: However, another single trigger can result:ДЄ 2.4MHz pulse train with a different duty cycle. So the code is simply:ДЄ 2.2MHz pulse train. The easiest and the most straightforward way to do GPIO Toggling is by using HAL_GPIO_TogglePin in STM32H7 MCU Package. HAL_GPIO_TogglePin HAL_GPIO_TogglePin with SysTick I also configured Timer 1, Timer 6 and DMA 1 as mentioned in the sections below.ДЄll the measurements are done with a Rohde & Schwarz RTB2004 300 MHz 2.5GS/s oscilloscope using a RT-ZP03 300 Mhz 10:1 probe. Very High here means 220 MHz when used with C=10pF and 100MHz when used with C=50pF (source: STM32H723ZG Datasheet). I enabled PB11 as GPIO Output and set maximum output speed to Very High. processor is running at 550MHz, busses are running at 275MHz and 137.5MHz. After creating the project, I disabled ETH (Ethernet), USART3, USB_OTG_HS and I configured the clock tree to maximum possible, e.g.

The MCU configuration is done in STM32CubeMX inside STM32CubeIDE. I am using STM32CubeIDE and STM32H7 HAL in STM32H7 MCU Package. Just to be clear, you should not generate a pulse train by GPIO toggling, this is just an experiment. This is I believe one of the most advanced and high performance MCUs in the market. SetupДЄll experiments in this post is done with a NUCLEO H723ZG board having an STM32H723ZG MCU.

Some of the ideas here are from a thread in STM32 forum. Since the MCU I am using is clocked at 550MHz, theoretically, assuming writing 1, writing 0, and then looping back each takes 1 cycle, this can be done at 550/3 ~= 183 MHz. Just as a comparison, I will also show pulse trains generated by using DMA instead of a while loop running on the processor, and by using a TIMER. The question in this post is as same as before, what is the maximum frequency of a pulse train (square wave) generated just by GPIO toggling, setting an output pin to high and low and repeating this forever ?

#STM32H7 NUCLEO SERIES#

Here is a second post on the same concept but this time using a much more powerful ARM Cortex-M7 based STM32H7 series MCU (STM32H723ZG) running at 550 MHz. Earlier this year I wrote a post titled micro:bit (nRF51, Cortex-M0) GPIO toggling.

Stm32h7 nucleo