Product forums

My development board is STM32MP257F-DK, and I want to use ST Edge AI Developer Cloud to convert. onnx models to. nb models. However, during the interface conversion of 'Optimize your model with STM32AI MPU Tool', an error message 'Error while generating optimized file. Generation does not contain any output' was reported. Afterwards, I will use the model provided by ST Model ZOO: st_yoloxn_d033w_025-416_qdq_int8_ort_detection_CCO_2017_Serson. onnx. I also encountered an error while optimizing. What is the reason behind it? A few days ago, I was able to successfully convert ssd_mobilenetv1_pt_coco_300_qdq_int8_ort_detection_CCO_2 The error message is as follows: 

Title edited to indicate that this is specific to J-Link.Hi,After updating to STM32CubeIDE 2.1.0 I get the following message after starting a debug session. Live expressions do not work.From the console:SEGGER J-Link GDB Server V9.14a Command Line Version JLinkARM.dll V9.14a (DLL compiled Jan 28 2026 17:11:52) Command line: -port 2331 -s -device STM32F769NI -endian little -speed 4000 -if swd -swoport 2332 -vd -rtos C:\ST\STM32CubeIDE_1.11.2\STM32CubeIDE\plugins\com.st.stm32cube.ide.mcu.externaltools.jlink.win32_2.5.200.202601091506\tools\bin\GDBServer\RTOSPlugin_FreeRTOS.dll -----GDB Server start settings----- GDBInit file: none GDB Server Listening port: 2331 SWO raw output listening port: 2332 Terminal I/O port: 2333 Accept remote connection: localhost only Generate logfile: off Verify download: on Init regs on start: off Silent mode: off Single run mode: on Target connection timeout: 0 ms ------J-Link related settings------ J-Link Host interface: USB J-Link script: none J-Lin

Whenever I am starting to make project  from scratch in STM32Cube AI studio, It gives complete structure with .ioc files as well as AI files, but as soon as I make changes in .ioc file and generate code after loading in CubeIDE, it changes the middlewares and build error comes . What is the solution for that? 1. Should I first make the .ioc project and then make project in STM32Cube AI studio ( Import the existing project from STM32CubeMX or STM32Cube AI studio)2. Another doubt is if I make project from preexisting .ioc, Do I need project in application mode for that as AI changes will work in application mode only ( but STM32N6 doesnt work without FSBL), what is your suggestion for Edge AI project building? The error was :Description Resource Path Location Type make: *** No rule to make target 'C:/Users/hp/.stm32cubeaistudio/workspace/Modified_STM32/Middlewares/ST/AI/Npu/ecloader.c', needed by 'Middlewares/ST/AI/Npu/ecloader.o'. Stop. Modified_STM32_Appli C/C++ Problem&

I've got some trouble with the CRC unit in an STM32L475.At first, I got wrong CRC results, but then it dawned on me that the compiled code didn't force only 8 bits at a time into CRC->DR. So I added what I've used elsewhere, to force the compiler to store only 8 bits in a 32-bit register.*(__IO uint8_t *)(CRC->DR) = Temp;Temp is a uint8_t variable. Below, it certainly looks like a single byte is transferred to CRC-DR (strb instruction). *(__IO uint8_t *)(CRC->DR) = Temp; 80013ea: 683b ldr r3, [r7, #0] 80013ec: 7018 strb r0, [r3, #0]However, after I modified my code that way, it would seem that nothing is calculated at all. No matter how many (non-zero) bytes I throw into CRC->DR, the output just stays at 0.If I revert back toCRC->DR = Temp;at least I get something, albeit an incorrect value as soon as I enter more than one byte.I initialise the CRC unit like this: CRC->POL = 0x04C11DB7; CRC->INIT = 0; CRC->CR = CRC_CR_RESET;Is there something blatantly obv

I’m choosing a microcontroller for a new STM32 design and I’d appreciate some recommendations.I don’t have particular requirements for compute performance, memory size, or ultra‑low power — just solid peripheral support and a Nucleo board available for development. Required peripherals: 1 × FDCAN (CAN FD) 3 × SPI (in practice only 1 SPI with multiple CS likely) 3 × I2C  Constraints: Prefer MCU in LQFP / QFN (no BGA) Compact (48 pins)  Development on a Nucleo board  I’ve looked at parts like STM32C552VET6 and STM32L552CCT6 as potential targets, but I can’t find a dedicated Nucleo board for them.   Any suggestion? thank you,   best regards,   Gianluca

Hi,I am using a STM32WB5MMGH6 BLE module and for some reason HAL_IncTick() is not being called. It is baffling because it was working perfectly well when I was recently working on it before switching to another project for a week.I stripped the project down to almost nothing and placed a HAL_Delay() right after the CubeMX-generated initialization code in main(). So no custom code could do anything to affect that behavior (like turning off interrupts etc). However, the delay never completes.The code is stuck inside the while loop in HAL_Delay() because HAL_IncTick() is never being called (i.e. the system tick is not incrementing) so the HAL_GetTick() always returns 0. __weak void HAL_Delay(uint32_t Delay) { uint32_t tickstart = HAL_GetTick(); uint32_t wait = Delay; /* Add a freq to guarantee minimum wait */ if (wait < HAL_MAX_DELAY) { wait += (uint32_t)(uwTickFreq); } while ((HAL_GetTick() - tickstart) < wait) { } } The Timebase Source is set to SysTick

I've implemeted a vision transformer for STM32N6 NPU.Repository: https://github.com/minchoCoin/stm32n6-transformerBatchMatmul operator is not yet supported on NPU and run on the CPU, but it is much faster than STM32H7(inference time was reduced by about 90% than STM32H7) because fully connected layer and convolutions is executed on NPU.Vision Transformer for STM32N6 NPU has three differences from the original ViT(A. Dosovitskiy et al., 2020)1. Patch Embedding (PE) is configured to be performed in the pre-processing process before model input: Because if the model has PE and self attention, it is not possible to interpret the model structure using STEdgeAI (I don't know why yet...)2. Remove bias parameter from Fully Connected layer or change fully connected layer to 1x1 Conv2D:  fully connected layer of onnx should have 2-dimensional input. however, fully connected layer of ViT has 3 dimensional input(batch,patch,embedding), so compiler removed the batch axis, and ma

Hi,Observed ProblemHAL_ADC_Start_DMA() returns HAL_OK, TIM2 starts, ADC begins converting (confirmed via ADC_ISR in debugger), but HPDMA1 Channel 0 immediately raises a DTE (Data Transfer Error), triggering the error callback. EnvironmentMCU: STM32N657X0HXQBoot mode: Two-stage boot — FSBL + XIP Application (external flash via XSPI2, mapped at 0x70100400)Toolchain: STM32CubeIDE / GCCHAL: STM32CubeN6 (latest)Security mode: Full Secure — all code runs in Secure Privileged mode, no NonSecure worldProject DescriptionThis is a minimal reproduction project containing only:ADC1 — single channel (PA8 / INP5), triggered by TIM2 TRGO, DMA circular modeHPDMA1 Channel 0 — linked-list circular mode, PERIPH_TO_MEMORY, ADC1 → AXISRAM2TIM2 — provides ADC trigger at fixed rate (Period=799, no prescaler)No ETH, no LwIP, no NPU — deliberately stripped down to isolate the DMA issue.Memory Layout (Linker Script)  ROM (XIP) : 0x70100400, 511K — code in external flashRAM (FLEXRAM) : 0x34000000,

Hello ST Community members,   STM32Cube AI Studio is ST’s next-gen desktop solution for optimizing and validating AI models running on STM32 MCUs. It brings a unified workflow, improved performance evaluation, and a modern user experience designed that will benefit both embedded engineers and data scientists. STM32Cube AI Studio is fully compatible with the STM32Cube ecosystem, allowing a smooth transition from your existing projects. Learn more: https://community.st.com/t5/developer-news/introducing-stm32cubeai-studio/ba-p/876445     Have a good day!

Hi I have FP32 TFLite model. I can able to compile quantised model with STM32 Cube AI studio. i am getting some error while compiling FP32 model with STM32 Cube AI studio. Please let me know the steps to build FP32 model and make the C files correspond to the model.  

I am using an STM32H7 MCU and configuring the system clock to 400 MHz using CubeMX. The clock tree shows SYSCLK = 400 MHz, and MCO2 is configured to output SYSCLK.However, when I debug:HAL_RCC_GetSysClockFreq() returns 64 MHzRCC->CFGR shows SWS = 0 (HSI is used)SW = 0, meaning the system is not switching to PLL It seems that the PLL is not being selected or is failing to lock, and the system remains on HSI. I have configured PLL and SYSCLK source in CubeMX, but the change is not reflected at runtime.HAL_RCC_GetSysClockFreq() RETURNS 64MHZ  

Hi, I'm using the image sensor evaluation kit with STEVAL-66GYMAI, STEVAL-EVK-U0I and the PC software of STSW-IMG501 to capture video stream. I would like to change the setting of MIPI lane from default dual-lane to single-lane operation. But, I failed to change it, so please tell me how I can complete to change the setting. I tried to change the MIPI lane setting by changing the register value of OIF_CTRL at 0x030c. I changed the register value via the PC software from 0x0082 to 0x0081 in accordance with the VD66GY manual. The ROI size, the frame rate and the MIPI data rate were set at 288x288, 12 FPS and 1500 Mbps, which seemed to be enough for single-lane operation. The frame rate was adjusted by changing LINE_LENGTH register. However, I couldn't get the video stream. The excution rate displayed on the top left of the PC software window was 0.5 in streaming, and any picutures were not shown on the PC software.  I could see the video stream with the default dual-l

Hi everyone,I’m having an issue connecting my custom board based on STM32H723ZGT6 to STM32Cube.AI Studio. The software cannot detect any COM port or target device.Setup:UART interface via USB-TTL converterUsing USART1 (PA9 TX / PA10 RX)TX/RX crossed correctlyWindows Device Manager detects the COM port normallyThe same COM port appears and in other software (e.g. Arduino IDE)  Dependency Software versions (all latest): STM32CubeProgrammer v2.22, path in setting also are correct.Additional info:I can connect to the MCU successfully using ST-LINK in STM32CubeProgrammerHowever, UART connection in CubeProgrammer fails with timeout Question:What is the correct way to make STM32Cube AI Studio detect and successfully connect to an STM32H723ZGT6 MCU on a custom board?Thanks for any advice.

Hi Support,I am unfamiliar with this project structure where the generated project contains separate boot/ and appli/ directories. I created the project using STM32CubeMX with the CMake toolchain selected, and noticed that two separate .elf files are produced after building.I am unsure which of these should be used when flashing the board using STM32CubeProgrammer. Currently, I am able to successfully flash the device using the .elf located in the /boot/ build directory, but I cannot flash the .elf located in /appli/.Could you clarify the purpose of the boot/ and appli/ directories and which firmware image is intended to be flashed? Additionally, is there a way to generate a more traditional single-application project structure (i.e., without separate boot and application directories) when targeting the STM32H7S3L8?Any guidance or suggestions would be greatly appreciated.

In the RM0506 rev. 1 file in chapter 14 (DDRPHYC), the DDR3L pins don't match the reference schematic in the examples-of-ddr-memory-routing-on-stm32mpus.zip file.In the STM32MP21XXAN_DDR3L-[No Variations]-A01_Schematic.PDF project, the DDR_A8-DDR_A31 pins have different functions.Which version is correct?RM0506STM32MP21XXAN_DDR3L-[No Variations]-A01_Schematic.PDF

Post updated by a ST moderator: 1- Title translated from Chinese to English. 2- Used </> button for the code. I am attempting to transplant the latest FreeRTOS kernel program on the basis of the project generated by CubeMX. When FreeRTOS uses HSI, there is no problem and it runs normally. However, when using HSE, an initialization timeout occurs. Without transplanting FreeRTOS, HSE can pass the initialization normally.My approach was to incorporate the xPortSysTickHandler, xPortPendSVHandler, and vPortSVCHandler required by FreeRTOS into the three functions in the it file. Then, I found that it was the xPortSysTickHandler() within the systick_handler() that was causing the issue. After commenting it out, the HSE could be initialized normally. So, I'm asking on the forum what the reason is and if there's a way to solve it. AI is not reliable. void SysTick_Handler(void) { /* USER CODE BEGIN SysTick_IRQn 0 */ /* USER CODE END SysTick_IRQn 0 */ HAL_IncTick(); /* USER CODE BEGIN

HiAs I experienced some problems with 2.1.0, I reverted to 2.0.0 and in that process I seem to have gotten the error:Program "gcc" not found in PATH - "[Discovery options] page in project properties"I followed the instructions found in the stm32CubeIDE user guide https://www.st.com/resource/en/user_manual/um2609-stm32cubeide-user-guide-stmicroelectronics.pdf, but the problem persists.Is there a way to manually clear this from any of the configuration XMLs perhaps?The code does still compile and creates binaries, so the error is just a nuisance.[Edit] spellingBest regardsMartin 

在官网下载的CLT工具链安装包，管理员运行后显示如图报错该怎么办？

So, when I was actually trying to work on the STM32F103CBT6TR MCU which is placed onto the custom board while I tried to toggle PA1 in it which only gave floating values then in debug, I found out it is going into Hard Fault Handler so then again I replaced them with new IC but again it is coming as same Hard Fault Handler. I also tried for pin 0 too. What can we do to solve this?

Split from How to make the FDCAN with DMA - which is solved.TL;DR: STM32s don't support DMA for CAN or FDCAN.Hi, thanks for the clarity.But, the doubt I'm having was, then how to handle that much data rate like eg.5mbps in FDCAN without any help of DMA?In my case, the CPU will be busy with doing other high priority works, so how can I handle the CAN data without the intervention of CPU.Kindly, suggest a way to resolve this. Thanks

I updated the STM3CubeProgrammer to version v2.22.0, and after restarting the app, it started showing a blank page. I was also not able to close the application directly. I had to end the task from the task manager. I uninstalled the latest version and installed v2.20.0, but it is also showing a blank page. Also, in STM32CubeIDE, I am not able to run and debug code. It gets stuck at  "Download in progress". Are both problems related?

Board: STM32N6570-DKIDE/Toolchain: Zephyr RTOS + west, Zephyr SDK 0.17.4Repository: stm32-hotspot/zephyr-stm32n6-ai-people-detectionOS: Windows 11I am trying to run the STM32N6 AI people detection application in flash boot mode using MCUboot RAM_LOAD as described in the README. The application runs correctly in debug/development mode (west debug), but does not boot at all when built and flashed for boot-from-flash mode.No output is seen on the UART console after the MCUboot banner — inference never starts.Steps to Reproduce:Following the README exactly:# 1. Build west build -b stm32n6570_dk --sysbuild app -DSB_CONFIG_MCUBOOT_MODE_RAM_LOAD=y -p # 2. Flash model weights (board in Development Mode: Boot 0: L and Boot 1: H) west flash-weights # 3. Flash MCUboot + app (board in Development Mode) west flash # 4. Move boot switches to Flash Boot mode (Boot 0: L and Boot 1: L) # 5. Power cycle ``` --- ### Observed Behaviour After power cycle in Flash Boot mode, the UART console is silent — n

Hello, I'm using the STM32N6570-DK board.I completed the TOUCHGFX project through this URL, and there was a DEBUG problem while I was making CUBEMX and CUBEIDE operating it.(https://community.st.com/t5/stm32-mcus/how-to-debug-stm32n6-touchgfx-applications-using-stm32cubeide/ta-p/845901)I'm not moving from FSBL to APPLI. I'm asking for your help on this part.Please check.

Hi, I tried to configure the USB Host for a project, but it’s difficult to configure, and I am not finding any online support available for the STM32H5 series controller. My requirement: I configured PB4 as an output, which is connected to a green LED. Whenever I plug the USB into my controller, the green light turns on, and whenever I remove the USB from the controller, the green light turns off. I’ve tried some methods, but they are not working. Can anyone please help me?      

I have been using a STM32F469i Disco board to interface with a 800x800 LCD in MIPI DSI video burst mode. In order for this to work from the DISCO board, I have an adapter PCB to connect from the DISCO board to the LCD.It took a while to get going, but the setup with the custom LCD and DISCO board is working fine.I have recently moved to a custom board; I have intentionally made the video dependent aspects of this board the same as the disco board so that I can run my software on either the disco or the custom board without making any changes.The custom board sometimes exhibits odd behaviour when rendering from the SDRAM via the LTDC.Enabling the test colour bars from the DSI always display ok without any problem.From SDRAM/LTDC, the horizontal start point shifts from the left side of the display to some arbitrary location across the width of the display. This occurs randomly, it can be running for hours and remain stable, other times it may shift once, then be stable for a time and may