TIM4 counter stops when ADC1 DMA less than 3 channels used.

KYou.1 · ‎2023-09-11

I know this is very weird behavior but it is quite easily reproduceable error.

I am running Windows 11 on a laptop.

CubeMX version : 6.9.1

CubeIDE version : 1.13.1

Board I am using is this. Looks like genuine STM32F103C8T6 chip.
I could upload USB_DEVICE successfully with this chip.

For the test, I only enabled TIM4 counter up mode with Global interrupt and ADC1 as DMA.

ADC channels are set as below.

I made a git repository here for review.

https://github.com/kyuhyong/stm32_adc_test

TIM4 on this project is not working when only 1 or 2 channels are enabled. (Only 1 ISR is called then stop)

If I enable ADC1 with more than 3 channels, TIM4 works as expected. (checkout to adc3 branch)

Very strange bug I guess. Any solutions?

TDK · ‎2023-09-11

> HAL_ADC_Start_DMA(&hadc1, (uint32_t*)adc_vals, 1);

It sounds like your CPU is swamped by interrupts and can't make progress in the main thread. Reduce the sample rate, or reduce the number of times the HT/TC interrupts are called by performing more than 1 conversion per DMA call.

If you feel a post has answered your question, please click "Accept as Solution".

View solution in original post

TDK · ‎2023-09-11

> HAL_ADC_Start_DMA(&hadc1, (uint32_t*)adc_vals, 1);

It sounds like your CPU is swamped by interrupts and can't make progress in the main thread. Reduce the sample rate, or reduce the number of times the HT/TC interrupts are called by performing more than 1 conversion per DMA call.

If you feel a post has answered your question, please click "Accept as Solution".

waclawek.jan · ‎2023-09-11

> Looks like genuine STM32F103C8T6 chip.

Unlikely.

> TIM4 on this project is not working when only 1 or 2 channels are enabled.

If ADC clock/sampling rate is high and DMA is circular, together with rather bloated Cube/HAL the DMA ISR may simply take up all the processing time.

Set a GPIO up at beginning of DMA ISR and set it down at the end, and observe using LA or oscilloscope.

JW

KYou.1 · ‎2023-09-11

Oh yes. Increasing sampling time from 13.5 cycles to 28.5 cycles solves the problem.
I never thought high sampling rate can cause CPU to halt rather then not updating ADC values.

Thanks!

KYou.1 · ‎2023-09-11

You think this chip is fake?

I have some experience with fake chips cannot handle USB CDC or HID device properly but this chip is working OK even when I setup as USB HID device.

waclawek.jan · ‎2023-09-11

> You think this chip is fake?

The linked discussion indicates, that it's very likely. I don't have personal experience.

JW