H7: TIM1-triggered DMA and TIM1-ADC-triggered DMA get out of sync at Debug Halt
Hi,
in a setup on STM32H725 where TIM1 ETR triggers several DMA activities, I observe an effect while debugging that I would rather like to avoid:
Looking at the NDTR of the three DMA, it seems that one extra ADC run (or at least the related DMA transfers) occurs each time the device is halted for debugging. I do not understand how this can happen.
Basic setup: TIM1 is started by ETR. Two DMA channels capture related register values to circular buffers and are triggered by TIM1 TRG and TIM1 CC4 (at CCR4=1). Combined ADC1+2 is triggered by TIM1 CC3 (at CCR3=89). The ADC results (2x 3x 16 bit) are written to a circular buffer via DMA. All DMA are initialized at the same time. The circular buffers are all of the same size.
All timers are halted during debugging (flags set in DBGMCU->APBxFZ1)
Observation: Normally (without any halt introduced debugger via SWD) the three DMA are perfectly in sync and stay in sync. For example, at each DMA Complete of the ADC channel, all NDTR are reset to the initial value (3072 for ADC where during each cycle 3 words are stored, and 1024 for the other two).
However, each halt using the debugger introduces an extra offset: at DMA Complete of the ADC channel, the other two get behind by +1 element. Now when ADC NDTR is 3072, the others are 1, or after next halt 2, and so on.
It seems that ADC DMA runs for one more time although it should be triggered only by the same events that trigger the others? I could understand a steady offset of 1 the other way round when looking to early (ADC behind others because it finishes later) but not this way, and especially not increasing with each halt??
Thanks for any ideas on this (and if more details are needed on the setup, I'll comment)
Kolja