Transfer time when using DMA and ADC

[LSoll.1]

Hi, I am using a STM32F407. I programmed one channel of ADC1 to start on the falling edge of TIM8, which has a frequency of 100 kHz. So far so good, the ADC performs a conversion every 10 us (100kHz), measured with the HAL_ADC_ConvCpltCallback function call.

The problem is when I want to use the DMA. When trying to transfer the data to memory with DMA, the time it takes increases to 120 us and if instead of converting 1 channel, I convert 2, the time goes to 60 us (the more channels I convert, the faster it is)

Could someone help me understand what is happening?

I leave the code below.

[TDK]

You're overloading the processor with interrupts. Convert more values per DMA call instead of just a few.

Having an interrupt at 100kHz is also likely swamping the main loop and preventing progress, especially since it's going through the HAL logic. There is no need to have HAL_TIM_PWM_PulseFinishedCallback defined. You can verify the pulse on a scope or logic analyzer.

Slow down the sample rate by an order of magnitude until you've verified your logic is correct.

[LSoll.1]

Hello!

I'm not sure that's the problem, I reduced the frequency of the PWM that fires the ADC to 10kHz and removed the HAL_TIM_PWM_PulseFinishedCallback() function and noticed that it enters the HAL_ADC_ConvCpltCallback() function every 1200us (833Hz). I don't understand the relationship between the number of channels that convert and the time it takes. I would have ensured that the more channels you want to convert, the longer it would take.

Inside the while(1) loop there is absolutely no going.

What I need is quite simple, a PWM No Output channel at 100kHz that will trigger the ADC on the falling edge and have the DMA transfer that data to an array. All at a frequency of 100k! But even though I reduce the sample rate, the rate at which DMA transfers data is less than PWM.

I can't understand what's going on!