STM32L433 Delay before Timer DMA starts

[JMala.3]

I am having a problem using Timer1 DMA to create a waveform used by one of our peripherals. I have set the timer to count up to 65K with a 1uS tick, and channel 4 as a comparator with a toggle output.I am generating the time between pulses in a table, then using DMA to have the timer generate the waveform, For example if I generate a table with 10, 15, and 25, the waveform goes high after 10uS, low after 15uS, and high again after 25uS.

The code below works, but there is a 65mS delay before the waveform starts. I suspect the timer is counting form 0 to 65K, then being reloaded, then the waveform starts. I need the waveform to start immediately. Any idea why this is happening?

void SendMessage(uint32_t message)
{
    CreatePeriodTable(message);
 
    MODULE_DATA_EN_HIGH;
 
    TIM1->CR1 &= ~TIM_CR1_CEN;  /* Disable the Timer */
    TIM1->CNT = 0;
    TIM1->CCR4 = 0;
    TIM1->CR1 |= TIM_CR1_CEN;     /* Enable the Timer */
 
    HAL_TIM_OC_Start_DMA(&htim1, TIM_CHANNEL_4, PeriodTable, TABLE_SIZE);
}

[Piranha]

It's happening because TIM_CR1_ARPE is enabled and can be solved by triggering TIM_EGR_UG before enabling the timer.

[waclawek.jan]

TIMx_CR1.ARPE does not influence direct writes to TIMx_CNT, so that's probably not the issue here.

After enabling the counter by setting TIMx_CR1.CEN it starts to count up, so by the time the Cube/HAL function enables DMA in TIMx_DIER, TIMx_CNT is already > 0 so it will match TIMx_CCR4 only after the wraparound.

One way to tackle this would be changing TIMx_CCR4 to 1 and moving the enabling of counter after calling the HAL_TIM_OC_Start_DMA() function.

JW