Issue with using Triger MUX with ADC ISR

[qingg]

Hi,

I'm using the i.MX7ULP evaluation board, IAR Embedded Workbench Arm 8.50.1, and J-Link Plus for a project. And I encountered several issues where the ACD interrupt is not called.

I used example trgmux_lptmr_trigger_lpit under .\boards\evkmcimx7ulp\driver_examples\trgmux\lptmr_trigger_lpit in NXP’s official SDK_2.9.0_EVK-MCIMX7ULP as a starting point for my project (will be referred to as “Example 1” below). I also added some code from lpadc_single_interrupt under .\boards\evkmcimx7ulp\driver_examples\lpadc\single_interrupt in the same SDK (will be referred to as “Example 2” below). I needed both functionalities of timer trigger and ADC interrupt.

Specifically:

• I moved the initialization portion of the code in Example 2's main() to a new function called AdcInit(). I then added a call to this function from DEMO_InitLptmrTrigger() which is called by main() in Example 1.
• I made the following changes to DEMO_InitLptmrTrigger() from Example 1:

mLpadcConfigStruct.FIFOWatermark = 15U; /* Set watermark as 15U. */
LPTMR_SetTimerPeriod(LPTMR0, 1600U);​

The first statement above is to get 16 samples before the ADC triggers an ISR. The second statement above is to set the sampling period to 100 microseconds (16 MHz / (1 s / 100 us)).

• I added ADC1_IRQHandler()/DEMO_LPADC_IRQ_HANDLER_FUNC() from Example 2 to Example 1
• I added my own code to ADC1_IRQHandler() that reads all 16 values from ADC buffer when ADC1_IRQHandler() is called (which means the ADC buffer is full with 16 values):

    g_LpadcConversionInProgressFlag = true;
    int i = 0;
    while (g_LpadcConversionInProgressFlag)
    {
        g_LpadcConversionInProgressFlag = LPADC_GetConvResult(DEMO_LPADC_BASE, &g_LpadcResultConfigStruct);
        local_data[i] = ((g_LpadcResultConfigStruct.convValue) >> 3U);
        i++;
    }

• I added my own code to ADC1_IRQHandler() that toggles a GPIO pin whenever ADC1_IRQHandler() is called:

    /* Toggle GPIO to PIN J11-1*/
    GPIO_PortToggle(BOARD_LED_GPIO, 1u << BOARD_LED_GPIO_PIN);
    SDK_ISR_EXIT_BARRIER;

• I removed function LPIT0_IRQHandler() in Example 1.

The resulting code is attached.

The only way to have ADC1_IRQHandler() called consistently, as observed through the GPIO toggle, is to reset the board by pressing the physical reset button.

If I stop debugging from IAR, and restart debugging, then ADC1_IRQHandler() is still called.

If I try to load another workspace, and then reload this workspace without resetting the board, then ADC1_IRQHandler() is not called.

If I reset the board and start debugging, ADC1_IRQHandler() is called again.

If I choose Reset in IAR's Debug menu, it will jump back to main(), and when I choose Go in IAR's Debug menu, ADC1_IRQHandler() is not called anymore.

In summary, either of these two things makes ADC1_IRQHandler() stop being called anymore: 1. Load another workspace and reload this workspace. 2. Reset in the debugging process so it jumps back to main() and continue debugging.

Can you help with this issue? Thank you!

[davidguinn]

Hi,

I'm Qing's colleague, I looked into the issue, a long story short, if the system is reset when the NVIC's ADC1 interrupt is active(i.e. NVIC_IABR1's ACTIVE59 is set to 1), then the interrupt will never be executed again.  The ADC1 active's bit is not cleared if there is a reset in middle of the interrupt(i.e. use the reset processor button in the IDE).  Once that active bit is set for an interrupt, according to ARM's v7 documentation:

If an interrupt is active when it is disabled, it remains in the active state until this is cleared by a reset or an exception return.

Therefore as a temporary work around, I added code to detect this issue and do a reset:

if(NVIC_GetActive(DEMO_LPADC_IRQn))
{
  NVIC_SystemReset();
}

This does fix the issue, the Active bit is cleared and the ADC interrupt is cleared.  Is there an alternative way to clear NVIC_IABR1 bit(s) than a system reset?

Thank you,

David