MSPM0C1104: Low power optimization - ADC pulls current high

R H

Part Number: MSPM0C1104

I tried optimizing the power consumption of my device and can get as low as ~4.5uA on my custom hardware in STANDBY1. When requesting a short ADC measurement. however, my current consumption jumps to over 1mA. I did narrow it down to a snippet where I shortly initialise my ADC and start the conversion, then after the conversion stop it again.

This is roughly what my low power mode initialisation does.

DL_TimerA_stopCounter(PWM_INST);
DL_TimerA_disablePower(PWM_INST);

DL_I2C_disableController(I2C_0_INST);
DL_I2C_disablePower(I2C_0_INST);

DL_VREF_disableInternalRef(VREF);
DL_VREF_disablePower(VREF);
DL_ADC12_disableConversions(ADC12_INST);
DL_ADC12_disablePower(ADC12_INST);

delay_cycles(POWER_STARTUP_DELAY);

This is my problematic code. Inserting a delay_cycles with ~300us or more wait time fixes the issue.

void adc_off(void) {
    DL_VREF_disableInternalRef(VREF);
    DL_VREF_disablePower(VREF);
    DL_ADC12_disableConversions(ADC12_INST);
    DL_ADC12_disablePower(ADC12_INST);
    delay_cycles(POWER_STARTUP_DELAY);
}

void adc_on(void){
    DL_VREF_enablePower(VREF);
    DL_VREF_enableInternalRef(VREF);
    DL_ADC12_enablePower(ADC12_INST);
    DL_ADC12_setStartAddress(ADC12_INST, DL_ADC12_SEQ_START_ADDR_00); // manual repeat of ADC sequence
    DL_ADC12_enableConversions(ADC12_INST);
    DL_ADC12_startConversion(ADC12_INST);
    delay_cycles(POWER_STARTUP_DELAY);
}

void my_conversion(void) {
    adc_on();
    while (interrupt_received == false) {;}
    delay_cycles(24*100*3); // fixes issue
    adc_off();
}

Why exactly could this occur? Why does a delay annul this issue? Why 300us? The time seems quite long to me for a 24MHz device.

When I remove the while loop and the delay, the issue also doesn't occur (no optimization). In that case, however, I assume that the registers have not been processed before turning ADC off again.

CTL0 = 0x05000000
CTL1 = 0x00010000
CTL2 = 0x03000000

4 months ago

0 Janz Bai 4 months ago

TI__Genius 15430 points

Hello RH,

I want to check some details about this issue:

It is normal that the current up when CPU and ADC is involved in to do conversion and calculation, etc.
The current you measured is the instantaneous value or the average value in a long time?
I think you use ADC interrupt to receive the ADC conversion result, right? So you need CPU involved in this process.
It is very strange that after you add a time delay, the power consumption down, because in the time length you do the time delay, the CPU is wake and ADC is also powered up. So I can't understand why the power consumption actually decreased.

BR,

Janz Bai

0 R H 4 months ago in reply to Janz Bai

Prodigy 80 points

Hello Janz Bai,

The 1mA current stays for an indefinite period of time but may be resolved with new ADC measurements it seems.
Yes, I am using ADC interrupts to notify my task of new available data.

I actually might have found a clue. It might be that I am disabling the ADC while a measurement is running. But would the ADC power down not resolve this? My ADC conversion runs for about 100us.

Best Regards,

0 R H 4 months ago in reply to R H

Prodigy 80 points

I should add that my ADC conversion runs as a sequence with 4 total conversions.

Adding stopConversion() at the top of my adc_off() function did nothing.

0 Janz Bai 4 months ago in reply to R H

TI__Genius 15430 points

Hello RH,

Could you please help to explain what is the meaning of "I am disabling the ADC while a measurement is running"?

You can try to reset whole ADC module before enter STANDBY mode, and then re-initialize ADC again after exit from STANDBY mode

And by the way, could you please show the ADC interrupt handle part code to me too? I want to check the interrupt handler setting when you use sequence

BR,

Janz Bai

0 R H 4 months ago in reply to Janz Bai

Prodigy 80 points

Hello Janz Bai,

by disabling the ADC I mean the adc_off function I posted initially. I am starting the measurement, wait for the interrupt to notify me that my measurement is done, then call adc_off. This could potentially happen during the measurement of the second measurement in the sequence.

volatile interrupt_received = false;

void ADC12_INST_IRQHandler(void)
{
    switch (DL_ADC12_getPendingInterrupt(ADC12_INST))
    {
    case DL_ADC12_IIDX_MEM0_RESULT_LOADED:
        interrupt_received = true;
        adc_data_flags |= 0x1;
        break;
    case DL_ADC12_IIDX_MEM1_RESULT_LOADED:
        adc_data_flags |= 0x2;
        break;
    case DL_ADC12_IIDX_MEM2_RESULT_LOADED:
        adc_data_flags |= 0x4;
        break;
    case DL_ADC12_IIDX_MEM3_RESULT_LOADED:
        adc_data_flags |= 0x8;
        DL_ADC12_stopConversion(ADC12_INST);
        DL_ADC12_setStartAddress(ADC12_INST, DL_ADC12_SEQ_START_ADDR_00); // manual repeat of ADC sequence
        break;
    default:
        break;
    }
}

12-bit ADC, no averaging, measuring on channel 2 of the SOT-16 (DYY) package. Using ULPCLK at 24MHz, 24 clock divider. Software triggers, internal VRef at 1.4V.

Best Regards

0 Janz Bai 4 months ago in reply to R H

TI__Genius 15430 points

Hello RH,

Let me check your code, and I will reply to you tomorrow (China time)

BR,

Janz Bai

0 R H 3 months ago in reply to Janz Bai

Prodigy 80 points

Hello Janz Bai,

I'm still awaiting your answer. Did you find anything?

Best Regards

+1 Janz Bai 3 months ago in reply to R H

TI__Genius 15430 points

Hello R H,

Sorry for the late response. I have checked your code. If you want to let ADC work normally, you need put this code "interrupt_received = true;" into the ADC interrupt case DL_ADC12_IIDX_MEM3_RESULT_LOADED: because if you want to disable ADC, you need for all the memory 0~3 is ready. But in your original code, you put the "interrupt_received = true;" in the case DL_ADC12_IIDX_MEM0_RESULT_LOADED: and you use

 while (interrupt_received == false) {;} in you code, this will cause that when the memory 0 is ready, you will disable ADC, then cause issue, because you use 
memory 0~3 in your code. But after you add the time delay, it will give enough time to let ADC complete the conversion of memory 1~3, then there will be no issue.

Best Regards,
Janz Bai

0 R H 3 months ago in reply to Janz Bai

Prodigy 80 points

Hello Janz Bai,

I want to only do the first ADC conversion. Why can I not stop the conversion with DL_ADC12_disableConversions(ADC12_INST) or DL_ADC12_stopConversion( ADC12_INST) without this issue?

void adc_off(void) {
    DL_ADC12_stopConversion(ADC12_INST);
    DL_VREF_disableInternalRef(VREF);
    DL_VREF_disablePower(VREF);
    DL_ADC12_disableConversions(ADC12_INST);
    DL_ADC12_disablePower(ADC12_INST);
    delay_cycles(POWER_STARTUP_DELAY);
}

According to the low power optimization guide this is the correct sequence.

Best Regards,

0 Janz Bai 3 months ago in reply to R H

TI__Genius 15430 points

Hello RH,

You can use the DL_ADC12_disableConversions(ADC12_INST) or DL_ADC12_stopConversion( ADC12_INST) to stop the ADC conversion, this is not the key point of this issue. I have helped you analysis the possible root cause in my last time response. In your application, you used the memory 0~memory 3 to store your conversion result, but you put the flag interrupt_received = true; in the memory0 load interrupt, it will cause "when memory 0 load, the intterupt_recieved will be true, then in your original code (with out time delay) like below figure, CPU will operate adc_off() directly soon which includes the ADC disable conversion. As I said, all of these occurs when memory 0 load, but at the same time, memory 1~memory 3 maybe not ready, that is the reason why you face the ADC usage issue.

And if you add the time delay in the my_conversion(), it will give time to let memory 1~memory 3 ready, then adc_off() operates, ADC runs correctly.

That is the reason why I ask you can put the flag interrupt_received = true in the memory 3 load intterupt.

Can you understanding now? If you have any questions about my explanation, feel free to let me know.

BR,

Janz Bai

0 R H 2 months ago in reply to Janz Bai

Prodigy 80 points

Hello Janz Bai,

I understand that if I put the interrupt_received = true statement into memory 3 load interrupt, I will not require the delay anymore.
The problem is that I wish to only do the conversion of memory 1 in this specific case. But if I stop the conversion after memory 0 load, the current consumption stays at 1mA where it should be <10uA.

What I do not understand is why stopping the conversion and powering down the ADC and VREF does not stop the power flow to the ADC.

Kind Regards,

Arm-based microcontrollers

Arm-based microcontrollers forum

MSPM0C1104: Low power optimization - ADC pulls current high