CPU does not wake up to run the Task after a post to the Task from an HWI.

Bryon,

Thanks for your very detailed problem description.  

It is good to hear that this is working as expected at the default MCLK and voltage level.  

When do you boost MCLK and VCORE?  In main()?

When you change MCLK are you also changing SMCLK?  I’m wondering if maybe a driver is assuming the slower SMCLK rate, and a faster rate is somehow affecting timing and the LPM transition.  This is just a guess.

If you are changing ACLK or SMCLK rates from the default, these need to be specified using the “ClockFreqs” module.  Are you doing this?  This module shows up in the “Device Specific Support” page in XGCONF, as “MSP430 Clock Frequencies”.   The values can be entered in the GUI, or by directly editing the application .cfg file, for example:

var ClockFreqs = xdc.useModule('ti.sysbios.family.msp430.ClockFreqs');
ClockFreqs.ACLK = 31250;
ClockFreqs.SMCLK = 400000;

A new MCLK frequency should be configured in the “Basic Runtime Options” page in XGCONF, or this can be specified in the .cfg file too, for example:

BIOS.cpuFreq.lo = 16384000;

Scott

Hi Scott,

Thanks again for looking into this issue.

Q: When do you boost MCLK and VCORE?  In main()?

A: I step VCORE up 1 level at a time using the PMM_setVCoreUp API and checking for STATUS_SUCCESS before boosting MCLK.  All of this is done within main prior to BIOS_start.

Q: When you change MCLK are you also changing SMCLK?

A: I am only changing MCLK also configured the new MCLK frequency in XGCONF.  SMCLK is kept at default 8192000 Hz.  I tried adding in the ClockFreqs module anyways even though ACLK and SMCLK are kept at default.  The issue persists.

I have several I2C devices hooked up to the MCU module.  I then tried dividing down the SMCLK by different factors and at 1024000 Hz the issue I was having for one particular I2C device was gone.  However, for a different one, it was present again.

Bryon

Hi Bryon,

OK, thanks. It seems that what you are doing should work.

I just looked at the device errata (http://www.ti.com/lit/er/slaz288l/slaz288l.pdf).  There are a few issues related to PMM and wakeup from LPM modes.  Do you think any of these might apply for your use case?  Are you using SVS?

Also, can you please post the code snippet you are using to boost MCLK and Vcore?

Thanks,
Scott

Hi Scott,

I had a look at the device errata and they do not appear to be related.  SVS is left at default and I believe it is enabled.  Below is a code snippet of how I am boosting Vcore and then MCLK.

	// Local variables.
	int i = (PMMCTL0 & (PMMCOREV0 | PMMCOREV1));	// Reads the current Vcore.

	// Increase the Vcore level by one and check if successful.
	for (i += 1; i <= Target_Level; i++) {
		if (STATUS_SUCCESS != PMM_setVCoreUp(i)) {
			return STATUS_FAIL;
		}
	}

	return STATUS_SUCCESS;

Once Vcore has been successfully set,

	UCS_clockSignalInit(UCS_MCLK, UCS_DCOCLK_SELECT, UCS_CLOCK_DIVIDER_1);

Thanks,

Bryon

Bryon,

OK, thanks for the additional information.

I need to discuss this with the device apps team to see if they have any ideas why the delay is needed.  I will get back to you as soon as I know more…

Scott