MSP430FR2433 SleepSeconds 18ua

[Jamesgrr 1]

Hi,

I'm working on a board with an FR2433, and also doing testing with the FR2433 Launchpad.

I am trying to troubleshoot high current using SleepSeconds()

With Suspend, I can get 1.5ua but sleep seconds is around 18ua, I believe it should be far lower based on the datasheet for LPM3. 

I am using a ucurrent gold so feel confident about the measurement.

I set all my pins (1-22) to OUTPUT LOW and have also tried with INPUT_PULLDOWN

Any assistance here would be greatly appreciated. Thankyou.

 

 

 

[Jamesgrr 1]

Using Driverlib without energia I am able to get VERY low consumption (less than .5ua)

Would love to know how this can be integrated with Energia SleepSeconds. Thanks. 

    WDT_A_hold(WDT_A_BASE);
    P2SEL0 |= BIT0 + BIT1;                  // P2.0: XOUT; P2.1: XI1

    //CSCTL4 = SELMS__DCOCLKDIV | SELA__XT1CLK;  // MCLK=SMCLK=DCO; ACLK=XT1
    // Port Configuration all un-used pins to output low
    P1OUT = 0x00;
    P2OUT = 0x00;
    P3OUT = 0x00;
    P1DIR = 0xff;
    P2DIR = 0xff;
    P3DIR = 0xff;

    PM5CTL0 &= ~LOCKLPM5;                   // Disable the GPIO power-on default high-impedance mode
                                            // to activate previously configured port settings

    PMM_turnOffRegulator();
    __bis_SR_register(LPM3_bits + GIE);     // Enter LPM3
    __no_operation();                       // For debug

 

[Jamesgrr 1]

I think messing with the power regulator is no good. My 'implementation' of it with Energia breaks SleepSeconds and the device restarts. Any assistance with getting current using SleepSeconds below 17ua would be greatly appreciated.

[NurseBob 111]

Have you looked at what's happening "under the hood" with CCS or similar? All things considered, I find it amazing that the current is as low as it is.  Energia is a great place to start and experiment, but it's not intended to replace coding that is far closer to the metal, er, silicon. Given it's heritage as a msp430 version of the Arduino, it's a very impressive piece of work.  But, to manage the supporting classes, and the ever-present loop, it is constrained if your're looking for ultra-low power apps.  You have already demonstrated that differential with your driverlib version.

[Jamesgrr 1]

Yes this is the conclusion I have come too. I believe Energia'  implementation of the Watchdog here is what causes the higher usage. It is very impressive that it manages to get even close. I think it is abit misleading to call SleepSeconds() true LPM3 though. 

It would be great to see some other implementation though; as the MSP430 is capable of sub 1ua sleep. 

Perhaps I will experiment with a library in the future to make this possible. 

Thanks for your response. 

[LIJsselstein 9]

Energia on the msp430fr2xxx-4xxx range falls back to using the REFO 32kHz clock when the external 32kHz crystal is not populated. REFO is a nice stable/accurate clock, but when you look it up in the datasheet (e.g. p24 in FR2433) you’ll notice the consumption is 15uA. See also line 139 in wiring.c.

So, to get really low power with Energia (~1uA in LPM3 with sleep/sleepSeconds), either customise wiring.c to use VLO clock instead of REFO (but you loose clock accuracy) or add external LF crystal. Using driverlib/CCS without ext. LF crystal is basically the same: lowest power=inaccurate clock, a bit less low power=accurate clock.

[Jamesgrr 1]

Thanks, this is indeed the case. 

I believe this is the section in question here.
It's abit weird because the comment above suggests that it should fall back to VLO? But actually uses REFO. Additionally, in 'MSP430FR2433.h" there is no define for setting ALCK to VLO, unlike the FR5xx boards.

/* If starting the XTAL timed out then fall back to VLO */
	if(!timeout) {
		/* ACLK = VLO = ~ 12 KHz */
		vlo_freq = 8000;
		/* Source ACLK from REFO */
		CSCTL4 |= SELA__REFOCLK;     
      
      

If I look at where SELA_REFOCLK is defined, I can see there is no way to direcctly do this

#define SELA__XT1CLK           (0x0000)       /* ACLK Source Select XT1CLK */
#define SELA__REFOCLK          (0x0100)       /* ACLK Source Select REFOCLK *

This is unlike the FR2355 which has 

#define SELA__XT1CLK                     (0x0000)        /* XT1CLK with divider (must be no more than 40 kHz) */
#define SELA__REFOCLK                    (0x0100)        /* REFO (internal 32-kHz clock source) */
#define SELA__REFOCLK_H                  (0x0001)
#define SELA__VLOCLK                     (0x0200)        /* VLO (internal 10-kHz clock source) */
#define SELA__VLOCLK_H                   (0x0002)
#define SELA__RESERVED                   (0x0300)        /* Reserved */
#define SELA__RESERVED_H                 (0x0003)

I have tried defining VLOCLK to use in Wiring. But this has been unsuccessful.

I am not too familiar with the 2355 so will need to see why this is different, but a cursory look at the datasheet doesn't spot anything too different in their Clock System. 

 

Any insights as to why these are different and how to use VLO for ACLK would be greatly appreciated.

(here is also the chance that I have dived into a rabbit hole here and am completely misunderstanding this)

[Jamesgrr 1]

Based on what I am learning here: https://e2e.ti.com/support/microcontrollers/msp430/f/166/p/749845/2768754#2768754

It appears that the "Enhanced Clock System" isn't available on this chip. Which is what limits ACLK to REFO.

Will update if I learn more about using it with the RTC module.

[LIJsselstein 9]

I believe that the documentation in wiring.c is wrong so I suggested a fix for that in this PR. I didn’t know that VLO could not be chosen for ACLK on the FR2433...so thanks for that.

This topic on E2E essentially mirrors this thread.

[Jamesgrr 1]

Yes, I have been asking around to try find out as much as I can about this. From Energia and Ti perspective.

 

Thanks for that PR.