AM625: [AM6202Q] Wakeup source from DeepSleep power mode

Hello Marwan

Could you please look into the below document.

Enabling Low Power Embedded Systems With AM62x Processors (ti.com)

Regards,

Sreenivasa

Hello Sreenivasa,

I have already looked into this white paper, still it didn't answer my questions.

Moreover, is it possible to transition from SeepSleep to Power Off based on some timers happening while being in DeepSleep?

Thanks,

Marwan

Hello Marwan

As i understand we are still evaluating the low power modes and we may have some additional documents coming.

Marwan Farrah said:

Moreover, is it possible to transition from SeepSleep to Power Off based on some timers happening while being in DeepSleep?

Can you please elaborate your requirement and what you meant by power off?

Regards,

Sreenivasa

Hi Sreenivasa,

I am speaking from SoC power modes point of view, so the question is could the SoC while being in DeepSleep power mode have some timer that would trigger the SoC to go to PowerOff power mode? or it needs at least to be in MCU Only mode in order to do such functionality and transition to PowerOff after a timer timeout?

Thanks,

Marwan

Hello Marwan

Curious to understand the usecase, are you using any device that supports the transition from DeepSleep to power off?

Regards,

Sreenivasa

Hi Sreenivasa,

The usecase to go to sleep is for later fast startup, but if the system has been dormant for long time, it shall shutdown to reduce power drastically.

Regards,

Marwan

Correct, but PMIC would be controlled from SoC, that's why I was asking if DeepSleep mode is capable to do that.

Thanks,

Marwan

Hi Marwan,

As Sreenivasa mentioned, AM62x doesn't support directly transition from DeepSleep mode to PowerOff mode. But I think there is a possible workaround: to transition DeepSleep mode to Running mode then to PowerOff mode.

When Linux needs to go to DeepSleep mode, use command "rtcwake -m mem -t <time-to-wake-up>" to put the system to DeepSleep mode. When it is resumed, Linux checks the wakeup event source, if it is RTC, it means the system has been in DeepSleep for awhile and need to shut down. Otherwise, it means the system is resumed by an external event, need to keep in running mode.

Please note that this is just a theory, it hasn't been validated, and is not supported in current processor SDK. There are at least 3 factors to consider:

- DM firmware provides an API to query the wakeup event source, see
  https://downloads.ti.com/tisci/esd/latest/2_tisci_msgs/pm/lpm.html#tisci-msg-lpm-wake-reason
  I believe it has been used in kernel, but kernel doesn't expose the information to user space;

- When the system is resumed by an external event, I am not sure how Linux handles the previous rtcwake command. Has RTC been gracefully cleanup up and ready for the next use?

- It is not implemented/supported in current Processor SDK, need a prototype to prove it is a solution, so it is a custom implementation on your side.