AM625: GPIO State Retention In Deep Sleep

Hi,

Which SDK are you currently using in your system?

Regards,

Nitika

the last 9.1 release, but we can use whatever newer release we want.

Hello Francesco,

Thank you for your response.

Please expect delayed responses as our assigned expert is out of office.

Thanks for your patience.

Regards,

Vaibhav

Hi Francesco,

It is expected that the GPIO state is the same before, during, & after Deep Sleep when using the default image.

I ran a simple test to verify this behavior on a TI EVM and the default image from SDK 9.1.

I set the GPIO pin voltage to High/Low or left hanging, then I entered Deep Sleep and continued to measure the voltage as it wakes up.

Using echo mem > /sys/power/state:

I used a MCU GPIO as a wake up source.

Using rtcwake -s 10 -m mem:

There are unexpected results when using RTC wake. After the device wakes up, the voltage moves to hanging for a moment then returns to the expected value. This might be caused by the GPIO driver needing a second to resume after Deep Sleep.

I will let the hardware expert comment further.

Thanks,
Anshu

Anshu Madwesh

Thanks for your tests, can you confirm that the GPIO you measured is on the main domain? Asking since the main domain stays in reset during deepsleep. For example, clock outputs from the main domain stop running.

Anshu Madwesh said:

After the device wakes up, the voltage moves to hanging for a moment then returns to the expected value. This might be caused by the GPIO driver needing a second to resume after Deep Sleep.

If I understand this correctly this is problematic, we need to have the GPIO to the exact state that was set before entering the deepsleep, any "glitch" will affect the functionality of our board and will require countermeasures in our hardware to prevent malfunctions, this is very important for us to understand.

Hi Francesco,

One detail that I forgot to mention the LPM device tree overlay was enabled: https://git.ti.com/cgit/ti-linux-kernel/ti-linux-kernel/tree/arch/arm64/boot/dts/ti/k3-am62x-sk-lpm-wkup-sources.dtso?h=ti-linux-6.1.y

fd said:

can you confirm that the GPIO you measured is on the main domain?

Yes this is correct. I was using GPIO0_42 of the Main Domain which is located on the User Expansion Connector on the SK-AM62B-P1. Here is the schematic: https://www.ti.com/tool/SK-AM62B-P1#design-files

fd said:

we need to have the GPIO to the exact state that was set before entering the deepsleep

Just so I understand correctly, are you using RTC wake up for Deep Sleep?

Best Regards,

Anshu

Anshu Madwesh said:

Just so I understand correctly, are you using RTC wake up for Deep Sleep?

Currently we are testing with RTC wake-up.

But we need to understand how this function works in general, this is a generic product, the wake-up could be from USB, uart, gpio, RTC,  ... 

Hi Francesco,

Thank you for the confirmation. I'll be discussing this with the Software Dev team. Will update this thread when I get new information.

Thanks for your patience,
Anshu