AM625: GPIO's impedance change in DeepSleep mode

OLEKSIY KASYAN

Part Number: AM625

Tool/software:

Hello support team,

We are encountering an issue with the GPIO behavior on our product based on the AM62 SoC.

When the SoC enters DeepSleep mode, the voltage level drops on the pins set to High. This voltage drop depends on the load/pull-down resistor attached to the SoC's pin, leading to problems.

For instance, a USB device (flash drive) is connected to USB0. The USB0_DRVVBUS pin of the SoC controls the USB power switch and has an external 10 kOhm pull-down resistor to the ground.

In active mode, the voltage on the USB0_DRVVBUS pin is 1.8V. However, when the SoC enters DeepSleep mode, the voltage on the USB0_DRVVBUS pin drops to 0.3-0.5V, causing the USB power switch to disconnect power from the attached USB device. This results in the SoC hanging, and it is impossible to recover from DeepSleep mode without resetting the SoC.

To solve this issue, I replaced the 10 kOhm pull-down resistor on the USB0_DRVVBUS pin with a 100 kOhm resistor. This resolved the SoC hang issue, but I have concerns regarding the reliability of this solution.

Measurements on the USB0_DRVVBUS pin with the 100 kOhm pull-down resistor show a voltage drop of 0.35V when the SoC enters DeepSleep mode. According to my calculations, the internal impedance of the USB0_DRVVBUS pin changes from almost 0 Ohms to 24 kOhms.

According to the configuration in the PADCONFIG register for the pin USB0_DRVVBUS it should retain its state when the SoC goes into the DeepSleep mode.

Identical behavior is valid also for the USB1_DRVVBUS pin and I suspect, for other GPIO-capable pins.

Is this behavior expected when the SoC enters DeepSleep mode?

What happens to the SoC's pins when it goes into DeepSleep mode?

I could not find any explanation of this behavior in the SoC's datasheet or reference manual.

Thank you.

over 1 year ago

0 Anshu Madwesh over 1 year ago

TI__Mastermind 18190 points

Hello,

Can you confirm that this is the same behavior that you're seeing with the GPIOs and Reset signals in these threads? Are there any notable differences?

https://e2e.ti.com/support/processors-group/processors/f/processors-forum/1361669/am625-gpio-output-state-in-deep-sleep
https://e2e.ti.com/support/processors-group/processors/f/processors-forum/1360238/am625-peripherals-goes-into-reset-state-when-processor-goes-into-deepsleep-mode

Thanks,

Anshu

0 fd over 1 year ago in reply to Anshu Madwesh

Intellectual 685 points

Anshu Madwesh

Anshu Madwesh said:
Can you confirm that this is the same behavior that you're seeing with the GPIOs and Reset signals in these threads? Are there any notable differences?

We do not know for sure. It's related to the same topic, but at the moment whether this is the same issue or another one is not clear.

One known difference is that USB0_DRVVBUS is not used as a GPIO.

0 Anshu Madwesh over 1 year ago in reply to fd

TI__Mastermind 18190 points

Hi fd,

Thanks for the clarification. If you could provide a schematic, that would be helpful in debugging this further.

Thanks,

Anshu

0 fd over 1 year ago in reply to Anshu Madwesh

Intellectual 685 points

Anshu Madwesh said:
Thanks for the clarification. If you could provide a schematic, that would be helpful in debugging this further.

USB_2_EN is connected with a 22ohm series resistor to USB1_DRVVBUS (F18)

0 Anshu Madwesh over 1 year ago in reply to fd

TI__Mastermind 18190 points

Thanks fd. There is a few other PADCONFIG settings that I'm working on testing. Will continue to experiment.

0 peaves over 1 year ago in reply to fd

TI__Guru 62855 points

The output buffer in each IO cells is disabled when the device enters Deep Sleep. The internal pull remains on to hold the signal in a known state when the device enters Deep Sleep. It can be configured as a pull-up or a pull-down. For this use case, using an external pull-down on the signal is problematic when the internal pull is trying to hold the DRV_VBUS signal high.

The DRV_VBUS pin is one of the few pins that has the internal pull turned on by default. The internal pull will be turned on and configured as a pull-down as soon as power is applied. This is done to ensure the VBUS is not sourced before software is ready. The external resistor could be removed and only use the internal pull to hold the VBUS switch off until software configures the IO and drives the output high to turn on VBUS. The internal pull could be changed from pull-down to pull-up at the same time. This would allow the internal pull to continue holding the signal high when going into Deep Sleep.

Regards,
Paul

0 fd over 1 year ago in reply to peaves

Intellectual 685 points

Hello peaves

peaves said:
This would allow the internal pull to continue holding the signal high when going into Deep Sleep.

Our system is a modular and general-purpose system. In general, the expectation is that the pins output state is preserved when entering deep sleep, including the output state and driver mode (whatever is push-pull, opendrain, internal pull up/down, and output buffer value).

Is this comment generic or this is specific to the DRVVBUS signals?

0 fd over 1 year ago in reply to peaves

Intellectual 685 points

Hello peaves

peaves said:
The output buffer in each IO cells is disabled when the device enters Deep Sleep.

Do you have any more details on this? Do I understand correctly that this implies that all the pins are configured as inputs when the device is in deep sleep? Can you share any reference to official TI documentation in this regard?

0 fd over 1 year ago in reply to fd

Intellectual 685 points

peaves

one more question, we would need some official documentation from TI in which this behavior is documented. Can you please provide a reference to it?

0 peaves over 1 year ago in reply to fd

TI__Guru 62855 points

I'm not aware of any documentation describing the Deep Sleep IO operation. I had to ask one of the systems engineers before answering your question. I'm currently on vacation and will not be in the office for a couple of weeks. I can try to find someone to begin creating a description.

Processors

Processors forum

AM625: GPIO's impedance change in DeepSleep mode