AGX Industrial not booting at negative temperatures

Hi Nvidia Team

We have a custom carrier board design, which we already used with the Jetson AGX Xavier Modules for quite some time. Now we started internal tests with an AGXi Module and saw in a temperature oven, that at low temperatures (around -25°C), our system is not booting with the industrial version. Checking the Debug Interface, we see sometimes just a few lines of output, and most of the time no output at all. The board supplies are powered on also if the device is not booting, so the CARRIER_PWR_ON Signal is active and stays active. Is there a software temperature trip level, that could keep the device from booting? What else could lead to a stop in the boot process of the device?
In other temperature regions, the module always boots without issues (-20°C to 70°C).
Any help in this topic is really appreciated.
Thank you.

Kind regards

Hi, it looks like issue of custom board at low temperature. AGXi module can work at -40C. Please check your custom carrier board.

Hi Trumany
The custom board starts with the Xavier AGX Module just fine at -25°C. What could be different between power-on of the AGX and the AGXi?

When the device is not booting, we could measure that CARRIER_PWR_ON is high and SYS_RESET_N is also high as described in the power on sequence. The carrier board voltages are also enabled and up.

Hi Trumany
We checked the Power-On Sequence and we cannot see any difference between a successful boot at higher temperatures and a non successful boot at very low temperatures. Will the Jetson Module release the SYS_RESET_N signal even if something before went wrong? Do you have other ideas what we could check?
Thank you.

Kind regards

We checked now if the system is not booting only when the AGXi is cold or only if our carrier board and the AGXi are cold. If the AGXi only is cold, the system still boots up normally. But now when pressing the reset button, the system does not boot until the reset is pressed again. Does this give you any indication what could go wrong?
Thanky you.

Hi, have no clue in your tests. Do you have another AGXi module to do cross-check? Basically AGXi module is validated at low temperature, so you should focus on your custom carrier board. It is hard to tell what caused the boot failure per your description. If you can use another module or validated carrier to do cross-check, that will be helpful.

Hi Trumany
We saw this behavior already with two AGXi modules and we tested a carrier board which was not booting with the industrial version with an AGX without seeing any problems. Resetting the module does only power off and on the Carrier board by the CARRIER_POWER_ON signal, right?
We sent our schematics to NVidia for checking. Let’s hope they find something.
Thank you.

Hi, where did you send the schematic to? Is there any feedback?

Hi, do you mean if your carrier is cold, the booting will fail no matter module is cold or not?

Have you checked all carrier power supplies to make sure no one is up before carrier_pwr_on asserted? As it needs to be guaranteed that between module and carrier, the shared IO pins status won’t be affected during power-on.

Hi Trumany
We got now one step further. When disconnecting the Goldcap attached to the VCC_RTC pin, the system always boots at low temperatures with the AGX industrial module. Do you have any explanation for this? Did you do your internal tests with a battery/goldcap connected to the DevKit?
Thank you.

We are checking internally, will update once available.

Hi, can you share the spec of “Goldcap”?

Hi Trumany
We have two types that we use (see the datasheets):
DHL-5R5D105T: https://www.elna.co.jp/en/capacitor/new/pdf/dhl_e.pdf
KW-5R5C105-R: https://www.eaton.com/content/dam/eaton/products/electronic-components/resources/data-sheet/eaton-kw-supercapacitors-coin-cells-data-sheet.pdf

Kind regards

Hi Trumany

The same boot problems occur when a battery is connected to the VCC_RTC Pin.
Battery Type is the following: https://www.renata.com/en-ch/product-detail/cr2477n/

Hi, have you probe the power on sequence? Can you share the waveform of power on sequence similar to the sequence listed in product design guide? Especially the timing between VIN_PWR_BAD and MODULE_PWR_ON? The VIN_PWR_BAD is expected to be de-asserted before others.

Hi Trumany
We checked the timing and there is no difference between a successful boot and a non-successful one. We measured the current on the VCC_RTC Pin and saw, that the industrial Xavier needs currents up to 0.4 mA (at 2.5V). This we were able to reproduce on the NVidia DevKit. The AGX Module however draws the expected current around 2uA. With the AGX Module we saw that the current changes the leading sign when it is turned on, meaning the Module charges the battery. Shouldn’t this feature be disabled by default? In the device tree it is disabled.
Do you have an explanation for this?

Hi, can you help double confirm if the power drawn is when nothing is connected but only a bbatt or the system is fully powered and bbatt is also there and drawing 0.4mA?

Hi Trumany

What do you mean with nothing connected? No module? Please explain what we have to double confirm.
There is another thread with the same topic: XAVIER Industrial RTC current consumption high
Did you internally reproduce the issue?

I mean no other power supply to the board?