We have been testing our Xavier AGX SOM-based system for temperature tests and came across the following behaviour in our system.
Test condition: The system is maintained at an ambient temperature of 70C in a power-off state. Once the chamber temperature settles, the power is applied to the system and boot-up doesn’t take place. There are no logs in the serial port in this condition. However, the system seems to be operational once the chamber is cooled down to room temperature. Our design follows NVIDIA DG suggested push-button controller circuitry which required CARRIER_PWR_ON from SOM to sustain carrier board power. In our case, the power LED turns ON and back off in a few seconds indicating the SOM may not have asserted the CARRIER PWR_ON signal during the 70C ambient condition.
We have confirmed this in at least 2 sets of units.
- Does the SOM prevent itself from booting up due to any thermal limits set by NVIDIA during boot-up?
- If so, could you please let us know the limits for the same? Also, do we have the controls to override the same?
- Are the limits the same for Xavier AGX Industrial and Commercial?
We would like to have feedback from NVIDIA for the above queries as soon as possible.
- The carrier board is powered using a 12V power supply. The power brick is placed outside the chamber and not subjected to high temperature.
- The system can boot up normally when the chamber is maintained at 25C. Keeping the system ON and raising the chamber temperature to 70C and maintaining it seems to operate fine when tested for a few hours. Both the carrier board and SOM seem to be working fine in this case.
- The temperature we tested for seems to be lower than the junction temperatures specified by NVIDIA in thermal design guide.