Jetson AGX Xavier boot issues

Autonomous Machines Jetson & Embedded Systems Jetson AGX Xavier
#1

We use L4T 32.4.3 to build custom BSP and using our custom carrier board.

One of our customer sis having below issue, can you please help?

There is a major fault in a computer with carrier board and NVIDIA AGX Xavier – it doesn’t start in some circumstances. It happens when computer works properly for some time and after shutting down (disconnecting the power supply) it is starting again.

Important information – computer works with power supply designed by our company. It’s supplied and it has connected 2 I/O signals (IOs form a SAM controller on carrier board which uses modules SPI lines). Mentioned disconnecting of supply is done by our power supply. Important condition for not starting of a computer is connecting to Ethernet network which sends video streams to the computer.

If computer wasn’t blocked then disconnection of Ethernet cable (during work) prevents from the emergence of a fault.

Time of proper action which is necessary to appearing of a fault is about 2 – 10 minutes.

Probably in some way it depends on temperature - time needed to emergence of a fault is longer when computer starts after long break than after some work. After blocking of a computer turning off and turning on doesn’t cause unlocking of a computer, because disconnection is done by turning off 12V power supply. During that time (blocked computer) Power Supply microcontroller has set states on I/O ports and CAN bus is working.

After turning off supply (power supply is also turned off) and disconnecting of Ethernet cable computer is unlocked during renewed starting of a supply.

We don’t know what is the real state of a computer so we can’t establish fundamental cause. We established that cutting I/O lines and CAN bus prevents from suspending of a computer.

According to our observation blocking of a computer is caused by intense Ethernet transmission during starting of the computer. Probably carrier board interprets it as an attempt of starting of a bootloader and there is no starting of booting of a system from a disk. At this moment system is blocked and computer is supplied with less number of amperes than during normal work.

We detected that when computer is blocked USER (Controlled by a GPIO) diodes switch on at the same moment as Power IN(indicates board is getting power)diodes. When computer is not blocked USER diode switch on 2 seconds later.

#2

Hi,

Sorry that I cannot understand your problem with the way you described your problem. We are not your co-worker so does not know HOW does your system look like.

It sounds like you have lots of causes to your issue. Please just list things one by one

  1. What is your exact issue

  2. What is your suspicious cause?

  3. What have you tried?

And please prevent to use some words like “computer”. If you are talking bout jetson, call it AGX or jetson. It is to prevent mislead with host computer.

Please describe what does “block” mean here. If you just want to say the jetson cannot boot, please just say it cannot boot. It is really confusing to see some statement like “blocking of a computer turning off and turning on”. It just make me wonder that you are saying that the board cannot boot up and also cannot shutdown???

#3

Lots of questions…hard to understand, but I can probably say something “general” which might be of use.

This sounds like the power supply is disconnected remotely via a command to something like a solid state relay. Do understand that required regulation must be quite good for the Jetson to start. If there is something like a solid state relay involved, then any added resistance may contribute to startup problems. You should add details related to any kind of relay or on/off switch implementation.

I don’t know why ethernet would matter, but it is interesting. Since ethernet doesn’t have an actual ground there couldn’t be a ground loop. Power supplies with some sort of spike or other protection can produce a current loop when off, but I don’t know of any case for this on ethernet unless there is a custom design with a DC ground involved. There are actually a lot of ways a system with power off can have unintended power leaks back into the system which is supposedly powered off. You might want to check out this video (this is a very good video, and quite possibly related to your issue):
https://www.toradex.com/webinars/backfeeding-unintentional-power-paths-hunting

#4