Xavier module power-on error

We developed an Xavier board by ourselves. During the test, we found that there was a certain probability that the Xavier module was abnormal in power on, and stuck in the system start-up stage.
The attachment is the debugging serial port printing information during normal startup and abnormal startup. In case of exception, the system prints “unable to handle kernel paging request at virtual address ffffc7bc9c98e9”,


and finally prompts that it needs to be restarted.
2
Would you like to inquire about the approximate reason for the problem?
abnormal-log.txt (94.0 KB)
normal-log.txt (88.4 KB)

when power-on error,the virtual address is flexible.

Hi,

Your post will be better served in the Jetson section of the forums, I will move it over for you.

Best,
Tom K

1 Like

Does this occur on every boot up ? whats the failure rate ? what version of Jetpack is this ?

The Jetpack version is"Jetpack 4.3 [L4T 32.3.1]".this error does not occur every time when device boot up,and the failure rate is 10%.

it seems to be a memory instability issue in the crash is varying at random addresses.
please check the voltage to ddr

Is the xavier DDR power supply related to the DDR_AP_1V1, DDR_VDD2_1v1_en, DDR_VDD2_1.1V, VDD_DDRQ in the figure below? The source of these power supplies SYS_VIN_MV stable operation with 8A load change rate within 200mV, SYS_VIN_HV 6A load and 6A load change rate within 200mV. Does the above power supply capacity meet the xavier requirements? Any other suggestions? thank。

have you made any changes to the CVM? I guess not. In that case if the voltage supply from the CVB (the board you have created) is fine, system should be stable.
Please elaborate what changes you have made to create your system.
Also, what is the input voltage and current you are giving to the board?

The CVM has not been changed in any way, and the carrier board mainly modifies the power topology. The input power supply of the carrier board is 12V/200W, and the output 5V/8A power supply via the LTM4626. The 12V and 5V are powered to the xavier module via the large MOSFET device FDMS6681Z. The 8A and 6A loads described above are validated on the output side of the FDMS6681.

The xavier power-up timing is controlled by an external MCU device and tested using an oscilloscope as shown in the figure below.

Sorry for the late response, have you managed to get issue resolved or still need the support? Thanks

The problem still exists, and the cause of the problem has not yet been located.
The product application temporarily detects the heartbeat of the IO output of the module through the external MCU device. If the IO pulse is not output after the timeout, the MCU will force the module to power off and restart.

Is the module CVM working fine with the Devkit or you just have the CVM?
What is the power profile you have chosen?
Can you boot just to initrd? I mean in the xml file, comment out system.img from app partition or flash with a dummy system.img
I believe the crash stack is not the same every time? Can you share more boot log logs?

Meanwhile we check with HW team on the waveform you have provided

CVM works fine on devkit. But no risk points are found on our own carrierboard,everything seems OK,We don’t know how to rectify and optimize.

thanks,please help answer the other questions.
Also, can you try with a 19v supply?

The carrier board allows a peak voltage of 20V, and 19V power supply may be risky. We can use 18V power supply for test verification.

Using the 18V power supply, the CVM boot failure problem still exists

There is no update from you for a period, assuming this is not an issue any more.
Hence we are closing this topic. If need further support, please open a new one.
Thanks

Hi, seems you need to probe and capture if any voltage drop/unstable when issue happen. Have you used oscilloscope to do that?