The PMIC on the Jetson Nano is very sensitive to moisture / EMP

Autonomous Machines Jetson & Embedded Systems Jetson Nano
1

I am currently testing the Jetson Nano production module (180-13448-DAAA-B01) on an Auvidea JN30B-POE-PD rev. 4 carrier board. The system reboots every time I subject one of the connectors on the carrier board to ESD. When searching for the root cause of why this happens, I found that when I spray just one drop of freeze spray directly on top of the PMIC (MAX77620A) the module reboots instantly. (The freeze spray leads to a tiny amount of condensation on the PMIC)
The module also resets if i subject the module to a small EMP above or below the PMIC.
This leads me to think, that maybe one of the pins on the PMIC is floating ?

2

Hi mst27jrt,

We are investigating this. Can you please provide the data on the exact EMP event above or below the PMIC?

3

Additional things need to check:

Can it repro on Jetson Dev-kit?
Is it the single case from one specific module?
Any swap test between module and carrier board?

4

The EMP event is generated by a H-field probe connected to a Piezoelectric-generator. (Gas igniter)
I have reproduced the issue on two Dev-kits (Both rev. A.02). They reboot on both the freeze spray test, and EMP test.
I have not swapped the modules, since according to Auvidea the dev-kit and the production modules are not compatible.

5

Hi mst27jrt,

the MAX77620A is located on the Nano module and not on the JN30. Some JST-GH connectors on the JN30 expose pins of the Nano without level conversion and protection. We need to find out which connectors you are using for your test.

Best regards,

Max

6

Hi mst27jrt, what kind of processing load are you running on the devices when this occurs? Also on the production module, do you have some kind of heatsink or TTP attached?

Can you link to some pictures of your setup? We haven’t been able to replicate the issue in our lab after similar testing.

7

We need to find out which connectors you are using for your test.
I have only tested ESD on the shield of the “External” connectors. (ie. Power, ethernet, HDMI and USB)

what kind of processing load are you running on the devices when this occurs?
The board is idle (2% CPU load according to “top”)

Also on the production module, do you have some kind of heatsink or TTP attached?
A heat-spreader is mounted on the module (with thermal paste), and the enclosure acts as heatsink.
https://auvidea.eu/product/70790/

8

This video shows that the Nvidia Jetson Nano Production module (rev. B01) mounted in a "JN30B Carrier Board for Jetson Nano Production Module Rev. 4 (PoE PD) " from Auvidea, reboots when subjected to only 2kV ESD on the shield of the Ethernet connector.

9

I don’t think this is a valid usecase and test scenario.
A reboot or unexpected reaction is to be expected if you test a naked board with no shielding concept at all for ESD.

10

The board also reboots when mounted in the Aluminium Case from Auvidea (Still at only 2kV ESD):

11

Ouch!
That should definately not happen (with a good design).
Is the aluminium case electrically WELL connected to the board or is it floating?
The GND-connected mounting holes of the PCB should have very good electrical connection to the case to ensure equal potential.
I would finally check if the power-supply can be excluded as cause for the reset.

12

It looks like issue of lack ESD diodes on Auvidea board.

13

The Auvidea board has ESD diodes on all external connectors.

This is what I think is the root cause :
The PMIC On the Jetson Nano module (MAX77620A) has a very high impedance 32.768 kHz crystal oscillator.
The PMIC also has some circuitry that monitors the oscillation of the crystal.
When the oscillation is disturbed or stopped completely, the PMIC will power-cycle the Nano module.

What I found that can cause the PMIC to power-cycle :

  1. Condensating moisture on the crystal oscillator pins.
  2. Noise introduced to any of the control signals to the PMIC.
  3. EMP near the crystal.

What we will do on our carrier board to mitigate this :

  1. Conformal coating on the PMIC and crystal should solve the moisture problem.
  2. Tracks connected to the PMIC control signals must be short/shielded to prevent them from picking up noise.
  3. EMP is not a real issue. So nothing will be done here.