Hi,
I’m using a Jetson Xavier for the robot we are building.
When doing only light stuff (like modifying some program through SSH), everything works fine. When launching the program, everything is still fine while the robot is in “idle” state (ie waiting for the user to press a start button). When the robot really starts (which involves “heavy” image processing (person tracking and generating depth maps with a ZED2 camera), then the Jetson shuts down, often almost immediately, and if not usually within the next couple of minutes.
When not launching heavy stuff, I can have the Jetson running all day without problem.
Do you have any idea where the problem might come from and how to solve it?
Some technical data on our robot that might help:
the robot is powered from a 6S (24V) Lipo battery
the Jetson is powered through a DC/DC converter (we tried 3 different ones : output 12V 5A; 12V 10A; 19V 5A)
3 capacitors are in parallel with the Jetson to help keeping the voltage stable (2200µF electrolytic, 47µF electrolytic, 0.1µF ceramic)
on one USB C port we connect the ZED2 camera, on the other one a USB hub with wifi tongle and 4G stick
We already went through quite some tests during the previous weeks, based on other posts on the forum, but without finding any solution :
initially we had the 24V to 12V 5A converter
we watched the supply voltage with a scope, which seemed quite stable, but still decided to add the capacitors to make sure
we swiched to a 12V 10A converter to make sure we are not limited by current
we did some tests on a lab power supply : at 12V we still had shut downs when starting to use the camera, but at 19V it seemed more stable
we changed to a 19V 5A power supply
So I’m starting to get quite desperate to find a way to get the jetson staying on.
Do you have any ideas how to get the Jetson Xavier staying on?
Any useful tests I can do? (if you know the precise requirements for the supply power, I can double check)?
Do you know of any DC/DC converter that works great with the Jetson Xavier, knowing I have a 6S Lipo as power source (25.2V when fully charged, it would be nice to be able to go down to 20V)
Thanks a lot in advance
Felix
PS : I found a “suspicious” part in /var/log/syslog, but couldn’t find any explanation about it, so I copy the few lines in case it might help :
Aug 12 09:31:54 boroboxavier0-desktop kernel: [ 2.018187] igbvf: Intel(R) Gigabit Virtual Function Network Driver - version 2.4.0-k
Aug 12 09:31:54 boroboxavier0-desktop kernel: [ 2.018195] igbvf: Copyright (c) 2009 - 2012 Intel Corporation.
Aug 12 09:31:54 boroboxavier0-desktop kernel: [ 2.018253] Intel(R) 10GbE PCI Express Linux Network Driver - version 4.6.4
Aug 12 09:31:54 boroboxavier0-desktop kernel: [ 2.018260] Copyright(c) 1999 - 2017 Intel Corporation.
Aug 12 09:31:54 boroboxavier0-desktop kernel: [ 2.026301] nct1008_nct72 4-004c: LOC shutdown limit 107
Aug 12 09:31:54 boroboxavier0-desktop kernel: [ 2.027244] nct1008_nct72 4-004c: EXT shutdown limit 109
Aug 12 09:31:54 boroboxavier0-desktop kernel: [ 2.036194] eqos 2490000.ether_qos: Setting local MAC: 0 4 4b cc 32 29
Aug 12 09:31:54 boroboxavier0-desktop kernel: [ 2.036343] libphy: dwc_phy: probed
Aug 12 09:31:54 boroboxavier0-desktop kernel: [ 2.038920] PPP generic driver version 2.4.2
Aug 12 09:31:54 boroboxavier0-desktop kernel: [ 2.039238] PPP BSD Compression module registered
Hi,
Almost identical problem is occurring with our Jetson-Xavier dev kit. While connecting Xavier to power supply with adapter and doing heavy work, there is no problem. But connecting the Xavier to our drone (similar battery with the fnilius’ robot), after heavy task is occurred, Xavier shut downs itself. We don’t use capacitors but we were about to buy one and test with capacitors. Seeing this post was good for us since fnilius did quite a test to find the cause.
Thanks for your answer.
I confirm, I’m using the starter kit.
I have to admit that unsoldering the resistor is not very reassuring. I suppose the processor won’t like it too much if we disable the protections.
I just checked the voltage with the new 19V regulator, and it turns out that one is very bad (I observed spicks of more than 1.5V downwards).
So maybe the solution is rather to find a good voltage regulator.
Is there any you know works well?
Otherwise, what exactly are the requirements? Is 12V good (easier to find) or should I go for 19V?
And what are exactly the constraints on the maximal ondulation allowed (the protection circuit is called dV/dt, but then only an absolute drop of -0.5V is mentionned.
19V is better than 12V as lower current will cause lower voltage drop on path. For heavy workload you can try new better 19V regulator. Removing R135 is validated for some heavy load application, it will not cause issue, you can try it.
Thanks,
so far, we didn’t managed to get a nice and stable voltage (the problem is that the DC/DC converter supplying the jetson is connected on the same battery as 4 big motors (250W each)).
Therefore, we decided to try the solution to remove the R135 resistor.
The issue is that, if I’m not mistaken, R135 is not on the accessible face of the starter kit. So I need to unmount the fan, which I didn’t managed yet (I removed the 5 screws on the bottom side (4 holding the 2 black plastic pieces and one directly on the PCB), but I still can’t remove the fan). Could you please explain to me how to dismount the Jetson Xavier to access R135?
I want to share information that might help someone when Jetson AGX DevKit spontaneously reboots by power issue.
We are using 3 cameras leopardimaging.com 4k 60fps + a lot of heavy cuda code.
So the load was 55W at peak.
Like " GPU:8861mW; CPU:9989mW; SOC:24026mW; CV:0mW; DDR:3545mW; SYS5V:4697mW; Total:51.118W;"
We used AC-DC to 12v.
And at maximum load, Jetson restarted when the voltage dropped from 12 to 11.5 V.
After much research, we increased the voltage to 14V and replaced all wires from AC-DC to Jetson with 1.5mm copper wires and soldered the 5.5 @ 2.5 connector.
The system is now stable.
