TX2 USB 3.0 camera crashes on Auvidea J90, while working on the dev kit carrier board.

We have a Twiga TV80 USB 3 interface board connected to a Sony FCB camera. The TX2 is set up on a J90 carrier with official Auvidea firmware. When we connected the camera to the J90 for the first time the stream worked for about 10 seconds afterwords it wouldn’t properly recognize the device. The device would be recognized and lost every other second or two. After that initial try we couldn’t stream the video even for a second. Rebooting connecting and disconnecting the usb cable would help at all. Sometimes the device isn’t even recognized.

When we take the TX2 off the J90 and install it onto the dev kit carrier board everything works fine.

We suspect that something is wrong with the DTB. Has anyone encountered a similar problem ?

I have no way to answer specifics, but the device tree must be modified for custom carrier boards. Unless you’ve installed the board support package for the J90 there is no chance that everything will work correctly on the J90. Conversely, if you’ve used the J90 BSP, then the dev kit carrier board would not function correctly. BSPs are specific to the JetPack/L4T/SDK Manager revision, and so even if you use the BSP from Auvidea, then this would fail if the BSP is for a different revision.

The fact that it works correctly on the dev kit says you do not really have the official Auvidea firmware running. Perhaps the BSP install was not actually correct, or perhaps something overwrote that install.

Are you certain you are using the correct BSP for your release, and that your BSP is still present? Mostly Auvidea would have to answer any details relating to this.

We have installed using the sdk manager. The installation wasn’t straight forward and we may have done something wrong. Is there a proper guideline for that ? Also can we verify somehow if the BSP is installed or not ? I have checked the DTB file in /boot/dtb it seems to be something from auvidea.

Also I have noticed that the J90 doesn’t have an RTC battery cell and has a dedicated pin for adding one. I have contacted them to know exactly how to do that and whether its necessary or not. Could the absence of the RTC cell result in such issues? The thing that bothers me the most is that we got video for a few seconds before it went totally dead.

I do not work on custom carrier boards, and have not used the J90. I couldn’t tell you specifically what to check for.

However, do know that in many cases device tree content in “/boot/” is no longer used, or else is used differently than in the past. In most cases you can expect the dtb to actually be signed and put into a partition. Sometimes the “/boot” content is just leftover and not really needed. In other cases, even if content there is used, earlier boot stages will see and use the partition content rather than the “/boot” content. This is because earlier boot stages which need this information do not have any ability to read ext4 file systems, and thus reading as binary data in partitions gets around that.

I would say to carefully read what Auvidea publishes as directions, making certain that versions are correct, and then to ask at each step where something fails. Some of that can be answered here, and I know Auvidea does monitor the forums, and so either they or someone here with J90 experience can answer.

So after some more testing, the story becomes more and more interesting. We operate in two cities 300kms apart from each other in the same country with the same time zone. We have 2 setups of the same system which are exactly the same. Both of them work when in one city and don’t when in the other. We switched them 2 times both of the times one worked the other didn’t. It seems that its dependent on some external factor but I cant quite figure out on what. We desperately try to do something to make the same issue arise in both cities but nothing seems to help. I dont even know if the BSP has anything to do with this anymore. Any ideas ?

One possibility is power supply stability. Not likely since the whole system does not crash, but mentioning.

Another possibility is noise or environment. For example RF noise, different cables, different temperature and humidity. Power supply is actually a subset of this.

Networking can actually change some issues. Sometimes bugs are related to network packet loss or latency (not an easy bug to find).

You may need to try to attach a serial console with logging to see what is going on. Keep in mind that even a small RPi could be placed there with a USB serial console logging…the system doing the logging does not need to be a full PC. Once a system boots I would not expect a lot of log lines, but if you worry about filling up the logging system, then you could save the serial console log to an SD card or USB thumb drive.