We want to install carrier boards on our aerial platforms for AI flight testing. The idea is to allow customers to develop in their lab on their own Jetson, bring it here, and plug it into our platform. I do not see any carrier boards available from NVIDIA. Do all the Jetson modules use the same pinout? Would this board work for multiple modules? Amazon.com: Alternative Solution for Jetson Xavier NX Developer Kit Directly Insert into Jetson Xavier NX Module Carrier Board Only @XYGStudy (Jetson-IO-Base-B) : Electronics
That board is from 3rd party, you may need to check the board specific to know that.
Regarding the Jetson module compatibility, please check Jetson FAQ | NVIDIA Developer
FYI, the developer kits are basically reference designs and all come with a carrier board with published schematic. Third parties do sell carrier boards, and when they do they use a “commercial” module instead of a dev kit module.
If the third party carrier board is an exact match to the electrical layout of the Jetson dev kit, then chances are the software is entirely the same.
When the layout changes in the carrier board, then things like GPIO will need different setup. That’s the firmware, and in this case, it is a change to the device tree. There is a lot of plug-n-play hardware out there which can “self describe” for drivers to deal with that hardware, but much of this is also simply connected at some physical address, and thus that type of hardware cannot self-describe. The device tree is what gets edited to tell the driver about such things.
When a third party provides a carrier board which differs from the dev kit you will likely find large parts of that device tree to be a match for the dev kit, but often new hardware (e.g., an SD card slot moving from the module to the carrier board is completely new hardware in combination with missing hardware) will be the reason for the edits. Third party manufacturers tend to do one of the following:
- Provide a patch to the NVIDIA flash software.
- Provide a rebranded version of the NVIDIA flash software which mainly differs in device tree and perhaps adding drivers.
- Or state that this carrier board uses the NVIDIA flash software and that one can just use that directly.
There are some differences among models, although many will have the same layout. If you’ve flashed a Jetson, then you have the flash software on your host PC. The location will be:
~/nvidia/nvidia_sdk/JetPack...version.../Linux_for_Tegra/
If you go there, then look at these files:
ls jetson*.conf
If you remove the “.conf” extension, then each of these are flash targets. This becomes more interesting when you realize these are symbolic links with a “human readable” name pointing at more technical names:
ls -l jetson*.conf
The more technical file name is just a combination of the module and the carrier board. There is also sometimes a reference in the file name to a revision.
The main .conf file will probably have its own content, but if you look closer, then each .conf file can include another file’s content. Eventually this abstracts to a separate reusable subset of content for the carrier board versus module. There are also sometimes differences for specialty flashes, e.g., flashing QSPI only on a dev kit which does not have eMMC.
You might consider a “snap in” connector for the full Jetson and not just the module. Then use the customer’s carrier board. For commercial sales purposes, other than distributors of dev kits, I don’t know of anyone who sells a final product based on a dev kit (there could be though). The warranty of a dev kit goes only to the original purchaser, while the warranty on a separately sold Jetson module is transferrable. If you were to build your own product (or if your customers were to build their own product), and then resell it (not as a dev kit being distributed), then you would be responsible for all warranty. Thus, to keep the module warranty, I think most resellers (other than distributing dev kits) would buy separate commercial modules and build or purchase the carrier board (which could be electrically a match for the dev kit and use NVIDIA’s software directly, or be custom and require a changed device tree for firmware).