I am using jetpack5.1.5 and a custom board. I encountered a problem when flashing, but the same board can be downloaded in jetpack5.1.2. The following is my experimental record:
Using bsp to download jetpack515, the download will time out. Using bsp to download jetpack512, the download is normal. The USB configuration of the two is the same.
Using sdkmanager to download jetpack515, the download will time out
sdkmanager currently does not have jetpack512, so use sdkmanager to download jetpack6.2.1, the download is normal
Observe the download process, the system first creates an image, and then puts the board into download mode. At this time, during normal download, the 0955:7035 NVIDIA Corp device can be seen on the host, but the timed download does not appear on the host. Have you changed the USB settings on jetpack5.1.5?
There is a logical error that it is not about what we changed something here. It is that you should always modify the device tree by following document for custom USB design. No one ever guaranteed sdkmanager could flash a custom board.
If you don’t do that, you just rely on some luck there to flash your board.
Sorry, my previous explanation may have misunderstood you, let me explain it again.
We have 2 versions of pcb.
Version v1.0: The hardware of the USB part is exactly the same as devkit.
Version v2.0: The fusb series chip is removed, and USB2.0 is used to directly connect Type-C and the core board.
The problem of being unable to download that I described before was experimented on v2.0. After our test, the download failure has nothing to do with the USB configuration, but seems to be related to eeprom.
We conducted a download experiment of Jetpack5.1.5 on the v1.0 hardware:
By default, we can successfully download the firmware of Jetpack5.1.5
Remove the fusb series chip on the pcb. At this time, both v1.0 and v2.0 are directly connected to orin and typec via USB2.0, and v1.0 can still download Jetpack5.1.5
After removing the eeprom on the board pcb, the device cannot download Jetpack5.1.5, and a timeout occurs.
Therefore, it can be determined that the download results are inconsistent due to different eeproms
In this case, I set cvb_eeprom_read_size=0x0, and I can download Jetpack5.1.5
If the eeprom is soldered at this time, cvb_eeprom_read_size=0, it still cannot be downloaded
When downloading failed, the xudc driver of the device seems to be unable to initialize normally.
Question 1: Jetpack5.1.5 seems to be unable to complete the flash when there is eeprom on the board and cvb_eeprom_read_size=0x100
Question 2: Does Jetpack5.1.5 have special requirements for eeprom when downloading?