Currently, I’m trying to receive the PDAF data produced by a sensor that can be embedded on the CSI data stream. I was able to determine that is possible accessing that data using LibArgus, and that made me proceed to take a look on the CSI/video capture Nvidia drivers to determine how and that information was gathered. I realize that the usage of such data was just available on the TX2 drivers not the TX1 drivers on Jetpack 3.1.
After researching a bit more I came across the following post:
Which actually also confirmed my suspicions of that support currently only available on TX2. I have the following questions:
According to the TRM, this should also be able to supported on TX1. Is there any particular limitation for not having that support on TX1 driver?
Are there any patches to add support EBD on TX1?
In case that it is possible to add, and there are no patches; could you provide a rough estimate on how much effort it could take to backport that support on TX1?
you don’t need patches to add EBD support.
please check your sensor datasheet, and configure correct embedded_metadata_height.
moreover, please look into the kernel driver about embedded metadata supporting.
thanks
r28.1_sources/kernel-4.4/drivers/media/platform/tegra/camera/sensor_common.c
Changes between vi2_fops.c and vi4_fops.c seem to be significant, could you confirm that support is currently only available for TX2? and if so, is there any particular reason for not suporting it TX1?
Thanks for the help, regarding my comment “I was able to determine that is possible accessing that data using LibArgus”, basically I was able to determine that the EBD was accessible from LibArgus and then researched how this might be possible using the current stack of drivers on the kernel, I was able to determine that t seems like it’s currently only possible for TX2.
Unless LibArgus uses a different stack of software than the current kernel drivers.
Is it possible for the VI HW/Driver to just pass the embedded data as part of the frame data to the V4L2 if we increase the frame height to include the embedded data lines. We don’t want the VI to extract the metadata, we want it to pass it along with the frame data.
yes, you’re able to pass through the signals within embedded metadata lines.
please refer to [Tegra TRM]-> [Video Input (VI)]-> [Embedded Data Formats] chapter for more details.
thanks