HI NXP
1. Download link for camera setup documentation
2. Where is the issue in the log below?
I would suggest you try to use libcamera to test.
export LIBCAMERA_PIPELINES_MATCH_LIST='nxp/neo,imx8-isi,uvcvideo'
gst-launch-1.0 libcamerasrc ! queue ! autovideosinkHI @danielchen
1.I've enabled media linking with the command (media-ctl -d /dev/media0 -l "'gc05a2 1-0037':0 -> 'csidev-4ad30000.csi':0[1]")
Hi, yrj
For BSP 6.12.20, I would suggest you refer to i.MX Linux Reference Manual Rev LF6.12.20_2.0.0.
You need to enable the media Link.
Before using GStreamer, you need to enable the media pipeline link between the sensor and CSI,
Please check the Linux Reference Manual, search "media-ctl" for more details.
HI @danielchen
1.BSP:imx-6.12.20-2.0.0.xml
2.This version is already very new. Is it necessary to use version 6.12.34?
which BSP version you are using? can you try the latest BSP version 6.12.34
HI @danielchen
1.Check the formats supported by /dev/video2, in case they do not meet expectations
2.The format is incorrect; it is suspected that the CSI and the camera are not compatible.
From the log file, the GC05A2 sensor at I2C address 1-0037 is detected and linked to the CSI interface.
gc05a2 1-0037 → csidev-4ad30000.csi → crossbar → mxc_isi.2 → /dev/video2
The format negotiation shows
-
- GC05A2 outputs
SGRBG10_1X10/2592x1944 - CSI and ISI are configured for
UYVY8_1X16/1920x1080
- GC05A2 outputs
This implies that format conversion is happening, likely via ISI, and the pipeline is correctly routed to /dev/video2.
Below fail log shows the requested format is not supported by the V4L2 device /dev/ivdeo2
Why is there an error in the log?
Hi @yrj
I would suggest you refer to UG10215: i.MX Camera Porting Guide.
https://www.nxp.com.cn/docs/en/user-guide/UG10215.pdf
This document provides an overview of the i..MX95 applications processor hardware camera subsystem, and then describes its corresponding software architecture with the Linux Kernel and User space. It also describes the porting instructions and how to configure the 3A algorithms.
Regards
Daniel
Regards
Daniel