Skip to content

cameras.md

Latest commit

 

History

History
112 lines (82 loc) · 3.21 KB

cameras.md

File metadata and controls

112 lines (82 loc) · 3.21 KB

Camera Support

This document covers tested camera sensors, available device tree overlays, and test commands.

Tested Cameras

Sensor Lanes Resolution Overlays Status
IMX477 2 4056x3040 single, dual Working
IMX219 2 3280x2464 single, dual, quad (PAB only) Working

IMX219 (Sony, 8MP)

Tested and working in 2-lane mode.

  • PAB: Tested on all 4 CSI ports (quad overlay)
  • JAJ / PAB_V3: Tested on both CSI ports (dual overlay)

Overlays

Overlay Filename Ports
Single tegra234-p3767-camera-p3768-ark-imx219-single.dtbo CAM0
Dual tegra234-p3767-camera-p3768-imx219-dual.dtbo CAM0 + CAM1
Quad tegra234-p3767-camera-p3768-ark-imx219-quad.dtbo All 4 ports (PAB only)

IMX477 (Sony Starvis, 12.3MP)

Tested and working in 2-lane mode on all carrier boards.

Overlays

Overlay Filename Ports
Single tegra234-p3767-camera-p3768-ark-imx477-single.dtbo CAM0
Dual tegra234-p3767-camera-p3768-imx477-dual.dtbo CAM0 + CAM1

4-Lane Mode (Not Working)

IMX477 4-lane overlays have been removed. While the Sony IMX477 sensor silicon supports 4 lanes, the RidgeRun nv_imx477 driver's 4-lane register initialization tables are incorrect. Getting correct values requires access to the Sony sensor NDA documentation. NVIDIA has acknowledged their own imx477-dual-4lane.dts reference overlay is broken.

2-lane mode provides full 12MP at 30fps which is sufficient for most use cases.

Installing a Camera Overlay

Build the overlay DTBs (from host):

./build_kernel.sh

Copy the overlay to the Jetson:

DTB_PATH="source_build/Linux_for_Tegra/source/kernel-devicetree/generic-dts/dtbs"
scp $DTB_PATH/<overlay>.dtbo jetson@192.168.55.1:~

On the Jetson, install and activate:

sudo mv <overlay>.dtbo /boot
sudo /opt/nvidia/jetson-io/config-by-hardware.py -n 2="Camera ARK IMX477 Single"
sudo reboot

List available overlays:

sudo /opt/nvidia/jetson-io/config-by-hardware.py -l

Test Commands

Verify Camera Detection

nvargus_nvraw --lps

v4l2-ctl

Capture 300 frames from /dev/video0:

v4l2-ctl --set-fmt-video=width=3840,height=2160,pixelformat=RG10 --stream-mmap --stream-count=300 -d /dev/video0

GStreamer

Install GStreamer (if not already present):

sudo apt-get install nvidia-jetpack -y

UDP h.264 stream (replace IP/port):

gst-launch-1.0 nvarguscamerasrc ! nvvidconv ! \
  x264enc key-int-max=15 bitrate=2500 tune=zerolatency speed-preset=ultrafast ! \
  video/x-h264,stream-format=byte-stream ! \
  rtph264pay config-interval=1 name=pay0 pt=96 ! \
  udpsink host=192.168.0.96 port=5600 sync=false

Record to file:

gst-launch-1.0 nvarguscamerasrc ! \
  'video/x-raw(memory:NVMM), width=3840, height=2160, format=NV12, framerate=30/1' ! \
  nvvidconv ! x264enc key-int-max=15 bitrate=2500 tune=zerolatency speed-preset=ultrafast ! \
  h264parse ! mp4mux ! filesink location=output.mp4

Select a specific sensor (for dual camera setups):

gst-launch-1.0 nvarguscamerasrc sensor-id=0 ...
gst-launch-1.0 nvarguscamerasrc sensor-id=1 ...