GX Camera on Firfly Boards

查看中文

1 Overview

The GX series cameras are designed for AI applications in the industrial field. They use the MIPI CSI-2 interface and are particularly suitable for embedded computing platforms. They feature a wide range of data formats and triggering capabilities, extremely low latency, high bandwidth, and reliable stability.

This article takes the Firefly's ROC-RK3588S-PC, ROC-RK3576-PC and ROC-RK3566-PC motherboards as examples to introduce how to connect GX series cameras to the RK3566/RK3568, RK3576 and RK3588S/RK3588 systems.

We provide drivers for the Linux operating system (taking Ubuntu as an example).

1.1 Camera Module List

Series	Model	Status
GX series	GX-MIPI-IMX662	Done

In addition, the driver for the V-by-One HS connection mode has been finished on the Ubuntu system.

2 Hardware Setup

The GX series and RAW series cameras require an ADP-MV2 adapter board to connect to the ROC-RK35xx-PC motherboard.

2.1 Connection of new ADP-MV2

2.1.1 Connection of GX series camera and ADP-MV2

The two are connected using 0.5 mm pitch*30P FFC cable with opposite-side contacts. The cable must be inserted with the silver contacts facing outside.

TOP	BOTTOM
ADP-MV2 to MV-MIPI-X	ADP-MV2 to MV-MIPI-X

2.1.2 Connection of RAW-MIPI-SC132M and ADP-MV2

The two are connected using 1.0 mm pitch*15P FFC cable with opposite-side contacts. The cable must be inserted with the silver contacts facing outside.

TOP	BOTTOM
ADP-MV2 to RAW-MIPI-SC132M	ADP-MV2 to RAW-MIPI-SC132M

2.1.3 Connection of other RAW series camera and ADP-MV2

The two are connected using 0.5 mm pitch*pin FFC cable with opposite-side contacts. The cable must be inserted with the silver contacts facing outside.

TOP	BOTTOM
ADP-MV2 to RAW series camera	ADP-MV2 to RAW series camera

2.1.4 Connection with Main board using ADP-MV2

The two are connected using 0.5mm pitch * 30P FFC coaxial wires, paying attention to the direction of the contact surfaces, silver contacts facing outside on the ADP-MV2 and facing inside on the RK board.

RK to ADP-MV2 and MV cam

2.2 V-by-One-HS-KIT Camera Connection Diagram

VBYONE connection with rk3588

3 Introduction to github repositories

https://github.com/veyeimaging/rk35xx_veye_bsp

https://github.com/veyeimaging/rk35xx_firefly

includes：

• driver source code
• i2c toolkits
• application demo

In addition, a compiled linux kernel installation package and Android image is provided in the releases.

4 Upgrade Firefly Ubuntu system

For the ROC-RK3566-PC,ROC-RK3576-PC and ROC-RK3588S-PC, we have provided an image of the release system.

Download the latest Ubuntu imge from https://github.com/veyeimaging/rk35xx_firefly/releases/ .

Refer to the Firefly documentation ROC-RK3588S-PC ROC-RK3566-PC ROC-RK3576-PC to burn in a standard system.

5 Check system status

5.1 Whether the camera is correctly recognized

After system update, reboot the main board.

Execute the following command on the main board to check if the camera is properly connected.

dmesg | grep gxcam

You can see the camera model and the camera version number probed.

A prompt as below indicates that the GX-MIPI-IMX662 camera is detected on the i2c-7 bus.

[6.667547] gxcam 7-003b: veye gx series camera driver version: 01.00.01

[6.781681] gxcam 7-003b: camera is: GX-MIPI-IMX662

[6.820210] gxcam 7-003b: Success to get gxcam endpoint data lanes, dts uses 2 lanes,will set to camera

[6.834597] gxcam 7-003b: gxcam_enum_controls success

[6.891209] rockchip-csi2-dphy csi2-dphy0: dphy0 matches m00_b_gxcam 7-003b:bus type 5

On the ROC-RK3588S-PC, the camera is mounted on i2c-7, with an i2c address of 0x3b.

On the ROC-RK3566-PC and ROC-RK3576-PC, the camera is mounted on i2c-4.

• Check the video0 device node:

ls /dev/video0

You should see:

video0

After successfully identifying the camera, the camera will be recognized as /dev/video0.

5.2 State Detection and Environment Variable Configuration

Here, We have provided two scripts that can automatically retrieve some information about the camera.

First, try using the probe_camera_info-rk.sh script. This script is used to detect the connected and successfully registered camera devices, retrieve the corresponding media device nodes, video device nodes, sub-device nodes, I²C buses, and device identifiers and other low-level information. After execution, it will generate an auto_camera_index.json file in the current directory and record the retrieved information in the file.

./probe_camera_info-rk.sh

cat auto_camera_index.json

[

  {

    "media_node": "/dev/media0",

    "video_node": "/dev/video0",

    "video_subnode": "/dev/v4l-subdev2",

    "media_entity_name": "m00_b_gxcam 7-003b",

    "i2c_bus": "7"

  }

]

By referring to the index information, we can see the "i2c_bus": "7" corresponding i2c_bus information, as well as the number of devices connected. The current index shows that only one device is connected, and the i2c_bus number is 7. If multiple devices are connected, the index information may include "i2c_bus": "10", "i2c_bus": "11" and so on. Then, using the gx_probe.sh script, if it is a multi-camera system, the i2c_bus can be executed based on the information read from the previous script, and the corresponding camera model, width, height, frame rate and other information can be configured in the environment variables.

Usage:

source ./gx_probe.sh 7

A typical output:

$ source ./gx_probe.sh 7

Found veye_gxcam camera on i2c-7.

Setenv CAMERAMODEL = GX-MIPI-IMX662

Setenv FPS = 60

Setenv WIDTH = 1920

Setenv HEIGHT = 1080

You can verify the environment variable output using:

echo $CAMERAMODEL

Note that these environment variables are only valid for the current session.

Important Notes:

• This script requires the mvcam driver version 1.1.06 or later.
• If your driver version is earlier than 1.1.06 or you need to use different width, height, or frame rate values, refer to the camera module manual and manually configure the following environment variables. Otherwise, subsequent programs may not function correctly.

Example:

export WIDTH=2432

export HEIGHT=2048

export FPS=50

5.3 Configuring  global variables

Based on the board model, configure the I2C_BUS global variable as follows:

• ROC-RK3588S-PC

export I2C_BUS=7

• ROC-RK3566-PC and ROC-RK3576-PC

export I2C_BUS=4

5.4 Using media-ctl to view topology

Using the media-ctl command can clearly display the current topography structure.

media-ctl -p -d /dev/media0

5.4.1 Link relationship

mv camera->rockchip-csi2-dphy0->rockchip-mipi-csi2->stream_cif_mipi_id0 - - ->DDR(/dev/video0)

The application can obtain images through the /dev/video0 node.

5.4.2 mv camera entity information

Taking the MV-MIPI-IMX296M as an example:

- entity 63: m00_b_mvcam 7-003b (1 pad, 1 link)

             type V4L2 subdev subtype Sensor flags 0

             device node name /dev/v4l-subdev2

        pad0: Source

                [fmt:Y8_1X8/1456x1088@100/6000 field:none]

                -> "rockchip-csi2-dphy0":0 [ENABLED]

You can see that:

• The complete name of this entity is: m00_b_mvcam 7-003b.(It is m00_b_mvcam 4-003bon ROC-RK3566-PC.)
• It is a V4L2 subdev (Sub-Device) Sensor.
• Its corresponding node is /dev/v4l-subdev2, which can be opened and configured by applications (such as v4l2-ctl).
• Its output format is [fmt:Y8_1X8/1456x1088@100/6000 field:none], where Y8_1X8 is a shorthand for a mbus-code, which will be listed in the next section of this article.
• The current resolution is 1456x1088.
• The current frame interval is 100/6000, which means the frame rate is 60.
• The data format output by the camera can be modified using the media-ctl command.

5.4.3 mbus-code list

MV series and RAW series cameras have different data format capabilities, which can be found in the data manual for each camera model.

Format on datasheet	mbus-code for media-ctl	FourCC pixelformat for v4l2-ctl
RAW8	Y8_1X8	GREY
RAW10	Y10_1X10	'Y10 '
RAW12	Y12_1X12	'Y12 '
UYVY	UYVY8_2X8	UYVY

6 Raw data format

The VICAP module of RK3588 supports two data saving formats, Compact and Noncompact RAW. You can modify the mode using the RKCIF_CMD_SET_CSI_MEMORY_MODE ioctl command of RKCIF. By default, the output is in Compact RAW format.

Compact raw and noncompact raw of rk3588 VICAP

6.1 Noncompact RAW

For pixel data with 10-bit depth or 12-bit depth, two bytes are always used to store one pixel. This storage method is convenient for software processing, but it has the disadvantage of occupying a large amount of space.

Depending on whether the effective data is stored in the high bits or low bits, it can be further divided into two types: high align and low align.

6.1.1 Noncompact RAW(high align)

Data is saved to the high bits, and the unused low bits are filled with 0. This is one of the data formats supported by RK VICAP.

6.1.2 Noncompact RAW(low align)

In Noncompact RAW (low align) format, data is saved to the low bits, and the unused high bits are filled with 0. The V4L2 standard 'Y10' (10-bit Greyscale) and 'Y12' (12-bit Greyscale) formats are both stored in this way.

The pixel_layer_convert conversion tool mentioned later in the article also converts Compact RAW to this storage format for easy display using image players.

6.2 Compact RAW

As shown above，there is no bit padding between pixels in this storage format.

6.3 Line stride

To facilitate fast operations on images, the system usually provides row-aligned buffer sizes for each line of data. RK3588 uses 256-byte alignment for this purpose.

line_stride = ALIGN_UP（image_width*bits_per_pixel/8,256）

For example, when the image width is 1456:

8bit depth，line_stride=1536

10bit depth，preferred_stride=2048

12bit depth，preferred_stride=2304

6.4 Format convert tool

We have written a small tool: pixel_layer_convert, which can easily convert Compact images to Noncompact (low align) images.

For example, the following command can convert a Compact RAW10 image with a width of 1456 to Noncompact RAW10 format:

./pixel_layer_convert -I R10C -i y10-1456x1088_0001.raw -o y10-1456x1088_0001_new.raw -w 1456

6.5 Raw data image player

We recommend using vooya as the player, which supports GREY, and unpacked image formats.

Also, y8 file can be used with this player: YUV Displayer Deluxe.

7 Application examples

7.1 Configure parameters using v4l2-ctl

$ v4l2-ctl -d /dev/v4l-subdev2 -L

User Controls

                   trigger_mode 0x00981901 (int)    : min=0 max=2 step=1 default=0 value=0 flags=volatile, execute-on-write

                    trigger_src 0x00981902 (int)    : min=0 max=1 step=1 default=1 value=1 flags=volatile, execute-on-write

                    soft_trgone 0x00981903 (button) : flags=write-only, execute-on-write

                     frame_rate 0x00981904 (int)    : min=1 max=60 step=1 default=60 value=60 flags=volatile, execute-on-write

                          roi_x 0x00981905 (int)    : min=0 max=1376 step=8 default=0 value=0

                          roi_y 0x00981906 (int)    : min=0 max=1024 step=4 default=0 value=0

Parameters can be set and get using the following methods.

v4l2-ctl --set-ctrl [ctrl_type]=[val]

v4l2-ctl --get-ctrl [ctrl_type]

All the above functions can be implemented using mv_mipi_i2c.sh.

Note that the above parameters cannot be modified during the capture process.

The following is an explanation of each parameter:

7.1.1 Trigger Mode

v4l2-ctl --set-ctrl trigger_mode=[0-2]

0：Video streaming mode

1：Normal trigger mode.

2：High-speed continuous trigger mode.

7.1.2 Trigger Source

v4l2-ctl --set-ctrl trigger_src=[0-1]

0: Software trigger mode.

1: Hardware trigger mode.

7.1.3 Software trigger command

v4l2-ctl --set-ctrl soft_trgone=1

7.1.4 Set frame rate

v4l2-ctl --set-ctrl frame_rate=[1-max]

The maximum frame rate is automatically updated as the resolution changed.

7.1.5 Set the starting position of the ROI

v4l2-ctl -d /dev/v4l-subdev2 --set-ctrl roi_x=0

v4l2-ctl -d /dev/v4l-subdev2 --set-ctrl roi_y=0

After setting the ROI starting position, you need to complete the full ROI configuration using the media-ctl command.

Note that setting the ROI may affect the maximum frame rate, and the ROI parameters need to meet the requirements specified in the camera manual.

7.2 Set image format using media-ctl

use the following command to configure the camera's data format, resolution, and frame rate using media-ctl:

media-ctl -d /dev/media0 --set-v4l2 '"m00_b_mvcam '"$I2C_BUS"'-003b":0[fmt:Y8_1X8/'"$WIDTH"'x'"$HEIGHT"'@1/'"$FPS"']'

Among them: "m00_b_mvcam '"$I2C_BUS"'-003b" refers to the complete name of the camera entity, Y8_1X8 is the mbus-code, '"$WIDTH"'x'"$HEIGHT"' indicates the resolution, 1/'"$FPS"' indicates the resolution frame rate.

The width and height here cooperate with the roi_x and roi_y of the v4l2-ctl command to form the ROI parameter.

For example, for MV-MIPI-IMX296M, the command after variable replacement would be:

media-ctl -d /dev/media0 --set-v4l2 '"m00_b_mvcam 7-003b":0[fmt:Y8_1X8/1456x1088@1/60 field:none]'

You can not only configure the data format, resolution, and frame rate in one command, but also modify them separately as needed.

7.3 Video Streaming mode

7.3.1 Set data format, resolution, frame rate

v4l2-ctl -d /dev/v4l-subdev2 --set-ctrl roi_x=0

v4l2-ctl -d /dev/v4l-subdev2 --set-ctrl roi_y=0

media-ctl -d /dev/media0 --set-v4l2 '"m00_b_mvcam '"$I2C_BUS"'-003b":0[fmt:Y8_1X8/'"$WIDTH"'x'"$HEIGHT"'@1/'"$FPS"']'

7.3.2 Frame rate statistics

In streaming mode, the following commands can be used for frame rate statistics：

v4l2-ctl --set-fmt-video=width=$WIDTH,height=$HEIGHT,pixelformat=GREY --stream-mmap --stream-count=-1 --stream-to=/dev/null

Or：

./yavta -c1000 --skip 0 -f Y8 -s ${WIDTH}x${HEIGHT} /dev/video0

7.3.3 Save image to file

• raw8

v4l2-ctl -d /dev/video0 --set-fmt-video=width=$WIDTH,height=$HEIGHT,pixelformat=GREY --stream-mmap --stream-count=1 --stream-to=y8-${WIDTH}x${HEIGHT}.raw

• raw10

v4l2-ctl -d /dev/video0 --set-fmt-video=width=$WIDTH,height=$HEIGHT,pixelformat='Y10 ' --stream-mmap --stream-count=1 --stream-to=y10-${WIDTH}x${HEIGHT}.raw

• raw12

v4l2-ctl -d /dev/video0 --set-fmt-video=width=$WIDTH,height=$HEIGHT,pixelformat='Y12 ' --stream-mmap --stream-count=1 --stream-to=y12-${WIDTH}x${HEIGHT}.raw

Please refer to the description in the previous section for the image format.

7.3.4 Example of yavta

7.3.4.1 Install yavta

git clone git://git.ideasonboard.org/yavta.git

cd yavta;make

7.3.4.2 Save image to file

After setting data format, resolution, frame rate，run:

./yavta -c1 -Fy8-${WIDTH}x${HEIGHT}.raw --skip 0 -f Y8 -s ${WIDTH}x${HEIGHT} /dev/video0

7.3.5 Example of import image to OpenCV

sudo apt install python3-opencv

See the samples directory on github for details.

python ./v4l2dev_2_opencv_show_grey.py --width 1456 --height 1088 --fps 60 --i2c 7

7.3.6 Example of gstreamer application

We provide several gstreamer routines that implement the preview function. See the samples directory on github for details.

7.4 Trigger mode

7.4.1 Set data format, resolution, frame rate

v4l2-ctl -d /dev/v4l-subdev2 --set-ctrl roi_x=0

v4l2-ctl -d /dev/v4l-subdev2 --set-ctrl roi_y=0

media-ctl -d /dev/media0 --set-v4l2 '"m00_b_mvcam '"$I2C_BUS"'-003b":0[fmt:Y8_1X8/'"$WIDTH"'x'"$HEIGHT"'@1/'"$FPS"']'

7.4.2 Software trigger mode

7.4.2.1 Set mode

v4l2-ctl -d /dev/v4l-subdev2 --set-ctrl trigger_mode=1

v4l2-ctl -d /dev/v4l-subdev2 --set-ctrl trigger_src=0

7.4.2.2 Start acquisition

v4l2-ctl -d /dev/video0 --set-fmt-video=width=$WIDTH,height=$HEIGHT,pixelformat=GREY --stream-mmap --stream-count=1 --stream-to=y8-${WIDTH}x${HEIGHT}.raw

7.4.2.3 Perform soft trigger operation

In other shell terminals, you can execute the following command multiple times for multiple triggers.

v4l2-ctl -d /dev/v4l-subdev2 --set-ctrl soft_trgone=1

7.4.3 Hardware trigger mode

7.4.3.1 Set mode

v4l2-ctl -d /dev/v4l-subdev2 --set-ctrl trigger_mode=1

v4l2-ctl -d /dev/v4l-subdev2 --set-ctrl trigger_src=1

The mv_mipi_i2c.sh script can be used to set rich trigger parameters.

7.4.3.2 Start acquisition

v4l2-ctl -d /dev/video0 --set-fmt-video=width=$WIDTH,height=$HEIGHT,pixelformat=GREY --stream-mmap --stream-count=1 --stream-to=y8-${WIDTH}x${HEIGHT}.raw

7.4.3.3 Perform hardware trigger operation

Connect the appropriate trigger signal to the trigger pin of the camera and trigger it.

8 i2c script for parameter configuration

We provide shell scripts to configure the parameters.

mv_mipi_i2c.sh user guide

9 Question Feedback

We are committed to providing richer possibilities for image applications on embedded platforms. Therefore, our software for embedded platforms is based on the principle of open source.

If you have any questions or suggestions about our existing software, please feel free to submit them to the forum or email our technical staff at xumm#csoneplus.com.

10 References

• ROC-RK3566-PC Manual

https://wiki.t-firefly.com/en/ROC-RK3566-PC/

• ROC-RK3588S-PC Manual

https://wiki.t-firefly.com/en/ROC-RK3588S-PC/

• ROC-RK3576-PC Manual

https://wiki.t-firefly.com/en/ROC-RK3576-PC/

• Firefly Linux User Guide

https://wiki.t-firefly.com/en/Firefly-Linux-Guide/index.html

11 Document History

• 2025-04-14

Add support for RK3576.

• 2025-03-23

Add description of mv_probe.sh.

• 2024-07-09

Add support for RAW-MIPI-SC535M.

• 2024-04-17

Support RK3566.

• 2024-03-10

Add pictures and descriptions of hardware connections for the new version of ADP-MV2.

• 2023-08-30

Add support for RAW-MIPI-IMX462M and RAW-MIPI-AR0234M.

• 2023-07-31

Support V-by-One on ubuntu system.

• 2023-04-12

Release 1st version.