@linoooooo 你好，我们的相机内部isp流程是没有整帧缓存的，相机本身引入的delay是毫秒之内。
我觉得你这个延时主要是v4l2取图缓存和显示缓存的缘故。不过说实话SDL库我们不太懂，怎样优化需要您自己研究一下。
至于相机采集后，您这边程序直接导入算法的话，延迟肯定要比这种测量方法小的多。

@israel_bolo
I see some warnings, but no errors. could you please continue and see if it compiles?

@otca
The camera is now successfully probed, which confirms that both the I2C and the enable pin are functioning correctly.
The current issue indicates that the I2C control channel is working, but the MIPI CSI data channel is not. Please verify whether the MIPI CSI interface pin assignment on this carrier board matches that of the official Orin NX development board.
The Device Tree Source (DTS) should be adjusted accordingly to match the carrier board.

@stiwana Sorry, we do have this in mind, but no actual plans or projected time of completion at this time.

@阿坎德 你好，有进一步的结果吗？

@veye_xumm now it works, thanks

@阿坎德 看现象感觉是内存泄漏。这个demo现在基本用途上是客户做一下出图验证和聚焦。所以只是个简单的例子。回头我们会安排时间跟进一下。

@megan
First of all, please go to the following link to update the driver of mvcam. The latest driver will set the lane number in the camera according to the configuration of the DTS.

https://github.com/veyeimaging/rk35xx_veye_bsp/tree/main/linux/drivers/kernel_v5.10

Secondly, the DTS needs to be configured according to your RK3588 board, and I can't do it directly for you. You can make analogies and modifications by referring to the DTS of the camera originally supported by your board.

@limengfei 可能是dts设置不正确。请根据你的硬件原理图正确设置dts。

@izaak In fact, in the Video streaming mode, it is impossible to achieve an exposure time of 1 second. I conducted a test in the 2 - lane raw8 format, as follows:

pi@raspberrypi:~/mvcamera/mv_tools_rpi $ ./mv_mipi_i2c_new.sh -w fps 0.01 w fps 0.01 pi@raspberrypi:~/mvcamera/mv_tools_rpi $ ./mv_mipi_i2c_new.sh -r fps r fps is 3.48 fps pi@raspberrypi:~/mvcamera/mv_tools_rpi $ ./mv_mipi_i2c_new.sh -w metime 1000000 w manual exp time is 1000000 us pi@raspberrypi:~/mvcamera/mv_tools_rpi $ ./mv_mipi_i2c_new.sh -r exptime r current exptime is 287129 us

It can be achieved in Pulse Trigger Mode. However, due to the excessive accumulated dark current, the signal noise will be very high. I don't think it has practical value.

@joe The RAW-MIPI-AR0234M is a MONO camera that has no color information and therefore does not support the UYVY data format. Please refer to this link below:
https://wiki.veye.cc/index.php/Mv_series_camera_appnotes_4_jetson#Configuring_global_variables

@veye_xumm 是的 非阻塞缓存后写图，已经解决了。抓拍的图像基本模拟一次就有一张抓拍，因为闪一次时间是1ms 所以只抓到一张也是合理的。

使用脚本可以，感谢

@aike 我们对接rk平台的经验中，只有一个厂家的板子遇到了给你这个完全一样的问题，就是用上次帖子中的方法解决的。该问题原因是此板厂提供的linux内核，调用veyecam2m_enum_frame_sizes时没有正确的传入fse->code参数，导致参数验证失败。

如果你这个方法解决不了问题，建议打开驱动中的DEBUG_PRINTK宏定义，查找一下哪个函数返回报错了。
如果只是打开这个宏定义不够，需要你手动在这些函数返回错误的地方增加打印信息，进行查找报错处。

@mv

For third-party carrier boards on the Orin Nano or Orin NX platform, we have indeed encountered issues with MIPI signal transmission quality. There isn't a good solution at the moment, so we recommend switching to an official NVIDIA board.

@veye_xumm Thank you, graet news

@paul-chang
这个只是一个提示信息，不影响驱动加载和后续测试的。
3f3fb167-a8b9-4a18-b05f-ced9201766d0-image.png
你可以看看有没有生成/dev/video0,如果没有，看看dtb是不是用对了。

@jihua 我们在rk平台没有遇到过这个问题。看ecc2这个错误提示的话，像是mipi信号劣化了。

@jgg You are welcome, will improve the wiki doc.

线下沟通，已解决。
解决方案是，在取图之前，通过system函数直接调用如下脚本：

export I2C_BUS=7 export WIDTH=2432 export HEIGHT=500 export FPS=50 v4l2-ctl -d /dev/v4l-subdev2 --set-ctrl roi_x=0 v4l2-ctl -d /dev/v4l-subdev2 --set-ctrl roi_y=548 media-ctl -d /dev/media0 --set-v4l2 '"m00_b_mvcam '"$I2C_BUS"'-003b":0[fmt:Y8_1X8/'"$WIDTH"'x'"$HEIGHT"'@1/'"$FPS"']'

然后再取图。
如果直接用mv_mipi_i2c,sh配置roi，那么相机内的参数直接被修改，这个数值无法实时反映到v4l2驱动层中。造成参数的错乱。