Dual MAX96712 8-Ch FHD Camera Porting via Argus API: Feasibility & CPU Load

Robotics & Edge Computing Jetson Systems Jetson Thor
1

Hello

We are designing a system to port 8 channels of FHD Bayer cameras using two MAX96712 deserializers. We intend to convert the Bayer format to YUV or RGB using the system’s internal pipeline.

The target device is Nvidia Thor T5000

Could you please clarify the following?

  • Can the Argus API fully support an 8-channel concurrent configuration with dual deserializers?

  • What would be the estimated CPU load for this setup? (Assuming the hardware ISP handles the debayering/format conversion).

Any guidance, documentation, or reference designs for dual-MAX96712 setups would be highly appreciated.

Thanks,

4

Dear @spshin ,
Are you using DRIVE AGX Thor or Jetson AGX Thor?

6

Hello SivaRamaKrishnaNV

Thank you for prompt response!

The AP is Jetson AGX Thor.

Software is jetpack

Thank you.

Best regard.

7

Get the detail information from below link.

8

Hello Shane

There is some unclear explaination.

I ported GMSL 8 cameras which have internal ISP. to thor

At this time, I would like to port GMSL 8 cameras which don’t have internal ISP so I want to use Argus API

Could you explain that is possible?

If possible, coud you estmate cpu load?

Thank you.

Best regard.

9

Argus must include internal ISP pipeline.

For your case you can use MMAPI but argus API.

Thanks

10

Hello Shane.

Thank you so much your prompt respose!

As you mentioned, I can use cusom isp alogrithm on userspace not Argus.

Is it correct?

If I do isp algorithm for 8 cam in userspace, It would have big cpu load.

I think we should use the cam which has internal isp block,instead of using non isp cam.

Could you let me know that how much cpu load may be used?

Thank you.

Best regard.

11

What’s I mean is you can use external ISP by v4l2-ctl without jetsonISP that bounding to Argus.

For the CPU usage that depend on the sensor output size.

Thanks

13

Hello Shane

Thank you for prompt response!

Is there HW accelated v4l2-ctl ISP interface on T5000?

If I have to use software ISP interface. I should make a decision to use tha cam with internal ISP

Thank you.

Best regard.!

14

No, but support DMA

15

Hello Shane

Thank you so much your prompt respose

Best regard.!

18

Hello Shane

Is there any way to port my 8 cams to nvarguscamerasrc so tht I can use HW accelated ISP algorithm like below

gst-launch-1.0 nvarguscamerasrc sensor-id=0 ! ‘video/x-raw(memory:NVMM), width=1920, height=1080’ ! nvvidconv ! xvimagesink

19

Check the programing guide to implement the kernel driver and device tree.

Thanks

20

Hello Shane

Thank you for your prompt response!!!

I would like to check below 2 items.

  1. Is it possible 8 gmsl raw10 cams to use nvarguscamerasrc?
  2. I ported 8 gmsl yuv cams to use v4l2src
    1. would it be almost same to port to use nvarguscamerasrc as to use v4l2src

Best regard.

Thank you.

21
  1. Yes
  2. Do you mean the driver or any others?
22

Hello Shane

Thank you for prompt response~~!!!

2.Do you mean the driver or any others?

→ I mean driver

Thank you.

Best regard!

23

Yes, it’s almost the same in kernel sensor driver.

Thanks

24

Thank you Shane

Please close case ~~!!!

25

Hello Shane

Before close case, I would like to ask one more question.

I have previously successfully ported a RAW10 camera sensor to a V4L2 device and verified data acquisition using GStreamer’s v4l2src plugin.

Now, I am planning to switch the pipeline to use nvarguscamerasrc for hardware-accelerated processing. Could you please clarify if nvarguscamerasrc can be used immediately after a standard V4L2 porting, or if additional modifications are required?

Specifically, I would appreciate guidance on what changes are necessary regarding the kernel driver (such as mandatory V4L2 controls), Device Tree configurations, and the ISP (Camera Core) calibration files to make it compatible with nvarguscamerasrc.

Thank you for your support.

Best regards

26

Yes, with correct device tree config should be able to run the nvarguscamerasrc.