Arducam IMX219 slot 5 not working on Jetson AGX Orin

Nvidia Jetson Cameras
1

Hey everyone, I’m running into a stubborn issue with my Arducam IMX219 Multi-Camera Kit on a Jetson AGX Orin. I’m trying to get all 6 cameras working in a 2x3 surveillance grid, but Camera 5 (sensor-id=5) refuses to cooperate.

All the cameras and cables work fine when I test them individually on other ports, so the hardware itself seems okay — it’s just that specific port/slot that won’t play ball.

I followed the official quick start guide here:
https://docs.arducam.com/Nvidia-Jetson-Camera/Nvidia-Jetson-Orin-Series/NVIDIA-Jetson-AGX-Orin/Quick-Start-Guide/

Here’s the Python script I’m using with OpenCV and GStreamer:

import cv2
import numpy as np

def create_pipeline(sensor_id, width=640, height=480, framerate=30):
    return (
        f"nvarguscamerasrc sensor-id={sensor_id} ! "
        f"video/x-raw(memory:NVMM), width={width}, height={height}, format=NV12, framerate={framerate}/1 ! "
        "nvvidconv ! video/x-raw, format=BGRx ! "
        "videoconvert ! video/x-raw, format=BGR ! appsink"
    )

def main():
    sensor_ids = [0, 1, 2, 3, 4, 5]
    cams = [cv2.VideoCapture(create_pipeline(sensor_id), cv2.CAP_GSTREAMER) for sensor_id in sensor_ids]
    for i, cap in enumerate(cams):
        if not cap.isOpened():
            print(f"Camera {sensor_ids[i]} could not be opened")
            return
    print("All cameras successfully opened. Press 'q' to exit.")
    while True:
        frames = []
        for i, cap in enumerate(cams):
            ret, frame = cap.read()
            if not ret or frame is None:
                print(f"Camera {sensor_ids[i]} is not providing a frame")
                frames.append(np.zeros((300, 400, 3), dtype=np.uint8))
            else:
                frames.append(cv2.resize(frame, (400, 300)))
        if len(frames) != len(sensor_ids):
            break
        row1 = np.hstack(frames[:3])
        row2 = np.hstack(frames[3:])
        combined_frame = np.vstack([row1, row2])
        cv2.imshow("Surveillance View (2x3 Grid)", combined_frame)
        if cv2.waitKey(1) & 0xFF == ord('q'):
            break
    for cap in cams:
        cap.release()
    cv2.destroyAllWindows()

if __name__ == "__main__":
    main()

It always fails right at the cap.isOpened() check for sensor-id=5. The other five cameras open and stream without any issues. Has anyone else seen a single port on this board just not work? Anything else I should try before assuming the board is defective?

2

Thanks for the detailed report and code. Before we jump to any hardware conclusions, let’s isolate whether the problem is in your Python/OpenCV pipeline or something lower-level. Can you try hitting sensor-id=5 directly with a bare GStreamer pipeline? This bypasses OpenCV entirely and goes straight to the display:

SENSOR_ID=5
FRAMERATE=21
width=4032
height=3040
gst-launch-1.0 nvarguscamerasrc sensor-id=$SENSOR_ID ! "video/x-raw(memory:NVMM),width=$width,height=$height,framerate=$FRAMERATE/1" ! nvvidconv ! nveglglessink -e

This runs at full resolution (4032x3040 @ 21fps) so we’re testing the full camera pipeline. Let me know what happens — does the stream open or do you get an error?

3

Alright, I went deep on isolation testing and here’s what I found:

  1. I physically swapped cameras around between ports — every single camera module works perfectly in slots 1 through 4, but no camera works in slot 5. Doesn’t matter which camera I put there.

  2. I also swapped the FFC ribbon cables around between all slots. Every cable works fine in slots 1-4. Slot 5 fails regardless of which cable is connected.

  3. Ran the GStreamer command you provided — works beautifully with every camera I plug into any slot… except when I use sensor-id=5 (the port in question). Same failure every time.

So after testing every combination of camera + cable + slot, the failure follows the slot, not the camera and not the cable. At this point I’m pretty convinced port/slot 5 on the camera board itself is defective. I’ve been at this for a while now and I can’t see any other variable left to test. What’s the best path forward here?

4

Good isolation work — that rules out a lot. Before we call it definitively though, one more thing: have you tried using the exact same FFC cable that you know works in another slot and moving it to slot 5? I know you mentioned swapping cables around, but I want to be absolutely sure the FFC in slot 5 isn’t the culprit here. Sometimes cables can look fine but have a subtle bend or contact issue that only manifests in a particular orientation.

5

Yep, already did that — I took a cable that works 100% in slot 1, moved it to slot 5, and the same failure occurs. Then moved it back to slot 1 and it works again. I repeated this with every cable I have. All of them work perfectly in every slot except slot 5. So unfortunately the FFC cables are definitely ruled out — the issue is tied to that specific port on the board no matter what camera or cable combination I use.

6

Thanks for confirming. At this stage we’ve systematically eliminated every variable:

  • All 6 camera modules work individually on other ports ✓
  • All FFC cables work individually on other ports ✓
  • The Python/OpenCV pipeline works for sensors 0-4 ✓
  • The raw GStreamer test works for sensors 0-4 but fails on sensor-id=5 ✓
  • Swapping known-good cameras and cables into slot 5 consistently fails ✓

This really only points to one conclusion: there’s a hardware fault on the multi-camera board at port/slot 5. It could be a bad MIPI connector, a trace issue, or a problem with the mux/switch IC servicing that port — but regardless, it’s not something fixable through software or configuration changes.

I’ve sent you details privately about how to proceed with getting a replacement board. Once you have a new unit, the same setup should work across all 6 ports without any of these issues. Let us know how it goes once the replacement is in hand.