• DepthAI
  • Calibration of modular OAK-D camera : Capture failed, unable to sync images!

Dear Experts,

We are rehearsing the calibration for our OAK FFC 4P + AR0234 based stereo camera and got following issue :

Capture failed, unable to sync images! Fix the argument minSyncTimestamp or (-mts). Set to: 0.05

Knowing that we only connected the FSIN pad on the FFC cable to the TRIGGER pad on the camera boards without configuring any GPIO for the frame/hardware sync. yet. Is that the cause ?

Best Regards,
Khang

We were finally able to calibrate our camera and get the noisy depth as below :

Will need to work more on the hardware sync.

    Hi l4es ,
    Do you trigger HW sync via GPIOs? What's the timestamp diff between left/right frame?

    • l4es replied to this.

      Hi erik.

      What's the timestamp diff between left/right frame?

      DepthAI version: 2.22.0.0
      DepthAI path: /home/hiep/anaconda3/envs/depthai/lib/python3.9/site-packages/depthai.cpython-39-x86_64-linux-gnu.so
      Connected cameras: ['CAM_A', 'CAM_B', 'CAM_C', 'CAM_D']
      -socket CAM_A : IMX477 4056 x 3040 focus:fixed - COLOR
      -socket CAM_B : AR0234 1920 x 1200 focus:fixed - COLOR MONO
      -socket CAM_C : AR0234 1920 x 1200 focus:fixed - COLOR MONO
      -socket CAM_D : AR0234 1920 x 1200 focus:fixed - COLOR MONO
      Cam name: {'CAM_A': 'IMX477', 'CAM_B': 'AR0234', 'CAM_C': 'AR0234', 'CAM_D': 'AR0234'}
      USB speed: SUPER


      left: 0:00:02.132263
      right: 0:00:02.132279
      camd: 0:00:02.142987


      left: 0:00:02.198767
      right: 0:00:02.198782
      camd: 0:00:02.209492


      left: 0:00:02.298522
      right: 0:00:02.298536
      camd: 0:00:02.309247


      left: 0:00:03.861366
      right: 0:00:03.861381
      camd: 0:00:03.83883


      We could see that left/right were software-synced in sub-ms but not the camd ?!?!

      Do you trigger HW sync via GPIOs?

      No, we haven't triggered HW sync via GPIOS yet. We followed your example but we haven't succeeded to sync the camd to left/right pair and the cameras seemed to freeze (waiting for the triggering pulses I guess). Note that I manually wired the FSIN2&FSIN3 with FSIN1&FSIN4 together instead of configuring the FSIN_MODE_SELECT pin in software.

      Update : I realized that your example was for generating a 1ms pulse on IO41 (FSIN_4LANE=FSIN1&FSIN4) which was connected to IO13 (COM_AUX_IO2 = FSIN2&FSIN3). I doubt that IO13 held the pulse to always LOW. If it is the case, how should I configure it correctly or modify the hardware by removing some resistor ?

      There seems to be an on-going ticket about the sync of AR0234 on OAK FFC 4P here : luxonis/depthai-python883

      • erik replied to this.

        Hi l4es ,
        I see, if it's a FW limitation then it would be best to continue the conversation on the ticket itself.

        • l4es replied to this.

          Hi erik,
          Thank you. Hope that it will be update soon.

          I've been seeing the sync error when calibrating my ar0234 cameras on my 3P board, they're arducam modules that don't have the extra pins connected to the module but instead exposed on the ribbon cable. Am I likely setting the same issue and should I follow the same steps to add external sync?

            kneave Yep that would be my guess, for stereo pairs you'd very much need hardware syncing, and current arducam AR0234 FFC boards unfortunately don't connect those lines.

            • l4es replied to this.

              Thanks Erik, I'll look at enabling that. I'm currently fighting the IMU which is no longer being detected so I'm not having much luck. By the time I get it working I may know it better than you do! 🤣

              • erik replied to this.

                Hi kneave ,
                Are you already on latest develop branch of depthai? As there were some issues with detection.

                I've been working from Main but switching now, pip has been giving me timeouts all day so it's taking a long time to switch. :/

                No luck with develop on my Windows PC, it wouldn't even detect the device was plugged in. Trying it on my Pi based robot now, though having to build depthai-python and the core as pip is proving useless for me today. If it still doesn't work, I'll start a new topic.

                Hi erik

                What did you mean by "and current arducam AR0234 FFC boards unfortunately don't connect those lines." ? Does it mean that even the OAK FFC AR0234 board disposes the TRIGGER pad, it is not physically connect to the sensor? And wiring the TRIGGER pad on the OAK FFC AR0234 board side to the FSIN pad on the FFC cable side has no effect ?

                Sorry no I mean they don't connect that line, they only dispose the TRIGGER on the PCB, so folks need to manually wire it.