Picamera2 capture_array hangs with Arducam Hawk-eye 64MP on Raspberry Pi 4

Raspberry Pi Cameras
1

Hi everyone, I have a Raspberry Pi 4 with an Arducam Hawk-eye 64MP camera and I’m hitting a hang when running the ArduCam/picamera2_examples/SavePng.py example. It freezes at:

RGB888 = picam2.capture_array(“main”)

The raspberrypi/picamera2/examples/capture_full_res.py script hangs the same way. A Raspberry Pi Camera Module V3 runs the exact same code and saves still images without any issue.

Here’s the code I’m testing:

import time
from picamera2 import Picamera2, Preview

picam2 = Picamera2()
picam2.start_preview(Preview.QTGL)

preview_config = picam2.create_preview_configuration()
capture_config = picam2.create_still_configuration()
picam2.configure(preview_config)

picam2.start()
time.sleep(2)

picam2.switch_mode_and_capture_file(capture_config, “test_full.jpg”)

It seems like any capture-related call after switching the configuration just hangs. My theory is that the internal capture method is waiting for a callback from the camera that never comes, maybe a missing callback in the Hawk-eye camera body or driver after a reconfigure.

There’s a GitHub issue that looks like the same setup (Pi 4 + Arducam Hawkeye 64MP) but it was closed without a real resolution: [BUG]picam2.switch_mode_and_capture_file freezes my program and camera · Issue #470 · raspberrypi/picamera2 · GitHub

Environment: Raspberry Pi 4, Arducam Hawk-eye 64MP, picamera2.

2

Can you make sure the Pi is set up with the latest Bullseye OS? The getting-started guide is here: https://www.raspberrypi.com/documentation/computers/camera_software.html#getting-started

With Bullseye, you can use the libcamera command to preview the camera directly. Also worth noting: with Bullseye there’s no longer a need to encrypt the IC before using the camera, it should work out of the box.

3

Thanks. I went back and tested a few more times: SavePng.py with the Hawk-eye always stops at RGB888 = picam2.capture_array(“main”), and capture_full_res.py hangs the same way. Swapped in the Camera Module V3 and the still image saves successfully every time. So it definitely looks specific to the Hawk-eye camera/driver combination.

4

Apologies, that previous reply about Bullseye setup wasn’t aimed at your actual issue, you can disregard it. I’ve reproduced the freeze you’re describing, and there is a temporary solution available. The problem comes down to camera driver/kernel version compatibility: the camera driver has to match the kernel version you’re currently running. Could you run:

uname -a

and post the output? Once I know your kernel version, I’ll send over the matching camera driver.

5

Here you go:

Linux raspberrypi 6.1.21-v8+ #1642 SMP PREEMPT Mon Apr 3 17:24:16 BST 2023 aarch64 GNU/Linux

6

Thanks. I’ve sent you a matching camera driver package through private message, it’s the low-speed PDAF IMX519 kernel driver built for kernel 6.1.21. Please check your messages and then:

  1. Comment out dtoverlay=64-mp in config.txt first.
  2. Extract the package.
  3. cd Release/
  4. ./install_driver.sh
  5. Reboot the Raspberry Pi so the driver takes effect.
  6. Run capture_full_res.py to check whether the freeze is gone.
7

That did it. I extracted and installed the driver from the package you sent, rebooted, and now I can switch the configuration and save files without any issue, the freeze is gone.

One thing though: the driver still looks like a provisional version rather than an official release. Is there any approximate timeline for when the official driver update will be released?

8

Great to hear it’s working now. To sum up for anyone else hitting this: the capture freeze on the 64MP Hawk-eye happens when the camera driver doesn’t match the running kernel, and the workaround is to install the matching low-speed PDAF IMX519 driver (after commenting out dtoverlay=64-mp) and reboot.

Regarding an official release, there’s no confirmed timeline yet. The driver I sent is still provisional, so keep using it for now and I’ll post an update when an official release is available.