Radxa Cubie A7Z – Radxa 4K Camera detected but no image output (need help + FFmpeg HW acceleration setup)

Hi everyone,

I recently bought a Radxa Cubie A7Z along with the official Radxa 4K Camera.

I followed the available setup instructions, including enabling the correct device tree overlay and loading the driver manually:

sudo modprobe vin_v4l2

After that, the camera appears correctly when I run:

v4l2-ctl --list-devices

It also shows up as a selectable camera inside tools like Cheese.

However, the problem is:

:red_exclamation_mark: The camera produces no image (black screen / no frames)

  • There is no video output in Cheese

Has anyone successfully used the Radxa 4K camera on the Cubie A7Z?

Are there known limitations, missing overlays, or drivers that must be patched?

I bought Cubie A7A. Just now, I want to buy a Radxa 13M 214. What the store service said shocked me a lot, saying that A7A‘s hardware can support this camera, but software can not support. They are doing the adaption. WTF! Please sale it after all tests is passed.

Really?! ..

Well than, i’m in the same boat then you guys.

Radxa 4k is detected but nothing works.

Is this really shocking? Probably Linux on 90% of ARM SBCs doesn’t support the hardware fully.

1 Like

It wouldn’t be if it wasn’t explicitly said on their products page “accessories for” and there it is, 3 different camera that doesn’t work but hey it’s nothing that weird.

Maybe only android supports it?

i was not able to detect the 4k camera until i use the r2 (rsdk) Pre-release Generic A733 Image

1 Like

i tried with r2(rsdk) image, still same issue,

Cubie A7A + Camera 4K + radxa-a733_bullseye_cli + cubie-a7a-radxa-camera-4k-415.dtbo
dmesg: imx415_mipi read retry 3, sensor_detect 0x0
v4l2-ctl: Bayer formats listed but no stream

Any suggestions please

Please refer to the tutorial on the Radxa documentation site for the commands used with the A7Z accessories.
The R2 version already supports these cameras: imx219, imx415, and imx214.
If you still encounter issues, you may want to check the ribbon cable.

Literally bought 2 camera (4k & the 13 AF camera) bought even extra 31pin ribbon cable.. it will only detect the camera but not output anything out of it.

Just out of curiosity, I don’t have the board, but it might help to know:

  • kernel version (uname -ra)
  • cli or kde?
  • try before update and try after update
  • Which command you are using

Documentation says you should use gstreamer:

DISPLAY=:0 gst-launch-1.0 v4l2src device=/dev/video1 en-awisp=1 en-largemode=1 ! video/x-raw,format=NV12,width=3840,height=2160,framerate=24/1  ! xvimagesink


xvimagesink renders the frame in XWindow , I’m not a KDE user but i think you need Xwayland to be able to display something.

Try to log a few more info from gstreamer;

DISPLAY=:0 GST_DEBUG_FILE=log.txt GST_DEBUG=4 gst-launch-1.0 v4l2src device=/dev/video1 en-awisp=1 en-largemode=1 ! video/x-raw,format=NV12,width=3840,height=2160,framerate=24/1  ! xvimagesink

post the log.txt here, don’t forget to format it.

So :slight_smile:

-Linux radxa-cubie-a7z 5.15.147-14-a733 #14 SMP PREEMPT Mon Jan 19 10:03:49 UTC 2026 aarch64 GNU/Linux
-kde

  • did and done, first thing i did and installing the overlays for the camera.

-Pretty much the cmd given, such has : gst-launch-1.0 v4l2src device=/dev/video1 io-mode=dmabuf num-buffers=10 ! video/x-raw,format=NV12,width=1920,height=1080 ! videoconvert ! jpegenc ! filesink location=~/test.jpg

and some variant of it. Resulting to the same issue.

for example :

radxa@radxa-cubie-a7z:~$ DISPLAY=:0 GST_DEBUG_FILE=log.txt GST_DEBUG=4 gst-launch-1.0 v4l2src device=/dev/video0 en-awis
p=1 en-largemode=1 ! video/x-raw,format=NV12,width=3840,height=2160,framerate=24/1 ! xvimagesink
Setting pipeline to PAUSED …
Pipeline is live and does not need PREROLL …
Pipeline is PREROLLED …
Setting pipeline to PLAYING …
New clock: GstSystemClock
[ISP]>>>>>>>>>>>>>>>>>>>> ISP VERSION INFO <<<<<<<<<<<<<<<<<<<
IPCORE: ISP602
branch: libisp-dev
commit: 665dc2e9613716c25abb6b64af145a73b989271e
date : Wed Dec 18 15:26:13 2024 +0800
author: [email protected]

[ISP]video device name is vin_video0
[AWIspApi]: F:awIspGetIspId, L:104, video0 → isp0
[ISP]STITCH_2IN1_LINNER
[ISP]ISP Set Sync Mode = 0x3, isp_sync_mode = 0x20003

And it hangs there.

Or by running it via the rsetup to test video0

(gst-plugin-scanner:8035): GLib-GObject-WARNING **: 12:59:27.194: cannot register existing type ‘GstRtpSrc’

(gst-plugin-scanner:8035): GLib-GObject-CRITICAL **: 12:59:27.195: g_type_add_interface_static: assertion ‘G_TYPE_IS_INSTANTIATABLE (instance_type)’ failed

(gst-plugin-scanner:8035): GLib-CRITICAL **: 12:59:27.195: g_once_init_leave: assertion ‘result != 0’ failed

(gst-plugin-scanner:8035): GStreamer-CRITICAL **: 12:59:27.195: gst_element_register: assertion ‘g_type_is_a (type, GST_TYPE_ELEMENT)’ failed

(gst-plugin-scanner:8035): GLib-GObject-WARNING **: 12:59:27.195: cannot register existing type ‘GstRtpSink’

(gst-plugin-scanner:8035): GLib-GObject-CRITICAL **: 12:59:27.195: g_type_add_interface_static: assertion ‘G_TYPE_IS_INSTANTIATABLE (instance_type)’ failed

(gst-plugin-scanner:8035): GLib-CRITICAL **: 12:59:27.195: g_once_init_leave: assertion ‘result != 0’ failed

(gst-plugin-scanner:8035): GStreamer-CRITICAL **: 12:59:27.195: gst_element_register: assertion ‘g_type_is_a (type, GST_TYPE_ELEMENT)’ failed

Setting pipeline to PAUSED …
Pipeline is live and does not need PREROLL …
Pipeline is PREROLLED …
Setting pipeline to PLAYING …
New clock: GstSystemClock
[ISP]>>>>>>>>>>>>>>>>>>>> ISP VERSION INFO <<<<<<<<<<<<<<<<<<<
IPCORE: ISP602
branch: libisp-dev
commit: 665dc2e9613716c25abb6b64af145a73b989271e
date : Wed Dec 18 15:26:13 2024 +0800
author: [email protected]

[ISP]video device name is vin_video0
[AWIspApi]: F:awIspGetIspId, L:104, video0 → isp0
[ISP]STITCH_2IN1_LINNER
[ISP]ISP Set Sync Mode = 0x3, isp_sync_mode = 0x20003

It seems to be the same error in both cases when trying to capture the image. Sorry, I can’t help beyond saying “triple check the ribbon cable at both ends”. The contacts are very thin and need to be aligned.

I know some time has passed, but my sense is that the Cubie+4K camera (possibly, any camera) is not yet viable. I’ve just managed to crash my kernel with v4l2-ctl -d /dev/video0 --all. Can’t really pinpoint why, but it’s repeatable. This is with the new r7/trixie CLI version, fully updated.

This is consistent with the other reactions: the camera is detected, it’s listed, but the drivers are just not ready for actual use.

It’s a shame – we need this, so it effectively rules the A733 Cubie’s out for us, and yet there’s a lot to like about the specifications. But if we can’t get a solid OS on it (and this one is not solid, filled with “not for production” kernel messages) – customers need to know in advance.

Hi morungos,

In allwinner’s kernel you can’t use v4l2-ctl in a standard way you do with your desktop computer.

You either use gstreamer or build a v4l2 user program which initializes the v4l2 input correctly.

The error reported here is possibly due to asking gstreamer to stream 4k from /dev/video0 and en-largemode=1 , a common mistake.

Correct usage:

/dev/video0 and en-largemode=0 for 1920x1080

/dev/video1 and en-largemode=1 for 3840x2160

1 Like

For what it’s worth, here’s the dmesg log:

Nov 13 21:07:39 radxa-cubie-a7a kernel: sunxi:vin:[INFO]: camera is on, close dfs

Nov 13 21:07:39 radxa-cubie-a7a kernel: sunxi:vin:[INFO]: camera is on, close dfs

Nov 13 21:07:39 radxa-cubie-a7a kernel: sunxi:vin:[INFO]: camera is on, close dfs

Nov 13 21:07:39 radxa-cubie-a7a kernel: sunxi:vin:[INFO]: camera is on, close dfs

Nov 13 21:07:52 radxa-cubie-a7a kernel: sunxi:vin:[INFO]: camera is on, close dfs

Nov 13 21:07:52 radxa-cubie-a7a kernel: sunxi:vin:[ERR]: sensor set dv timings error!

Nov 13 21:07:52 radxa-cubie-a7a kernel: sunxi:vin:[ERR]: sensor dv timings cap error!

Nov 13 21:07:52 radxa-cubie-a7a kernel: sunxi:vin:[WARN]: sensor g_pixelaspect fail!

Nov 13 21:07:52 radxa-cubie-a7a kernel: sunxi:vin:[WARN]: sensor g_pixelaspect fail!

Nov 13 21:07:52 radxa-cubie-a7a kernel: sunxi:vin:[ERR]: v4l2 sub device scaler get_selection error!

Nov 13 21:07:52 radxa-cubie-a7a kernel: sunxi:vin:[ERR]: v4l2 sub device scaler get_selection error!

Nov 13 21:07:52 radxa-cubie-a7a kernel: sunxi:vin:[ERR]: v4l2 sub device scaler get_selection error!

Nov 13 21:07:52 radxa-cubie-a7a kernel: sunxi:vin:[ERR]: v4l2 sub device scaler get_selection error!

Nov 13 21:07:52 radxa-cubie-a7a kernel: sunxi:vin:[ERR]: v4l2 sub device scaler get_selection error!

Nov 13 21:07:52 radxa-cubie-a7a kernel: sunxi:vin:[ERR]: v4l2 sub device scaler get_selection error!

Nov 13 21:07:52 radxa-cubie-a7a kernel: sunxi:vin:[ERR]: v4l2 sub device scaler get_selection error!

Nov 13 21:07:52 radxa-cubie-a7a kernel: sunxi:vin:[ERR]: v4l2 sub device scaler get_selection error!

Nov 13 21:07:53 radxa-cubie-a7a kernel: sunxi:vin:[WARN]: v4l2 sub device g_parm fail!

Nov 13 21:07:53 radxa-cubie-a7a kernel: Unable to handle kernel NULL pointer dereference at virtual address 00000000000000b0

Nov 13 21:07:53 radxa-cubie-a7a kernel: Mem abort info:

Nov 13 21:07:53 radxa-cubie-a7a kernel:   ESR = 0x0000000096000004

Nov 13 21:07:53 radxa-cubie-a7a kernel:   EC = 0x25: DABT (current EL), IL = 32 bits

Nov 13 21:07:53 radxa-cubie-a7a kernel:   SET = 0, FnV = 0

Nov 13 21:07:53 radxa-cubie-a7a kernel:   EA = 0, S1PTW = 0

Nov 13 21:07:53 radxa-cubie-a7a kernel:   FSC = 0x04: level 0 translation fault

Nov 13 21:07:53 radxa-cubie-a7a kernel: Data abort info:

Nov 13 21:07:53 radxa-cubie-a7a kernel:   ISV = 0, ISS = 0x00000004, ISS2 = 0x00000000

Nov 13 21:07:53 radxa-cubie-a7a kernel:   CM = 0, WnR = 0, TnD = 0, TagAccess = 0

Nov 13 21:07:53 radxa-cubie-a7a kernel:   GCS = 0, Overlay = 0, DirtyBit = 0, Xs = 0

Nov 13 21:07:53 radxa-cubie-a7a kernel: user pgtable: 4k pages, 48-bit VAs, pgdp=000000010b7d2000

Nov 13 21:07:53 radxa-cubie-a7a kernel: [00000000000000b0] pgd=0000000000000000, p4d=0000000000000000

Nov 13 21:07:53 radxa-cubie-a7a kernel: Internal error: Oops: 0000000096000004 [#1] SMP

I am not using gstreamer at all. I am simply using v4l2 standard utilities to query the standard v4l2 ioctls. v4l2-ctl is a standard v4l2 user program; arguably it is the standard user program. This is not an error: it is taking out the whole kernel. That cannot be my problem, ever. The backtrace shows this is a kernel memory error.

I think the original question asked by the OP was, more or less: has anyone ever got a Cubie and 4K camera to work? I don’t see one positive response. I would love to have one, but I don’t.

In other words, you can’t use v4l2-ctl without crashing the kernel for the reason on previous post.

If you just want to know if the camera works on Cubie A733, the answer is yes. I have my Radxa Cubie A7S (Cubie A733) with imx415 and imx214 working just fine.

If you ask me how, i tell you, use gstreamer or build an application , search my posts here in the forum.

You can try this:

sudo gst-launch-1.0 v4l2src device=/dev/video1 en-awisp=1 en-largemode=1 ! video/x-raw,format=BGR,width=3840,height=2160 ! jpegenc ! multifilesink location=file_3840x2160_%d.jpg

If you just want to stream videos using WebRTC you can follow these instructions here: https://youtube.com/shorts/yKWK1oklRYw?is=c0vjMd7VV512wfCv

That is useful, and I will concede that it is possible to spin up a gstreamer pipeline that can generate an MJPEG. However….

  1. I get a ton of noisy warnings and apparent errors along the way, log will follow
  2. After a couple of successful runs, the exact same command now simply hangs, even after a reboot

I repeat: this is not a usable software setup yet. It could be, but it is not. And this is with the stock device, latest stock OS, and its stock accessory. I have used cameras on at least six devices recently; this is the only one that is even close to this problematic. This is not convincing me that this is viable for developing code, let alone production.

My aim is to get libcamera running as a V4L2 front end, but I could live with standard V4L2. I have standard V4L2 C code that works, but it is in development still. gstreamer is not an option because I need UVC and uvcsink is in an awful state. And I am not going to even start working on my code if I cannot get something as basic as capture using standard tools working repeatably and consistently.

Log:

[ISP]>>>>>>>>>>>>>>>>>>>> ISP VERSION INFO <<<<<<<<<<<<<<<<<<<
IPCORE: ISP602
branch: libisp-dev
commit: 665dc2e9613716c25abb6b64af145a73b989271e
date  : Wed Dec 18 15:26:13 2024 +0800
author: <[email protected]>
--------------------------------------------------------

[ISP]video device name is vin_video1
[AWIspApi]: F:awIspGetIspId, L:104, video1 --> isp1
[ISP]STITCH_2IN1_LINNER
[ISP]ISP Set Sync Mode = 0x3, isp_sync_mode = 0x20003
[ISP]tdm_rx device name is sunxi_tdm_rx.0
[ISP]open isp device[0] success!
[ISP]get isp_ctx from /mnt/isp1_1920_1080_30_0_imx415_mipi_ctx_saved.bin success!!!
[ISP]detect /mnt/ success.
[ISP]find imx415_mipi_1920_1080_30_0 [imx415_mipi_1080p_isp_cfg] isp config
[ISP]create isp0 server thread!
[ISP_ERR]isp_dev_start, line: 717,unable to subscribe to tdm event: Inappropriate ioctl for device (25).
[ISP]tdm_rx device name is sunxi_tdm_rx.1
[ISP]open isp device[1] success!
[ISP_WARN]VIDIOC_VIN_GET_SENSOR_OTP_INFO return error:Invalid argument
[ISP]ISP1 OTP Disable
[ISP]get isp_ctx from /mnt/isp1_1920_1080_30_0_imx415_mipi_ctx_saved.bin success!!!
[ISP]detect /mnt/ success.
[ISP]find imx415_mipi_1920_1080_30_0 [imx415_mipi_1080p_isp_cfg] isp config
[ISP]create isp1 server thread!
[ISP_ERR]isp_dev_start, line: 717,unable to subscribe to tdm event: Inappropriate ioctl for device (25).
[ISP_ERR]events_loop, line: 162,isp0 event select timeout
[ISP_ERR]events_loop, line: 162,isp0 event select timeout
[ ... repeat at infinitum ... ]

Oh one extra thing: the video was helpful, but it’s worth noting that bullseye is now EOL by debian, so the standard images <= r6 no longer update unless you manually edit the sources –> archive.debian.org. So I’ve used r7. I don’t know if r6 is better, but I’m not going to lock into a system without security patches.