I am experiencing troubles in make it work imx477 with OAK-4P-POE. The problem happens within the Arducam ircut (`https://www.uctronics.com/camera-modules/camera-for-raspberry-pi/high-quality-camera-raspberry-pi-12mp-imx477/arducam-12mp-imx477-ir-cut-filter-auto-switch-camera-for-jetson-nano-xaviernx-b0274.html` ), used in this blog article ( ). Result of our tests: * OAK - Using two camera, test without additional rpi-adapter: No-camera detect * OAK- Test within a single camera: No-camera detect * OAK -Test with a official RPI HQ IMX477 camera: Camera works * CM4 board : Test of of 477p arducam ir-cut: They work good as expected Then we performed tests within the official RPI-Adapter that is declared working for both OAK-4P-POE and imx477. We see that the chain of FFC and adapter limit the number of connected lanes to 2. We test on OAK : * OAK with official RPI-adapter (just 2 lane result connected) `` help="List frame metadata (seqno, timestamp, exp, iso etc). Can also toggle with `\`") `` `DepthAI version: 2.29.0.0` `DepthAI path: /home/marco/oaz/depthai/depthai/lib/python3.12/site-packages/depthai.cpython-312-x86_64-linux-gnu.so` `Enabled cameras:` ` cama : color` ` camd : color` `CAM: cama` `CAM: camd` `Connected cameras:` ` -socket CAM_A : IMX477 4056 x 3040 focus:auto - COLOR` ` -socket CAM_D : IMX477 4056 x 3040 focus:auto - COLOR` `USB speed: UNKNOWN` `IR drivers: []` `Cam: cama camd [host | capture timestamp]` `FPS:[14442C1011B736D200] [192.168.178.76] [1732638482.801] [host] [warning] Monitor thread (device: 14442C1011B736D200 [192.168.178.76]) - ping was missed, closing the device connection` `FPS: 0.00| 0.00 0.00| 0.00 Communication exception - possible device error/misconfiguration. Original message 'Couldn't read data from stream: 'camd' (X_LINK_ERROR)'` `Exiting cleanly` Then we check again any possible reference and we find the following information that let us think can impact on our test: Arducam make difference between imx477 base version and 477p ( ) We find three types: 1. IMX477 2. 477M The 477M and IMX477 have exactly the same optical size and pixels, and the frame rate at different resolutions is also the same. We designed the Mini Size version to solve the the problem of taking more space. 3. 477P we see that there is no advice on same resolution and frame rates, so I suppose that arducam make a modified firmware and change minimal parameter. Cause we are on the 3rd case we think that there are some minimal change related to this customization realized by arducam. The 477p cameras on rpi cm4 results in: > > `$ libcamera-vid --list-camera` > > > > `Available cameras` > > > > `-----------------` > > > > `0 : imx477 [4056x3040 12-bit RGGB] (/base/soc/i2c0mux/i2c@1/imx477@1a)` > > > > ` Modes: 'SRGGB10_CSI2P' : 1332x990 [120.05 fps - (696, 528)/2664x1980 crop]` > > > > ` 'SRGGB12_CSI2P' : 2028x1080 [50.03 fps - (0, 440)/4056x2160 crop]` > > > > ` 2028x1520 [40.01 fps - (0, 0)/4056x3040 crop]` > > > > ` 4056x3040 [10.00 fps - (0, 0)/4056x3040 crop]` > > > > `1 : imx477 [4056x3040 12-bit RGGB] (/base/soc/i2c0mux/i2c@0/imx477@1a)` > > > > ` Modes: 'SRGGB10_CSI2P' : 1332x990 [120.05 fps - (696, 528)/2664x1980 crop]` > > > > ` 'SRGGB12_CSI2P' : 2028x1080 [50.03 fps - (0, 440)/4056x2160 crop]` > > > > ` 2028x1520 [40.01 fps - (0, 0)/4056x3040 crop]` > > > > ` 4056x3040 [10.00 fps - (0, 0)/4056x3040 crop]` Any solutions?

Ok, found right cable. We already did it and no detection. > I retest it now with 1 Ircut on CAMD: > > CAM is power and i can listen clearly the ir-cut filter switch if i cover fotoresistor. > But it does not detect any cam > > [upl-image-preview uuid=0112bdc1-734e-48b1-bde1-06942119f331 url=https://discuss.luxonis.com/assets/files/2024-11-29/1732897592-300984-image.png alt={TEXT?}] > > > `$ depthai cam_test` > > `` /home/pop/PY/.venv/lib/python3.12/site-packages/depthai_cli/cam_test.py:109: SyntaxWarning: invalid escape sequence '\`' `` > > `` help="List frame metadata (seqno, timestamp, exp, iso etc). Can also toggle with `\`") `` > > `DepthAI version: 2.29.0.0` > > `DepthAI path: /home/pop/PY/.venv/lib/python3.12/site-packages/depthai.cpython-312-x86_64-linux-gnu.so` > > `No cameras found!`

Unable to use arducam 477p (ircut version) in OAK-4P-POE.

oaz

I am experiencing troubles in make it work imx477 with OAK-4P-POE.

The problem happens within the Arducam ircut (https://www.uctronics.com/camera-modules/camera-for-raspberry-pi/high-quality-camera-raspberry-pi-12mp-imx477/arducam-12mp-imx477-ir-cut-filter-auto-switch-camera-for-jetson-nano-xaviernx-b0274.html ), used in this blog article ( https://discuss.luxonis.com/blog/4922-oak-cameras-for-ndvi-perception ).

Result of our tests:

OAK - Using two camera, test without additional rpi-adapter:
No-camera detect
OAK- Test within a single camera:
No-camera detect
OAK -Test with a official RPI HQ IMX477 camera:
Camera works
CM4 board : Test of of 477p arducam ir-cut:
They work good as expected

Then we performed tests within the official RPI-Adapter that is declared working for both OAK-4P-POE and imx477.
We see that the chain of FFC and adapter limit the number of connected lanes to 2.
We test on OAK :

OAK with official RPI-adapter (just 2 lane result connected)
help="List frame metadata (seqno, timestamp, exp, iso etc). Can also toggle with `\`")

DepthAI version: 2.29.0.0

DepthAI path: /home/marco/oaz/depthai/depthai/lib/python3.12/site-packages/depthai.cpython-312-x86_64-linux-gnu.so

Enabled cameras:

cama : color

camd : color

CAM: cama

CAM: camd

Connected cameras:

-socket CAM_A : IMX477 4056 x 3040 focus:auto - COLOR

-socket CAM_D : IMX477 4056 x 3040 focus:auto - COLOR

USB speed: UNKNOWN

IR drivers: []

Cam: cama camd [host | capture timestamp]

FPS:[14442C1011B736D200] [192.168.178.76] [1732638482.801] [host] [warning] Monitor thread (device: 14442C1011B736D200 [192.168.178.76]) - ping was missed, closing the device connection

FPS: 0.00| 0.00 0.00| 0.00 Communication exception - possible device error/misconfiguration. Original message 'Couldn't read data from stream: 'camd' (X_LINK_ERROR)'

Exiting cleanly

Then we check again any possible reference and we find the following information that let us think can impact on our test:
Arducam make difference between imx477 base version and 477p (https://docs.arducam.com/Raspberry-Pi-Camera/Native-camera/12MP-IMX477/#imx477-camera )

We find three types:

IMX477
477M
The 477M and IMX477 have exactly the same optical size and pixels, and the frame rate at different resolutions is also the same. We designed the Mini Size version to solve the the problem of taking more space.
477P
we see that there is no advice on same resolution and frame rates, so I suppose that arducam make a modified firmware and change minimal parameter.

Cause we are on the 3rd case we think that there are some minimal change related to this customization realized by arducam.

The 477p cameras on rpi cm4 results in:

$ libcamera-vid --list-camera

Available cameras

-----------------

0 : imx477 [4056x3040 12-bit RGGB] (/base/soc/i2c0mux/i2c@1/imx477@1a)

Modes: 'SRGGB10_CSI2P' : 1332x990 [120.05 fps - (696, 528)/2664x1980 crop]

'SRGGB12_CSI2P' : 2028x1080 [50.03 fps - (0, 440)/4056x2160 crop]

2028x1520 [40.01 fps - (0, 0)/4056x3040 crop]

4056x3040 [10.00 fps - (0, 0)/4056x3040 crop]

1 : imx477 [4056x3040 12-bit RGGB] (/base/soc/i2c0mux/i2c@0/imx477@1a)

Modes: 'SRGGB10_CSI2P' : 1332x990 [120.05 fps - (696, 528)/2664x1980 crop]

'SRGGB12_CSI2P' : 2028x1080 [50.03 fps - (0, 440)/4056x2160 crop]

2028x1520 [40.01 fps - (0, 0)/4056x3040 crop]

4056x3040 [10.00 fps - (0, 0)/4056x3040 crop]

Any solutions?

jakaskerl

oaz
The variant we have in the office is a little bit different (w/o CS mount,..) but the underlying sensor and CBA board are the same.
I tested with 4P (r5m1e5) --UC 796 cable -- UC517 (imx 477 - though detected as 577.)
I have no issues with streaming.

You shouldn't need a RPI adapter since the cable is 22pin to 26pin which should work out of the box. Can you test that?

EDIT: from tests on 4P-POE, it only works on 4L sockets. I get the same error as you on CAMB and C.

Thanks,
Jaka

oaz

Ok, found right cable.
We already did it and no detection.

I retest it now with 1 Ircut on CAMD:

CAM is power and i can listen clearly the ir-cut filter switch if i cover fotoresistor.
But it does not detect any cam

$ depthai cam_test

/home/pop/PY/.venv/lib/python3.12/site-packages/depthai_cli/cam_test.py:109: SyntaxWarning: invalid escape sequence '\`'

help="List frame metadata (seqno, timestamp, exp, iso etc). Can also toggle with `\`")

DepthAI version: 2.29.0.0

DepthAI path: /home/pop/PY/.venv/lib/python3.12/site-packages/depthai.cpython-312-x86_64-linux-gnu.so

No cameras found!

oaz

oaz Can it conflict with calibration in flash written with other camera?
Do I have to delete calibration?

oaz

Official RPI camera via uc-796 +rpi-cam-adapter on camD

Lets do a mixed test:

Then on CamD the official HQcam via adapter and on CamA the ir-cut.

.venv/lib/python3.12/site-packages/depthai_cli/cam_test.py:109: SyntaxWarning: invalid escape sequence '\`'

help="List frame metadata (seqno, timestamp, exp, iso etc). Can also toggle with `\`")

DepthAI version: 2.29.0.0

DepthAI path: /home/pop/PY/OAK/.venv/lib/python3.12/site-packages/depthai.cpython-312-x86_64-linux-gnu.so

Enabled cameras:

camd : color

CAM: camd

Connected cameras:

-socket CAM_D : IMX477 4056 x 3040 focus:fixed - COLOR

USB speed: UNKNOWN

IR drivers: []

[]

Cam: camd [host | capture timestamp]

FPS: 0.00| 0.00 [14442C1011B736D200] [192.168.178.76] [1732900441.987] [host] [warning] Monitor thread (device: 14442C1011B736D200 [192.168.178.76]) - ping was missed, closing the device connection

FPS: 0.00| 0.00 Communication exception - possible device error/misconfiguration. Original message 'Couldn't read data from stream: 'camd' (X_LINK_ERROR)'

Exiting cleanly

[2024-11-29 18:14:17.169] [depthai] [error] Device with id 14442C1011B736D200 has crashed. Crash dump logs are stored in: /home/pop/PY/OAK/.cache/depthai/crashdumps/08db14601bbb81545245df81c59cf79051a7538b/crash_dump.json - please report to developers.

on crash_dump i pick some relevant part :
RTEMS_FATAL_SOURCE_ASSERT
Fatal error on MSS CPU: trap: 00, address: 00000000",
"errorSource": "RTEMS_FATAL_SOURCE_INVALID_HEAP_FREE",

Finally again the ir-cut alone on camA:
$ depthai cam_test

/home/pop/PY/OAK/.venv/lib/python3.12/site-packages/depthai_cli/cam_test.py:109: SyntaxWarning: invalid escape sequence '\`'

help="List frame metadata (seqno, timestamp, exp, iso etc). Can also toggle with `\`")

DepthAI version: 2.29.0.0

DepthAI path: /home/pop/PY/OAK/.venv/lib/python3.12/site-packages/depthai.cpython-312-x86_64-linux-gnu.so

No cameras found!

HERE some information about bootloader and library version
DepthAI Library Version: 2.29.0.0

Found device '192.168.178.76', MxId: '14442C1011B736D200', State: 'BOOTLOADER'

DeviceInfo(name=192.168.178.76, mxid=14442C1011B736D200, X_LINK_BOOTLOADER, X_LINK_TCP_IP, X_LINK_MYRIAD_X, X_LINK_SUCCESS)

/home/pop/PY/OAK/print_info.py:22: DeprecationWarning: Use constructor taking 'UsbSpeed' instead

with dai.Device(dai.Pipeline(), infos[0]) as device:

MxId: 14442C1011B736D200

USB Speed: UsbSpeed.UNKNOWN

Connected Cameras: []

Available Camera Sensors: {}

Product Name: OAK-FFC-4P-POE, Board Name: NG2093

DepthAI Library Version: 2.29.0.0

jakaskerl

oaz
I'm slightly confused now. You want to make IR-CUT work but is not detected in any case? (talking about the small camera, that is connected to the CAM_A socket). The HQ RPI camera is the one with a CS lens, on CAM_D. This one works with the RPI adapter, as expected. Please correct me if I'm wrong. The link you sent is for the CS mount camera which is the one on the CAM_D. Or were lenses swapped?

If that is the case, can you check if cable is correctly connected on the IRCUT. In some variants the connectors are top down, which is often a source of confusion.

Thanks,
Jaka

oaz

jakaskerl I just swapped m12-to-cs with a pure m12 with ir-cut but it is the same camera as in CamD.
All the cams have been sucessfully tested with raspberry, but no any success test on oak-4p-poe.
We did double check to cables.
The only way we can see camera are detected (but not working, just cam detect) is by using rpi-adapter like u can see with the CAMA in this photo.

jakaskerl

oaz
Ok, I'll sync with the team on the best course of action.

Thanks,
Jaka

oaz

jakaskerl Let me be more precise. The cam with m12 (CAMA) is ir-cut camera ardulink, it is not exact the same we see on CAMA (the official RPI), but it is the correct arducam that they sell with cs mount .
thank you Jak

Luxonis-Alex

Hi @oaz , sorry for the late reply here...
I checked the camera module PCB from our blog post (https://discuss.luxonis.com/blog/4922-oak-cameras-for-ndvi-perception), and it looks a bit different, as in this picture:

See the unpopulated IC in the bottom-right corner of the image above.
On your module that chip is present, and I'm afraid it's an MCU interposed between the I2C bus on the 22-pin connector and the I2C of the sensor, implementing the PiVariety protocol (not sure if 477P stands for this), that unfortunately we don't support at this time with depthai:

Would it be possible to post a close-up photo?

oaz

Luxonis-Alex
I confirm the presence of the MCU with green arrow.

That is not nice for us, considering that we ordered the camera following the link from the NDVI post.
So there is no viable solution to make our setup work on oak?

Luxonis-Alex

oaz
I believe it may be possible to bypass the MCU, even without desoldering it, by a PCB rework to move just the resistors in that group of 6 for rerouting the I2C bus SCL and SDA pins to go to the sensor directly.

The pins on the connector are (from https://www.arducam.com/raspberry-pi-camera-pinout/):

...
19 - GND
20 - SCL
21 - SDA
22 - VCC (3.3V)

Possible to post a close-up photo of your camera module PCB?

oaz

We got 2 instead of 4 soldered resistor.

Are the resistors to be placed zero ohm?

In that case we can easily test it.
Can u confirm we just need to make the two resistor in the vertical of the T (group of 6) .,

Luxonis-Alex

oaz
Yes, I think these 2 resistors at the bottom (that are populated on our variant, and missing on yours) are the link resistors between connector/host bus and sensor bus. You can definitely try populating those with zero-Ohm, or just some wires or solder blobs.
I'm not sure if that's enough though, but worth giving it a try afterwards to see if the camera is recognized/streaming.

If things still don't work, it might be possible that the MCU is interfering, and best would be to remove it as well as in our variant, but that's more difficult. Another option could be to keep the MCU in reset, if we can identify its pinout. Possible to check the markings on the MCU?
Unfortunately we don't have the design docs of the Arducam camera module, and I also don't have these modules with me to give it a try.